Direct Methanol Fuel Cell systems in portable electronics A ...

Direct Methanol Fuel Cell systems in portable electronics

A metrics-based conceptualization approach

Bas FLIPSEN

i



Proefschrift

ter verkrijging van de graad van doctor

aan de Technische Universiteit Delft,

op gezag van de Rector Magnificus prof.ir. K.C.A.M. Luyben,

voorzitter van het College voor Promoties,

in het openbaar te verdedigen op dinsdag 14 december 2010 om 15:00 uur

door

Sebastiaan Frederik Johan FLIPSEN

ingenieur Luchtvaart en Ruimtevaart

ingenieur Advanced Industrial Design Engineering

geboren te Vlierden.

ii

Dit proefschrift is goedgekeurd door de promotoren:

Prof.dr.ir. J.C. Brezet

Prof.dr. C. Spitas

Copromotor: Dr. J.S.M. Vergeest

Samenstelling promotiecommissie:

Rector Magnificus voorzitter

Prof.dr.ir. J.C. Brezet Technische Universiteit Delft, promotor

Prof.dr. C. Spitas Technische Universiteit Delft, promotor

Dr. J.S.M. Vergeest Technische Universiteit Delft, copromotor

Prof.dr.ir. Geraedts Technische Universiteit Delft

Prof.dr. A.H.M.E. Reinders Technische Universiteit Delft, Universiteit Twente

Prof.dr.ir. A. van Keulen Technische Universiteit Delft

Dr. P. Li The University of Arizona, Tucson



Bas Flipsen

Thesis Delft University of Technology, Delft, the Netherlands

Product Engineering research group, publication nr. 2

ISBN 978-90-5155-069-6

Cover design by René Smeets, Gerard Nijenhuis and Bas Flipsen

Layout by René Smeets KineticVision

Printed by VSSD in Delft (the Netherlands)

Distributed by Product Engineering

E: [email protected]

T: +31-15-2789398

Copyright © 2010 by Bas Flipsen. All rights reserved. No part of this publication may be

reproduced, stored in a retrieval system, or transmitted, in any form or by any means,

electronic, mechanical, photocopying, recording, or otherwise, without the prior written

permission of the author.

iii

Contents

Contents ..............................................................................................iii

Summary .............................................................................................. v

Samenvatting....................................................................................... ix

Symbols, acronyms and definitions ...................................................xiii

1. Introduction......................................................................................1 1.1 Problem definition .......................................................................2 1.2 Goal of the thesis.........................................................................3 1.3 Research methodology & structure of the thesis...........................3

2. Alternative power sources in context.................................................7 2.1 Fields of application.....................................................................7 2.2 Power and energy systems compared .........................................14 2.3 Conclusions...............................................................................27

3. Power source design and selection in practice, state of the art ........29 3.1 Approach...................................................................................30 3.2 Tools and Methods in the Field ..................................................31 3.3 Case study review......................................................................39 3.4 Conclusions...............................................................................45

4. Research Questions.........................................................................49 4.1 Answers to the initial research question.....................................49 4.2 Redefined problem definition .....................................................51 4.3 Main research questions............................................................52 4.4 Methodology ..............................................................................53

5. From batteries to DMFC power systems...........................................55 5.1 Introduction to batteries ............................................................55 5.2 DMFC explained ........................................................................61 5.3 Comparison...............................................................................66 5.4 Discussion.................................................................................84 5.5 Conclusions...............................................................................87

6. The user and their choice for a power source...................................91 6.1 Approach...................................................................................92 6.2 User acceptance and their differentiating properties...................94

iv

6.3 Results...................................................................................... 97 6.4 Discussion .............................................................................. 100 6.5 Conclusions ............................................................................ 104

7. Introduction to different orders of modeling..................................107 7.1 Orders of modeling .................................................................. 108 7.2 Differentiating properties......................................................... 112 7.3 Conclusions ............................................................................ 117

8. First order model: a heuristic approach to modeling a DMFC power

source ..........................................................................................119 8.1 Case study of a MP3 player...................................................... 120 8.2 Design of the DMFC hybrid system.......................................... 128 8.3 Design of the ‘scaled’ DMFC hybrid system.............................. 138 8.4 Discussion of the design, redesign and the model .................... 148 8.5 Modified heuristic models for a DMFC hybrid power system..... 153 8.6 Evaluation of the newly developed model ................................. 156 8.7 Conclusions ............................................................................ 159

9. Second order model: database-driven metrics-based design............161 9.1 “Automated design” approach.................................................. 162 9.2 Presentation of the optimization algorithm............................... 169 9.3 Component selection: MP3 player ............................................ 174 9.4 First test run: initial algorithm ................................................ 180 9.5 Second test run: improved algorithm ....................................... 187 9.6 Third test run: applying the evolutionary algorithm ................. 190 9.7 Discussion on the results ........................................................ 196 9.8 Conclusions ............................................................................ 201 9.9 Recommendations ................................................................... 203

10. Conclusions.................................................................................205 10.1 Answer to the main question ................................................... 205 10.2 Scientific and technological relevance ...................................... 213 10.3 Generalization of the models.................................................... 215 10.4 Recommendations for future research ..................................... 216

References.........................................................................................221

A Modeling of the open cell Voltage for a DMFC.................................231

B Mathematica files..........................................................................237 B.1 Initial algorithm....................................................................... 237 B.2 Evolutionary algorithm ............................................................ 248 B.3 Input tables (dbase) ................................................................. 254

Acknowledgements...........................................................................259

Curriculum Vitae ...............................................................................261

v

Summary

It is impossible to imagine life today without portable electronics like the

laptop computer, PDA, tablet pc’s and cell phone. All these products are

equipped with batteries which grant them grid independence and all-round

portability. The numbers are huge, for instance in 2009 more than 180

million smart phones have been sold, and this will be more than 400 million

by 2014. Connectivity to the internet and more functionality make these

devices long for longer run times and higher power use. At the moment

intensively used smartphones have to be recharged quite often, and an

alternative power system could be a solution to increase the time in between

recharges. This thesis concerns about such an alternative and studies the

techno-economical feasibility of Direct Methanol Fuel Cell (DMFC) systems

applied in portable electronic devices. To test the feasibility different

mathematical models are developed, which can be used during the

conceptual phase of the design phase. Instead of a conventional qualitative

approach, a quantitative comparison and selection is pursued, which may

count as ground-braking in this domain.

Based on an explorative research, data for different alternative power

systems in the application field of portable electronic devices are identified.

This data is then used to compare different power sources and energy

carriers with each other on three basic properties: power, energy and costs.

The DMFC is pinpointed as a power system which could outperform the

lithium-ion battery. Potentially, this system can store two to seven times the

amount of energy contained in current lithium-ion rechargeable batteries. To

test the feasibility of DFMC systems in portable applications a literature

research is conducted into available tools and methods. During the

conceptualization phase of the design process different tools are found which

can be used to simply identify the opportunity of different power systems.

Other more enumerative tools can be used during later phases of the design

vi

process, like embodiment and engineering where specification of the device

is already laid out. Tools which could identify a power system but which can

also used for designing a power system during the specification and

conceptualization phase are not found.

To research this lack of design tools during conceptualization the main

research question (RQ) is defined. The RQ deals with the feasibility of DMFC

systems for portable electronics and how to identify the opportunities of

DMFC systems in early phases of the design process. Differentiating

properties between DMFCs and rechargeable batteries are researched via a

techno-economical comparison and a research into the willingness of the

user to buy a fuel-cell powered cell-phone and laptop computer. In

comparison with the lithium-ion battery, a DMFC is an energy dense system,

especially applicable in low-power/long endurance applications like smoke

detectors. Based on a simple zero-order model the DMFC could be a factor

two to three smaller and six times more lightweight in this application field.

In the application fields as the laptop computer and cell-phone the DMFC

should be extended with an extra intermediate accumulator, leading to a so-

called DMFC-hybrid system, or in short the ‘DMFC system’. The zero-order

model used to identify the application field is based on basic energy and

power densities of three parts of the DMFC system, the fuel cell stack, the

bill of products and fuel tank. To increase accuracy a first-order

mathematical model is proposed. The model is constructed by designing a

DMFC system for a flash-drive MP3 player, the Samsung YP-Z5F. This model

breaks the DMFC system into five parts, the three presented earlier, plus the

empty space and the intermediate accumulator. The coefficients, efficiencies

and constants derived during this research by design project are based on

heuristic data from literature and prototype designs. To test the model, two

commercially available DMFC systems (the SFC Jenny and SFJ Efoy 2200)

are evaluated. Results from this validation show that the first-order model

estimates the volume of the five parts correctly, but approximations are

presented over a wide-range of well to not-so good system designs. Well-

engineered systems, like the Jenny, are positioned in the optimal estimate

zone, and conventionally designed systems (the Efoy) find itself in the middle

of the estimates. This type of models is well useable for the product designer

when a low amount of data is available, and can be designated as rules-of-

vii

thumb. Still, it is advised to test more commercially available DMFC systems

to fine-tune the constants, coefficients and efficiencies.

To research the design and specifications of the DFMC system more in depth

a second-order model is proposed. The model consists of an algorithm which

(i) selects and designs components, (ii) optimizes the selected component-set

to a minimizing objective function, and (iii) visualizes the results in a 3D

design. To test the algorithm a program is written in Mathematica which

evaluates multiple structural variants of the selected components.

Component selection is based on the objective function taking volume,

weight and costs into account. Once a fitting component-set has been

chosen, the algorithm places the components in 3D space and optimizes the

space for minimized volume. Two algorithms have been applied to optimize

for volume: the Random Walk (RW) method and an Evolutionary Algorithm

(EA), which guides the solution to a local minimum. The RW method is very

enumerative and converging to an optimum takes up a long time (more than

8 hours). The evolutionary approach reached convergence within a low

amount of calculations and time (15-30 minutes), and with better

performance. It converges to a lower value of the objective function, and

being quicker it is more useable for testing the feasibility of a DMFC system

applied in a certain application. The algorithm is tested on the previously

described MP3 player, and convergence was reached with a minimal value of

the boundary volume twice as large as the benchmarked battery. The DMFC

system is thus twice as large as the benchmarked lithium-ion battery. The

results also show that weight is less of an issue, because weight has halved

compared to the benchmark. The sales price is difficult to estimate because

not all necessary data is available. Based on the preliminary results the price

will be at least twice as high as the benchmarked battery, making this fuel

cell system not feasible within this application. Improvements in fuel cell

efficiency and a higher concentration of methanol could result in a feasible

solution. The specification of the DMFC system equals that of the

benchmarked battery; however the economical feasibility is still an issue. It

can be expected that the consumer is willing to pay more for improvement in

comfort, as can be seen in the transition from NiMH batteries to the Li-ion

battery. Improvement in weight probably is not enough.

viii

It is recommended to test the algorithm even further by evaluating a wide

range of applications. The database of components should therefore be

updated and extended with more commercially available components. The

initial pricing for the DMFC system is very high, but also the running costs

of methanol are high. This could be of subordinate concern when comfort is

improved, e.g. by increasing run time.

The first order model uses a minimal amount of input data and is therefore

an interesting tool for concept developers. The second order model can be

used to fine-tune the first order model.

Although this thesis is a small step in quantitative testing the feasibility of

concepts, it can be a significant stepping stone for industrial design

engineers. Making use of the strength of computers is finally within the

reach of the concept designer. We only have to grab it.

ix

Samenvatting

Tegenwoordig is een leven zonder draagbare elektronica zoals een laptop,

tablet pc of mobiele telefoon ondenkbaar. Deze producten bevatten allemaal

een batterij die het apparaat draagbaar en onafhankelijk maakt van het

stopcontact. De aantallen zijn immens; in 2009 zijn er 180 miljoen

smartphones verkocht en de verwachting is dat dit in 2014 minstens 400

miljoen worden. Door draadloos internet en andere functies is de batterij van

dergelijke apparaten snel leeg en is er een grote vraag naar langere

gebruikstijden en meer vermogen. Een alternatieve energiebron zou de ‘tijd

tussen opladen’ kunnen verlengen. Deze thesis gaat over dergelijke

alternatieven, en onderzoekt de technisch economische haalbaarheid van

Direct Methanol Brandstofcellen (DMFC) toegepast in draagbare elektronica.

Om de haalbaarheid te testen, zijn er verschillende mathematische modellen

ontwikkeld die gebruikt kunnen worden tijdens de conceptfase van het

ontwerpproces. In plaats van de conventionele kwalitatieve aanpak wordt in

deze thesis een kwantitatieve benadering nagestreefd. Dit zou baanbrekend

kunnen zijn binnen dit domein.

Middels een exploratief onderzoek is veel data gegenereerd over alternatieve

energiebronnen toegepast in kleine apparaten. Deze data is gebruikt om de

verschillende energiesystemen en vermogensgeneratoren met elkaar te

vergelijken op basis van drie eigenschappen: vermogen, energie en kosten.

De conclusie uit dit onderzoek is dat de DMFC de prestaties van de lithium

ion batterij kan benaderen en zelfs overtreffen. Middels een

literatuuronderzoek is gezocht naar gereedschappen en methoden die de

ontwerper kunnen ondersteunen in zijn zoektocht naar een alternatieve

energiebron. Er zijn enkele gereedschappen gevonden die de mogelijkheden

van bepaalde energiebronnen identificeren tijdens conceptualisatie. Voor

latere fasen in het ontwerpproces zijn er vooral numerieke tools gevonden. In

deze fasen zijn de eisen aan het product al vastgesteld en de specificaties

x

van het product bekend. Er zijn geen tools gevonden welke naast het

identificeren van de mogelijkheid ook nog bijdraagt aan de conceptualisatie

ervan, tijdens de conceptfase van het ontwerpproces.

Om dit gemis te ondervangen is de hoofdvraag opgesteld. De hoofdvraag

betreft de haalbaarheid van DMFC systemen in draagbare elektronica, en

hoe de kansen van dit systeem in een vroeg stadium geïdentificeerd kan

worden. Door middel van twee onderzoeken zijn de verschillen tussen de

DMFC en de oplaadbare batterij uitgezocht. Allereerst is er een technologisch

economisch vergelijk gemaakt tussen de twee energiebronnen. Vervolgens is

er een gebruikerstest uitgevoerd waarin de bereidheid van de consument is

getest om de stap te maken van batterij naar DMFC. In vergelijk met de

lithium ion batterij heeft de DMFC een hoge energiedichtheid, en is deze

vooral geschikt voor apparaten met laag vermogen en lange gebruiksduur

specificaties. Gebaseerd op een simpel nulde orde model is gebleken dat de

DMFC twee tot drie keer kleiner en tot zeker zes keer lichter kan zijn dan de

lithium ion batterij. Wanneer de DMFC wordt toegepast in een laptop of

mobiele telefoon moet een extra batterij worden toegevoegd; dit om

piekvermogens op te vangen. De combinatie van DMFC met batterij leidt tot

een DMFC hybride systeem, afgekort ‘DMFC systeem’. Het ontwikkelde nulde

orde model is gebaseerd op de energie- en vermogensdichtheid van de drie

belangrijkste onderdelen van de DMFC, de brandstofcel, de tank en alle

andere componenten zoals de pompen (stuklijst). Om de nauwkeurigheid te

vergroten is dit model verbeterd naar een eerste orde mathematisch model.

Door middel van een design case, de Samsung YP-Z5F MP3 speler, is het

nulde orde model verder uitgewerkt. Het ontstane eerste orde model is uiteen

te rafelen in vijf onderdelen, naast de eerder gemelde drie onderdelen, wordt

nu ook het loze ruimte en de batterij meegenomen. De gebruikte

coëfficiënten, rendementen en constanten zijn afkomstig uit literatuur en

prototype data. Om het model te testen zijn twee commercieel beschikbare

DMFC systemen geëvalueerd, de SFC Jenny en Efoy 2200. De resultaten van

de evaluatie laten zien dat het eerste orde model de volumes van de vijf

onderdelen goed schat. De uitkomst laat een breed spectrum zien van

mogelijke oplossingen met in het uiterste de goed ontworpen DMFC

systemen zoals de SFC Jenny en in het midden van het spectrum de

conventioneel ontwerpen systemen zoals de SFC Efoy 2200. Dit soort

xi

modellen, ook wel vuistregels genoemd, zijn erg bruikbaar voor de

industrieel ontwerper tijdens de conceptfase waarin weinig data (specs)

beschikbaar is. Toch is het te adviseren om het model te fine-tunen door het

model met meer commercieel beschikbare DMFC system te evalueren.

Om de specificaties van een DMFC systeem beter te kunnen ontwerpen is

het tweede orde model ontwikkeld. Dit model bestaat uit een model dat (i)

componenten uit een database selecteert of ontwerpt, (ii) de geselecteerde

componentenset optimaliseert naar de te minimaliseren doelfunctie, en (iii)

het resultaat visualiseert in een 3D ontwerp. Om het tweede orde model te

testen is er een programma geschreven in Mathematica die gebruikt maakt

van een optimalisatiealgoritme. Dit programma evalueert meerdere

structurele varianten uit een geselecteerde componentenset. De

componentselectie is gebaseerd op een doelfunctie, die rekening houdt met

het volume, het gewicht en de prijs van de componentenset. Wanneer er een

geschikte componentenset is gekozen plaatst het algoritme de componenten

ten opzichte van elkaar in een 3D ruimte. Het algoritme minimaliseert

vervolgens het productvolume door de plaatsing van de componenten te

variëren. Twee algoritmes zijn toegepast: de Random Walk (RW) methode en

een Evolutionair Algoritme (EA), welke de oplossing meer naar een lokaal

minimum richt. De RW methode is erg rekenintensief en convergeert pas na

veel berekeningen en tijd (8 uur) naar een minimaal volume. De EA echter

convergeert sneller (15-30 minuten) en zelfs naar een lagere waarde van het

volume. Het algoritme is getest op het ontwerp van een DMFC systeem voor

de eerder gebruikte casestudie, de MP3 speler. Het algoritme convergeert

naar een twee keer zo groot systeem vergeleken met de uitgangssituatie, de

lithium ion batterij. De resultaten laten ook zien dat het systeem twee maal

lichter is dan de batterij. Gewicht zal dan ook geen opstakel vormen voor de

haalbaarheid van het systeem. De prijs van het systeem is moeilijk in te

schatten daar niet alle data beschikbaar is. Gebaseerd op initiële

schattingen zal de prijs twee maal hoger zijn dan de uitgangssituatie. De

haalbaarheid van dit systeem komt hiermee dus in het geding. Verbeteringen

in het brandstofcelrendement en hogere concentraties methanol zullen

bijdragen aan een haalbaar systeem. Bij vergelijkbare energie en

vermogensspecificaties van de batterij, zal het DMFC systeem twee maal

groter en twee maal duurder worden. Uit vergelijkbare transities van de

NiMH batterij naar de Li-ion batterij is gebleken dat de consument wel

xii

bereidt is meer te betalen voor verbeterde prestaties, maar of een

gewichtswinst alleen genoeg is valt te betwijfelen.

Het wordt aanbevolen om het algoritme te testen in een breder

toepassingsgebied. De componentendatabase zou hierbij aangevuld moeten

worden met nieuwe commerciële componenten en regelmatig hernieuwd

moeten worden. De initiële prijs van het brandstofcelsysteem is hoog en de

kosten tijdens gebruik zijn door de prijs van methanol ook hoog. Dit kan,

zoals al eerder is gemeld, van ondergeschikt belang zijn als meer comfort

wordt geboden door bijvoorbeeld een langere gebruikstijd. Het eerste orde

model maakt gebruik van een minimale hoeveelheid inputdata, en is daarom

een interessante tool voor tijdens de conceptontwikkeling. Met behulp van

het tweede orde model kan het eerste orde model verbeterd worden.

Hoewel dit proefschrift een kleine stap is in het kwantitatief testen van

concepten, kan het een belangrijke opstap zijn voor industrieel ontwerpers.

De kracht van computers is eindelijk binnen het bereik van de

conceptontwerper en we hoeven deze kans alleen maar aan te grijpen.

xiii

Symbols, acronyms and definitions

Symbol Definition Dimensions

A Surface area mm2

Acell Active surface area mm2

c Cost €

C Capacity of an accumulator Wh

C Charge or discharge rate A Ah-1, C

ces Empty space coefficient -

cfp Flat pack coefficient -

cm can Canisters’ weight coefficient -

cv can Canisters’ volume coefficient -

cmeoh Methanol to total fuel fraction -

D Design Vector -

E Amount of electric energy Wh

E0 Standard cell potential V

Eday Energy needed for a single day Wh

F Faradays constant C mol-1

F(D) Objective function -

fsp Self Pumping Frequency Hz

g(D) Design constraint -

h Height mm

h(D) Behavioral constraint -

i Current density mA cm2

I Current A

ic Cross-over current density mA cm2

j(D) Discrete value constraint -

l Length mm

m Mass kg

M Molar mass kg mol-1

xiv

m Mass-transfer overvoltage constant mV

M(D) Dimensionless optimization metric -

ma Platinum loading on the anode mg cm-2

m Mass flow g min-1

maccu Accumulator mass kg

mc Platinum loading on the cathode mg cm-2

mfc Fuel cell mass kg

mfc Fuel cell mass kg

mfc-system Fuel cell system mass

mtotal Total mass kg

N Methanol concentration mol L-1

n Mass-transfer overvoltage constant cm2 mA-1

ne Amount of electrons per mole mol-1

P Power W

Pmean Mean power W

Ppeak Peak power W

px Specific power of component x W kg-1

(pρ)x Power density of component x W L-1

Qmax Maximum flow rate mL s-1

R Molar gas constant J K-1 mol-1

rau Auxiliary power to net power ratio -

rcloud Cloud vector consisting or only translation

vectors

-

rr Rotation vector -

rt Neighboring desing vector -

rt Translation vector -

SCE Specific Cost per unit of energy or power € Wh-1

SCP Specific Cost per unit of power € W-1

SPR Self pumping ratio min-1

T Temperature K

ttank Thickness of the tank mm

ux Specific energy of component x Wh kg-1

(uρ)x Energy density of component x Wh L-1

V Volume flow ml min-1

xv

V Volume mm3, L

Vaccu Accumulator volume mm3, L

Vbop BOP volume mm3, L

Vcell Working Cell Voltage V

Vempty Empty space volume mm3, L

Vfc Fuel cell volume mm3, L

VOC Open Cell Voltage V

Vtotal Total volume mm3, L

w Width mm

x Amount of moles used/produced in reaction -

z Number of electrons -

α Radius factor -

αa Charge transfer coefficient at the anode -

αc Charge transfer coefficient at the cathode -

Δgf Molar Gibbs free energy kJ mol-1

Δhf Lower heating value MJ kg-1 MJ kg-

1

Δtcycle Single cycle runtime s

Δtrun Runtime on one charge s

ΔVact Activation losses V

ΔVmass Mass transport losses V

ΔVx-over Cross-over losses V

ΔVΩ Ohmic losses V

ηbop BOP efficiency %

ηDC/DC DC to DC convertor efficiency %

ηfc Fuel cell efficiency %

ηsys System efficiency %

θ Angle of rotation vector -

λ Preference factor -

λstoch Stochiometric ratio at the anode or cathode -

ρ Density kg L-1

xvi

AC Alternating Current

AFC Alkaline Fuel Cell

BIOS Basic Input/Output System

BOP Bill Of Products

CAD Computer Aided Design

CES Cambridge Engineering Selector

CFRP Carbon Fiber Reinforced Plastics

CPU Central Processor Unit

DC Direct Current

DG Distributed Generation

DLC Double Layer Capacitor

DMFC Direct Methanol Fuel Cell

DOD Depth Of Discharge

DOF Degrees Of Freedom

DSM Design Structure Matrix

DVD Digital Video Disk

EA Evolutionary Algorithm

EDLC Electric/Electrochemical Double Layer Capacitor

EHD Electro HydroDynamic

EM Electro Magnetic

EMT Environmental Monitor Test bed

FC Fuel Cell

FEM Finite Element Method

GA Genetic Algorithm

GDL Gas Diffusion Layer

GPS Global Positioning System

HHV Higher Heating Value

xvii

HOGA Hybrid Optimization by Genetic Algorithms

HOMER Hybrid Optimization Model for Electric

Renewables

HS High Strength (steel)

HSDPA High Speed Downlink Packet Access

HT High Tenacity

IC Integrate Circuit

ICE Internal Combustion Engine

JPL Jet Propulsion Lab

LCC Life Cycle Costs

LCD Liquid Crystal Display

LED Light Emitting Diode

LHV Lower Heating Value

Li-ion Lithium ion

Li-poly Lithium polymer

LPG Liquefied Petroleum Gas

MCFC Molten Carbonate Fuel Cell

MEA Membrane Electrode Assembly

MEMS Micro Electro Mechanical Systems

MP3 MPEG layer 3

MPEG Moving Picture Experts Group

MS Microsoft

MTI Mechanical Technology, Incorporated

NASA National Aeronautics and Space Administration

NiCD Nickel Cadmium

NiMH Nickel Metal Hydride

NiTi Nickel Titanium

NPC Net Present Cost

xviii

OLED Organic Light Emitting Display

PAFC Phosphoric Acid Fuel Cell

Pb Lead

PC Personal Computer

PCB Printed Circuit Board

PDA Personal Digital Assistant

PEFC Polymer Exchange Membrane Fuel Cell

PEM FC Polymer Exchange Membrane Fuel Cell

POWER Power Optimization for Wireless Energy

Requirement

PV PhotoVoltaic

RF Radio Frequency

RQ Research Question

RTG Radioisotope Thermoelectric Generator

RW Random Walk

SPR Self Pumping Ratio

SFC Smart Fuel Cell

SOC State Of Charge

SOFC Solid Oxide Fuel Cell

SPR Self Pumping Ratio

TE Thermo Electric

TU Delft University of Technology Delft

VOC Open Cell Voltage

WiFi Wireless Fidelity

WIMS-ERC Wireless Integrated Microsystems Engineering

Research Center

WMA Windows Media Audio

xix

Power system A combination of a power source with an energy container

Power source Electric power generator

Energy Electric energy contained

Cell Electrochemical unit

Battery A device that converts the chemical energy contained in

its active materials directly into energy by means of an

electrochemical oxidation-reduction (redox) reaction. A

battery refers to one or more cells connected in series or

parallel.

1

1 Introduction

The past 10 years more portable electronics have entered our lives. All these

electronic devices are powered by a battery and mainly the rechargeable

Lithium Ion battery. In 2009 more than 180 million smart phones have been

sold and this will be more than 400 million in 2014 [1]. Connectivity to the

internet makes these devices long for longer run times and higher power use

[2-4]. At the moment intensively used smart phones have to be recharged

quite often, and alternative power systems could be a solution to increasing

the run time.

This Chapter is an introduction to the research executed over the past 6

years and years before. During my professional career before the academic

career at TU Delft, I have been working on alternative power systems and the

application of it in portable products. At TNO industrial Technology we have

developed product concepts using a range of power systems, from wind-up

‘electric’ toothbrushes [5] to fuel cells in portable electronics [6, 7], where we

were in search for application fields for fuel cells [8] to finding applications

for butane/propane canisters [9]. In that period I noticed a need from the

product designer (mainly academia) for knowledge-based design, indication

the direction of the design process, making us of mathematical modeling

during early phases of the design process. This thesis is about this early

modeling, making use of mathematical calculations (with and without

computers) in early phases of the design process, applied in the field of

alternative power systems.

The problem and the goal of the research are defined and Section 1.1 and

1.2. The Chapter finishes with the structure of the thesis, which can be used

as the story line for this thesis, Section 1.3.

2

1.1 Problem definition In the design and engineering of risky and expensive products, like aircraft,

the principle of ‘first time right’ is one of the key factors in surviving the

competition. Predicting the feasibility of these products are thus of great

importance during every phase of the design process. For every design phase,

different tools as structural analysis, design for performance and costs are

the main drivers. Analytical tools and heuristic models are developed and

used to predict the performance, physical properties and life-cycle costing as

good as possible [10]. For less expensive products, the designer has to revert

to more trial-and-error methods and reading up on literature. Time, and

thus costs, is limiting the designer into getting in the deep of new

technologies. For the consumer industry alternative power systems are thus

less visible and designers are not known with the state-of-the-art

developments in the field.

Successful compact design of a portable electronic product depends on

proper selection of components and a sound matching of the power system.

A large part of the product is defined during early phases in the design

process, during which issues are addressed and important choices are made.

During these phases the product design engineer wants to size the

application and has to make important and sometimes irreversible choices,

which in a later stage are difficult and often costly to alter (principle of

Pareto). On the other hand the selection of the power system often takes

place late in the design process [11]. Generally the designer takes this

component of the shelf when the application is already designed or even

prototyped. “Power supplies are thus frequently an afterthought” [12]. Tools

and methods to make select a power system in early phases of the design

process are limited available. Because of the time-constraints, low visibility

of alternatives, and the lack of knowledge the designer looks merely at

primary or secondary batteries. Thus opportunities for short-term but

especially long-term developments in portable electronics are in that way

overlooked. The availability of analytical tools to support the designer in

making this choice is not available.

3

1.2 Goal of the thesis Three problems have been addressed in the previous section, (i) a problem

initiated by the growth of power-hungry electronic devices longing for longer

run times, (ii) the visibility of alternative power systems when designing

portable electronic products, and (iii) the limited availability of tools and

methods to make a selection of the power system or even design a power

system during the preliminary design phases.

In this thesis the field of alternative power systems is explored in search for

the existing alternatives for the rechargeable battery (mainly lithium-ion)

used in portable electronics. This thesis is thus started with an initial

research question:

“Which power systems can compete with the commonly used rechargeable

lithium based battery in the application field of portable electronic devices,

and how can a systematic approach help product designers select power

systems during the preliminary design phase?”

After definition of the context in Chapter 2 and analyzing the state of the art

tools and methods in Chapter 3, the initial question as described above will

be answered in Chapter 4. In the same chapter the main research question

is defined which is as follows:

“Are direct methanol fuel cell systems feasible for portable electronics and

can we identify the opportunities of DMFC systems in early phases of the

design process?”

The main question and its accompanying sub-questions will be answered in

the following Chapters 5 to 10.

1.3 Research methodology & structure of the thesis

To answer the initial research question this thesis will first explore the field

of alternative power systems. A large amount of data about alternative power

systems is acquired during my professional career, and is completed with an

4

exploratory research, described in Part I of this thesis. Part I is about the

context in which the research is conducted. First the application field for

batteries and its alternatives is defined in Chapter 2.1. This chapter will

define the bandwidth in which this research is conducted. Second a

quantitative performance data, physical data and cost data has been

gathered about different power systems which are presented in Chapter 2.2.

The field of power systems can be divided into two parts, the power

generators and the energy carriers. Two power systems (defined as a

combination of a power generator and energy carrier) are indicated as

potentially runner-ups for the commonly used lithium based rechargeable

batteries.

Because the interest is not only on alternative power systems but also on

selection and using mathematical modeling during conceptual design phase,

a literature review has been made on the tools and methods for selecting or

reviewing power systems. In Chapter 3 the design process is elaborated on,

and different tools and methods are pinpointed which can be used during

different phases of the design process.

In Chapter 4 the initial research question will be answered and the research

is more focused on the Direct Methanol Fuel Cell (DMFC) power system in

Part II of the thesis. In Chapter 5 a technological comparison is made of the

DMFC power system and the rechargeable battery, in search for

differentiating properties. Because one of the goals is a systematic approach

towards selecting/designing power systems, the focus in this chapter is on

using metrics for design and evaluation of power systems. In consumer

electronics it is not only about technology, but the user is also involved by

researching their preferences when buying a portable electronic device

(Chapter 6). This chapter results in quantitative data about the preference of

the user, making it possible to use this data for the selection phase of the

design process.

Data acquired during Part II is used in Part III where metric-based selection,

design and concept evaluation is researched. Four orders of modeling are

introduced in Chapter 7 as the basis for the rest of this thesis. Three orders

(zero, first and second-order modeling) are evaluated by means of Design

5

Research. A case-study from literature is used evaluating the zero-order

model and produce the first-order model (Chapter 8). This order of modeling

makes use of simple equations to test the feasibility of DMFC power systems

on three basic properties: volume, weight and costs. To improve accuracy of

the model a second-order model is proposed making use of the computer as

a generator and evaluator of multiple design architectures, or structural

variants (Chapter 9). The second-order of modeling is used as the stepping

stone for the third-order model, not evaluated within this thesis. The thesis

will finish with a concluding chapter where the research question is

evaluated, answered and, when needed, discussed.

To get an instant overview of the research the structure of the thesis

described in Figure 1.

Par

t I

Par

t III

P

art I

I

2. Context

3. Power source design and selection in practice

4. Main research Questions & methodology

5. DFMC as an alternative for lithium ion batteries

6. The user and DFMC systems

7. Approach to a metric-based design methodology

8. First order model, a heuristic approach

9. Second order model, database driven parametric d i

10. Conclusions & recommendations

1. Introduction

Figure 1: Structure of the thesis.

7

2 Alternative power sources in context1

Two of the problems stated in the introduction were growth of electronic

devices which are more power hungry, now and in the near future, and the

low visibility of alternative power systems. To make alternative power

sources more comparable to each other an overview is made of alternatives

to the rechargeable batteries, Section 2.2. This comparison is based on basic

metrics as power, energy, price, weight and volume and has been executed

in the end of 2004.

Before this overview is presented, Section 2.1 gives the definition of “portable

electronic devices”, and thus the limits of the search field in Section 2.2.

Section 2.3 will draw the conclusions about the application field for different

power sources and the feasibility of some power sources in the field of

portable electronic devices. The chapter will finish with Section 2.3 giving an

overview of definitions used in this thesis.

2.1 Fields of application

2.1.1 What are portable electronic devices?

For the research described in this thesis the focus will be on “portable

electronic devices”. We limit the search field physically to products “which

can be carried with one arm for a longer period of time and which are not

voluminous”. Furthermore these devices “should be powered by batteries or

would potentially be interesting to power without electric cords (electronics)”.

1 Parts of this chapter are presented at the International Power Sources Symposium

(2004) and published in the Journal of Power Sources (2006), “Power Sources Compared: the ultimate truth?”162(2).

8

These vague terms, as “to voluminous” and “carried with one arm”, can be

brought back to numbers by looking at the power range of available portable

electronic devices and the energy it has to carry. The products can be

differentiated by power consumption as pictured in Figure 2. The smallest

power module consists of a couple of sensors and light buttons in the range

of 10ths of milli-Watts. The range of flash-memory MP3 players and hard

drive players work within 25 to 250mW and the maximum power

consumption of standard mobile phones and PDA's is around 0.5 to 1.2W

[14]. Smart phones, combining wireless communication with other office

functions have higher power consumption, because of WiFi, GPS and HSDPA

modules, a faster CPU and larger OLED displays [14, 15]. A larger power

consumer is the laptop computer ranging from 8-30W [16].

Figure 2: Power range of different consumer electronics.

Table 1 shows that portable electronic devices can, besides power range, also

differentiate by other parameters like available energy, volume, mass or the

working voltage. The working voltage is often a result of the type and the

number of batteries used. The mass and volume are the result of type of

battery, the power consumption and the runtime, but are limited by the

design and the designer.

9

Table 1: Overview of different consumer products and the differentiating metrics of the battery.

Batt type P (W)

E (Wh)

V (mm3)

m (g)

V (V)

Sensors button 0.010 0.15 700 2.2 1.5–15

Flashdrive MP3 AA 0.025 1.1 5,000 11 1.5

Harddrive MP3 li-ion/poly 0.250 2.9 6,600 13 7.2

Mobile phone li-ion/poly 0.800 3 30,000 50 2.4-7.2

PDA li-ion/poly 1.50 3 11,000 26 3-6

Portable DVD players

li-ion/poly 9–20 50 375,000 - 7.4-12

Laptop computers li-ion/poly 10–30 50 250,000 400 4-14.4

Electric bike Pb/NiMH 500 192-360 - 1500 24-36

In this thesis a distinction is made in product groups by dividing them to

power and capacity. Four groups can be defined by different power range:

1. Micro power: 0.01 to 0.1 W

2. Low power: 0.1 to 10 W

3. Medium power 10 to 100 W

4. High power: 100-1000 W

The same thing can be done with capacity, which can be divided as:

1. Low capacity: 0.1 to 1 Wh,

2. Medium capacity: 1-10Wh,

3. High capacity: 10-100 Wh

4. Ultra high capacity: 100-1000Wh systems.

If power and capacity is combined Table 2 can be made. To prove the

combinations are genuine and other combinations are either rare or not

existing in any product group, a product group was searched for, for every

power-capacity combination. The table shows that most products stay within

a bandwidth of power-capacity and less extremes can be found. The

extremes are found in products with low power and high endurance

applications like the ad-hoc sensor for data-networking and measuring

buoys. The measuring buoy works at powers in the range of 1W and have to

10

work for at least 6 months [17]. These applications are not intended to be

portable by hand and will be left out of the research scope. Other products

like high-power short-endurance applications can be found in portable

power drills (hammer drills), which are normally powered by the grid, or

handicap cars and electric bikes. Within this research scope the latter two

applications are not seen as portable.

Table 2: Power energy combinations for different application fields.

0.001-0.1W 0.1-10W 10-100W 100-1000W <1Wh

Sensors Flashdrive mp3

1-10Wh Fire

detector

Cell phone PDA Harddrive mp3

Screw driver Video camera

10-100Wh

Future smartphone

Laptop comp. DVD player

Power drill

>100Wh Car battery

Electric bike

Handicap car

2.1.2 Batteries for portable electronic devices

Besides the fast increasing number of portable electronic devices in the

world [1, 18], the functionality of the devices is increasing as well. Cell

phones get more functionality every year and new platforms like the iPad are

developed, which brings high-computing capabilities to portable sizes.

Mobile phones and laptop computers get smaller by the years but the power-

source volume as a percentage of the total applications volume is consistent

over the years, in between 10 to 30% as pictured in Figure 3 for the period

1995 to 2003. The large volume of the battery makes this one of the main

components defining the products volume and form.

11

Figure 3: Trend towards smaller cellular phones and batteries over a period of 1995 to 2003 (n=63).

For portable electronic products there is a trend towards lighter phones and

lighter batteries (Figure 4). The lower limit is probably reached at 90 to 100

grams for the phone and 20 to 25 grams for its battery. In 2010 the weight of

the cell phone is still in between 80 g for “call-only” cell phones with small

display to 110 g for cell phones with larger displays and more functionality

[19]. Note that almost 25% of the weight of the cellular phone consists of the

power source. More complicated electronics using larger displays often have

less effective space for batteries. The weight of the PDA and its battery has

been constant over the period of 1995 to 2003 (~160 g for the PDA and ~22 g

for its battery). In the last 5 years PDA’s have been taken over by smart-

phones which weight in in between 124 g [20] and 137g [21], slightly more

lightweight than the PDA in 2003. The batteries for PDA’s and laptop

computers account for 15% of the total mass. For laptop computers a larger

diversity can be seen resulting in light- but also heavyweight laptop

computers, each with their own application field.

12

Figure 4: Trend towards lighter cellular phones and batteries over a period of 1995 to

2003 (n=63).

Another important trend in battery design is the trend for thinner batteries.

Thinner batteries mean thinner products. You can see this trend especially

in batteries for low power applications like mobile phones (Figure 5) with a

thickness of less than 5mm in 2003. The thickness of PDA batteries hasn’t

changed meaningfully in between 1995 and 2003, with a median of ~10mm

(n=58). Based on the current thickness of smart phones the current batteries

used in these devices are even thinner than 10mm [20, 21]. Laptop

computers have a median thickness of ~20mm (n=103). Generally the laptop

computer uses cylindrical batteries compared which are, because of their

construction, more energy dense than the flat batteries used in lower-power

applications like cell phones and smart phones.

13

Figure 5: Thickness trend in the thickness of batteries applied in different cellular

phones (n=63). The Motorola Razr V3 is maybe the thinnest available cellular phone at the moment (13mm when folded), with a battery thickness of less than 5mm

(33.5x49x4.8 mm).

2.1.3 Definition of portable electronic devices

We can now define “portable” in portable electronic devices as products

which can be carried easily with one arm, imposing no limitation to

autonomy, weight and volume. The volume of the device is limited by a

handbag size, and weight of the heaviest laptop computer (6.2L and 4.5kg

[22]). The power source of a portable electronic device should be integrated

in the device, weighing less than 0.5kg and having a volume of less than

0.5L.

Applications with low-power and long endurance performance specifications

are mostly not intended to be portable in a carrying kind of way, and those

non-portable products are left out of the scope of this research. In the quick

application scan no portable products are found with the defined needs. In

the future the rise for this kind of products is not dismissed and thus taken

into account during this research. Applications with high-power and short-

endurance performance specifications, like the high-power drill, will be taken

into account. Ride-able products, like electric bikes, are left out of the scope

of this research. Medical and other implants, which belong to unique

14

contexts and pose special constraints, do not belong to the considered group

and are not considered in this study.

2.2 Power and energy systems compared Alternatives to the lithium based battery can be found in several research

projects discussing the possibilities of (alternative) power sources on a

theoretical basis [23-26]. Although these research projects give an excellent

theoretical overview, our concern is directed towards more practical values.

Within the literature reviewed we have not found studies describing

applicable data on power sources, and a literature study has been executed

to compare the different power generating and energy containing systems

with each other.

This section gives an introduction in a large number of alternative power

sources and energy systems. Practical data of all these power sources and

energy systems are accumulated from literature, specification sheets and

internet sources. All data is brought back to its basic normalized parameters

for prize, power and energy contained.

2.2.1 Approach

A desk search has been executed to acquire power source specifics. We

defined the following specifics to be of importance for power sources: power

output, voltage, dimensions, weight, conversion efficiency (mechanical to

electric) and retail price per item. For energy systems the defined specifics

are nominal voltage, ampere-hour capacity, dimensions, weight and retail

price per item.

Data is gathered in five ways. At first an internet search has been carried out

to find price, power and energy specifics for different commercially available

energy systems. More specific data was found in the second search field:

specification sheets from manufacturers manufacturing power-generators or

storage devices. Third search field: storage devices, especially batteries used

in cellular phones and laptop computers, were measured by students at

Delft University of Technology [22, 27]. The energy specifics for these

batteries were looked after on the battery itself or via specification sheets on

the internet. Fourth field of research consisted of acquiring data from

15

scientific papers. A number of case studies [28-30] describe power specifics

of for instance fuel cells. When no data was available from literature it was

generated by means of calculations based on theory and practical conversion

efficiencies. The potential energy available for e.g. fuels are calculated based

on the lower heating value and literature-based conversion efficiencies. Note

that his research has been conducted in 2004. Data is not updated, unless

mentioned otherwise.

To compare the power sources to each other general parameters are used

like the power and energy density (p and u) and the specific power and

energy (pρ and uρ). Also two new parameters are defined to give more insight

in the cost-price for a power source or energy storage device2, namely the

Specific Cost (SC) of power and energy:

retailP

ouput

cSC

P= (1)

retailE

cSC

E= (2)

Batteries on the one hand are both a power generator and energy containing

system, but for most alternative systems the power and energy demand is

separated in a module which produces power and a module containing

energy. To make a fare comparison these two parts have been separated in

the comparison and in Section 2.2.2 power sources are compared and in

Section 2.2.3 energy storage systems. The combination of a power source

with a energy container defines a power system for a portable electronic

device. Batteries are not taken into account as a power source but are

compared as an energy container.

2.2.2 Power sources compared

In Table 3 an overview is given of the median of the power density, specific

power and specific cost. All data is based on the power module itself without

taking auxiliary systems into account. Figures 6 and 7 show the results of

specific power, power density and specific cost in a graphical form.

2 The SCE is based on the initial price and not life cycle costs.

16

Table 3: Power sources compared on the median of the power density (pρ), specific power (p) and specific cost (SCP), range is in between brackets.

Median pρ

(W L-1)

Median p (W kg-1)

Median SCP

(€ W-1)

TE (n=4) [31-33] 625 (351-701) 74 (63-185) 11.33 (7.71-18.40)

Piëzo (n=24) [34] 20 (5-29) 2.6 (0.6-3.7) 24,000 (10-75E3)

EM gen. (n=4) [35-37] 106 (3-362) 33 (22-43) 6.85 (5.00-8.00)

PV (n=34) [37-42] 3.6 (1.6-136) 8.3 (2.8-54) 9.2 (5.69-230.88)

DMFC (n=16) [28, 29, 38-44] 8.8 (1-77) 12 (1.5-60) 111.84

PEM FC (n=46) [44, 46, 49-51] 50 (1-306) 20 (1.3-122) 52.78 (5.70-1,160)

4-stroke ICE (n=15) [52-59] 54 (28-834) 162 (100-750) 0.25 (0.13-0.53)

2-stroke ICE (n=51) [45-52] 1651 (882-2,933)

1241 (788-2,265) 0.11 (0.06-3.47)

Human body (n=12) * 2.8 (2.1-3.5) 2.6 (1.9-3.2) 0.04

* Based on a healthy male adult cycling for 10min; price is based on a loaf of bread and

a cycle efficiency of 25% [53-55].

0

1

10

100

1,000

10,000

Pie

zo PV

Hum

an p

ower

DM

FC

PE

MF

C

Ele

ctro

mag

netic

The

rmo

elec

tric

4-st

roke

com

bust

ion

2-st

roke

com

bust

ion

Sp

ecif

ic P

ow

er (

W k

g-1

)

(a) Specific power range

17

.

0

1

10

100

1,000

10,000

Pie

zo PV

Hum

an p

ower

DM

FC

PE

MF

C

Ele

ctro

mag

netic

The

rmo

elec

tric

4-st

roke

com

bust

ion

2-st

roke

com

bust

ion

Po

wer

den

sity

(W

dm

-3)

(b) Power density.

€ 0.01

€ 0.10

€ 1.00

€ 10.00

€ 100.00

€ 1,000.00

€ 10,000.00

€ 100,000.00

Pie

zo PV

Hum

an p

ower

DM

FC

PE

MF

C

Ele

ctro

mag

netic

The

rmo

elec

tric

4-st

roke

com

bust

ion

2-st

roke

com

bust

ion

Sp

ecif

ic C

ost

(E

UR

W-1

)

(c) Specific cost.

Figure 6: The range of the data acquired for: (a) the specific power range, (b) the power

density researched and (c) the specific cost (median is plotted as a dot).

18

The power sources can be divided into four groups:

1. Electricity conversion based on temperature, force and torque or

radiation, like Thermo Electric elements, Piëzo elements, Electro

Magnetic conversion technologies and Photo Voltaic cells.

2. Direct chemical combustion-to-electricity, like fuel cells

3. Indirect chemical combustion-to-electricity, like internal

combustion engines who needs an extra mechanical-to-

electricity convertor.

4. Human power

0

1

10

100

1000

10000

0 1 10 100 1000 10000

Specific Power (W dm-3)

Po

wer

den

sit

y (

W k

g-1

)

Combustion engine

PEM fuel cell

DMFC

Photovaltic cells

Electro Magnetic Generator

Thermo Electric Generator

Human body

combustion engines

PEM fuel cell

Direct Methanol fuel cell

Electro magnetic gen.

Photovoltaic cells

Thermo electric gen.

Human body

Figure 7: Log-log plot of the power specifics for the different power sources based on data

used in Table 3

19

Figure 7 shows all power sources found plotted in a graphical form. The

figure shows a log-log plot of the specific power versus the power density of

all power sources. Batteries are not included because these products are

besides a power source also a energy storage device, and could best be

compared with power sources combined with energy storage means (like

fuels).

The internal combustion engines deliver the most power per unit of mass

and per unit of volume and are the most cost effective power source. The

data is mainly based on Internal Combustion Engines (ICE) found in RF-

controlled planes. The electro-magnetic generators and thermo-electric

elements are also very small and lightweight power generators, and also very

cost effective. The direct chemical combustion to electricity generators, the

Direct Methanol and PEM fuel cells deliver a fairly good amount of power per

unit volume and weight, but are very costly at the moment.

2.2.3 Energy storage systems compared

To compare the data acquired by the previous research fields, all data is

transformed to the general parameters. In Table 5 an overview is given of the

median of all available data. The range, if available, is in between brackets.

All data is based on the energy system itself including the casing but without

auxiliary systems. All energy densities and specific energies are reduced to

electric energy per unit of mass or volume. For fuel-based storage system,

the conversion efficiency is taken into account which is described in Table 4.

Figure 8 and Figure 9 present the specific energy, energy density and

specific cost in a graphical form.

Table 4: Efficiencies of different conversion systems.

efficiency (%)

Hydrogen to electricity 27%

Hydrogen to electricity (compressed) 9%

Internal combustion to electricity 5%

Methanol to electricity 17%

20

Energy storage systems can be divided into three groups:

1. Fuel or gas based energy containers like hydrogen, LPG, petrol and

methanol.

2. Combination of electrochemical cells, used to convert chemical

energy into electricity, like primary and secondary batteries, but also

voltage generators like capacitors (ultra, super and boost cap).

3. Mechanical storage like flywheels and springs.

Table 5: Energy storage systems compared on the median of the energy density, specific energy and specific cost, the range is in between brackets.

Median uρ

(Wh L-1)

Median u

(Wh kg-1)

Median SCE

(€ Wh-1)

Hydrogen (ambient)* 1 8.89 5.25

Hydrogen (300bar)* 55 2.96 15.75

LPG** 319 639 1.94

Petrol** 500 694 2.5

Methanol*** 830 1051 0.22

Alkaline (n=18) [63-68] 438 (217-514) 162 (91-192) 0.4 (0.07-8.50)

Primary lithium (n=9) [63-68] 565 (260-708) 248 (99-306) 2.07 (1.21-13.32)

Zinc-Air (n=7) [63-68] 1.138 (219-1.496) 350 (182-434) 3.06 (0.90-10.34)

Recharg alk. (n=4) [26, 64, 69-78] 308 (278-342) 117 (105-132) 0.65 (0.27-1.58)

Lead-acid (n=31) [27, 56-66] 80 (59-101) 32 (24-39) 0.5 (0.20-1.79)

NiCD (n=47) [26, 64, 69-78] 86 (42-156) 37 (11-75) 3 (1.33-25.00)

Nickel MH (n=76)[27, 56-66] 147 (31-327) 53 (15-90) 3.77 (1.61-143.23)

Lithium-ion (n=63)[27, 56-66] 212 (20-1,210) 121 (11-660) 9.45 (2.77-59.99)

Lithium-poly (n=6)[27, 56-66] 200 (115-237) 135 (103-146) 19.03 (17.30-22.21)

Ultra cap (n=25) [57, 67-70] 0.3 (0.01-4.1) 1.7 (0.3-21.4) 3.00E6 (0.5E6-71E6)

Power spring (n=8)**** 0.05 (0.01-0.56) 0.03 (0.01-0.66) 3.37E5 (0.85E5-40E5)

Flywheel Carbon FRP***** 330 220 250.92

Flywheel HS steel***** 240 30 15.97

Flywheel cast iron***** 30 5 77.43

* Based on heating value and a conversion via PEM fuel cell (conversion efficiency of the

fuel cell plus the power use of auxiliaries: 27% for the uncompressed hydrogen and 9% for compressed hydrogen)

** Based on heating value and a conversion via combustion and [71] *** Based on heating value of methanol, and a conversion via DMFC (conversion efficiency

of the fuel cell plus power use of auxiliaries: 17%) **** Calculation based on 8 cases: CFRP, E-glass NiTi and Austenitic steel; material data [72] ***** Based on material data from [72]

21

Table 5 shows the energy densities of ambient hydrogen is the bottleneck for

this energy storage system. The gaseous fuel has to be compressed to at

least 300 bars before it is competitive with other storage systems like

secondary batteries. Other fuel based energy storage media have a very high

energy density and specific energy, and are very cost effective compared to

secondary batteries (especially lithium based batteries). As shown in the

table but also in Figure 8 the primary batteries have higher energy density

than fuel based storage systems, but per unit of mass these storage devices

are worse.

0.01

0.10

1.00

10.00

100.00

1,000.00

10,000.00

Hyd

roge

n

Hyd

roge

n @

300

bar

LPG

Met

hano

lP

etro

lA

lkal

ine

cilin

der

Lith

ium

cili

nder

Zin

c A

irA

lkal

ine

Lood

zuur

NiC

dN

iMH

Li-io

nLi

-pol

yU

tlra

/ bo

ost

caps

Pow

er s

prin

gC

arbo

n F

ibre

(60

%vo

l)H

igh

Str

engt

h S

teel

Cas

t Ir

onLe

ad a

lloys

Sp

ecif

ic e

ner

gy

(Wh

kg

-1)

(a) Specific energy

22

0.01

0.10

1.00

10.00

100.00

1,000.00

10,000.00

Hyd

roge

n

Hyd

roge

n @

300

bar

LPG

Met

hano

lP

etro

lA

lkal

ine

cilin

der

Lith

ium

cili

nder

Zin

c A

irA

lkal

ine

Lood

zuur

NiC

dN

iMH

Li-io

nLi

-pol

yU

tlra

/ bo

ost

caps

Pow

er s

prin

gC

arbo

n F

ibre

(60

%vo

l)H

igh

Str

engt

h S

teel

Cas

t Ir

onLe

ad a

lloys

En

erg

y d

ensi

ty (

Wh

dm

-3)

(b) Energy density

€ 0.01

€ 0.10

€ 1.00

€ 10.00

€ 100.00

€ 1,000.00

€ 10,000.00

€ 100,000.00

€ 1,000,000.00

€ 10,000,000.00

Hyd

roge

n

Hyd

roge

n @

300

bar

LPG

Met

hano

lP

etro

lA

lkal

ine

cilin

der

Lith

ium

cili

nder

Zin

c A

irA

lkal

ine

Lood

zuur

NiC

dN

iMH

Li-io

nLi

-pol

yU

tlra

/ bo

ost

caps

Pow

er s

prin

gC

arbo

n F

ibre

(60

%vo

l)H

igh

Str

engt

h S

teel

Cas

t Ir

onLe

ad a

lloys

Sp

ec

ific

co

st

(EU

R k

Wh

-1)

(c) Specific cost

Figure 8: The range of the data acquired for specific energy, energy density and specific cost (median is plotted as a dot).

23

0

0

1

10

100

1,000

10,000

100,000

0.0001 0.001 0.01 0.1 1 10 100 1000 10000

Electric Power (W)

Pri

ce

(EU

R W

-1)

PEM fuel celltwo-stroke combustion enginefour-stroke combustion engineDMFCPhotovaltaic cellsElectro magnetic dynamoPiezo generatorThermo Electric generatorHuman power engine

Piezo gen.

Two-stroke CE

Four-stroke CE

PEM fuel cell

DMFC

PV cells

DynamoTE gen.

Human power

€1

€10

€100

€1.000

€10.000

€0,10

Figure 9: Comparison of the power sources based on retail price.

2.2.4 Discussion

Table 3 gives a good overview of the potential of some power sources and

why some might not be commercially successful. For instance the piëzo

materials can be used to generate power for portable electronics. The retail

price for this power generator is more than €30,000 per delivered Watt. The

retail price for the piëzo-generator is based on mass-produced piëzo material

by Piëzo Systems Inc. [34]. Piëzo-generator as a power source will not be

commercially interesting applied in portable electronic devices.

Figure 6 shows that the high-power combustion-engine stands out. The

engines have a power density of more than 1.6kW L-1 (>1.2kW kg-1). Values

are based on commercially available mini-engines used in model airplanes. It

shows that the potential of this power source is very high as a power source

for portable electronic devices, but also in mechanical actuated devices like

24

high-power drills. They seem to be very promising as an alternative power

source for portable applications. Why is this power source not applied more

often in portable electronic products? Reasons are its high noise output,

vibration, the toxic exhaust fumes and instability of electrical power from the

system. The engine is a spinning mass at high angular velocities and is

subject to angular velocity variation when load changes [25]. Besides output

the engine needs an air intake as well as cooling. The smallest existing

combustion engine is the Cox Tee Dee .010 two-stroke engine has a

mechanical power output of 15W at a speed of 32,000rpm, encapsulated in a

volume of 37.5x38x7mm3 and weighing only 14grams [45]. This production

engine is still to large to be placed in for instance a laptop computer. MEMS

and nanotechnology could lead the way to miniaturizing the engines more so

it could be installed in portable electronics [47]. The high noise output can

be decreased by muffler systems. The implication of miniaturization on noise

output is unknown at the moment.

The next interesting power source is the Polymer Exchange Membrane fuel

cell (PEM FC), fueled by hydrogen, and its more inefficient sister, the Direct

Methanol Fuel Cell (DMFC). A major problem in this area is the storage of

hydrogen. New developments could improve the total energy density of a

PEM power system but do not seem to be commercially available in the

upcoming years. The DMFC doesn’t have this problem. Methanol is easy to

refill when power is needed. The DMFC systems of today are fueled with a

3%wt methanol water mixture. Improvements are reported already by

Toshiba and MTI fuel cells which claim to have boosted the concentration up

to 100%. This means a higher energy density of the total system and a

smaller tank needed. This makes DMFC systems more interesting for

applying in portable electronics. The technology is still in its research phase

and industrial application still seems to be a long way.

When looking at costs the most cost effective power source is the human

body when cycling (Figure 7). The human body is only generating mechanical

output which has to be converted to electricity. Dynamos, thermo-electric

elements and piëzo generators are good conversion technologies. The

dynamo is a very cost-effective power converter. On the other hand PV cells

and thermo-electric generators are very cost-effective when compared to fuel

25

cells. This is probably because more effort in the past years is applied on

this field. For instance PV cells are under investigation since 1950 at the Bell

labs for use in space applications. At the moment large factories have been

build by Siemens, Shell and BP to mass-produce the cells on the roll. PEM

fuel cells on the contrary are under research since 1960 at General Electric

for the use in the NASA Gemini project. They got industrial attention since

the 1980’s by the car companies. Micro Direct Methanol fuel-cells were first

presented by Robert Hockaday in the beginning of the 1990’s [73]. A lot of

work is done since to decrease the price of fuel cells by lowering the amount

of platinum needed for the membrane and making it mass-produceable.

Tsuchiya and Kobayashi [74] analyzed the mass-production cost structure of

PEM fuel cells for automobiles (50 kW) by a learning-curve model. A

moderate scenario shows that production cost will decrease from €1.46 per

Watt for the production of 40 cars to about €0.14 per Watt in 2010 for the

production of 50,000 cars (a cut down of a factor 10 over ten years). This

price is in the same order as that for commercially available micro-

combustion engines. If these numbers are applicable to low-power fuel cells

in the range of 10 mW to 100 W has to be seen.

For DMFC systems the costs of materials used for fabricating are very high

(especially the high cost of platinum electro catalysts). According to [75] the

production costs for a direct methanol fuel cell could be $5 per generated

Watt, for a 20W and 60Wh power system. In 2004 the Smart Fuel Cell SFC-

A25 cost around €112 per generated Watt, but in 2010 the prices dropped to

€53 per generated Watt [76] for the next generation DMFC systems from

Smart Fuel Cell, the EFOY 2200 (90W) system.

Hydrogen seems to be a good option to be the next energy carrier for a wide

range of applications (Figure 8). The storage of hydrogen in small volume is

still a big issue researched on. Compression or other ways to decrease

volume of hydrogen is necessary. Problem with compression is the high

pressures needed ranging from 250 bars to 600 bars. These high

compression rates call for a high tension casing like steel or fibers.

Alternatives like nanotubes and liquefied hydrogen at low temperatures still

seems to very far away especially when applying it to consumer electronics.

26

Other carbon-hydrogen liquids like methanol are less energy dense but are

more practical in use. In combination with a fuel cell these energy carriers

are a great opportunity for powering portable electronic devices.

Petrol and other carbon-hydrogen liquids are able to carry lots of energy per

unit of volume and mass when using combustion-engine technology to

convert from fuel-to-mechanical-to-electricity. When carbon-hydrogen

liquids are directly converted to electricity, for instance by means of a DMFC,

the energy characteristics seems to improve because overall system

efficiency of the DMFC is higher than the combustion engine generator. In

comparison with the lithium-ion battery the methanol fuel-cell combination

could improve energy densities up to a magnitude of 3, and specific energies

up to 7.

The next interesting energy storage system is the Zinc-Air battery. This

system is a battery which, like fuel cells, needs air on the cathode side. A

proven technology used for instance in hearing aids, and lately also as a

back-up charger for cellular phones and PDA’s [77]. Problem with this

battery type is the single use only, but it can still compete with hydrogen or

methanol as an energy carrier, without the advantage of quick refill.

Rechargeable batteries have improved very fast the last decades. Large

improvement steps have been made from the first Nickel-Cadmium (NiCd)

cell to the energy dense lithium ion battery (li-ion). As can be seen in Figure

8 (a) and (b) the lithium-ion battery has a high energy density and specific

energy, but is very costly when compared to other battery systems, Figure 8

(c). Apparently the price of the battery is of subordinate concern to the

improvement in convenience and decrease of volume and mass.

Within the group of mechanical storage devices the flywheel and especially

the Carbon Fiber Resin Polymer has high potential energy specifics.

Unfortunately the price is still very high and can only decrease when the

price of carbon fibers decreases. The price now is based only on mean

material price for CFRP filled with 60%vol HT-carbon which is in between

€188 and €314 per kg [72]. Because the energy stored in flywheels is related

linear with mass and to the square of the angular velocity, smaller flywheels

can only store significant energy when rotation speeds are high. The

rotational speed is limited by the strength and density of the material. High

27

rotation speeds introduce gyroscopic forces complicating movement of the

total system. Implementing two contra-rotating flywheels will annul the total

angular momentum, but introduces significant design problems, like a-

synchronic acceleration [25]. Flywheels created from High Tension steel

seems to be a cheaper alternative but cannot compete with the energy

density of rechargeable batteries in general.

The worst energy storage system presented in the previous figures is

capacitors. Their energy storage specifics are both very low in comparison

with batteries, and their retail price is too high to be of interest as the

primary energy storage medium. On the other hand capacitors are not made

for energy storage but to boost power output. They are well-known for their

high pulsed power output, for the use in hybrid systems like in electric

buses but also to increase run-time of portable electronics [78]. They are

potentially interesting as an intermediate power source generating high

instant power for hybrid energy systems.

2.3 Conclusions This chapters’ goal was to compare and pinpoint the power systems which

can compete with lithium based rechargeable batteries in the field of

portable electronic devices. We have defined portable electronic devices in

the first section as ‘devices using electric energy, with an integrated power

source and carriable in once bag’. The size of the device has been limited to a

large laptop computer size (<6.2L) and weight (<4.5kg). Devices which are

implantable and applications, mainly in the field of low power, which are not

intended to be carried around are left out of the scope of this research. This

does not mean all low-power long-endurance applications are out of the loop.

At the moment no products satisfying this definition have been found, but in

the future the rise for these kinds of products are not dismissed and thus

taken into account during this research.

From the previous desk search it can be concluded that we can pinpoint two

power source-fuel combinations which can compete with the lithium based

rechargeable battery: (i) the high power-dense combustion engine in

combination with the high energy-dense carbon fuels as petrol, and (ii) fuel

28

cells chemically burning high-energy dense (carbon) hydrogen-based fuels

direct to electricity.

Besides the high power density and energy density of internal combustion

engine systems, the system is very cost-effective. Problem with the internal

combustion engine is its high noise and toxics output. Internal-combustion

power sources are thus of interest for outside applications where noise is

less of an issue than the lack of electric grid. To produce electricity the

internal combustion engine has to be coupled with a generator, making the

overall efficiency lower. Applications using the direct mechanical output of

the engine are of more interest. Feasible applications can be found in

construction work environments, like for high-power rotary-hammer drills.

The second interesting power-source is the fuel cell. Polymer exchange

membrane fuel cells make use of hydrogen as its fuel. Hydrogen is mass-

wise a very energy-dense fuel, but very poor volume-wise (specific energy). To

decrease volume hydrogen has to be compressed at high rates up to 600bars,

making this fuel less interesting for portable and small applications. When

combining the fuel cell with the more convenient and still high energy dense

methanol fuel, a possible successor to the lithium ion battery can be defined.

Because of the high energy-dense fuel, this power system could be a factor 7

smaller than currently lithium-ion batteries. Note that the fuel-cell and

auxiliaries are not taken in to account in this factor. It still has to be seen in

which cases the total system will be smaller than the lithium-ion battery.

The overviews given by Figure 6 to 9 compare the different power sources

and the energy carriers with each other. These figures are only indicative,

and only present the power generator and energy carrier on its own. It is still

not feasible to compare a power system, a combination of power generator

with an energy carrier, to the rechargeable battery. This is because with

most alternatives, especially the fuel-cells and internal combustion engine

systems, are a combination of a power source with a certain fuel. To make

alternative power systems more visible to the designer, new models have to

be developed, in which a specific power source energy combination (power

system) is compared with rechargeable batteries.

29

3 Power source design and selection in practice, state of the art3

In Chapter 2 the potential competitors for lithium-based rechargeable

batteries have been defined. The second part of the initial research question

is about how the designer can select or design a power system during the

preliminary design phase. This chapter will be about answering this question.

According to [80] a design of an industrial product is defined by four pieces

of information: (i) the dimensions of the parts and the product, (ii) the

production technology used to produce the parts, (iii) the materials used in

the parts, and (iv) the way the parts are put together. Portable products not

only consist of mechanical parts and components but more and more of

electronics and control engineering. A fifth piece should be added to the

previous list which involves the design of the software. Based on the results

from Chapter 2 the fuel cell system is indicated as a potential successor to

the lithium-ion battery powering portable electronic devices. The fuel-cell

system is besides an industrial product consisting of several components

placed in space, also a mechatronic product where software and power

management have great influence on the design of the system. The design of

a fuel cell system which is engineered by engineering designers in which

“form follows function” is generally the case. In this thesis the design of

software and power management is not taken into account and the focus is

on mechanical design.

To produce the correct drawings for fabricating the product the product

designer has to analyze and define the problem. The goal of the product or

3 Parts of this chapter are presented at the Knowledge Foundations’ Small Fuel Cells

2009 conference in Orlando FL (2009): “Selection of power sources for portable applications”.

30

the function it has to fulfill is thought through, he has to draw multiple

solutions to the problem, think and rethink the solutions and even do

calculations for testing of the feasibility or optimizing the system. The

thinking process and methods used during the design process are well

described in design handbooks as [81] and discussed in literature [82, 83].

In general the systematic design approach is described by (i) laying out a

specification, followed by (ii) exploration and evaluation of potential useful

concepts. Based on this evaluation (iii) the most promising concepts are

elaborated by producing potentially embodiments, which are evaluated and

(iv) the most promising embodiment is elaborated further to produce the

final detailed design.

The choice for an alternative power source, like the fuel cell system, can be

part of the strategy chosen by the company based on e.g. environmental

issues. When this is the cases the realized product will be equipped with this

alternative power source because it was one of the goals of the project. In

most cases the designer is not directed towards one solution and he/she is

open for alternatives to the commonly used battery. The designer has to

evaluate, compare and choose for a battery or, if visible, an alternative power

source during the design process. To do this he can make use of tools and

methods to make the optimal choice.

3.1 Approach This chapter reviews tools and methods for choosing or designing power

sources applied in portable electronic devices. The goal of this chapter

consists of two parts: first is to identify available tools and methods, and

detect when those tools are applied during the design process. Second we

want to know how real-life designers evaluate different optional power source

concepts when designing a product.

A literature review is conducted in Section 3.2, where specific tools and

methods for choosing or designing power sources are searched for. This

review is conducted by desk searching for tools on the internet and scientific

literature found on Scopus and ScienceDirect [84, 85].

31

Section 3.3 reviews case-studies, subdivided in three case-studies from

literature and two interviews with experts. The case-studies and expert

interviews are conducted to explore the mere pragmatic design approach

when designing power sources. Three case-studies from literature are

reviewed and a life case-study is evaluated by interviewing two experts in the

field. The interviews are held with a questionnaire a leading. The questions

asked were subdivided into a general introduction in which the research was

explained, followed by the design process the interviewed goes through when

designing an application powered by a portable power source. The design

freedom is discussed and in which phase of the developing process the main

focus is, when designing the power source. After designing the designer has

to make choices, how the design engineer makes these choices is questioned

in the part about decision making. The questionnaire is ended with examples

from practice.

Two of the case studies focus their attention to designing direct methanol

fuel cells, but the other two also evaluate other alternative power sources

during the design process. This chapter will end with conclusions in Section

3.4.

3.2 Tools and Methods in the Field A literature review is done on tools and methods in which different power

and energy sources are compared, selected or even designed. Four different

search and selection type of tools can be distinguished in literature, which

will be explained more in detail in the remainder of this section:

• Power source comparison charts (Section 3.2.1), which can be used

to identify power-source opportunities for different power sources.

• Power source selection for a specific application (Section 3.2.2),

especially interesting for the designer, designing portable electronics

with specific performance characteristics.

• Power source design tools (Section 3.2.3), which help the designer in

dimensioning the power source.

• Optimization tools (Section 3.2.4), where all possible configurations

of power-sources and energy-containers are evaluated and optimized

for single or multi objectives.

32

3.2.1 Application selection tools

Simple charts giving an overview different power sources can be used to

identify the opportunity of a specific power source. One of the most used

graphs is the Ragone plot [86-88]. An example is plotted in Figure 10. The

plots show regions related to specific applications by energy and power

requirements. The Ragone plots are usually presented in a log-log plot which

shows the limits in available power and energy of a specific power system,

and also the optimum working region given by the part of the curve where

both energy and power are high. The specific form of a Ragone curve

depends on the internal losses and leakage properties of the power system

[87]. The plots are generally obtained by measurement at a constant power

load connected to the power system but sometimes derived by means of

mathematical models. The variables are normalized power and energy

specifications of the power source, either on basis of weight or volume.

Adding the specific power and energy specifications of a particular

application will give the designer feedback on possible power sources

matching those.

Figure 10: An example of a Ragone plot [89].

3.2.2 Power source selection tools

In the previous chapter different alternative power sources have been

compared to each other based on three main aspects: (i) size, (ii) weight, and

33

(iii) costs. A large database of commercially available power and energy

sources (and combination thereof) is used to compare them as presented in

the previous chapter. These graphs can be used to compare different power

sources but also to make alternative power sources more visible to the

designer.

The ‘Database and Selection Method for Portable Power Sources’ from [11] is

based on the ‘free search’ selection strategy for materials and processes from

[90], and is implemented in the Cambridge Engineering Selector (CES) [72].

The CES selection tool is widely used by product designers to screen

thousands of materials and production processes on their applicability for a

specific product, part or function.

Figure 11: An example of the output of the CES based power-source selection method [11].

The search engine of CES is a ‘free search’ selection method, one of three

possible selection strategies available [90]:

• Free search is based on quantitative analysis of performance and

normalized characteristics. So called ‘design indices’ are used to

direct the designers in their search for the optimal choice.

• Questionnaire based search is based on expertise-capture, guides

the uninformed user through a more or less structured set of

34

decisions, using built in expertise to compensate for the lack of it in

the user.

• Inductive reasoning and analogy search is based on a library of

previously solved problems or ‘cases’. An analysis of its features, a

solution and an assessment of the degree of success of this solution

is given, guiding product designers in product development or

improvement.

The power-source selector is mainly based on basic characteristics as mass

m, volume V, and cost c of existing components. In Figure 11 an example is

given of the output of the power-source selection-method. Besides size and

mass, a selection can be made manually based on performance parameters,

like average voltage and shelf life, and derived properties like energy per unit

of cost and energy density. This tool is interesting for exploring optional

power sources when conceptualizing a new or existing product. It makes

alternative power sources also visible to the user, besides optional primary

or secondary batteries.

In [91] van Gennip developed the PowerQuest, a database driven

power/energy source selection tool for designers during the concept phase.

The focus in 2006 was merely on photovoltaic cells, and it was intended to

be extended with other alternative power sources. The input for this online

tool consists of mean power draw (in Watts), a simple usage profile during

the day, the irradiation, and more specifications, like working voltage, when

known. This tool is used during the initial conceptualization phase of the

design process. It tries to give the user feedback on the possible match,

energy-wise, between the photovoltaic cell and the application the user

wants to evaluate, and returns with the design consequences when using a

photovoltaic cell, for instance the physical needed sun-irradiated surface

area.

3.2.3 Designing alternative power sources

A third problem designers have to cope with is the dimensions of a new

power source. Especially component volume and size is of great importance

when giving form to an application. For primary batteries a limited number

35

of standard forms exist, making the choice more easy when designing. When

the lithium-ion rechargeable battery was introduced the form was not

standardized. Cell phones and other flat electronic devices directed the form

from cylindrical to a thin flat prismatic form, the cheaper pouch cell or even

to the less energy-dense flexible cell. In general the battery is available in

prismatic form. A simple approach to dimensioning a battery during the

concept phase is by taking only three main constraints into account: (i)

energy it has to contain (Wh), (ii) charge and discharge currents (C), and (iii)

cycle life of the battery.

The procedure can be almost as easy as dividing the energy the application

needs with the energy-density of the specific battery (Wh L-1). The knowledge

to make the choice can be derived from an expert, a handbook or

specification sheets from battery manufacturers. This procedure is well

implemented in the CES based power-source selection-tool described in

previous sections.

Giving form to multi-component power sources, like fuel-cell hybrids and

other hybrid power-sources, is not as easy as giving form to stand alone

batteries. When minimizing size, mass or costs in hybrid power source

design one specific optimization tool is of interest, namely ‘Power

Optimization for Wireless Energy Requirement’ (POWER), developed at the

University of Michigan’s Wireless Integrated Microsystems Engineering

Research Center (WIMS-ERC), described in [92]. Factors influencing

selection of the power supply can be summarized as follows [12]:

• Electro chemistry: cell potential, discharge/charge profile, capacity

and lifetime;

• Geometry: surface area, volume and mass;

• Environment: temperature, pressure and exposure.

To overcome some of the limitations of electro chemistry, size, and shape,

often two or more power supplies are used within the same system. Typically

a hybrid power supply combines a high-power dense element, with a high

energy dense element. Different studies have shown that improved

performance is possible with hybrid supplies [93].

36

In [92] the systematic method is described for the selection and design of a

power systems for a WIMS Environmental Monitor Test bed (EMT). The

method focuses on selecting and designing power supplies for

microelectronic devices which are prescribed after design is nearly complete.

The methodology constraints are operating temperature, energy/power

density, and energy/power density. Further requirements/constraints on

rechargeability, mass, volume and lifetime are allowed in selecting the

appropriate battery electro chemistries or configurations like parallel, series

and combinations thereof. The algorithm separately evaluates results of

three strategies (approaches) to system design, specifying either:

• A single aggregate power profile,

• A power system designed to satisfy several power ranges (from

micro-, milli-, and Watts), or

• A power system designed to be housed within specified spaces

within the application, with device constraints on volume and

surface area.

The algorithm is implemented in MATLAB code and presented in [102, 104].

The batteries used in the application are selected from a database

comprising a large amount of commercially available primary and secondary

batteries. The algorithm described in the papers has the following stepwise

outline:

• A time segmented power profile of the system and its components is

introduced. Due to the impracticality of mapping small fluctuations,

data is coarsened before inputted into POWER;

• The energy, weighted power, energy density and weighted power

density of the device power-system are calculated4;

• The system is broken up in sub devices classified in power range

(from micro- to milli-Watts);

• A restriction of electro chemistry based on environmental

constraints is executed;

• For all devices in the system the voltage and current are computed.

4 The nominal voltage of the cell is the operating or rated voltage of the cell specified

by the manufacturer

37

• Capacity fade is estimated based on discharge current and cycle

number, using expressions from literature;

• Based on conditional statements, like the maximum number of

batteries used, the total number of batteries required to fulfill the

power profile is calculated; cells can be placed in combinations in

series and/or parallel according to energy, voltage and current

factors;

• Selection and ranking is done based on mass requirements, energy

density and battery lifetime;

• Results are compared on mass, volume surface area and the number

of cells in the configuration;

• Additionally the costs can be entered as a constraint.

The methodology focuses on micro-power system combining single primary

and secondary batteries. The application field is implantable electronics

where volume and weight of the power source is of great importance.

3.2.4 Optimization tools

In the paper written by [94] hybrid power systems, combining larger power

systems with energy storage systems, are optimized for element sizing. The

optimizing is based on costs and not on size. So-called ‘energy-hubs’ are

introduced, an optimal hub layout of a power system described as in Figure

12. Converters link inputs and outputs through coupling factors which can

be considered to be the converter’s steady-state energy efficiency. In general

a “system approach” is applied, and the costs of the system is optimized

based on normalized investment cost per hub element, e.g. a wind generator

costs approximately $2 per Watt, and normalized input data, e.g. from

available wind power.

Other optimization tools use dynamic models to optimize hybrid electric

energy generators with mostly two objectives in mind, namely the efficiency

of the system and the unmet load. These optimization tools are focused on

the total Net Present Cost (NPC) throughout the lifespan for both off-grid and

grid-connected power systems for remote, stand-alone, and distributed

generation (DG) applications. Enumerative methods like HOMER [95]

38

evaluate all possible solutions requiring excessively high calculation time. To

reduce time heuristic techniques, like genetic and evolutionary algorithms

are applied, an example is the Hybrid Optimization by Genetic Algorithms

(HOGA) by [96].

inputs outputs

P1 P2 P3

L1 L2 L3

energy hub

Figure 12: General energy hub diagram [94].

3.2.5 Discussion

A few methods and tools have been found which could support the product

designer in making a choice for a power-source in different phases of design.

The method described by [92] is generally used in the embodiment or

detailing phase of the design process, when all constraints (application

characteristics) are defined and a power source is chosen “off the shelf”.

The comparison methods described in Chapter 2, by Ragone plots or

database-driven comparison charts [11] are very useful during the start of

the concept phase, when only the constraints/requirements are known. The

methods indicate the possibilities and opportunities of new power sources

and energy storage devices. After ‘screening’ power sources, the next step

should be sizing of the chosen power/energy system. In the method of Fu,

Lu. et al. the mass and volume parameters are simplified. For instance the

mass of the fuel cell is equaled to the mass of the fuel, which in general

cases a good estimate, but gives problems for the fuel cell system because

the components needed to convert the potential fuel energy to electricity (e.g.

with fuel-cells and pumps) enclose a large part of the total volume and

weight, but are not taken into account.

39

The tools described in [97] are used in larger scale power/energy systems

like wind-power in combination with alternative storage systems. The model

consists of larger blocks of devices. All combinations of power sources and

energy storage systems are evaluated and the optimal combination is chosen.

The method is a system block approach with only a single objective,

minimizing costs.

3.3 Case study review Two experts from Jet Propulsion Lab (JPL) in the field of power source design

have been interviewed [98, 99] to discuss the process of developing a power

system for interterrestrial missions. The interviewed have been selected on

basis of their expertise in the field of power source development.

Besides the interviews three cases-studies have been evaluated reported in

two journal papers and one graduate reports. The following papers/reports

have been reviewed:

1. Icardi et al. “Compact direct methanol fuel cells for portable

application” [100].

2. Bruce Lin, “Conceptual design and modeling of a fuel cell scooter for

urban Asia” [101].

3. Rohrschneider et al., “Flight system options for a long-duration mars

airplane”, [102, 103].

The reports and papers are reviewed on the practical design approach, tools

and methods used, and the main problems which had to be overcome.

3.3.1 JPL interviews

The Jet Propulsion Lab (JPL) is visited by the author on June 11, 2009, and

is situated in Pasadena, California, United States of America. The author has

interviewed two members of the engineering staff at this laboratory, Thomas5

a senior member of the engineering staff [98] and Paul, a senior power

system engineer [99]. Both work in the field of power source engineering for

5 Only the first name is used to indicate the interviewed.

40

extraterrestrial applications like the Mars rover and satellites going out of

the Earths orbit.

JPL is the research laboratory of NASA, since 1958 when the first satellite,

Explorer 1, was send in to space. Since than JPL has send spacecraft to

every planet in the solar system form Mercury to Neptune. JPL engineers

design new space exploration satellites and system parts. Power source and

energy storage devices are a main research field, because these exploration

devices have to work under extreme conditions and have to work for a long

period of time. Costs, durability, mechanical stability and mass are the key

drivers for designing these energy generators.

Thomas is senior manager of a crew who is addressed when a power source

for a mission has to be identified. As an expert he has the right system in his

mind when asked based on simple input variables like cycle life. The focus

when designing a new power source is on optimization for mass, “costs, I

don’t really worry about, it’s not my job”. He uses in-house developed tools

made with Microsoft Excel, Mathcad and does his design work in CAD.

During the identification stage (conceptualization) there is fairly no top-level

programming, FEM or thermodynamic programming involved.

The interview continued with Paul who builds a great variety of satellites

going around the solar system, space, orbit and land. This means the power

source has to work in a variety of environments and thermal design is of

great importance. The most simple, cheapest and lightest option is the solar

array with batteries. The first choice Paul makes is the choice between the

power source, a primary battery, solar array, radioisotope thermoelectric

generator (RTG) or fuel cell. As means of energy storage there are two options,

secondary battery or fuels. Note that the designers specify the primary

battery as a power source and the secondary battery as a energy storage

device. Paul defines a energy storage as an object which can be recharged,

and “you cannot recharge a power source!”, When selecting a power source

or energy storage device it is useful to think in the time domain, from

milliseconds (capacitors), months (primary battery), years (fuel cells) and

more years (RTG or solar array). Solar arrays are capable of working over a

fairly wide time domain as long as the eclipses are not to long. Hybrid

41

systems, or as Paul calls it cascade systems, are used when short bursts of

power is needed and the system sleeps for a long period. A capacitor is

mainly added to stabilize the electric main power bus. To select or design the

power source a general engineering approach is followed. The process is

iterative and cannot be seen as linear:

1. First the mission requirements are set up and system ideas are

developed. In this stage already the limitations of the system is

scanned and the limits are poked.

2. Set up of a block diagram of the system, design its architecture

based on reducing risks, costs and mass. The optimization mode is

mainly focused on meeting the requirements of the mission, “where

doe we go, how long does it take, which direction am I pointing”

3. Write down your assumptions and what the key drivers are

4. Make up a cost estimation and a planning

5. Put together a list of equipment with its specifications, and set up a

list of hardware you need (breadboards)

6. Establish procurement costs, schedule and apply workforce

7. Estimate total costs, and make an analogy with comparable projects

8. Pushback: “I can do it that with this system when we change the

requirements (e.g. changing the point of direction of the satellite), or

introduce tricks like turning solar arrays from the sun to get the

same power on Jupiter as on earth”,

9. Minimize for cost, mass, mechanical issues (o.a. rigidness), “we

cannot have a floppy designs” and to a lesser extend size.

10. Always pushing the limits of the power source, also outside the

boundary of the specification sheets, and the requirements, e.g. by

changing the requirements.

JPL has the facility to apply concurrent engineering in their design process,

which is different than the described “linear” process of different handbooks.

This is interesting when pushing the limits of technology and changing the

requirements, and quick decision making is essential. The facility of JPL

consists of a room with people involved in the project, in different fields of

expertise, like thermal designers, mechanical designers, mission experts,

telecom and system designers. Thomas and his crew use in-house made

42

tools, T-max implemented in Microsoft Excel, to make the right choice

quickly. T-max is a knowledge based reference tool, which uses technology

tables, and is used to implement parametric design. Paul calls it “Technology

table driven parametric design”. The input is load profiles and

sensor/actuators chosen, and its feasibility is tested. The database metrics

have to be updated every year by experts. To keep up to date with new

technologies Thomas is visiting conferences, reviews new technology in

house by testing the product or component, and uses specification sheets.

The technology experts are questioned to the bone for quantifiable

performance characteristics.

3.3.2 Compact DMFC for portable application

Icardi et al. [100] describe a standard systematical engineering approach

towards designing a 250Watt DMFC system for weather stations, medical

devices, signal units, gas sensors and security cameras. First they model the

system in Matlab/Simulink to evaluate the heat and mass fluxes and

pressure drops. The main system components are described in a block-

diagram, with all of there components and input and output flows of gasses

(air and exhaust fumes) and fluids (fuel and water).

The goal of the model was to choose the optimal combination of components

by evaluating setups varying the membranes (A or B), the number of cells

(30 and 40 cell stacks), with or without insulation, and for different

methanol streams. Optimal performance was not specified in the paper. After

modeling the DMFC system and choosing the optimal combination of

components and layout, the design was not initiated. Next step would be

embodiment and engineering of the outputted system.

3.3.3 Conceptual design and modeling of a fuel cell scooter for

urban Asia

Bruce Lin made a conceptual design of a fuel cell scooter for urban Asia in

his thesis report in 2004 [101]. Chapter 4 of this thesis describes how,

utilizing the technology described in the previous chapters, a fuel cell scooter

is designed. First the requirements are defined and performance is modeled

in Matlab. The model consists of system model of the fuel cell stack and

43

auxiliary systems. Basic assumptions are made to obtain overall measure of

performance, for instance parasitic power is a linear function of gross power,

and conservative assumptions are made where data is poorly known.

When improving the systems performance and costs, Lin considers a high-

pressure fuel cell operation and hybridizes the fuel-cell with a battery for

peak powers. The method of attack of this design is a standard engineering

approach, where first the performance requirements are described. Next a

vehicle model is made, where the load at the wheels is defined, and

inefficiencies of the system parts is described. Based on drive-cycles or load

profiles from literature a normalized drive cycle is described. A fuel-cell

system model was setup, where the number of cells in the stack (small or

large cell) and thus the working power-density of the stack, and the

polarization curve were defining the system performance. Maximum

efficiency and a cheap fuel cell were leading in the optimization.

3.3.4 Flight system options for a long-duration mars airplane

Rohrschneider et al. [102] describe the comparison of five different

propulsion systems to be used in a long-duration Mars airplane: a

bipropellant rocket, a battery powered propeller, a DMFC powered propeller,

and beamed solar and microwave powered propeller system.

The approach of Rohrschneider starts with the definition of the goal

obtaining as many new measurements as possible during the mission flight.

The goal of the systems optimization is to get as much data as possible

during one flight. The problem is attacked in two different ways, by

extending the range or endurance of the airplane, or by keeping the range

and endurance fixed while increasing payload. The first solution of extending

the endurance is chosen which can be done by either changing the

technologies used or by finding a combination of the design variables that

produce a better solution.

After the goal and approach is defined a model of the mission is made,

defined in the form of a Design Structure Matrix (DSM) [104]. The different

analyses are broken down according to their traditional disciplines, of which

44

one is the propulsion trade off. Five different systems are modeled according

to density equations more specified by curve-fitting of performance

specifications and physical variables like size, weight and form. The following

power systems were traded of: a bipropellant rocket, a battery powered prop,

a DMFC power prop, a beamed solar powered prop and a beamed microwave

powered prop.

The focus of the optimization is airplane mass within a constrained volume.

Three values are taken into account for choosing the right airplane for the

job: mass of the airplane but also the rocket (mission gross mass), life cycle

costs (LCC) of the power system and total project, and range (velocity,

endurance). Different airplane sizes are calculated by means of a scale-factor

(running from 0.5 to 1.75). Every size of airplane is calculated and the

endurance is limited by velocity (for larger airplanes) or by volume (for the

rocket powered airplane). In the end a graph is produced in which the

launch vehicle can be chosen for the Mars airplane based on scaled size, the

mission gross mass and for the mission LCC.

To make the choice between the different power sources of the airplane,

ModelCenterTM [105] was used to calculate the gross mass, cruise velocity

(endurance) and the LCC. Varying the scale and the configuration of the

airplane resulted in different optimal architectures/power systems for

different missions.

3.3.5 Discussion

In general a system-engineering approach is used by the interviewed persons.

First the requirements are stated, next a system model is made, and all

components are stated and pushed to its limits by means of simulation and

testing of components. Weight and costs seems to be the key drivers for the

decision making. Volume was often less of an issue, and thus left out of the

decision equation. Especially in a concurrent development environment

there is no time to think about a problem. Engineers in the project-room

expect a quick answer for there instant change of the requirements or added

requirement. This means instant evaluation of the power system is a must,

which puts a lot of pressure on the common knowledge of the power-source

45

engineers. Simple tools are used driven by “technology tables” and

parametric design. For concurrent engineering environments quick

calculations are a must.

In general the case studies also show a standard systematical-engineering

approach towards designing power sources. First requirements are setup,

followed by a system model, modeling the performance of the system. Only

Icardi used a block-diagram, but it is assumed the others did the same

without reporting this. After modeling the system Icardi stopped discussing

the design process for the rest of the paper. The modeling part is seen as the

design. The main objective of this design was performance of the system.

Bruce Lin also used the standard approach towards designing the scooter.

First the requirements were set, and an analytical model of the load and

fuel-cell system was produced. The load is matched with the systems

performance, resulting in a size, weight and initial costs of the power system.

Simple equations were used to generate physical dimensions of the

components of the power system and energy storage.

Rohrschneider developed a Mars airplane and took the total gross costs of

the mission into account of the design. Different power systems of the

airplane have been evaluated which, as a part, is taken into account of the

gross performance. The overall project was scheduled in a Design Structure

Matrix. The trade-off between the different airplane propulsion systems was

based on three objectives: mass within a constrained volume, cruise velocity

(endurance) and LCC. Scaling the airplane gave different results.

3.4 Conclusions The goal of this chapter is to identify available tools, when those tools are

applied in the design process, and how real-life designers evaluate different

optional concepts. Different tools have been found in the literature search

which can be used by the product designer supporting him/her in the

selection process of the power system, during the concept phase of the

design process, but also during embodiment. The usability of the tools are

discussed in the previous section and conclusions are drawn in this section.

46

In Figure 13 an overview is given of all tools and methods described in the

previous sections of this chapter, on the time line of the systematic design

approach. PowerQuest, the CES method and working with Ragone plots are

used during the specification (identification) and concept design phase.

During concept/embodiment design system-modeling is generally used to

evaluate different concepts and make a choice based on a single or multiple

objectives. The POWER and HOMER tools are specifically build to evaluate

battery concepts during embodiment/engineering phase of the design

process, where the application specification is already laid out. The T-max

tool used by the JPL experts is a tool which evaluates the consequences for

the power-source instantly when changes are made in the design or

requirements of the application. This tool can thus be used to evaluate

concepts quickly giving instant consequences for the embodiment of the

design (and vice versa).

POWER

Specification Concept Embodiment Detailing

Powerquest

T-Max

DSM

Ragoon plot

CES

Matlab/Simulink

HOMER, HOGA

Energy hubs

Figure 13: Tools and methods during the design process.

The interviewed experts showed the need for tools where changes in the

design or requirements, or when adding new requirements, are instantly

evaluated. Table-driven parametric tools, like their in-house made T-max

47

software, are used by the experts. These tools are based on expert knowledge,

specification sheets of new technologies and components, and in-house

evaluation of the components. Especially in a concurrent design

environment instant evaluation tools like T-max are a must for the engineers,

meaning instant feedback of the consequences of changes in the design or

requirements.

In general a systematic approach is used to bring an idea, via specifications,

conceptualization to embodiment and finally an engineered product. For

every phase of this engineering design process different tools are available.

There are tools which (could) help the designer in identifying opportunities

for alternative power sources like the CES method, Ragone plots and

PowerQuest. These identification tools are based on normalized figures like

energy density, and therefore not always correct. The numbers outputted

from these programs cannot be used to evaluate concepts but merely

identifies the opportunity. Improvement in these tools is needed giving more

than just initiatives or making alternative power sources visible. Tools, like

T-max , which evaluate the power system instantly when changes in the

design are made, are a must for the concurrent design engineer.

Generalized and instant evaluation tools should be developed to give the

designer improved basis for their concept choice.

In the case-studies the tools made were primarily used for the goal of their

single project. The tools were therefore made in software made for modeling

and simulation, like Matlab/Simulink. The goals of the case-study models

came down to optimization of system performance and minimizing (initial)

costs. The Matlab/Simulink tools simulate only one specific setup, and

optimization is mainly done by changing parameters by hand.

49

4 Research Questions

4.1 Answers to the initial research question The initial research question defined in Chapter 1 has been researched in

Chapter 2 and 3:

“Which power systems can compete with the commonly used

rechargeable lithium-based battery in the application field of

portable electronic devices, and how can a systematic approach help

product designers select appropriate power systems during the

preliminary design phase?”

In Chapter 2 a diversity of competitors for the rechargeable battery has been

explored. A thorough search to power systems and energy carriers has been

presented and is evaluated. From the desk search it can be concluded that

there are two power system-fuel combinations of interest as an alternative

for the lithium based rechargeable battery:

(i) The high power-dense internal combustion engine (ICE) in

combination with the high energy-dense carbon fuels as petrol,

and

(ii) Fuel cells chemically burning high-energy dense carbon and

hydrogen-based fuels direct to electricity.

Based on power density of the engine, the high energy density of carbon-

based fuels and the low-price of power-system the internal combustion

engine seems to be the best alternative power system. Because of high noise

and toxics output the internal combustion engine is especially of interest for

outside applications, where a high power output is required like in

50

construction work environments. The second interesting power system, the

fuel cell system, is a clean, quiet and energy dense option. In low-power

application the low-temperature fuel cells are of most interest, resulting in

two options the hydrogen based fuel cell (PEM cells) and fuel cells using a

liquid fuel like methanol (DMFC). Hydrogen is complicated because of its low

energy density, and thus needing high pressure vessels to make the system

small. Methanol on the other hand is liquid, has a high energy density, and

is convenient in use, making the fuel a competitive energy carrier. Therefore

the research in this thesis will focus further on this promising alternative for

the lithium-based rechargeable batteries.

In the same chapter the field of portable electronic applications has been

explored and defined as devices using electric energy, with an integrated

power system and carriable in once bag. The size of the device has been

limited to the largest found [22] laptop computer size (<6.2L) and weight

(<4.5kg). Minimum values for these two properties have not been defined.

Devices which are implantable and low-power long-endurance applications

which are not intended to be carried around are left out of the scope of this

research. This does not mean all low-power long-endurance applications are

out of the loop. At the moment no products satisfying this definition have

been found, but in the future the rise for these kinds of products are not

dismissed and thus taken into account during this research.

To explore the systematic approach helping product designers selecting a

power system during the preliminary design phase a state-of-the-art

literature review and case-study review is executed in Chapter 3. Different

tools and methods have been found which could help the product designer

in his/her search for a power system during all phases of the design process.

Four different search and selection type of tools can be distinguished:

comparison charts, selection tools, design tools and optimization tools. There

are tools which (could) help the designer in identifying opportunities for

alternative power systems. These identification tools are based on

normalized figures like energy density, and therefore not always correct. The

numbers outputted from these programs cannot be used to evaluate

concepts but merely identifies the opportunity. Improvement in these tools is

needed giving more than just initiatives or making alternative power systems

51

visible. Tools, which evaluate the power system instantly when changes in

the design are made, are a must for the concurrent design engineer.

Generalized and instant evaluation tools should be developed to give the

designer improved basis for their concept choice. Based on the results from

the case-study review a generalized approach of the engineering designer is

to produce its own evaluations tools in mathematical programs as

Matlab/Simulink. Optimization is done by evaluating different setups by

changing different parameters by hand.

4.2 Redefined problem definition In general the systematic design approach is described by [82, 83]: Laying

out a specification, followed by exploration and evaluation of potential useful

concepts, and based on this evaluation the most promising concepts are

elaborated by producing potentially embodiments, which are evaluated. The

most promising embodiment is elaborated further to produce the final

detailed design. Every phase in this design process can be broken down in

four stages, namely analysis, heuresis, evaluation and the stage in which a

choice has to be made [82]. During the analysis the intellectual activity is

very low, while the heuresis stage is intellectually challenging for the

designer [106]. The latter two stages show less strenuous intellectual activity.

The measured high strain during the heuresis phase is probably caused by

the lack of vision on a clear end to the process: ideas keep coming and there

is no guarantee that the discovered ideas are the best, or at least comparable

to the best. This makes this stage in a phase of the design process a good

place to help the designer in making up their mind.

Usually a power system is chosen at the very end of the design process [11]

instead of at the begin phase of conceptualization. Successful, compact

design of a portable application depends on proper selection of components

and also a matching power/energy system. A large part of the product is

defined in the concept phase, and decisions in this phase are merely first

assumptions. During this phase the product design engineer wants to size

the application and has to make important and sometimes irreversible

choices. The engineer makes use of knowledge developed at school and

university, experts, experience from past projects, or mother wit. Because

the power system is one of the main components, taking up 10 to 30% of the

52

applications volume, definition of the component is of great importance.

However power-system selection tools during concept phase of the design

process are available but not commonly used or lack detailed output for

sizing and dimensioning the power system and thus give an accurate

feasibility of the power system alternative.

Internal combustion engines are very power dense and the combination with

the high energy-dense petrol fuels is very promising. Drawback of this

system is its low system efficiency, its high noise output and output of

exhaust fumes. The DMFC system in combination with the energy-dense

methanol-fuel seems to be a good runner-up. This power system has a good

power-dense power-generator in combination with a practical fuel, methanol.

Problem only is how to get the designer to identify this power system as an

alternative and the differences between a fuel cell system and a battery

implies that the researcher/designer has to design the power system instead

of merely select it like a battery.

4.3 Main research questions DMFC systems seems to be a good follow-up on batteries and this thesis will

focus further on DMFC power systems and test the feasibility in portable

electronics. It will explore the process of designing and evaluating fuel cell

systems during the concept phase of design. The main research question is

as follows:



design process?”

In the context of this scientific engineering study the question implies on the

technological feasibility, visibility for the designer and user acceptance. To

answer the main question several sub-questions have been formulated:

1. Which technological, physical and economical properties differentiate the

DMFC power system from the lithium-based rechargeable battery in the

field of portable electronic devices?

53

2. Which demands with respect to DMFC power systems in portable

electronic devices are most significant from a user point of view, and

how can we use this in a design method?

3. How can we identify the opportunities of DMFC power systems for

portable electronic devices, during early phases of the design process?

4. What are the basic properties that help the designer to model a DMFC

power system from a techno-economical perspective?

5. Which algorithms can be formulated to give the product designer not

only insight in the opportunity of DMFC power systems, but also help

the designer in an appropriate systematic approach

4.4 Methodology As already described in the introduction the focus of Part II is to investigate

the differentiating properties from a technological and user point of view. The

results from these studies, described in Chapter 5 and 6, are used to define

different orders of modeling (Chapter 7), which is an introduction to Part III:

metric-based selection, design and concept evaluation during the conceptual

phase of the design process.

4.4.1 Approach

In the continuation of this thesis the technology of DMFC is compared to the

lithium-ion battery in Chapter 5. The comparison is based on primary

properties, like volume, weight, costs and life cycle specifics. The

technological review is followed by a user acceptance test where the

consumer is assessed in Chapter 6 on the acceptance of DMFC systems as

an alternative for the lithium ion battery. The user is tested by means of a

conjoint het analysis, where the differentiating properties of DMFC and

lithium-ion batteries are used as basis, like volume weight, energy density,

costs and period of charging. Three orders of models are proposed in

Chapter 7, followed by exploring only the first two orders, a heuristic

approach (Chapter 8) and a database driven parametric design approach

(Chapter 9). Both models are metric-based identifying tools and are for use

during the conceptual design phase.

54

4.4.2 Generalization

The research approach chosen delivers first insights in the potential

application of DMFC power systems in portable electronic devices, and of a

potential systematic design approach, which is tested via simulation and the

creation of several engineered product designs. Given the fact that the

research activities took place within an office environment, the findings have

to be considered as indicative in the first place. Also only one electronic

product has served as a case study, which in theory limits the generalization

of the results. Although validation has been sought in peer reviews and

presenting interim results of the study in journal papers and at conferences,

the study and its results should be considered as mainly of an explorative

nature.

4.4.3 Validity and reliability

Validity of the study refers to the measures taken to check the accuracy of

the results by following certain transparent procedures. Within this thesis

different methods are used to acquire data around the scope of the research:

data acquisitions by means of data mining (Chapter 2), interviews and case-

study reviews (Chapter 3), user testing with the conjoint analysis (Chapter 6),

and design research with the help of predictive models (Chapters 5, 8 and 9).

All measures described by Creswell [107] and Baarde and De Goede [108]

have been taken into account to assure validity which will be discussed in

Chapter 10. For instance the data acquired, mined and presented in Chapter

2 is based on different sources and triangulation is used to make a concise

and valid dataset.

The reliability of the data is guaranteed by using a systematic approach to

data collection and evaluation of the data by using statistics (Chapter 2), and

for the user studies (Chapter 6) the protocol was followed belonging to the

method used. The reliability of the first-order predictive model presented in

Chapter 8 is validated by means of two cases from the field. To ensure better

reliability of the models more designs over a range of applications, have to be

evaluated which is discussed in the recommendations of Chapter 10.

55

5 From batteries to DMFC power systems

Within the field of alternative power systems, Chapter 2 defines DMFC

systems as the potential successor for the lithium ion based battery when

applied in portable electronic devices. To make a fair comparison and define

differentiating properties between a DMFC system and a lithium ion battery,

this chapter will give an introduction to secondary batteries in Section 5.1,

with the main focus on lithium based batteries, and the direct methanol fuel

cell system (Section 5.2). In Section 5.3 the difference between the DMFC

and lithium-ion battery is discussed on basis of volume, weight, costs and

life cycle specifics. The results are discussed in Section 5.4. Section 5.5 will

give conclusions on the comparison. The goal of this chapter is to describe

the differentiating properties between the DMFC system and the lithium ion

battery and lay down the first steps towards a metric-based design method.

In Chapter 7 the differentiating properties described in this chapter will be

evaluated for a set of metrics used to choose or design a DMFC system.

5.1 Introduction to batteries A power-system used in portable electronics consists of a Voltage generator

in combination with an energy storage device. In battery systems this is

combined in one device. The basis of batteries is in the use of electrode

materials and use of electrolyte. The active materials in the anode and

cathode define the voltage difference, where, during the discharge, electrons

are produced or oxidized at the anode (negative electrode) routed via an

alternative route, through an external load, to the cathode (positive

electrode). There the electrons are accepted and the cathode material is

reduced [109]. The electric circuit is completed in the electrolyte by the flow

of anions (negative ions) and cations (positive ions) to the anode and cathode,

respectively (Figure 14).

56

An

ode

Ca

thod

e

flow of anions

flow of cations

Electrolyte

− +

e

Figure 14: Electrochemical operation of a discharged cell.

Batteries come in different shapes: cylindrical, button size, prismatic and

pouch cells. The latter two are most commonly used sizes in portable

electronics where thickness is a constraint. In these applications a large

battery surface area is used with a small thickness (Figure 15). The

prismatic cells have a hard casing in contrary with the Pouch cells which is

kept together with a heat-sealable foil, making the battery having the highest

packaging efficiency amongst battery packs [110].

Figure 15: The Nokia BL-5C lithium ion battery.

57

5.1.1 Theory

“A battery is a device that converts the chemical energy contained in its

active materials directly into electric energy by means of an electrochemical

oxidation-reduction (redox) reaction.” [109]. Two types are feasible, the

primary, single use, battery and the secondary, rechargeable, battery. In the

case of the latter battery type the battery is recharged through reversal of the

process.

The term “battery” refers to one or more “cells”, or electrochemical units,

connected in series or parallel. A cell consists of three major components: 1)

the anode or negative electrode, 2) the cathode or positive electrode, and 3)

the electrolyte or ion conductor (Figure 16).

cathode(NiOOH)0.49V

anode(Cd)

-0.81V

+

_

e

e

Figure 16: Basic principle of a Nickel Cadmium cell.

The combination of electrodes defines the cell voltage and capacity of the cell.

Research in battery development is focusing on the most advantageous

combination of anode and cathode materials, which is referred to in the

name of the cell, e.g. Nickel (NiOOH) Cadmium (Cd). Besides performance

other aspects are taken into account like costs of the materials, difficulty in

handling and other deficiencies. The anode is selected with efficiency, high

Coulombic output (Ah g-1), good conductivity, stability, ease of fabrication

and low costs in mind. Metals are mainly used for the anode and because

58

lithium being the lightest metal available, this is predominantly used since

the beginning of the nineties [109].

The working principle of a battery is explained by means of an example of

the NiCad battery (Figure 16). During discharge cadmium metal is oxidized

at the anode to cadmium hydroxide plus two electrons. At the cathode the

electrons are accepted and nickel oxide is reduced to nickel hydroxide. The

electrons are directed externally via an electric circuit resulting in an

electron flow, or current. The electric circuit is completed in the electrolyte

by the flow of negative and positive ions to the anode and cathode

respectively.

Anode: Cd + 2 OH- Cd(OH)2 + 2e-

Cathode: NiOOH + H2O + e- OH- + Ni(OH)2

Overall: Cd + NiOOH + 2 H2O Cd(OH)2 + 2Ni(OH)2

During charge the current flow is reversed and oxidation takes place at the

positive electrode and reduction at the negative electrode. The reactions as

described above are hereby reversed.

5.1.2 Theoretical performance specs and practical values

The theoretical voltage of a cell can be determined by the type of electrode

materials contained in the cell. It can be calculated from the free-energy data

obtained experimentally. As presented in Figure 16 the nickel cadmium

battery has a positive nickel electrode potential, or reduction potential of

0.49V and a negative cadmium electrode potential, oxidation potential of -

0.81V. The standard cell potential can now be calculated by taking the

difference between the cells potentials, in this case E0 = 0.49+0.81 = 1.30V.

For lithium based low-temperature rechargeable cells the theoretical

potential ranges in between 3.50 to 4.10 for Lithium-ion and

Lithium/Manganese dioxide cells, respectively [109].

The capacity of the battery is determined by the amount of active materials

(electrodes) in the cell and expressed as the total quantity of electricity

involved in the electrochemical reaction, defined in terms of coulombs or

59

ampere hours. The theoretical capacity of electrochemical cells can be

calculated by means of the electrochemical equivalent weight (Ah g-1) of the

anode and the cathode. The equivalent weight of cadmium is 1.34 Ah g-1 and

for nickel this is equal to 0.292, making the overall theoretical capacity equal

to 0.182 Ah g-1.

For lithium based rechargeable batteries the theoretical capacity is equal to

0.100 Ah g-1 for Li-ion cells, to 0.286 Ah g-1 for the Lithium/Manganese

dioxide cell.

The theoretical energy density can now be calculated by multiplying the

capacity with the cells voltage. All specifications of different, commonly used,

rechargeable cells are described in Table 6.

Table 6: specifications of different rechargeable cells, based on [109].

Theoretical values Practical values

Voltage V

Capacity C

Spec. energy

u

Voltage V

Spec. energy

u

Energy density

uρ

Battery type

Anode

Cathode

(V) (Ah kg-1) (Wh kg-1) (V) (Wh kg-1) (Wh L-1)

NiCD Cd NiOOH 1.30 181 235 1.2 35 100

NiMh MH NiOOH 1.35 178 240 1.2 75 240

Lithium-ion LixC6 Li(i-x)CoO2 4.10 100 410 4.10 150 400

Li-Manganese dioxide

Li MnO2 3.50 286 1001 3.00 120 265

Besides theoretical values the practical values of different battery systems

are also described in Table 6. These numbers can be used for an initial

choice of battery when designing an application needing a secondary battery.

For portable equipment the selection between a primary or secondary battery

is not clear-cut and influenced by many factors, like use time and contextual

requirements. If power requirements are low and the application will be used

infrequently over a long period of time, the primary battery is the better

choice. When the application is used frequently, life-cycle costs and the user

“convenience”, like freedom from a charger, should be taken into account. In

general the designer can use the following rule of thumb taken from [111]: “A

product should work on a replaceable energy source for a month at least to

60

be attractive to the consumer”, meaning that primary batteries are best for

delivering power over a period longer than a month, either frequent or

infrequent, and a secondary battery is the main choice when the application

has to be “recharged” more often than every month.

5.1.3 Specs, different sizes and applications

The theoretical voltage and capacity is defined by the type and amount of

active materials used. The value is higher than the practical values, as can

be seen in Table 6. In practice only a fraction of the theoretical values is

realized, because of the need for electrolyte and passive components like the

container, separators and current collectors. This need for more passive

components contribute to the volume and weight of the battery, making the

practical values lower than the theoretical. In general only 25 to 35% of all

material used in batteries is active [109].

Batteries and cells are available in four forms, the cylindrical (spiral wound

or bobbin construction), button, prismatic and as pouch cell, Figure 17.

When a high energy density is preferred, mostly cylindrical cells are used.

This is because of their low surface-to-volume factor and thus low lost of

space because of the need for the seal and other construction materials. The

surface-to-volume factor also influences the high rate performance. Higher

ratios result in better heat dissipation, thus higher possible discharge rates.

Flatpack designs have a high surface-to-volume ratio making these designs

perfect for higher discharge rates.

Figure 17: four forms available for primary and secondary batteries, from left to right, the

cylinder, button, prismatic and pouch cell.

For laptop computers mostly cylindrical cells are used where discharge rates

(C) are low but long user-time is preferred. For smaller applications like

PDA’s, smart phones and cell phones thickness of the battery is guiding and

mostly thinner prismatic or pouch cells are used. Pouch cells uses, in

61

contrast to prismatic cells, heat-sealable foil, achieving a packaging

efficiency of 90 to 95% [110], making this cell besides high energy dense,

also easy to form.

5.2 DMFC explained Fuel cells convert a fuel (for instance hydrogen) in electricity via a chemical

process (see Figure 18). Depending on the type of fuel cell the conversion

efficiency is between 20 to 90%. When hydrogen is used as the fuel the only

emission is water, making it a clean energy storage system.

anode cathode

membrane

6H+

CH3OH + H2O

CO2 3H2O

1.5O2

6e-

Figure 18: basic principle of a direct methanol fuel cell.

There are a range of fuel cell types all with their specific characteristics and

developments, in different stages of development. Different fuel cell types can

be broadly grouped based on their working temperature, Figure 19. In the

high temperature range (600 – 800 degrees Celsius) solid oxide fuel cells

(SOFC), and molton carbonate fuel cells (MCFC) have a very high efficiency

(55 - 70%), and are mainly used for stationary electricity generation. In the

low temperature range (80 – 200 degrees Celsius) Proton Exchange

Membrane (PEM FC) and Direct Methanol Fuel Cell (DMFC) are the most

promising. Efficiencies vary from 20 – 50%.

62

Molten Carbonate

Solid Oxide FC

Polymer Exchange Membrane FC

Direct Methanol

Phosphoric Acid FC

Alkaline FC

Regenerative FC

Working temperature 50 100 150 200 250 … 400 600 800 1000

Figure 19: A temperature differentiating overview of the commonly used fuel cells system. DMFC systems are compact power sources capable of converting a chemical

composition in a fuel to electric energy. The fuel cell combusts fuel atoms

(CH3OH) into H+ protons and electrons e-. The H+ protons pass the fuel cell

membrane, often the Nafion 117, while the electrons cannot pass this

membrane. The electrons are routed around by short-circuiting the anode

with the cathode (Figure 18) by adding a load [112]. A current will flow giving

power to any kind of application.

5.2.1 Theory

Fuel cells are like batteries, electrochemical cells that convert chemical

energy directly into electricity. The main difference between fuel cells and

batteries is that the active materials are not an integral part of the device,

but are fluids or gases fed to the electrodes of the cell from an external

source. As long as the active materials are fed into the cell the power source

can deliver a current, a voltage, and thus power. For a PEM cells and DMFCs

the “active materials” are hydrogen and oxygen or methanol/water and

oxygen, respectively. As a result the power performance depends on the fuel

used and the energy contained by a fuel cell system depends on the amount

of fuel in the container.

The fuel cell can best be compared with primary batteries because in most

cases the fuel cell does not produce the fuel when “recharged”. Most fuel

cells are reversible, but because of efficiency problems this is mostly not the

case. Fuel cells can be classified into two categories: the direct systems,

63

where fuel can directly react in the fuel cell, and indirect systems where first

fuel is converted by means of reforming to a hydrogen-rich gas which on its

turn is fed into the fuel cell.

Figure 18 shows the basic principle of a fuel cell. In this case the Direct

Methanol Fuel Cell. In this fuel cell the methanol/water mix (CH3OH + H2O)

is the active material fed into the anode, the fuel electrode, and air/oxygen

(1.5 O2), is fed into the cathode, the oxygen electrode. The cell produces

water (3 H2O) plus carbon dioxide (CO2) as a by product. The overall cell

reaction in the DMFC is represented by the following equation:

H2O + CH3OH + 1.5 O2 CO2 + 3 H2O

During this reaction 6 electrons are produced, which are routed bypassing

the cell. The membrane is the electrolyte of the cell which only conducts H+

ions (cations). In DMFCs the membrane is the same as for hydrogen fueled

PEM cells, often the Nafion 117 membrane.

5.2.2 Theoretical performance specs and practical values

Just like with batteries cells, the Voltage depends on the active materials

used. The theoretical reversible open cell voltage E for DMFCs can be

calculated with the molar Gibbs free energy:

fg

EzF

−Δ= (3)

Where the molar Gibbs free energy fgΔ for methanol is equal to -698,2 kJ

mol-1 [112], F is Faradays number equal to 96485 C mol-1, and z is the

number of electrons passing from the anode to the cathode. In the case of

DMFCs the number of electrons is equal to 6 for each molecule of methanol

(see Figure 18). This will result in a theoretical open cell Voltage equal to

1.21V. In Table 7 all data is described for both the PEM cell and the DMFC,

the most commonly used fuel cell type for portable applications.

Just like for batteries the capacity of a “fuel cell system” depends on the

amount of active material used by the cell. For DMFCs methanol and oxygen

64

are the active materials and the theoretical energy contained in the system

can be calculated by multiplying the amount of fuel, in kg, with the lower

heating value LHV of the fuel fhΔ which is equal to 19.93 MJ kg-1, or 15.75

MJ L-1.

Table 7: specifications of different low-temperature fuel cells (<100 degrees C), based on [112] and [13].

Theoretical values Practical values Specific energy

ufuel Energy density

ufuelρ

Fuel

cell type

Anode CathodeOC

voltage E0

(V)

(MJ kg-1)

(kWh kg-1)

OC voltage

VOC (V)

(MJ kg-1)

(kWh kg-1)

PEM FC 2H2 O2 1.16 120 33.3 - 32.4 9.0 DMFC CH3OH+H20O2 1.21 19.93 5.54 0.72 3.4 0.94

In practice the performance of fuel cells (the power generator) is lower than

the theoretical values. In practice the theoretical open cell Voltage is never

reached for both DMFC and PEM cells. Based on a multiple regression

analyses executed on four DMFC cases describing 47, changing fuel and air

flow, membrane surface area, platinum loading on anode and cathode,

working at different temperatures, and different current flows [113-116], an

equation is extrapolated which produces the open cell Voltage as a function

of cell temperature T, the methanol concentrated in water N and the

platinum loading on the cathode mc (Appendix A)

457 1.58 10.8 18.6OC cV T N m= + − + (4)

With this equation a fuel-cell performance model can be defined with the

open cell voltage as the maximum voltage and the different losses deducted

according to [112, 117, 118]:

( )cell OC act x over massV V V V V VΩ −= − Δ − Δ + Δ − Δ (5)

Where:

0.19V i r iΩΔ = ⋅ = ⋅ (6)

( ) ( )log 0.2295 logactV A i T iΔ = = (7)

65

( ) ( )log 0.2295 log 0.25x over cV A i T i−Δ = = (8)

( ) ( )exp 0.0211exp 0.04massV m ni iΔ = = (9)

Figure 20 shows the analytical model together with polarization curves from

more recent design cases described in literature [119-121].

Figure 20: Polarization curves of different measured Nafion membranes taken from [119-121]. The dashed line represents the curve calculated according the equation described

above.

5.2.3 Architecture and component overview of the DMFC

system

The fuel cell is only a power generator. If the fuel cell is coupled with a tank

and other auxiliary components it becomes a power system. To generate a

controlled power output two types of systems can be categorized: the passive

and the active DMFC. A passive fuel cell system generates power without the

use of other powered components. The tank with fuel (normally a 3%

solution of methanol in water, N=1), is fed directly to the anode, with for

instance wicks, and the cathode is open to the outside air [122-124]. Passive

systems’ advantage over active systems is the lack of energy consuming

auxiliary components, making the system simple, cheap and requiring little

maintenance. The drawback of passive systems is among others their low

efficiency and its sensitivity to direction.

66

Active system on the other hand depend on auxiliary components as pumps,

sensors and other actuators to generate a power which can follow the load

and make efficient use of the methanol available in the tank. Active systems

consist of at least the fuel cell, the tank, an air pump and a fuel pump.

Figure 21 gives an overview of all components in an active DMFC system

described in a functional layout. Note that not all functions are essential and

not always need to be actively powered. The fuel-cell system boundaries are

described by the grey field.

Fuel Tank

anod

e

Methanol feed Mixing

chamber (1M)

Water feed

Air

cath

ode

CO2 venting

Boost converter

Load

gaseous flow fluid flow electric flow information flow sensors (flow, methanol, power (Vi), temperature)

Air filter H2O condns.Humidifyer Heat exch.

+ -

+ -

H2O condns.CO2 filter

fuel mix

fuel mix

CO2 Air

CH3OH

H2O

Clean Air or O2

fuel mix +

CO2

H2O+ Air

H2O

m

p

f m

p

f

on/

off

μC intermediate accumulator

Fuel Cell System

Figure 21: Functional analysis of a direct methanol fuel cell system.

5.3 Comparison “Among all types of fuel cells, direct methanol fuel cells have exhibited the

greatest potential to replace lithium ion batteries for portable and micro

power sources” [122]. The energy density of methanol is high compared to

batteries available and the amount of time to “recharge” a DMFC power

source, by replacing the fuel cartridge, is instant compared to recharging a

lithium ion battery.

67

To show the differences in lithium ion batteries and DMFCs a comparison is

made based on performance specifications as power and energy for a certain

amount of volume and weight, and initial pricing and life cycle costing of

both systems. First a preliminary model is presented which is used to make

quick metric-based feasibility studies for DMFC systems in different portable

electronic devices.

5.3.1 Preliminary model

A fuel-cell power system consists of different components. The system can be

broken down into three parts:

1. the fuel cell, which converts the methanol mix into electricity,

2. the fuel tank, which contains the methanol and water in a mixture

or separately,

3. auxiliary units, like pumps, electronics, plumbing, and wiring, called

the Bill Of Products (BOP).

Besides the components, the systems volume consists also of empty space or

in other words ‘dead space’ which does not add to weight. In [29] a design of

a fuel cell system is presented, which produces the electricity for charging

the ImpresTM battery of a Motorola cell phone6. The volume and weight of the

systems can be broken down into three main contributors, the fuel cell, the

BOP, and fuel tank. The total volume and weight of the 2 Watt system is

0.69 kilograms and 0.92 liters. In Table 8 the break-down in the main parts

is listed.

Table 8: Volume and weight of the three major parts of the Motorola DMFC charger. Volume

(mm3)

Weight

(g)

Fuel cell stack 87,000 180

Fuel tank 413,000 155

BOP 416,000 155

6 The Motorola fuel cell systems is an actively fueled micro-DMFC (graphite version),

with a peak power of 6.7WP, an efficiency of 20%, containing 336Wh of electric energy prototype

68

Based on these figures a first approach to an analytical model is proposed in

which the volume and weight of a fuel cell system is sized.

The volume of a DMFC system can generally be based on two variables, peak

power (P) and the amount energy of energy (Wh) has to contain. It is

assumed that dimensions of the fuel cell stack and the BOP is based on the

power specifics (P) and the dimensions of the fuel tank are only based on

energy specifics (E). Empty space is included in the BOP volume part. The

following analytical model of the volume in liters is than derived:

( ) ( ) ( )can

fc fuel bop

peak peak

sys vfc fuel bop

V V V V

P PE

p c u pρ η ρ ρ

= + +

= + + (10)

Where: (pρ)fc = 77.3W L-1

(pρ)fuel = 4373Wh L-1

(pρ)bop = 16.1W L-1

ηsys = 20%

cv can = 0.85

The energy density of methanol is based on its Lower Heating Value. The

system efficiency ηsys and volumetric canister coefficient cv can are based on

figures from [29]. BOP is the largest contributor to the total systems volume

(~45%). Less than 20% of the BOPs volume is claimed by the electronics, air

pump, mixing chamber, fuel pumps and others, meaning more than 80% of

the total BOP volume is claimed by empty space, electrical interconnects and

plumbing [29, 125], equaling 35% of the total systems volume.

For weight the same assumptions are made as for volume. Empty space does

not add an increase of the systems weight and is thus not included (note:

theoretically air is not weightless, but assumed to be in this case). The

following analytical model can now be derived:

69

can

fc fuel bop

peak peak

fc sys m fuel bop

m m m m

P PE

p c u pη

= + +

= + + (11)

Where: pfc = 37.2W kg-1

ufuel = 5536Wh kg-1

pbop = 43.2W kg-1

ηsys = 20%

cm can = 0.80

The specific energy of methanol is based on its Lower Heating Value.

These simple models are to compare the DMFC system with other power

sources, like the lithium ion battery, by means of a so-called Ragone plot, as

figured in Figure 22 where the performance specifications of the Sony

3834450A8 [126] prismatic lithium ion battery (4.2V, 7Ah) is compared with

the performance specifications of the DMFC system, described above.

With the help of the simple models derived above an optimization can be

made, for instance when the efficiency of the system is improved from 20%

to 25%, and when empty space is neglected, this will influence volume and

weight.

70

a) Specific power versus specific energy

b) Power density versus energy density.

Figure 22: Ragone plots for volume (a) and weight (b) of the derived DMFC systems compared with the Sony lithium ion battery [126].

71

5.3.2 Volume specifications

The volume of cell phones and their batteries have decreased with a factor 2

over the past 10 years. In contrast to battery weight the battery-volume

percentage in mobile phones was constant over the past 10 years at

20%±10%(SD) of its total volume. The volume of PDA’s has decreased over

the period from 1995 to 2004 to ~150 cm3, but their batteries have increased

with a factor 1.5 in volume. This results in an increase of the batteries

volume in PDA’s to 20-27% of the total volume. For laptop computers the

batteries volume ranges in between 6 and 11% of the total volume. The

magnitude has increased slightly over the past 10 years.

The theoretical energy density of Methanol based on the lower heating value

is 4373 Wh L-1. At a system efficiency of 20% the ‘practical’ energy density of

methanol will be around 875 Wh L-1, an improvement of almost a factor 2

compared to the Sony lithium-ion battery described in Table 9.

Table 9: Overview of the characteristics for one primary (alkaline), several rechargeable

batteries and the prototype DMFC from Toshiba. Alkaline

Duracell*

MN1500

NiCad

Sanyo

N3US bulk

NiMH

Sony**

NHAAB4E

Li-ion

Sony

383450A8

Li-poly

Varta

easypack

1000

DMFC

Toshiba***

Dev. phase**** C. C. C. I. C. P.

Shape Cyl. Cyl. Cyl. Prism. Prism. Prism.

Capacity (mWh) 2850 1000 2500 830 1070 2000

Weight (g) 24 23 30 14.3 23 8.5

Dimensions (mm3) 50x∅14.2 50x∅4.2 50x∅14.2 50x35x3.8 62x35x5 22x56x9.1

Energy dens. (Wh L-1) 428 151 379 475 365 178

Spec. Energy (Wh kg-1) 143 52 100 221 172 235

Cycle life 80% (-) 1 300-700 400-500 400-500 >500 -

Normal cell Voltage (V) 1.5 1.2 1.2 3.7 3.7 0.4

* Duracell AA battery, MN1500.

** Sony’s NHAAB4E Rechargeable NI-MH AA Battery Pack, NHAAB4E.

*** Based on a 20h usage at 100mW the energy needed is 2Wh. The tank contains 2cc of

methanol at a 99.5% concentration, containing an amount of energy equal to 0.256MJ =

71.1Wh. The overall efficiency of the system is than equal to ~2.8%.

**** C=commercial, I=introduction, P=prototype phase

72

This table also shows the energy density increase over the past 10 to 15

years in battery technology. The prototype DMFC system of Toshiba [43]

shows that, in contrast to its weight specifics, the volume specifics of the

DMFC system is worse than all other primary or secondary battery presented.

They are even worse when compared to state-of-the-art commercial lithium

ion batteries.

Based on the analytical model an estimate is made for the fuel cell system

volume when applied in different applications. The data is presented in Table

10 and is visualized in Figure 23. From these figures it can be concluded

that BOP is one of the largest contributor to the volume of the total system.

It must be noted that more than 86% of the total BOP volume of the

Motorola design was dead space, electrical interconnects and plumbing, and

only 14% is claimed by the electronics, air pump, mixing chamber, fuel

pumps and others (Figure 24). Thus, diminishing dead space is one of the

key issues to increase energy and power density.

Table 10: Comparison of real power and energy characteristics of the existing battery (left four columns) and the estimated size according to Equation 10 when a DMFC system is

used (right two columns). Power

(W)

Battery

(Wh)

Volume

(cm3)

Energy

density

(Wh L-1)

Estimated

FC volume

(cm3)

Estimated

Energy

density

(Wh L-1)

Sensors 0.010 0.15 0.7 214 0.93 160

Flashdrive MP3 0.025 1.1 5 220 3.23 341

Harddrive MP3 0.25 2.9 6.6 439 22.3 130

Mobile phone 0.80 3 10 300 63.7 47

PDA 1.50 3.5 16.3 215 117 30

Laptop computers 30 65.1 333 195 2331 28

73

Figure 23: Estimated improvement in volume for a (active) fuel-cell system applied in different consumer products (estimated value is pictured as a percentage of the real

value).

Figure 24: 1Watt DMFC system inside a modified Motorola IMPRES charge housing [29].

When looking at sensors and the flash-drive MP3 player, the DMFC system

will be an improvement compared to the standard battery (factor 2). For

these power systems the fuel cell is only a small part of the total power

system.

74

For power systems with power consumption of 800mW and up, the DMFC

will not improve volume characteristics (even when BOP is neglected). In this

application field key research issues are:

• Miniaturize the fuel cell by improving power density (W cm-2),

• improving the fuel cell efficiency (%),

• bringing down peak power by means of power management and

hybridizing with a battery,

• decreasing BOPs volume by diminishing dead space in the system,

• and miniaturize auxiliaries like pumps and blowers and their

electrical interconnects and plumbing.

A hybrid system could probably improve the volumetric characteristics of the

power system drastically. Based on the preliminary model an increase of the

systems efficiency to at least 27% will make the fuel-cell system’s volume

competitive with the standard battery, for application in devices as the cell

phone, PDA and laptop.

In general fuel cells outperform the battery in areas that need high energy

and lower peak power. Compared to the fuel cell the battery is a high power-

dense energy-source. For high power requiring systems a hybrid system

would be a better solution. This is also under scribed by a presentation from

the Energy Technology Lab from Motorola Labs in 2003 [42, 127].

5.3.3 Weight specifications

The importance of weight in portable electronics is clear. Cellular phones for

instance are carried in ones pocket and a heavy device will result in irritation

by the consumer. DMFC systems have the opportunity to be more

lightweight than batteries. The theoretical specific energy of methanol based

on the lower heating value is 5536Wh kg-1. At a DMFC systems’ efficiency of

20% the “practical” specific energy of methanol could be 1107Wh kg-1, which

is 5 times higher than the Sony lithium-ion battery described in Table 9.

75

Figure 25: Theoretical and practical specific energy for different primary and secondary

batteries, and the PEM and DMFC fuel cell [43].

The practical value of the lithium-ion battery is 37% of its theoretical [109],

and the energy density of lithium-ion and polymer batteries increases with 5

to 10% per year according to [128]. The energy density is overtaking the

density of alkaline batteries. Broussely and Archdale [129] extrapolated their

findings for the lithium-ion battery from 1995 to 2004, and according to

them the maximum attainable will probably be around 275Wh kg-1, almost

~55% of theoretical. Linden [109] notes that “the actual energy that is

available form a battery under practical, but close to optimum, discharge

conditions is only about 25 to 35% of the theoretical energy of the active

materials”.

Toshiba developed a micro-fuel cell [38] that, besides equal weight specifics

to the lithium-ion battery, has more potential to improve. The design reached

76

only 14% of the ‘practical’ value in 2004 as described above, thus open for

improvement.

To obtain a good overview of the opportunities for DMFC systems, the weight

of a DMFC power-system is estimated for different applications in the range

of sensors (10mW) to laptop computers (100W). Again with data acquired

from the Motorola design case the fuel cell [29], the BOP, and fuel tank

weight is estimated, as presented in Section 5.3.1. In Table 11 these

estimates are presented for DMFC power system applied in several

applications. The results are also presented in Figure 26 and the estimated

mass is given as a percentage of the real battery mass.

Table 11: Comparison of real power (left four columns) and energy characteristics of the battery, and the estimated weight according to Equation 11 when a DMFC system is

used (right two columns). Peak

Power

(W)

Capacity

(Wh)

Weight

(g)

Specific

Energy

(Wh kg-1)

Estimated

system

weight

(g)

Estimated

specific

energy

(Wh kg-1)

Sensors 0.010 0.15 2.2 68 0.66 228

Flashdrive MP3 0.025 1.1 11 100 2.40 459

Harddrive MP3 0.25 2.9 13 223 15.53 187

Mobile phone 0.80 3 20 150 43.13 70

PDA 1.50 3.5 26 135 78.66 44

Laptop computers 30 65.1 480 136 1568.01 42

State-of-the-art DMFC systems are mainly fueled with low percentage

methanol-water solutions. Basic DMFC systems are fueled with a 3 to 6%m

methanol in water solution, but literature describes improvements of 30% to

99% [43]. Motorola uses an active mixing system which is a great part of the

total auxiliary system. It is expected that 100% solution is feasible in a short

period of time, meaning no water has to be carried as dead weight.

77

Figure 26: Estimated improvement in mass of a (active) fuel-cell system applied in

different consumer products (the estimated value is pictured in a percentage of the real value).

It can be concluded that for sensors, flash-drive MP3 players and the hard-

drive MP3 player, the active-fueled fuel-cell system would have potential to

be an improvement in comparison with the used battery. In this application

field a factor 2 to 3 mass reduction is achievable with state-of-the-art

actively-fueled fuel-cells. In general half of the weight is occupied by the fuel

cell and the other half by BOP, affecting the total weight greatly. Fuel only

takes up a small part (~10%) of the total weight in this breakdown.

For power systems above 800 mW an active platform will probably not

realize an improvement in comparison with the battery. Mass will increase

even when the electronics are not taken into account. Efficiency of the fuel

cell could decrease overall weight, especially the weight of the fuel cell. For

laptop computers an efficiency increase of the power system (electronics

excluded) should be high (more than 35%) before it can compete with the

lithium ion battery’s weight characteristics.

78

For these “high power” applications the following key issues have to be

resolved:

• Mass reduction of the fuel cell self,

• mass reduction of the auxiliaries,

• integration of components on the electronics,

• improving the fuel cells efficiency or increase power density (W cm-2),

• system optimization by hybridizing the power system,

• and decrease peak power of the application.

5.3.4 Costs specifications

Rechargeable batteries energy densities have improved very fast the last

decades. Large performance improvements have been made between the

NiCad cell and the Lithium-ion battery (since 1990). The Lithium-ion battery

is high energy dense, but is very costly when compared to other battery

systems, see Figure 27. Apparently the price of the battery is of subordinate

concern to the improvement in convenience and decrease of volume. Data for

this figure is derived from Table 5 in Chapter 2, and the price for the grid

(€0.21 kWh-1), methanol (€2.99 L-1 for the EFOY 10L canister), petrol (€1.56

L-1) and LPG (€0.76 L-1) is updated to regular prices in the Netherlands in

2010.

Besides costs per kWh the initial costs of fuel-cell systems are higher than

lithium-ion batteries (Figure 28). Again, data for this figure is derived from

Table 3 in Chapter 2. When we have a look at the cycle costs for the use of

the system, the electric grid is the cheapest of all but rising steadily the last

10 years (used for rechargeable batteries), followed by the alkaline battery

and methanol used for direct methanol fuel cells, which is five times more

expensive than the electric grid.

For DMFC systems the costs of materials used in fabricating are very high

(especially the high cost of platinum electro catalysts). According to Dyer [75]

the production costs for a direct methanol fuel cell could be $5 per generated

Watt, for a 20W / 60Wh power system. A research done by the Darnell group

in 2003 [130] shows that fuel cell systems for mobile phones and laptop

79

computers should be around $4.15–$4.60 per generated Watt to be

competitive with lithium-polymer batteries and the pricing should be closer

to $2.81–$3.71 per generated Watt when competing with lithium-ion

batteries. In 2004 the Smart Fuel Cell SFC-A25 commercial price was €112

per generated Watt, but in 2010 the price is halved to €53 per generated

Watt [76] for the next generation DMFC systems from Smart Fuel Cell, the

EFOY 2200 (90W) system7. Prices are dropping but based on the figures

given by the Darnell Group it has to drop a factor 10 more before it gets

competitive with lithium-ion batteries.

Figure 27: Specific cost per kWh for the fuel and specific energy systems, updated for

2010.

To get a better impression on the costs of fuel cells versus its competitors the

life cycle costs are compared in Figure 29 for a 1W power generator. The

figure shows that the DMFC fuel cell system is very expensive throughout its

life-cycle. This is due to its high initial price but more because of the high

price for methanol this power system seems to be commercially of low

7 The EFOY 2200 sale price is EUR4799.

80

interest. The price for alkaline batteries is very low and for an alkaline AA

battery the prices dropped from €1.74 in 2004 to €1.25 in 2010. According

to this analysis the most cost-effective power system is the alkaline battery.

The life-cycle costs of the rechargeable batteries depend much on the local

electricity price, which increases slowly the last years and strongly depends

on the region you are in. In the Netherlands the electricity price is quite high

compared to other countries in Europe [131].

Figure 28: Specific costs per Watt for different power systems, data from 2004.

The life-cycle costs for the consumer are for both the rechargeable batteries

and a DMFC system higher than a regular alkaline battery. Figure 29 shows

that alkaline batteries are cost effective even at higher energy needs, so why

should the consumer make the transition to a DMFC system. One AA

alkaline battery can hold 2.4Wh of energy. Taking the same size (7.7cm3) for

a methanol canister, it can hold a practical 6.4Wh (at a conversion efficiency

of 17%). This means the number of cartridge replacements for the methanol-

cartridge is 2.7 times less than replacement of the alkaline battery. A

lithium-ion battery can contain up to 1.6Wh in the volume of an AA battery,

which means the “time in-between charges” could be increased with a factor

3.9 when a DMFC system is used. This could be a unique-selling point for

81

the DMFC system used in more energy-hungry devices like smartphones and

portable tablet pc’s.

The initial price of DMFC systems has dropped a factor 2 over the past 6

years, but is still a factor 5 times higher compared to the lithium ion battery.

Initial prices have to drop, but life-cycle costs even more to make the DMFC

power system cost-effective compared to the lithium-ion battery.

Figure 29: Life cycle costs of different energy systems (comparison based on a 1Watt AA

battery), prices have been updated for 2010.

Why are the initial costs of fuel cell systems so high? If we break down the

material costs of the fuel cell we can distinguish the following major

contributors [42]:

• material cost for the membrane and catalyst

• micro fluidic components, such as pumps, the mixer and valves

• fuel cell stack

• electronics

• fuel

82

When we buy a Membrane Electrode Assembly (MEA) at Fuelcellstore [132]

the price is around $105 per generated Watt8. According to the Darnell

group the MEA (GDL, membrane, electrodes, bipolar plate, gaskets and

other) accounts for 40% of the total DMFC costs and 60% can be carried

back to other components and assembly. Total retail price is than equal to

$263 per generated Watt.

The material costs for the membrane and catalysts are significant. The price

for Platinum at the moment is $63.50 per gram [133]. The platinum load on

the Motorola fuel cell is 8 mg Pt cm-2 for the cathode and 10 mg Pt:Ru (1:1)

cm-2 for the anode [29]. The stack has an active area of 110cm2 (5x5cm2, 6

cells) resulting in a price for only platinum used of $125, or $18.65 per

generated Watt. The price for platinum has doubled since the crisis in 2008,

and scarcity is an issue for the future. This means the platinum loading has

to come down, or new catalysts have to be discovered to bring down the

initial price. Acta S.p.A (“Acta”) announced that it had filed a patent

application for a HYPERMEC catalyst for electrolyzers which offers very high

efficiency and yet contains no expensive platinum [71].

Besides material costs the fuel cell development is still in its infant phase.

The Smart Fuel Cell power units are only produced in small series and

focusing on niche-markets like generators for campervans. The development

to improve power and energy densities are still going on. On the other hand

the production of lithium ion batteries is full grown and less costly due to

scale and low assembly effort. The fuel-cell power system consists of a

number of components which increase assembly time and thus cost. This

calls for integration of components, making the system less complex.

According to Motorola the “concept of system-on-chip or power-on-substrate

is the driving force to reduce the complexity of fuel cell assembly and the

manufacturing cost. Highly integrated fuel cell systems with active and

passive micro fluidic components can be achieved by utilizing emerging

MEMS technology” [47].

8 Based on 5 layer DMFC MEA 25cm2 with GDL, $116.90. The Motorola cell has 6

cells and delivers 6.7W, resulting in a price of about $105 per generated Watt.

83

Overall the initial costs of DMFC systems should be cut down, but more

urgent is a needed price-drop of methanol. The costs can be cut down by:

• Lowering or diminishing platinum loading

• Lowering material use and especially dedicated materials like Nafion

• Scaling up production

• Integrating components

• and make use of mass-production manufacturing like injection

molding and etching (PCB)

5.3.5 Life cycle specifics

The times a battery or power system could be recharged or reused depends

on the technical life span of these power sources. For instance the NiCd

battery is troubled with its memory effect bringing back the technical life

span. If charged at regular intervals and always to the max the battery

lifetime will be extended. NiCd batteries can be recharged in between 300 to

700 times and NiMH batteries almost 400 to 500 cycles [134]. Lithium

batteries need a more complex charging procedure because of safety reasons.

The number of recharges is limited to approximately 500 cycles but

improvements are on the way: Toshiba’s claims 1000 cycles [135]. Battery

cycle life is limiting the maximum technical life span a product can be used.

The technical life span for a battery powered laptop computer is in between

1,500 and 2,500 hours (3 to 5 times the number of cycles). The economical

life of portable electronics is two to three years. The technical life span of the

battery will in general outlive the economical life. The degradation rate for

most application batteries is in between 2 to 10V h-1.

DMFC systems have, theoretically, an infinite number of cycles and its

technological life span is limited by material or membrane defect. Knights et

al. [136] describe the “aging mechanisms and lifetime of PEFC and DMFC”

modules. When using load cycling strategy the degradation rate of a DMFC

is 13 V h-1 for almost 2000 hours straight. For DMFC systems the

degradation is clearly higher than those for batteries, typically in the range

of 10 to 25 V h-1. When we only take this cell-performance degradation into

account, the technical life span is more than 60% less than for standard

84

rechargeable batteries. Applied in a laptop computer this will result in a

failure within 600 to 1,000 hours, shorter than the economical life of the

electronic device. The cycle life of fuel cells is thus still an issue to be solved.

5.4 Discussion A tangible trend towards lighter portable electronic devices can be seen.

Battery weight is decreasing at a steady pace, based on the opportunities

introduced by new rechargeable battery types. The specific energy is

increasing from 50Wh kg-1 for NiCd batteries to more than 200Wh kg-1 for

lithium based batteries. For smaller products like the cell phone this

increase in specific energy is mainly used to shrink the battery and make the

device less heavy. Over the years the amount of energy contained in the

battery of a cell phone varied in between 2,000 and 4,000mWh (Figure 30),

en converges to a mean 3,000mWh.

The amount of energy contained by larger power hungry portables, like the

laptop computer, increases with the years (Figure 31). This shows the need

for more energy in the same volume (battery volume percentage over the

total volume is constant over the years) is one of the major goals for laptop

computers.

At the moment a trend is set towards smart phones and tablet-like pc’s

always connected to the internet, like the smartphones and the iPad from

Apple. These devices are more power hungry than the cell phones described

in this chapter. The user uses its smartphones not only for making phone-

calls but more and more for internet applications, which results in a

recharge of the device more often (every other day). For these devices the

need for more energy contained in a smaller volume is, like for laptop

computers, urging.

In Section 5.3.1 an analytical weight model a simple analytical model was

constructed for volume. When applying this model for guessing the volume

of portable electronic devices, the DMFC volume for products in between

10mW and 250W is smaller than the battery used (Figure 23). Again it must

be noted that the model was based on a sole DMFC system without an

auxiliary battery accounting for peak power. The DMFC-hybrid can also

85

improve volume specifics of the fuel cell stack and the BOP equal to the

factor calculated by previous equation.

Figure 30: Amount of energy contained in a battery for a cell phone (trend over 1995-

2003).

Figure 31: Amount of energy contained in a battery for a laptop computer (trend over

1991-2004).

86

Lithium ion batteries are reaching its boundaries of 275Wh kg-1 (Figure 25)

which is almost 55% of the theoretical maximum. The Toshiba micro-DMFC

has only reached 14% of its theoretical maximum, meaning there is a

potential for both DMFC and PEM fuel cells to top the maximum attainable

for lithium based batteries.

Just like volume a quick guesstimate is made for the weight of a DMFC

power system powering different portable devices over a range of 10mW to

100 W. The preliminary results show that the DMFC system will probably

improve the mass specifics of smaller devices like the flash-drive MP3

players and sensors (<250mW) and not the larger devices like the cell phone,

PDA’s and laptop computers. The model described is based on a sole DMFC

system without an auxiliary battery and not on a DMFC hybrid. When

applying a DMFC-hybrid the weight could be reduced even more. The weight

of the fuel cell stack and the BOP can be reduced, at the cost of adding a

small battery, with a factor equal to:

peak

mean

P

P (12)

Analyzing Figure 23 (estimated volume) and Figure 26 (estimated weight), it

shows that weight is less of a problem than volume. Especially when

comparing the density of a DMFC system with that of the lithium ion and

lithium polymer battery. The density of both these batteries is respectively

1.60 kg L-1 and 1.40 kg L-1. Pure methanol on the other hand has a density

equal to 0.79 kg L-1, and combined with water the density of the fuel for a

DMFC system will always be lower than 1 kg L-1. Thus at same volume the

methanol fuel cell will have the advantage over these batteries and be lighter.

Based on this assumption we can postulate that if volume is improved this

will have a even more positive effect on mass.

The price for methanol is high and does not result in a lower life-cycle costs

compared to rechargeable batteries (Figure 27). The initial price is also still

very high compared to the lithium ion battery, but the lower number of

“recharges” could mean a unique selling point for this power system. The

87

initial price has to be cut down by at least a factor 2, and the methanol-price

should be a factor 4 times lower before it gets economically competitive with

the grid-price, and thus with the rechargeable lithium-ion battery.

5.5 Conclusions The goal of this chapter is to answer the first sub-question: Which

technological, physical and economical properties differentiate the DMFC

power system from the lithium-based rechargeable battery in the field of

portable electronic devices? Three main technological properties have been

defined, whereby most power systems are compared with: volume, weight

and (initial) costs.

The volume of cell phone batteries have decreased over the past 10 years,

but in contrast to the battery weight the volume-percentage was kept

constant (cell phones got smaller and the battery followed this trend). On the

other hand for PDAs and laptop computers the battery-volume percentage

has increased by a factor 1.5. The increase in improved energy specifics is

thus mainly used to shrink the battery and make the device less heavy. The

power source in a portable electronic device is an important part defining the

form (for 20% to 30%) and weight (for 15 to 25%) of the device. This means

selecting this component defines largely the form and weight of the device.

Thus decreasing the power source in volume and weight gives the designer

more design freedom.

Theoretically the volume of DMFC systems has an improvement potential of

a factor 2, at a systems efficiency of 20%, over the current lithium-ion

technology. Prototypes are still lagging behind like in the case of the Toshiba

micro DMFC system [43]. Problems with this prototype is the large volume

needed for the fuel cell stack and other auxiliary components, only 7% is

active material, and the low systems efficiency of 2.8%. Weight is of a lesser

issue than volume. DMFC systems have the opportunity to be a factor 3

times more lightweight than lithium ion battery technology. Based on these

figures we can postulate that improving volume specifics of the fuel-cell

system will have an even more positive effect on weight specifics.

88

Initial costs and life-cycle costs of a fuel cell system are high compared to

primary and secondary batteries. The initial price has to be cut down by at

least a factor 5 before it competes with the lithium-ion battery. The

methanol-price should also be cut down with a factor 4, making it

competitive with the grid-price, and thus with the rechargeable lithium-ion

battery. On the other hand the price for lithium ion batteries is high

compared to low-cost rechargeable batteries as the nickel metal hydrate

battery. Apparently the price of the battery is of subordinate concern to the

improvement in convenience (lesser recharges) or decrease in volume. The

DMFC system is an improvement in this field and has to be replaced a factor

2.7 times less than an alkaline battery. Compared to the lithium-ion battery

the DMFC system has to be “recharged” a factor 3.9 times less.

In general fuel cells outperform the battery in areas that need high energy

and lower peak power. Compared to the fuel cell every battery is a high

power-dense energy-source. For low-power and medium power devices, like

the cell phone and laptop computer, a hybrid system will be a better solution.

In this system the fuel cell will only deliver mean power and an extra

intermediate accumulator will take care of peak-power demands. At the cost

of adding a small battery, the volume and weight improvements in the fuel-

cell stack and the BOP is high (Figure 22, group [B]). To test the

improvement a case study is proposed, for a low-power application, with

long-endurance performance-specifications.

Based on the simplified analytical model of volume and weight the

application field can be found mainly in low-power applications with a long

runtime like sensors, MP3 players (without display), and smoke detectors.

Higher power applications are less applicable to be equipped with DMFC

systems because the fuel cell stack gets to large. In between high-power and

the micro-power devices the sole DMFC transits to a DMFC hybrid power

system (DMFC system), which concedes itself in between the MP3 player

(with display) and the cell phone. Volume is hereby the most constraining

parameter. To define this border, and the application field, even better, the

volume and weight of the three parts, fuel cell, BOP and tank, should be

defined better in the model.

89

The main properties found in this chapter will be evaluated and ordered by

importance with a user research. The user will be tested in his/hers

willingness to buy a fuel-cell powered portable, based on the properties

defined in this chapter and other user-properties.

91

6 The user and their choice for a power source

For most users of electronic applications the power source in the application

is less important than the functionalities of their application. One important

property comes always forward when buying the application, the time the

application will work on one charge. The consumer is used to dealing with

charging the batteries, often done overnight. Fuel cells on the other hand

have different properties than batteries. As described in Chapter 2 methanol

has a high energy density compared to rechargeable batteries, which results

in a potentially longer run-time for DMFC power systems which could be an

incentive to transfer from batteries to methanol.

Besides increase in runtime also several physical properties will change in

the product. Instead of recharging the battery a cartridge has to be replaced

or the tank is refilled with a syringe. The fuel has to be bought in advance,

which is on one hand costly compared to the ‘free’ energy from the grid. On

the other hand the cartridges have to be carried along. That may be a hassle,

but instant ‘recharging’ is the opportunity.

To find the answer to the second research question about the properties and

its significance from a user point of view a conjoint analysis has been

executed. The conjoint analysis consists of a user acceptance study,

discerning the properties of consumer products likely to affect the transition

from batteries to fuel cell powered systems, Section 6.1. Five differentiating

properties have been selected in Section 6.2 (including the three defined in

Chapter 5) and a conjoint analysis is executed to prioritize these five

properties. In Sections 6.3 and 6.4 the results are presented and discussed.

This chapter will end with conclusions in Section 6.5. In Chapter 7 the

92

differentiating properties from the user point of view will be evaluated further

and metrics for choosing or designing DMFC systems will be defined.

6.1 Approach The approach of this research consists of four parts. First a literature study

is done to find the differentiating properties between the lithium ion battery

and the potential DMFC system. Second the conjoint analysis is setup where

the differentiating properties are examined more in detail.

6.1.1 Differentiating properties

A literature study has been done to find out what properties affect

consumers in buying or using a specific application like a laptop computer, a

PDA or cell phone. More general we want to know “under what conditions

the consumer wants to make the change towards a new energy system” and

“which differentiating properties influence this choice and in which priority”.

Section 6.2 will present this part.

6.1.2 Conjoint analysis

Based on these differentiating properties a conjoint analysis is executed. The

conjoint analysis is a method to find out about several properties of a

product at the same time. By applying the method it is possible to find out

about priorities of properties, as well as the preferred value of each property

itself [137]. The participant is given a pile of in total 16 cards per specified

product. Each card represents a product by describing a couple of properties

of that product (Figure 32).

It is important that the participants fully understand the effect of each

property. Thus prior to the assignment the participant is informed about the

goal, process and what is expected of him/her. If necessary examples are

used. The participant is asked to prioritize the set of cards from best to

worse. This line up of cards is entered in a statistical program, in our case

SPSS [138]. With the help of this program the order of importance of the

properties is calculated for only the participant and for the whole group of

participants. Also the participants preferred values for each property is given.

In this way it is possible to find out about the importance of the properties in

93

relation to each other, and to predict whether a product with certain

properties will be acceptable for the consumer.

Figure 32: Example of four product cards (in Dutch) for the cell phone and laptop

computer.

6.1.3 Research subjects

Two products have been chosen in the broad range of portable electronic

devices available, Chapter 2, namely the laptop computer and cell phone.

The laptop computer is a larger consumer product with medium-power

performance specifications which has in general a use time of 3 to 8 hours.

The cell phone on the other hand is a small product with low-power

specifications and works for a longer time, 1 to 7 days, in between chargers.

At third optional applications, the PDA/smart phone is not chosen as one of

94

the subjects. In 2004, the year of this research, the PDA was a common

business application but not common as a consumer electronic. Most

participants did not have a PDA or smart phone and thus this device was left

out of the study.

6.1.4 Participants

The research subjects of the project are the laptop computer and the cell

phone. Both are common consumer products and consumers are likely to

put different demands on each of them due to their size and applications. All

test participants own a laptop computer and a cell phone and are well

known with these applications. The analysis was conducted with 21 test

participants. The participants were mainly recruited amongst students and

employees of the Delft University of Technology (TUD), and therefore this

group cannot be seen as an indicative group.

6.2 User acceptance and their differentiating properties

A literature study has been done to find out which differentiating properties

can influence the choice of the consumer for a certain power system or

product in which a certain power system is applied. Five different

perspectives on portable electronics and its power source is researched for.

6.2.1 Instant recharging

If we look at the DMFC system as a black box with same performance

specifications as the recharge battery, the biggest difference between those

energy sources is the way energy is inserted in the system. For rechargeable

batteries this is done by recharging the system by connecting it to the

electric grid by means of a wire. For DMFC systems the “battery” is

recharged by a cartridge or a fueling system. Fueling a tank is instantly

compared to charging a battery takes up at least 1 hour. In [75], Dyer sees

instant recharging of the portable as an advantage. Charging a battery of a

cell-phone still takes up at least half an hour, while replacing a cartridge is

done in a matter of seconds. At what costs will this advantage be worth for

the consumer?

95

6.2.2 Energy

In “Whither Fuel Cells ...” [111], DeMuro tries to find an indication of the

amount of energy needed before the fuel cell is felt as an improvement to

regular rechargeable batteries. He makes a comparison between the

rechargeable (secondary) battery and the single use (primary) battery. The

fuel cell is one of the latter, because the user has to buy cartridges.

Conclusion from this paper was that a primary energy container will be more

attractive over a rechargeable when one cycle of the product in use equals

more than one month. A traditional cell phone will use up 12Wh in a month,

which matches with 12mL of methanol. Taking packaging into account the

total volume of the cartridge will be around 14mL. This is in the same order

as the volume of a standard cell-phone battery. Is the consumer willing to

pay extra for more energy in the same compartment, and what is the

minimal runtime?

6.2.3 Costs

Two different costs have to be taken into account, (i) the initial costs and (ii)

the cycle costs. For the initial costs the main question is what price is the

consumer willing to pay for the fuel cell system as a whole? The initial

pricing for a fuel-cell powered device will be higher, as described in Chapter

5, than for the battery powered counterpart.

Cycle costs, or the costs made for recharging the power source, are different

though. Charging from the grid is seen by the consumer as ‘free’, but costs

around €0.21 per kWh in the Netherlands (2010). A fully charged battery

contains an amount of energy equal to 0.3Wh of energy, which means the

costs for one charge is equal to €0.012 (240 hours standby or 6 hours talk

time). For a standard cell phone this equals to a time-between-charges of a

small week. The production price of methanol is €0.14 per liter. Taking a

fuel-cell system efficiency of 25% into account, 1mL of methanol will produce

around 1Wh of electric energy, equaling a cartridge price of €0.14 per kWh.

When buying methanol in the shops the price is higher than the production

price mentioned (~€2.95 per liter [76]). The price for a cartridge will thus be

ranging between €0.14 and €2.95 per kWh.

96

6.2.4 Volume

In [139] Mangione and Smith discuss the importance of form factor in mobile

computing, and postulates that a small and handy size is one of the most

important factors is for market success. In [140], Chapin draws the same

conclusion that consumers prefer small volume electronics. According to [75]

the fuel cell cannot compete with batteries when size does matter and a

small amount of energy is needed. The fuel-cell plus auxiliaries will be a

large part of the total volume breakdown, and the benefit will start at an

energy demand of 20Wh.

6.2.5 Weight

Chapin [140] has tested four electronic products on density. He made four

models with increasing density. Test participants had to judge the quality of

the product, the level of technology and the durability of the product. The

result showed that a lighter product was judged positive at all three

properties. Based on cell-phone batteries the mean density of lithium based

batteries is 1.40-1.60kg L-1 (Chapter 2). On the other hand methanol has a

density equal to 0.79kg L-1. Thus, at same volume the methanol fuel cell will

always have the advantage over the lithium-ion battery and be lighter.

6.2.6 Differentiating properties

With the literature described in this section five differentiating properties can

be defined when replacing the rechargeable battery form portable application

with a fuel cell. These differentiating properties are used as to describe the

product:

1. Purchase price (in Dutch: “aanschafprijs”), this is the initial price

the consumer has to pay for the product including the power source.

2. Charging or cartridge (in Dutch: “opladen of cartridge”) and the price

the consumer has to pay for charging (in time) or refilling the

cartridge (in costs)

3. Use time per charge (in Dutch: “gebruikstijd per lading”)

4. Weight (in Dutch: “gewicht”) of the total product

5. Volume (in Dutch: “volume”) of the total product

97

Within the property ‘charge or cartridge’ the test participants could choose

between a charging system equal to what they are used to, or a system with

fuel cartridges that can be replaced instantly. For the first option the time it

needs to charge is noted and for the latter option the extra costs. It is

assumed the cost for the grid is seen as ‘free’ by the participant, so no extra

note is given about price. When the participant asks about the availability of

the cartridges, it is told that they can be bought as easy as penlight batteries.

Each property has three or four different possible values, as can be seen in

Table 12. To get a better feeling about the weight and volume of the different

products described on the cards physical models are used made out of

polystyrene foam and given the proper weight.

Table 12: Overview of all property values used in the test.

Cell phone Laptop

purchase price €100 €1000 €200 €2000 €300 €3000 charge or cartridge charge 1 hour

charge 6 hours cartridge 5 cent cartridge 50 cent

charge 1 hour charge 4 hours cartridge 10 cent cartridge 1 Euro

time of use 3 days 3 hours 1 week 8 hours 1 month 24 hours weight 70 g

100 g 140 g

2.4 kg 3.2 kg 4.0 kg

volume 75 mL 3 L 105 mL 4 L 150 mL 5 L

6.3 Results The output from the SPSS conjoint-analysis consists of a table with three

columns, the averaged importance score, the utility estimate and the

correlations between observed and estimated preferences. The latter column

gives a measure for the quality of reproduction of the empirical data by the

results of the conjoint analysis, and will not be discussed in this chapter.

98

6.3.1 Averaged importance score

The averaged importance score gives the importance values relative to each

other for the different properties. For the cell phone and the laptop computer

their order became as depicted in Figures 33 and 34.

Figure 33: Averaged importance score of the five different properties for the cell phone.

Figure 34: Averaged importance score of the five different properties for the laptop

computer.

99

‘Volume’ turns out to be by far the most important factor for choosing a cell

phone. ‘Purchase price’, ‘time of use’ and ‘charge or cartridge’ seems is of

equal importance. The ‘weight’ of the system is considered the least

important for the tested participants. The score on ‘charge or cartridge’ could

differ because in the conjoint analysis of variables with more different values

(in this case four instead of three) tend to get a higher rating.

For a laptop computer ‘volume’ is also the most important factor, but

‘purchase price’ is of almost equal importance. ‘Charging or cartridge’ and

‘time of use’ is less important, and ‘weight’ is also considered least important

for this group of test participants.

6.3.2 Utility estimate

The average importance scores give only averages of importance of the

different properties. The conjoint analysis also outputted the utility estimate,

which is a figure of appreciation for every value of the property. Figure 35

shows the utility estimates for the cell phone and Figure 36 show those for

the laptop computer. Figures above zero are positively appreciated by the

participants, and vice versa, and the higher the number, the more a value is

appreciated.

Figure 35: Overview of the utility estimate for the cell phone.

100

Figure 36: Overview of the utility estimate for the laptop computer.

6.4 Discussion The output from the conjoint analysis gives us an indication for the

appreciation of a certain property compared the other researched (averaged

importance score). Also within the property a preference for a value is

outputted by means of the utility estimate. In this section the different

properties are discussed referring to the previous sections output data,

described in Figures 33 to 36, but also to the interviews with the

participants.

6.4.1 Purchase price

For both the cell phone as the laptop computer purchase price is the second

important factor. Looking at the consumer preference it is clear that all

participants put the initial purchase price in the same order from less

expensive (€100 for the cell phone) to more expensive (€300). All values of

this property has a negative appreciation by the participants, but price is not

often considered as an issue because in the Netherlands it is common to get

a phone for ‘free’ when bough in combination with a contract. For the laptop

computer the purchase price is of more importance, almost equal to volume

101

property. As with cell phones all values of this property are appreciated

negatively for the laptop computer.

6.4.2 Charge or cartridge

The preference values for this property are low (in between -1 and 1),

meaning there was not much agreement between the participants on the

choice of a charging system or a cartridge system (at extra costs). Looking at

the utility estimates, the participants all show a preference for charging over

a cheap cartridge for the cell phone and a cheap cartridge over charging for a

short period of time for the laptop computer. A few participants had a clear

liking for one system over the other. Those who preferred the charging

mainly though they would either forget about buying fuel-cartridges or it was

too much of a fuss buying them. Many participants mentioned they did the

charging overnight, so 6 hours was no problem. Those participants

preferring the cartridge system liked the idea of being independent of the

electric grid, or disliked the charging time. The cartridge price of €0.50 for a

full-charge of the cell phone was often seen as rather expensive, especially

when it was in combination with a low time-to-use. When the cartridge could

result in a use-time of a month, this high price was seldom a problem. Many

participants calculated the life-cycle costs to get a better feeling of the total

costs.

6.4.3 Time of use

For both the cell phone as the laptop computer all participants agreed the

longer use-time in between charges the better. However not every participant

thought it equally important. Three days was often to short, even if their cell

phone lasted only for three days. A use-time of a week was seen as an

improvement and most acceptable. A use-time of one month was equally

appreciated as the weeks use time. For the laptop computer 8 hours of use-

time was seen as a big improvement, because you would be able to spend a

whole working day away from the electric grid. The difference between 8 or

24 hours was not seen as an improvement. Three to four hours is seen by

most of the participants as inconvenient. A user time of a full working day (8

hours) is preferred by the participants for the laptop computer. For the cell-

phone a user time of one week is greatly appreciated.

102

6.4.4 Volume

For the cell phone the property of volume is seen as the most important

factor compared to the other properties. Mainly male participants preferred

the smallest volume: “it has to fit easily into my pocket” is a quote of one of

the participants. For the laptop computer this property also is the most

important factor. All participants agreed that a smaller volume is preferred

to the larger volume. Five liters was seen as a bulky laptop computer, four as

an average, and three liters would easily fit into a bag. When we compare

volume preference numbers for the cell phone in Figure 35 to that of the

laptop computer in Figure 36, it shows that for the cell phone a small

volume has a high preference score (>2), while for the laptop computer this

number is negative (< -2.4).

6.4.5 Weight

Compared to the other properties weight was seen as the less important

factor influencing there decision to buy the product. Not all participants

agreed on ‘the lighter cell phone, the better’. Three participants preferred a

heavier cell phone (100 g) over the lightest cell phone (70 g). The 140 grams

cell phone was to heavy for every participant. Especially the two female

participants did not see weight as an issue, because they mainly carried

their cell phone in their bags instead of a pocket. For the laptop computer

the weight of the system also less important compared to the other

properties. In this product all participants preferred a lighter laptop

computer over a heavier laptop computer.

6.4.6 ‘Time of use’ combined with ‘charge or cartridge’

During the card layout by the participants, the participants often took a

longer time to consider the combination of charging, cartridge use and the

time of use (or time in between charges). To test the combination of these

two properties the test results were re-evaluated with twelve cards instead of

16. The twelve evaluated product cards share the same properties of weight,

volume and purchase price while the properties of ‘time of use’ and ‘charge

or cartridge’ varied. The sequence of these, for each the cell phone and

laptop computer, twelve cards is entered in SPSS for a new conjoint analysis.

The results are shown in Figure 37 for the cell phone and Figure 38 for the

103

laptop computer. The order of preference for the user time stays the same

with a short charging time appreciated more than the cheap cartridge for the

cell phone. On the other hand, for the laptop computer the cheap cartridge is

appreciated more than a short charging time.

Figure 37: Overview of the utility estimate for the combination of charge or cartridge and

the use time for the cell phone.

Figure 38: Overview of the utility estimate for the combination of charge or cartridge and

the use time, for the laptop computer.

104

6.5 Conclusions A conjoint analysis is executed to test the preference of future consumers on

their willingness to buy a fuel-cell powered cell phone or laptop computer.

Five properties have been compared: purchase price, charge or cartridge (at

an extra cost), time of use, weight and volume. In this section these

properties will be elaborated on. Because the choice for charge or cartridge is

strongly dependent on the improved time of use, these properties have been

tested in combination.

The participants in this test see volume of both products as the most

important property of the product. The participants have the lowest issue

with weight. A small increase in weight of the cell phone would not matter

very much to the participants, but an increase in volume would. This

preference is especially strong for the smaller product compared to the larger

product, the laptop computer.

Based on the combination test of ‘time of use’ and ‘charge or cartridge’ the

participants prefer quick charging over a cheap instant charge, even if this

results in longer use times. In contrast with the laptop computer where a

cheap cartridge is preferred over short charge periods. This preference is

stronger for larger products where use-time is a bigger issue, than for

smaller products. In general it can be stated that cartridges, and thus fuel-

cell power systems, are more appreciated when use-time is an issue, and the

cartridge grand the user more time-between-charges.

A longer time of use is considered quite a big advantage, especially for laptop

computers. The present day laptop battery run time does not live up to the

needs (8 hours). Bringing an extra cartridge in is hardly a problem because

the laptop is often carried in a laptop bag, which also contains the charger or

an extra battery. The price of the cartridge is also not seen as big issue for

larger products like the laptop computer. For the cell phone this advantage

is not explicit. It can be concluded that especially for larger portable

products the consumer is more willing to pay extra for longer use time. For

smaller products this advantage is not explicit.

105

To answer our research question stated in Chapter 4, we have identified five

discerning properties which are not equally important to the user. Volume is

the most important and weight is the lowest importance. The results from

this study can and will be used to evaluate power system designs, where

volume has the most influence in decision making and mass the lowest. In

Chapter 7 this will be explained more in detail.

107

7 Introduction to different orders of modeling

The goal of this chapter is to define metrics which can be used in different

orders of modeling, exploring the feasibility of DMFC systems applied in a

portable electronic device commonly equipped with a rechargeable battery.

The models proposed in this thesis are meant to be used during early phases

of the design process, when no or few information is available. Because

‘feasibility’ is such a vague term we have to define this more by introducing

orders of modeling. Section 7.1 describe three orders of modeling which are

proposed to be used by an product designer for evaluating DMFC power

systems as an opportunity for a specific portable electronic device.

In Chapter 4 a comparison is made between a DMFC system and the

lithium-ion battery from a technological point of view. From this perspective

especially volume, weight and price seems to be the most important

properties where the DMFC differs most with the lithium-ion battery. In

Chapter 5 we have made the same comparison but now based on the user

preference. Five properties, initial price, recharge or cartridge, use time,

weight and volume, have been evaluated by means of a conjoint analysis.

The research showed that volume was one of the most important properties

influencing the user choice in buying the device. To answer research

question 3, as defined in Chapter 4, the designer point of view will be

evaluated in Section 7.2. The “set of considerations when choosing a battery

for a portable application” [109] is evaluated and completed with the

selection of a DMFC power system. The main differentiating metrics are

defined in Section 7.3, which are going to be used in the first and second-

order model proposed in Chapters 8 and 9. This chapter proposes the use of

108

mathematical modeling for the identification of DMFC power systems in

portable electronic devices during early phases of the design process.

7.1 Orders of modeling In this section we define four orders of modeling, from basic comparison

based tables and figures (zero order), through the heuristic approach (first

order) to the automated design (second and third order). The first two

approaches use normalized equations and are to be used without the

computer. The higher the order of modeling the higher the accuracy, but also

the complexity (Figure 39). For higher order models the computer is used to

evaluate multiple designs on a one or more objectives. The second order

model optimizes for ‘basic’ properties, while the third order model also

includes other ‘specific’ properties. Section 7.2 will formulate the basic and

specific differentiating properties for the DMFC system compared with

batteries.

Zero-order model

First-order model

Second-order model

Third-order model

Complexity

Rule-of-thumb Enumerative

Computer

Acc

urac

y

Low

H

igh

Figure 39: Complexity and accuracy of the four proposed orders of modeling.

7.1.1 Zero order model, comparison

In Chapter 2 a comparison between power sources is made, based on

normalized values for energy, power and costs. This type of model we like to

109

call the ‘zero order model’. This model compares power sources and energy

systems, and gives good indication what the overall opportunity is for a

specific power system. Although power systems are made visible to the

designer with simple comparing plots, this method does not identify an

opportunity for a specific power source applied in a portable electronic device

which exists or has ‘to be designed’.

7.1.2 First order model, the heuristic approach

The next order model, the ‘first order model’, will help the designer in

evaluating the feasibility of a power source (in our case the DMFC power

system) in comparison with other power sources (in our case the lithium-ion

battery). In Chapter 5 a first approach towards such a first order model, the

preliminary model, is described. Based on a case-study by Motorola [29, 125]

the weight and volume of a DMFC system is calculated by breaking down the

system into three major contributors, the fuel cell stack, BOP (including

empty space) and the fuel tank. This preliminary first-order model gives good

feedback to the designer whether a DMFC system is feasible and how large it

will be. Problems with the first-order model developed in Chapter 4 are (i) its

bad scalability, when applied on a range of applications9, (ii) only basic

properties, like volume and weight, are taken into account, and (iii) there is

no feedback to the designer about the amount of improvement compared to

the standard used battery.

To improve the model a breakdown in components should be made and the

three main parts of the fuel-cell system have to be redefined in normalized

functions. Furthermore an objective function has to be defined to give the

designer direct insight if the DMFC system is an improvement compared to

the standard used battery.

This type of modeling is generally used in conceptual design, and known as

a “rule of thumb”, where equations give the designer preliminary insight in a

9 For instance the contribution of empty space to the total volume is great according

to the model, but in bigger systems the contribution would a lot smaller and will probably not be large percentage of the BOP (as is described in the preliminary model in Chapter 4).

110

specific problem or identify opportunities for specific technologies. This type

of modeling can best be used when no computer is available.

The input data in the model is moderate, but still needs knowledge of the

designed or to be designed device. User profile is needed which is used to

define a load profile. The designer has to know how high the load will be

when the user uses a certain function. Development of the load profile in the

form of a power-to-time function or graph is needed to give a good estimate

of the DMFC hybrid physical performance. The designer has to fall back on

its electronics knowledge and describe the power needed for every function of

the application. This can be done by adding quantitative data (as maximum

and nominal power) to all solutions for every function in the morphological

chart. Summing all power data for every function will give the designer

insight in the potential power needs of the product.

The preliminary model used in Chapter 5 is evaluated resulting in the first-

order model described in Chapter 8. This model is evaluated with two

commercially available DMFC power systems from Smart Fuel Cell [141,

142]. The results show the accuracy of the estimated volume of the three

parts of the fuel-cell system when compared with actual commercially

available DMFC power systems. The mean absolute error is 115% for a very

compact fuel-cell system and 15% for a conventionally designed fuel-cell

system. The large error for the compact fuel-cell system is mainly caused by

wrongly estimating the volume for the BOP. Compact design depends mainly

on the 3D insight of the designer. The BOP can thus be estimated more

proper by evaluating different structural variants, which the designer will

evaluate. The designer is limited by the number of structural variants

he/she can evaluate within a period of time. Furthermore the objective(s) to

evaluate these structural variants are not known or well-defined by the

designer. To solve the problem of evaluating a limited amount of designs, the

computer can be introduced to evaluate more-or-less unlimited amount of

structural variants. The second-order model introduces computer-based

evaluation of multiple designs.

111

7.1.3 Second order model, structural variants

The first order model described in the previous section gives a quick

impression of the feasibility of a DMFC power system applied in a portable

electronic device. The first order model is based on normalized parameters

for basic properties. Problems described are the (i) accuracy of the model, (ii)

the low amount of properties to optimize for, and (iii) the lack of a single

metric, showing the product designer the improvement opportunity in a

single figure.

One part of design and engineering is the act of sizing, dimensioning and

selecting the detailed elements of the design. This part of the design process

can be used to test the feasibility of different design concepts, by using the

computer to evaluate multiple design solutions on one or more parameters.

In general the designer can only evaluate a limited amount of structural

variants (often less than five) on a limited amount of objectives (often only

volume), with a low amount of differentiating components. Within these

structural variants there are variables which specify the main proportions,

like weight, dimensions and costs, but also other details of the application.

To test the feasibility of different concept design, and not only structural

variants, a “quantified optimum design” method can be used. Quantified

parameters are used to evaluate a concept with the help of the computer

which can evaluate a greater deal of concepts in less time than a designer

could.

For the second-order model the accuracy could be improved by evaluating

multiple structural variants with the help of the computer. All structural

variants are generated by the computer and are build up from standard

building blocks, representing all basic components needed in a fuel-cell

system. New metrics are introduced for all basic properties, and an objective

function is defined with whom the design is evaluated with.

In Chapter 9 the second order model is proposed using the computer to

evaluate multiple structural variants and optimizes for the basic properties.

For the second order model the number of metrics is limited to the basic

112

three properties. Components are chosen from a database consisting of

geometrical and non-geometrical data.

7.1.4 Third order model, taking extra variables into account

Besides the basic properties used in the second order model, other, more

specific, properties can be taken into account while evaluating different

structural variants. Specific properties are defined in Section 7.2.

The second order model represents all components in its most basic form,

the parallelepiped. Their actual form is different and to contribute to a more

accurate solution the components should be represented by other

geometrical forms than only parallelepipeds. Also interconnections between

components, like electric wiring and tubing, is not taken into account in the

second-order model and should also be introduced in the third-order model.

7.2 Differentiating properties In Chapter 5 and 6 the difference between DMFC systems and lithium ion

batteries is explored from a technological and a user point of view. The main

differences between the two power systems are volume, weight, charge or

cartridge, use-time and costs. The choice for a power source on the other

hand is not only defined by these properties and will be investigated in this

section. In the Handbook of Batteries [109] a “set of considerations when

choosing a battery for a specific application” is described. In this section

these considerations are discussed for the use in the first, second and third-

order model (Section 7.1). The evaluation will be based on the following

constraints:

• What are the ‘basic’ considerations, and what considerations are

‘specific’?

• Is the consideration quantifiable and formed to a property, or is it

Boolean?

113

According to [109] the following considerations are important and influence

the selection of a battery:

1. Type of battery: primary, secondary or reserve system

2. Electrochemical system: matching of the advantages and

disadvantages and of the battery characteristics with major

equipment requirements

3. Voltage: nominal or operating voltage, maximum and minimum

permissible voltages, voltage regulation, profile of discharge curve,

start-up time, voltage delay

4. Load current and profile: constant current, constant resistance, or

constant power; or others; value of load current or profile, single

valued or variable load, pulsed load

5. Duty cycle: continuous or intermittent, cycling schedule if

intermittent

6. Temperature requirements: temperature range over which operation

is required

7. Service life: length of time operation is required

8. Physical requirements: like mass and volume, size, shape, weight;

terminals

9. Shelf life: active/reserve battery system; state of charge during

storage; storage time a function of temperature, humidity and other

conditions

10. Charge-discharge cycle (if rechargeable): float or cycling service; life

or cycle requirement; availability and characteristics of charging

source; charging efficiency

11. Environmental conditions: vibration, shock spin, acceleration, etc.;

atmospheric conditions (pressure, humidity, etc.)

12. Safety and reliability: permissible variability, failure rates; freedom

from out-gassing or leakage; use of potentially hazardous or toxic

components; type of effluent or signature gases or liquids, high

temperature, etc.; operation under severe or potentially hazardous

conditions; environmentally friendly

13. Unusual or stringent operating conditions: very long-term or extreme-

temperature storage, standby, or operation; high reliability for

special applications; rapid activation for reserve batteries, no voltage

114

delay; special packaging for batteries (pressure vessels, etc.);

unusual mechanical requirements, e.g., high shock or acceleration,

nonmagnetic

14. Maintenance and resupply: ease of battery acquisition, accessible

distribution; ease of battery replacement; available charge facilities;

special transportation, recovery, or disposal procedures required

15. Cost: initial cost; operating or life-cycle cost; use of critical or exotic

(costly) materials.

In the following part all the considerations are discussed and evaluated for

admittance in the first, second or third-order model. An summarization of

this discussion can be found in Table 13.

Consideration 1 is the first choice the designer wants to make. As discussed

in the previous Chapter the DMFC power system is often seen as an

alternative for the secondary system, but is best typed as a replaceable

energy system, thus as an alternative for the primary battery. The result

from all orders of modeling should be an indication which power system is

the better option for the case-study to be evaluated. This consideration

should thus be the result of all-order models.

Within this thesis it is assumed that the DMFC system complies with the

requirements for major equipment, and thus consideration 2 will not be

taken into account.

Unlike some primary and some secondary batteries, fuel cells do not have a

flat discharge curve, but a very steep voltage drop at increasing C-rate.

Within this thesis it is assumed that the DMFC power system complies with

the voltage requirements of the application. Consideration 3 will thus not be

taken into account for the zero to secondary model, but is proposed to be

added to the third-order model.

Consideration 4 and 5 consider the load-profile and the cycling of the

application. As described in Chapter 5 the load-profile is guiding for

choosing the system components and will be used as input in the first,

second and third order model.

115

Temperature (consideration 6) requirements of the system are of great

importance but not differentiating compared the lithium ion battery. The

working temperature range of fuel cell systems is equal or even better to that

of battery systems. Besides that the application field for both power sources

requires a moderate working environment of temperatures between 0 to 40

degrees in which they are functioning well and reliable.

Service life of the power source system is of great importance, because the

system should work flawlessly for at least the economical life expectance of

the application. The economical life expectancy of cell phones is low.

According to Geyer and Blass [143] “the number of end-of-use handsets

increased even faster than this since cell phone lifetimes have been

decreasing, from 3 years in 1991 to 18 months by 2002 and probably even

less today.”. For computers the economical lifetime is longer than for the cell

phone, nearing three years. Especially the technological lifetime of the

battery is bringing the lifetime of laptop computers down to 2 to 5 years

[144]. Consideration 7 is thus an important consideration and should be

included when basic considerations are well satisfied. For this purpose an

approximation should be developed for the technological lifetime of the

power source system and its components. This will not be considered within

this thesis.

As described in Chapter 5 and 6, the physical requirements like weight and

geometrical form are important considerations for the user and for the

designer who wants to choose a power system. The physical properties

(consideration 8) like weight and volume are thus considerations belonging

to the basic considerations, and should be applied in all orders of modeling.

Shelf life is important when the application is on the shelf for a long period

of time, and is only applicable for reserve batteries. DMFC power systems

applied in portable electronic devices do not belong in this group, and

consideration 9 is thus not taken into account in this thesis.

Because the DMFC system is not ‘rechargeable’ but should be refilled or

replaced like primary batteries, consideration 10 is not taken into account.

116

Environmental considerations (consideration 11) are difficult to quantify

reliability in one figure, and thus will not be taken into account in this thesis.

One could imagine that the safety of using high concentrations of methanol

(often of AA quality 99.5%) brings safety issues like poisoning. The safety

issues concerning consideration 12 are difficult to quantify and, within this

thesis, not taken into account. The environmental issues about power

systems is becoming a more urgent problem within the design of products

and systems. Quantifying the life-cycle environmental impact of the DMFC

system is complex and strongly depends on the input and reliability of

underlying data. This issue is not addressed within this thesis.

Basically all portable electronic devices made for consumer use are made for

moderate operating conditions, leading to consideration 13 not taken into

account within this thesis.

Consideration 14 is difficult to quantify and thus not taken into account

within this thesis.

Consideration 15 describes the initial costs and the life cycle costs, which is

differentiating because of the high initial price of fuel cells and the lower life

cycle costs of the fuel used in fuel cell systems. This consideration is

quantifiable and is an important factor, as described in Section 5.3.4 and

5.3.5, to the success or failure of the DMFC power system, and will be

included as one of the main considerations.

117

Table 13: Summary of all considerations and the use in the different orders of modeling. Consideration 1st order

model

2nd order

model

3rd order

model

1. type of battery Result

2. electrochemical system Not taken into account

3. voltage X

4. load profile Basic input

5. duty cycle Basic input

6. temperature X

7. service life X

8. physical requirements X X X

9. shelf life X

10. charge-discharge curve Not taken into account

11. environmental conditions Not taken into account

12. safety and reliability X

13. unusual operation cond. Not taken into account

14. maintenance and resupply Not taken into account

15. cost X X X

7.3 Conclusions The goal of this chapter was to define models which can identify the

opportunities of DMFC power systems during early phases of the design

process. Four orders of modeling are proposed to identify DMFC power

system in Section 7.1. The final part of this thesis will investigate these

models.

The zero and first-order models can be used without the use of computers

and is based on normalized properties. The second-order model introduces

an automated process focused on automating the process of designing and

evaluating structural variants. Optimization of the system is based on the

basic properties. A third-order model is proposed which should include more

than the basic properties, but also more specific optimization properties.

To complete the search for metrics the designers point of view in selecting a

power source is evaluated by means of the method described in the

Handbook of Batteries [109]. The list of considerations is evaluated and two

118

basic considerations are filtered out: physical properties and costs. The basic

properties to be used in the further development of the first and second-

order model are thus weight, volume and costs. Other, more specific

considerations, and thus properties, are proposed to be evaluated in the

third-order model. Specific, quantifiable, properties which can be evaluated

are: voltage, temperature, service life and safety and reliability.

119

8 First order model: a heuristic approach to modeling a DMFC power source10

As described in Chapter 7 this thesis will evaluate different order of models.

In this chapter the first-order model will be proposed and evaluated. The aim

of this model is to give the designer more insight in the feasibility of DMFC

power systems applied in portable electronic applications using simple

equations (part of research question 5). In Chapter 5 a preliminary model is

used to identify the opportunities in the application field for DMFC power

systems. The preliminary model is based on the design of a case study

presented by Motorola [29, 125]. When applied to different portable

electronic devices the model estimates the volume and weight specifics of a

DMFC power system matching the used battery. Products over de whole

range of portable products were evaluated and the results defined a group of

products in which a DMFC power system could be a technological feasible

alternative. Of all physical constraints volume was the most constraining

property. The preliminary model used was of low accuracy, because it did

not take an intermediate accumulator into account to take care of peak

powers, and the scaling of BOP seems to be off-scale. To make a more

accurate definition of feasible products, the volume and weight of the three

parts, fuel cell, BOP and tank, should be defined more accurate.

This chapter evaluates the preliminary model, described in Chapter 5, by

designing two DMFC power systems powering a Samsung YP5 MP3 player

(Section 8.1). Two approaches are of interest, the ‘standard engineering

approach’, based on commercially available components, and a design

approach based on ‘scaled components’. Both design approaches give a good

10 Parts of this chapter have been presented ASME fuel cell science and technology

conference in Neport Beach, CA (2009) and is published in the Journal of Fuel Cell Science and Technology (2010), “Designing micro fuel cells for portable products”.

120

impression on short-term and long-term feasibility respectively. In Section

8.2 a first design is made using commercially available components.

Unfortunately the components are scarcely available in this power-range en

thus a second design is made where all components are scaled based on

their performance characteristics (Section 8.3). The accuracy of the

preliminary model is tested by means of these designs (Section 8.4) and a

new first-order model is proposed in Section 8.5. Instead of three parts, the

improved DMFC power source model uses more specified parts. This first-

order model is evaluated in Section 8.6 by two commercially available DMFC

systems produced by Smart Fuel Cell AG, the Jenny and Efoy-2200. This

chapter will finish with conclusions on further improvement of the model in

Section 8.7.

8.1 Case study of a MP3 player

8.1.1 Application field

Possible application fields for DMFC power systems are identified with the

help of a Ragone plot (Figure 10) [28, 137, 138]. The power and energy

specifications of different portable electronic devices are laid over the Ragone

plot and three fields of interest can be pinpointed:

[A] High power, short endurance: for applications demanding high power

boosts (>1C). In this case the battery will be the best solution. The

graph shows that the fuel cell system is not going to be an

improvement, even when the fuel-cell system is optimized

[B] High power, normal endurance: for applications demanding normal

discharge characteristics (0.01<C<1) a DMFC in combination with a

battery will be the best solution to improve volume and weight

characteristics. The hybrid power source uses best of both worlds:

high ‘energy density’ of the methanol and high ‘power density’ of the

lithium ion battery.

[C] Low power, long endurance: for applications which have to work a

very long period at very low discharge (C<0.01), e.g. smoke detectors

or Wireless Sensor Network nodes. Stand alone fuel cells seem to be

the best power source for this quadrant.

121

Figure 40: Fields of opportunity for DMFC, based on [29, 125], compared to lithium ion

battery [126].

The plot shows that application field [A] and [B] are of interest for the fuel-

cell only system and a DMFC hybrid power source respectively. Application

field [B] is a field where the endurance of low-power portable consumer

electronics could be increased by adding an extra intermediate accumulator

to the fuel cell. In this chapter the flash drive MP3 player is pinpointed to be

of great interest because it fits this description. The Samsung YP-Z5F flash-

drive MP3 player [146] is choosing as the case study for this thesis, because

this MP3 player had the longest stated runtime of 36 hours at the time

(2007).

8.1.2 Test setup

To get an impression on the runtimes and power profile of the MP3 player

the Samsung Z5F is tested under different conditions (Figure 41). First a

durability test is conducted. Second the different power modes are measured:

122

(i) standby, (ii) play MP3 (56, 128kbs, etc), (iii) light on/off and (iv) pictures.

These measurements give the nominal, mean and peak power output of the

fuel cell system to be designed. From these power settings a choice is made

between a hybrid system (fuel cell + battery) or standalone fuel cell. In the

following section the MP3 player and the test setup is described.

Figure 41: The insides of the Samsung Z5F (millimeter).

The power draw of the battery depends on different aspects of play mode: the

music’s volume, lights-on or off, MP3 or WMA, screen on or off and music

bitrate. All tests are executed at a measure rate of 1 second. In this chapter

the Samsung MP3 player is subjected to these different tests. The battery is

tested to a durability test, a bitrate and compression test, a display test, a

loudness test and a power down and startup test. These tests are described

and conclusions are drawn at the end. Finally this chapter ends with the

conclusions on the tests and some requirements for the synthesis of the fuel

cell system.

The power draw of the battery depends on different aspects of play mode as

described in above. All tests are executed with a Grant SQ800 data logger at

a measure rate of 1 second. Figure 42 shows the test setup. The current is

123

measured over a resistor of 1Ω and the voltage is measured over the MP3

player.

V

3.8V 820mAh

V

1Ω

+ -

Sensitive Voltage measurement

Working Voltage measurement

Load Battery

Figure 42: Test setup to measure the current draw and working voltage for the Samsung

Z5F MP3 player.

8.1.3 Durability test

In Figure 43 the durability test results are shown. The test is executed while

playing the U2 album “All that you can't leave behind” at 128kbps and in

shuffle mode. After 46.5 hours the MP3 player ended playing. The mean

power draw during this continuous play mode is 70mW. The current draw

ranged in between 17 and 22mA and the voltage ranged from 3.4 to 4.1V.

The maximum calculated capacity of the battery is 864mAh or 3315mWh.

0

0,5

1

1,5

2

2,5

3

3,5

4

4,5

28-06-060:00

28-06-0612:00

29-06-060:00

29-06-0612:00

30-06-060:00

30-06-0612:00

1-07-060:00

time

Vo

ltag

e (V

)

0

5

10

15

20

25C

urr

ent

(mA

)

Voltage (V)

Current (mA)

Figure 43: Voltage and current use during a full discharge in continuous play mode,

without interruption.

124

0,0

100,0

200,0

300,0

400,0

500,0

600,0

0 50 100 150 200 250

time (sec)

pow

er u

se (m

W)

WMA 128kBs

MP3 128kBs

Figure 44: Power draw of two different compression formats WMA and MP3, both at

128kBs.

8.1.4 Bit-rate and file compression test

From test executed with bitrates ranging from 56kbps to 256kbps no

difference was found in the power draw. At every bit-rate the power draw for

MP3’s was 72mW (with lights out, 50% loudness). However a large difference

is found when different compression formats are used. In Figure 44 the

power draw is depicted for the same song but differently compressed to

either a MP3 or WMA file, played at 128kbps. When both the backlight and

the screen were set to off the power drawn is 72 and 99mW for respectively

the MP3 and WMA file. In general the power draw of the system is largely

prescribed by the backlights and the LCD screen (see Table 14).

Table 14: The difference of in mean, minimum and maximum power draw for the same MP3 and WMA file.

MP3 128kBs WMA 128kBs

Start (everything on) 264 (186-363) mW 321 (245-519) mW

Lights off 236 (156-450) mW 230 (169-428) mW

Screen off 72 (61-92) mW 99 (78-191) mW

8.1.5 Display and loudness test

To get a better insight in the power draw of the LED lights and the LCD

screen, a luminance test has been executed. For this test the screen

125

luminance is raised from lights off (0%) to maximum luminance (100%),

when screen was on and the music was set to off. The power draw of the

LEDs seems to be linear to the percentage of luminance and ranges from

146mW at 0% to 381mW at 100% luminance. From these figures it can be

concluded that the power draw of the LCD screen plus internal components

is equal to 146mW and the backlights will take up a mean 235mW at

maximum luminance.

To get an impression of the influence of sound level a loudness test is

executed. The song (128kbps, MP3) is repeated at 5 different sound levels

(0%, 25%, 50%, 75% and at 100%). When lights and screen are set to off, the

power consumption of the first four sound levels did not change much,

~73mW. When the sound level was boosted to 100% the power consumption

increased to approximately 114mW, with lights and screen set to off.

8.1.6 Start-up and power-down test

Peak powers are very large, and the maximum peak power was measured

during power down: 867mW for almost 2 seconds. The startup and the

power down of the system are shown in Figure 45. When the system shuts

down it goes into standby mode, still draining approximately 10mW

continuously. After 24 hours of inactivity the system completely shuts down

and doesn’t drain the battery anymore. Startup from standby mode is

different and shorter in time than a full start-up.

-

100

200

300

400

500

600

700

800

900

1.000

1 21 41 61 81 101 121

time (sec)

po

we

r u

se

(m

W)

start up 5,82mWh

lights on 3,17mWh

lights off 1,26mWh

power down1,27mWh st

an

db

y

Figure 45: Start-up and power-down of the MP3-player.

126

8.1.7 Live user scenario’s

When designing a fuel cell system two parameters are very important: the

mean and maximum peak power draw of the load. Based on user profiles

these parameters can be chosen. Assuming the power drain per function can

be summed, the following equation can be used for simulating the dynamic

power draw of the MP3 player:

( )standby screen backlight musicE P P P P dt= + + + (13)

Where: Pstandby = 10mW

Pscreen = 136mW (for both on and off)

Pbacklight = 235 light intensityI

Pmusic =

61-104mW for MP3

89-130mW for WMA (at 75% and 100% loudness)

0

50

100

150

200

250

300

350

0 30 60 90 120 150 180 210 240

time (s)

load

(m

W)

Pow er MP3 128kBs

Pow er WMA 128kBs

mean pow er MP3

mean pow er WMA

Figure 46: The coarsened load curve.

In this case we are going to assume the worst case scenario, namely an

“intensive user”, listening to one song at a time. After listening to it he/she is

using the display to skip to the next number or picture. The load profile of

127

Figure 44 is used and coarsened as visualized in Figure 46. This load curve

has been repeated for at least 2 hours, equaling 30 cycles. Figure 47 shows a

pie-chart diagram of the power breakdown of the MP3 player.

backlight (@100%); 235

LCD; 136

added when using WMA; 30

loudness @100%); 104

Standby; 10

backlight (@100%)

LCD

loudness (@100%)

added when using WMA

Standby

Figure 47: mean power use of different functions in the MP3 player.

As can be seen from the Figure 46 the WMA files are more demanding than

the MP3 files. The nominal power drain for 128kBs WMA files is 150mW and

the maximum power draw is 321mW. Between songs the user is probably

playing with the MP3 player for ~10 seconds at maximum power drain of

321mW. The load characteristic which this standardized user is demanding

is quantized in Table 15.

Table 15: The load curve in numbers as visualized in Figure 46. Period

(sec)

PMP3

(mW)

PWMA

(mW)

EMP3

(mWs)

EWMA

(mWs)

Playing 10 264 321 2640 3210

Start 30 264 321 7920 9600

Lights off 30 236 230 7080 6900

Screen off 180 72 99 12960 17820

Total 240 - - 30600 37560

Mean - 122 150 - -

128

8.2 Design of the DMFC hybrid system

8.2.1 Program of requirements

Now that the physical boundaries and the power characteristics of the

product are known a design has to be made to verify if a fuel cell system is

feasible. The goal of this design exercise is to test the feasibility of a fuel-cell

power-system which fits into the volume of an existing battery compartment

and fulfills the performance need of an intensive user. The fuel cell system

should be small, have an energy density which equals or exceeds that of a

standard lithium-ion rechargeable battery, and is made out of commercially

available components. The following requirements are the main drivers for

this design:

• The power-system plus the fuel tank must fit in the battery

compartment of the Samsung YP-5Z MP3 player (66x33x4mm) and

weighs less than 21g.

• the system peak power has to be at least 868mW, maximum of

900mW

• The power-system should work for an intensive user, requiring a

user profile equal to that of Figure 46, for at least 2 hours or 30

cycles.

• The runtime of the power system should equal or exceed that of the

existing lithium-polymer battery (17h for intensive user, 46,5h

maximum).

• The energy available for an intensive user should be equal to 3,1Wh.

• The power and energy density of the system should be equal or

exceed 100W L-1 and 348Wh L-1.

• The power system module should be easy to assemble and

disassemble

8.2.2 Fuel cell model

The theoretical reversible open cell voltage (E) can be calculated and is for

methanol fueled fuel cells equal to 1,206V. In practice this value will never

be reached and the practical open cell voltage (VOC) is used more often. The

value of the open cell voltage is not constant but depends on the working

129

temperature, the methanol concentration, cathode loading and other

parameters. A multiple regression analysis is executed taking 47 measured

points from 4 different DMFC cases [113-116], as described in Appendix A.

This analysis resulted in the following function describing the open cell

voltage VOC (in mV) as a function of temperature (in K), methanol

concentration (in mol dm-3), and cathode loading (in mg cm-2):

457 1.58 10.8 18.6OC cathV T N m= + − + (14)

Other parameters, like active surface area, fuel flow, cathode anode loading,

and air flow, have been evaluated and have no significant influence on the

open cell voltage. The practical working voltage V can be described as the

open cell voltage minus the Ohmic losses ΔVΩ, the activation losses ΔVact, the

crossover and internal currents ΔVcross, and the mass transportation and

concentration losses ΔVmass, which can be modeled as [112]:

( )OC act x over massV V V V V VΩ −= − Δ − Δ + − Δ (15)

Or:

( ) ( )ln expOC cV V ir A i i m ni= − − + + (16)

Where:

i = current density variable in mA cm-2

r = Area Specific Resistance [117] 0.19kΩ cm2 at 120°

C

ic = crossover current density ~ 0.25i

A = combined slope of the Tafel line for the anode and

cathode A = ( ) ( )2 2

RT RTF Fa cα α+

m 2.11E-2 mV

n

= Mass-transfer overvoltage constants (based on the

Ballard PEMFC [112]) 4.00E-2 cm2 mA-1

αa = charge transfer coefficient at the anode [118] 0.239

αc = charge transfer coefficient at the cathode [118] 0.875

F = Faraday’s constant 96.485 C mol-1

R = Molar gas constant 8.314 J K-1 mol-1

130

The practical V-i curve is shown in Figure 48. With the fuel-cell model

described above the size for the different components can now be set up.

This will be described in detail in the following section.

0

50

100

150

200

250

300

350

400

450

500

0 50 100 150 200 250

current density (mA/cm2)

Vo

ltag

e (m

V)

0

5

10

15

20

25

30

35

40

45

50

Po

wer

(m

W/c

m2)

Figure 48: V-i characteristics used in the fuel cell model at T=298K, mc=2mg cm-2 and

N=1 mol dm-3 (~4%v/v).

8.2.3 System design

In Figure 49 an overview is given of all components, mass flows and the

interconnections. An intermediate accumulator is needed to take care of

peak-power load which probably results in a decrease of the fuel cell volume,

weight and cost. In the following paragraphs the main components will be

discussed and sized. Within this thesis only the components which fall in

the fuel-cell system boundaries, as pictured in Figure 48, are taken into

account. The conversion from a low-voltage to higher/lower working voltage

is not investigated but assumptions have been made, which are discussed in

Section 8.4.2. The physical properties of fuel cells are based on the model as

described in the previous section. The physical properties of all other

components like the pumps and the tank are based on a mass-flow and fuel

131

cell performance model as described by [112] and will be discussed in

Section 8.2.4 to 8.2.8.

In our case the system is a parallel/series hybrid. The fuel cell delivers a

constant power output. The battery is charged when the load is low and

delivers power when peak power is needed. In this case the fuel cell can be

sized based on mean power instead of maximum power output.

Fuel Tank

anod

e

Methanol feed Mixing

chamber (1M)

Water feed

Air

cath

ode

CO2 venting

Boost converter

Load

gaseous flow fluid flow electric flow information flow sensors (flow, methanol, power (Vi), temperature)

Air filter H2O condns.Humidifyer Heat exch.

+ -

+ -

H2O condns.CO2 filter

fuel mix

fuel mix

CO2 Air

CH3OH

H2O

Clean Air or O2

fuel mix +

CO2

H2O+ Air

H2O

m

p

f m

p

f

on/

off

μC intermediateaccumulator

Fuel Cell system

Figure 49: functional overview of all components in the DMFC system.

8.2.4 Sizing of the fuel-cell flat-pack

The requirement for the fuel cell is a repetitive load-profile as described in

Figure 46. The mean output power of the fuel cell should deliver is at least

150mW. Taking charge efficiency, overall Ohmic losses (~90%) and power

using auxiliaries (-10%) into account the power output of the fuel cell should

be around 185mW. The basic volume requirement and specifically the

maximum thickness of the power system results in a flat pack architecture

of the cells. The number of cells needed is at least three, to acquire a

workable voltage of 0,7-1V. Higher number of cells probably will increase

132

costs and surface area. The fuel cell output voltage will be boosted to a

working voltage of 3,8V. In Table 16 an overview is given of the general

specifics of the fuel-cells flat-pack needed to fulfill the load. Every fuel cell

membrane is placed in between two endplates made out of injection-molded

carbon-filled polymers (see Figure 50d). The mass flows needed to fulfill to

the output power can be found in Table 17.

Table 16: General specifics of the fuel-cell flat-pack calculated to deliver a constant

power of 185mW. Fuel cell membrane Nafion117

Cathode loading mc 2 mg cm-2

Anode loading ma 4 mg cm-2

Active surface area A 3x 14x14mm2

End plates 6 plates 22x22x1.5mm3

Number of cells 3

Nominal output voltage V 0.76V

Open Cell voltage VOC 1.58V

Nominal current density i 125mA cm-2

Cell end temperature T 302K

Methanol concentration N 1 Mol L-1

Fuel cell efficiency 21%

Table 17: Calculated in- and outgoing mass flows in the fuel-cell power-system producing a constant systems power output of 150mW [112, 147].

Anode IN Anode OUT

Total CH3OH 31 μL min-1 CH3OH 28 μL min-1

Total H2O 783 μL min-1 H2O 766 μL min-1

CO2 1679 μL min-1

Cathode IN Cathode OUT

Air flow 26.8 mL min-1 CH3OH x-over 0.6 μL min-1

of which O2 flow 5.07 mL min-1 Air flow 24.0 mL min-1

of which O2 flow 2.54 mL min-1

H2O fluid 4 μL min-1

H2O vapor 1 μL min-1

H2O osmotic 16 μL min-1

133

Figure 50: Overview of the size for the main components. From left to right: a) available pumps, b) available intermediate accumulators, c) the water and methanol tanks, d) the

final fuel cell design, and e) the PCB.

8.2.5 Sizing the intermediate accumulator

The list of requirements demanded that the power-system should be able to

follow the load profile as described in Figure 46. In the first 10 and the

following 60 seconds of this load curve the power needed will be realized by

the fuel cell (150mW) plus power from the added accumulator (171mW). The

following 30 seconds will be powered by the fuel cell (150mW) and the

battery (80mW). In the final 3 minutes the fuel cell will power the MP3 player

on its own (99mW), and the remainder power (51mW) will charge the battery

to its nominal capacity (80-100% SOC). The minimum capacity of the battery

needed to fulfill the load of this 4 minute song is approximately 2.6mWh.

During the 30 seconds startup the system uses 700mW. To deal with this

load a minimum of 6mWh is needed.

Options to accumulate this amount of energy are a Nickel Metal Hydrate

(NiMH) button-cell, a lithium rechargeable button cell, or a capacitor (see

Figure 50b). Five important factors will influence the choice: the technical

lifespan, the ability to charge and discharge at high currents (C), the working

voltage, the available capacity and its volume. Both the Lion as the NiMH

battery have a short life-span when discharged to 0-10%SOC. To increase

life-span the batteries should be discharged to 90%SOC, resulting in a

minimum capacity of 26mWh for both the batteries. Lithium based button-

cells are mostly Lithium Vanadium or Aluminum-Manganese batteries

working at a voltage of 3V. The rated (dis)charge rates (<0,1C) of all lithium

ion batteries is the limiting value. A high capacity battery or battery-pack is

needed to comply with the large (dis)charge currents. State-of-the-art lithium

button cells don’t seem to be applicable for intermediate accumulation.

134

NiMH batteries on the other hand are more able to handle high (dis)charge

currents. When complying with the requirements described above two Varta

V40HR NiMH [56] batteries are capable. The specifications can be found in

Table 18 Two batteries of this type are used in the design of the power-

system. The third option, using a capacitor, is good because of its high

power density and long cycle-life, but can not compete with NiMH batteries

on its volume. Furthermore the start-up specifics demand a long on-the-

shelf-life and capacitors tend to have a short self discharge period.

8.2.6 Sizing the fuel tank

In Table 17 the in- and outgoing mass flows are calculated needed to fulfill a

constant power delivery of 185mW. The fuel cell consumes about 3.1μL min-1

(including cross-over) of methanol and 1.4μL min-1 of water. According to the

list of requirements the power-system should be able to produce 3.1Wh of

energy. The amount of methanol needed to fulfill this requirement is 3.75mL.

In our design we use a methanol tank of 4mL. Water is also consumed, in

total 1.5mL, and a tank of the same size as the methanol tank should be

able to comply with the requirements. Because of osmotic drag, almost

16.7mL of water is dragged through too the cathode during one take. If this

water is not recycled a water tank of at least 18mL is needed. If the water is

fully or partially recycled the water tank will take up less space. To decrease

volume we assume water will be recycled as much as possible. The tanks are

made of flexible plastics which is blow-molded in its final shape (Figure 50c).

8.2.7 Fuel and air pumps

Fuel is introduced into the fuel cell by a pump. The flow rate of the fuel

mixture is approximately 0.82mL min-1 (Table 17). Three micro pumps

found comply with this requirement, the HNP mzr®-2521 micro annular gear

pump [148], the ThinXXS MDP1304 piezo actuated micro diaphragm pump

[149] and the Bartels Mikrotechnik MP5 piezo actuated micro diaphragm

pump [150]. Both the HNP and ThinXXS pumps are quite bulky compared to

the Bartels MP5 (see a). The MP5 complies with the required low thickness of

4mm. Further specifics of this pump can be found in Table 18.

135

Air supply to the fuel cell can be done passively, semi-active by means of a

blower or active by means of an air pump. To decrease volume and add more

control to the cells performance an active air pump is chosen. One Bartels

MP5 pump can only power an air flow of 15mL min-1. To comply with

minimum of 26.8mL min-1 air flow required, two MP5 pumps are needed. For

this design two pumps are linked in parallel.

8.2.8 Other components

The largest components have been defined above. Other smaller components

needed to make the system work are a methanol sensor, water recycle

system and a water-methanol mixer (e.g. from ISSYS [151]). Furthermore

some temperature and flow sensors could be added to improve the

performance of the fuel cell system. The processor used in the MP3 player

could be used to control the system. We assume this is possible in this

design.

8.2.9 Final design of the power system in the MP3 player

The specifications of the main components are summarized in Table 18. An

intermediate accumulator is needed to take care of the peak power load

resulting a in a decrease of the fuel cell volume, weight and system costs.

Table 18: Specifications of the main components used in the fuel cell power system.

Component Type Specifications Dimensions

Fuel cell Nafion 117 185mW 3x 14x14mm2

Capacitor Varta V40HR

NiMH

2x 1.2V; 20mAh Ø11.5x5.4mm3

Methanol tank 4000mm3

Water tank

Blow molded 3.1Wh; ~4mL;

t=0.5mm 4000mm3

Fuel pump Bartels MP5 50μl/min -

5ml/min

14x14x3.5mm3

Air pump Bartels MP5 50μl/min -

15ml/min

2x

14x14x3.5mm3

The design of the system is based on a constant power output of the fuel cell

system of 150mW. Taking charge efficiency of the battery, Ohmic losses

136

(~90%) and power using auxiliaries (~10%) into account, the power output of

the fuel cells should be around 185mW.

The fuel-cell hybrid can power a flash-drive MP3 player. A CAD model of the

power system is made to get more insight in the size of all components they

are drawn in CAD. Different structural variants have been evaluated and the

final assembly is shown in Figure 51. Electronic interconnections haven’t

been modeled. The CAD model shows the system cannot fit into the space

available in the Samsung MP3 player. The power systems size is

35x86x8.1mm3 (24,400mm3) and the total volume of all main components

alone is equal to 16,300mm3. The volume breakdown of the system is shown

Figure 52.

The list of requirements state that the volume should not exceed 8,700mm3,

at equal performance specifics. The total amount of energy the designed

power systems delivers is equal to that of the used battery, resulting in an

energy density of 135Wh L-1 and a power density of 37W L-1, when taking

empty space into account (Table 19). This is almost 3 times less than the

available lithium ion polymer battery used in the MP3 player. To decrease

volume and increase energy density, the efficiency at low temperature

operation (303K) should be higher, empty space should be diminished and

finally all components should be smaller en sized according to the

performance specified.

Table 19: Comparison between the lithium polymer battery used and the designed power system.

Lithium polymer

battery

DMFC power

system design

Fuel cell (mm3) - 4,600

Fuel tank (mm3) - 7,800

BOP (excl. empty space) (mm3) - 3,700

Empty space (mm3) 8,300

Total volume (mm3) 8,700 24,400

Energy (Wh) 3.1 3.3

Energy density (Wh L-1) 348 127

Peak power density (W L-1) 103 37

137

Figure 51: The fuel-cell power-system assembly.

Figure 52: Volume breakdown of the fuel-cell power-system at a total volume of

24,400mm3.

138

8.3 Design of the ‘scaled’ DMFC hybrid system As discussed in the previous section, the energy and power density could be

increased by increasing the fuel-efficiency and by using scaled down

components. In this section the fuel cell membrane efficiency is compared to

state of the art efficiencies described in literature. Higher efficiencies result

in a smaller fuel-cell flat-pack and less fuel needed.

8.3.1 Resizing the fuel-cell flat-pack

The Nafion 117 membrane is normally used in Direct Methanol Fuel cells.

The efficiency of the Membrane Electrode Assembly (MEA) is quite high

compared to its competitors. The cell performance can generally be

characterized with polarization and power curves. In the model for the

conventional design the membrane is dimensioned for use at a constant

nominal load of 185mW needed to charge/discharge the battery and

auxiliaries like the pumps.

Figure 53 and Figure 54 show the polarization and power curves for the cell

used in the model. These figures are based on the model described in [147].

The figures also show other curves taken from recent literature [131-133].

Figure 53: Polarization curves of Nafion 117 MEA’s found in literature [131-133], plus

the analytical model (dashed line) [147].

139

Figure 54: Power curves of Nafion 117 MEA’s found in literature [119-121], plus the

analytical model (dashed line) [147]

At least three membranes are needed to acquire a workable voltage of

0.761V. The nominal current is 125mA cm-2 and the membrane efficiency in

this case is equal to 21%. The operating temperature of the cell is calculated

using a thermodynamic model made for the power system. The final working

temperature will not exceed 303K. Most curves found in literature are based

on high temperature operation (333-353K), while the temperature of the

micro-scale fuel cell will probably not be more than 310K.

Temperature has a great influence on the performance of the cell. Different

polarization and power curves are found in recent literature [119-121],

Figure 53 and Figure 54. Besides the direction of the polarization curve, the

modeled polarization curve fits the current measured low-temperature

curves from literature quite good. Probably the Ohmic losses are higher in

real setups than assumed in the model.

The cell developed by [120] has a slightly better performance curve than our

model. The efficiency of the cell is slightly better. At a nominal current

density of 125mA cm-2 the Padhy cell has an efficiency of 29% instead of

21% for the model. The improved performance of the Padhy cell is mainly

140

due to more efficient SS316 endplates with modified serpentine flow field.

Assuming this efficiency can be met in the renewed design the active area of

one cell could be decreased from 14x14 to 12x12mm2.

8.3.2 Resizing the fuel tanks

The previously described efficiency increase also influences the amount of

fuel needed to comply with the demand. The volume of the methanol tanks

will decrease with a factor 1.4 compared to the conventional design. In the

new design only 2.7mL of methanol and 1mL of water is consumed to fulfill

the requirement of 3.1Wh of energy. Again it is assumed that the water

produced by the cell is recycled as much as possible. For the final redesign

the water tank is as large as the methanol tank, both with a storage volume

of 2.79mL (~3.17Wh).

8.3.3 Resizing the intermediate accumulator and trends for the

future

Commercially available rechargeable button cells, as used in the

conventional design, are mainly produced for application in low-power

backup power for personal-computer real-time clocks and BIOS

configuration data. The low discharge characteristics make them not very

useful for application in the fuel-cell battery hybrid design. In general the

battery used in the conventional design is used as a peak-power generator.

Per cycle the battery discharges within 1 minute and is charged in 3 minutes.

This 4 minute charge/discharge-profile of the battery demands the following

for the battery used:

1. high discharge rate, up to 60C

2. high charge rate, up to 20C

3. long cycle life, more than 16,000 cycles

The demands above cannot be met with commercially available rechargeable

button cells. This is mainly because of the low rated (dis)charge rate of

rechargeable button cells and their low cycle life of 1,000 cycles. By taking a

larger cell the charge and discharge rate will decrease, but this also results

in a larger and bulky battery. In the conventional design two Varta V40HR

141

NiMh button cells were used which could comply with the charge and

discharge demands and still be small. For the chosen load-profile all

rechargeable batteries will reach their expected maximum number of cycles

at 1,000 cycles (10%DOD), meaning in this case after 3 months. Alternatives

are either using an other intermediate accumulator, like a capacitor, or

changing the load-profile of the battery to a longer cycle period.

Alternative accumulators are high capacity capacitors, or Electric

/Electrochemical Double Layer Capacitors (EDLC). The electrochemical DLC

is also designated as super capacitor, ultra capacitor or pseudo capacitor,

and is a unique electrical storage device which can store more energy than

conventional capacitors and offer high power density, Table 20. Most

commercially available Electrochemical DLCs have a low specific energy

below 10Wh kg-1 compared to rechargeable batteries.

In Table 20 an overview is given of commercially available super capacitors

and two prototypes. Most of the super capacitors are very voluminous,

because of the interesting application field of hybrid and electric vehicles

(0.05 to 0.547L). The smallest in size is 0.0016 Liters, Maxwell PC5 [69] and

has an energy density of 2.13Wh L-1.

Table 20: Overview of commercially available supercapacitors and two prototypes taken

from [81, 163-165] Voltage

V

(V)

Capacity

C

(F)

Capacity

E

(Wh)

Weight

m

(kg)

Volume

V

(L)

Spec.

energy

u

(Wh kg-1)

Specific

power

p

(W kg-1)

Energy

density

uρ

(Wh L-1)

Power

density

pρ

(W L-1)

Maxwell 2.7 350 0.26 0.06 0.05 4.4 1068 5.28 1282

Maxwell PC10 2.5 10 0.009 0.0063 0.0034 1.37 660 2.57 1240

Maxwell PC5 2.5 4 0.003 0.004 0.0016 0.84 470 2.13 1185

Ness 2.7 1800 1.35 0.38 0.277 3.6 975 4.94 1338

Asahi Glass 2.7 1375 1.03 0.21 0.151 4.9 390 6.81 542

Panasonic 2.5 1200 0.77 0.34 0.245 2.3 514 3.19 713

Power syst. 2.7 1350 1.01 0.21 0.15 4.9 650 6.86 910

Fuji Heavy

ind.-hybrid

3.8 1800 2.67 0.232 0.143 9.2 1025 14.93 1663

APowerCap 2.7 55 0.04 0.008 0.006 5.0 8300 7.0 11620

Skeleton 2.7 1422 1.07 0.201 0.137 5.3 2929 7.8 4311

142

Most developments are directed to increasing the specific energy as can be

seen in the prototype carbon/carbon devices in Table 20. The specific energy

is projected to increase from 4 to 5Wh kg-1 [152]. For our design we are

interested in the energy density (per unit of volume) instead of the specific

energy (per unit of mass). To compete with the NiMh batteries chosen in the

conventional design, the super capacitor has to have a energy density of at

least 9.3Wh L-1. As shown in Table 20 all commercially available super-

capacitors have an energy density lower than 7, but for the Power system

(2.7V) and the Fuji Heavy Industry Hybrid (3.8V). According to press releases

in 2005 [153] Fuji even prototyped a li-ion hybrid capacitor with energy

density boosted up to 27Wh L-1. This capacitor uses li-ion battery electrodes.

Applying high energy dense super capacitors in our application will not only

decrease volume but also solve the problem of cycle life. According to Burke

[152] it is unlikely that capacitors “using battery-like electrodes will have a

cycle life comparable to the carbon/carbon devices (greater than 500,000

cycles) for deep discharges. However, a cycle life of 100,000 deep cycles

seems possible by proper design of the battery-like electrode.” For our

application this maximum number of deep cycles is sufficient.

In the conventional design a comparison is made between the Varta V40HR

NiMh battery, Varta MC621 Li-ion battery and Maxwell PC10 super capacitor.

The PC10 capacitor has an energy density of 2.57Wh L-1 (capacity of

8.6mWh), and is more bulky than two NiMh button cells from Varta. When

the Fuji Heavy Industry hybrid capacitor is scaled down to the required

energy demands (10.6mWh) a capacitor with a volume of 710mm3 is feasible.

Based on the form of the PC10 the capacitor could have the following

dimensions 13.6x10.8x4.8 mm3.

8.3.4 Liquid pump

The fuel pump chosen in the conventional design was a Bartels MP5 micro

pump [150]. When the fuel cells have to deliver a constant power of 185mW

the fuel flow should be 814 μL min-1 (Table 21). The maximum liquid flow of

the Bartels MP5 pump is in between 4.5mL min-1 and 5mL min-1 [150]. The

143

liquid flow needed is less than 1mL min-1 meaning the used component is a

factor 5.5 times to powerful.

In Vishal et al. [154] and Laser et al. [155] an overview is given of different

techniques to pump small amounts of liquids. In general the micro pumps

considered in literature are classified in mechanical pumps, like rotary and

vibrating diaphragm pumps, electro kinetic micro pumps, magneto kinetic

micro pumps, phase change micro pumps, and other novel techniques.

In [155] the term Self Pumping Frequency fsp is introduced which is an

interesting metric ratio of maximum flow rate to package size:

maxsp

package

Qf

V= (17)

If we compare piëzo-electromechanical diaphragm pumps to electromagnetic

micro pumps the self pumping frequency of the latter is much lower. For

both commercially and laboratory pumps the volume of the electromagnetic

pump is a factor 3 to 7 larger than the piëzo actuated pump (Table 21).

When volume is of great importance, like in the design of the micro fuel cell

system described, piëzo actuated diaphragm pumps seem to be more

interesting. Reviewing other pumps under development, like the Electro

Hydrodynamic (EHD) and Electro Osmotic pump, a big improvement in

minimizing the pumps volume is achievable.

Table 21: Overview of different commercially and laboratory micro pumps [148-150, 156-159].

Company Type Year Availa-bility

Pump technique actuator

Volume(mm3)

Max liquid flow (μL min-1)

SPR (min-1)

Power rat. (mW min L-1)

Bartels MP5 2006 C diaphragm, piezo 686 5,000 7.3 40

thinXXS MDP1304 2007 C diaphragm, piezo 1383 7,000 5.1 33

Schwarzer SPV125ZL 2009 C diaphragm, piezo 20535 30,000 14.6 1.7

Bohm et al. - 1999 L diaphragm, piezo 288 1,900 6.6 -

Bohm et al. - 1999 L diaphragm, EM 1,000 2,100 2.1 0.24

HNP MZR-2521 2006 C annular gear, EM 9,955 9,000 0.9 333

Richter et al. - 1991 L EHD, injection 6.84 14,000 2,047 -

Chen et al. - 2000 L ElectroOsmotic 0.003 15 4,386 1.4

144

For middle long period the piëzo diaphragm pump will be the right choice. At

the Fraunhofer Institut IZM a lot of development is going on in the field of

piëzo actuated diaphragm micro pumps [160-162]. In Table 22 an overview

is given of the micro pumps developed in the laboratories of Fraunhofer

Institut IZM. From personal communication with Fraunhofer [163] it became

clear that the high flow pump is being industrialized. The Self Pumping Ratio

(SPR) of these micro pumps is a factor 5 to 10 higher than the commercially

available pumps. When miniaturizing piëzo pumps to the volume flow of

need scale effects have to be taken into account. The self pumping frequency

of the two commercially available piëzo pumps are 5.1 to 7.3 with a mean of

6.3 (Table 21).

Table 22: Overview of new developments in micro pumps from Fraunhofer Institut IZM [13, 160, 161]

Type Year AvailabilityCom./Lab.

Pump technique actuator

Volume(mm3)

Max liquid flow(μL min-1)

SPR (min-1)

Power ratio (mW min L-1)

standard 2007 L diaphragm, piezo 54 2,000 38 -

high flow 2009 L diaphragm, piezo 54 4,000 74.1 -

high pressure 2009 L diaphragm, piezo 54 400 7.4 -

micro 2009 L diaphragm, piezo 18 1,000 55.6 25

high perform. 2009 L diaphragm, piezo 2827 12,000 4.2 -

For the liquid pump needed in our design this means the pump should be a

factor 5.5 smaller than the chosen Bartels MP5 micro pump in the

conventional design. This results in a volume of approximately 125mm3. The

Bartels pump dimensions are equal to 3.5x14x14mm3. Because the

deflection of the piëzo element will decrease with decreasing width and depth,

scaling will be equal in all three directions resulting in a pump with the

following dimensions: 2.0x7.9x7.9mm3. If we take the developments of IZM

into account the fuel pump could even be an extra factor 5 to 10 smaller,

resulting in pump dimensions of 12.5-25mm3. Take in mind the pump

dimensions published by Fraunhofer are without connectors and housing.

8.3.5 Air pump

When the fuel cells have to deliver a constant power of 185mW the airflow

should be 26.8mL min-1 (Table 23). The maximum airflow of a MP5 pump is

15mL min-1, so at least two pumps were needed to fulfill the required flow. In

145

the conventional design two Bartels MP5 pumps were used. At the time of

designing the fuel cell system in 2007 [147] Bartels announced the MP6,

combining two piëzo actuators inside one single housing and a maximum

gas flow of 20mL min-1 [150], stored in a housing of 30x15x3.8 (1710) mm3.

Alternatives for the piëzo-actuated air pump are passive airflow or rotating

electromagnetic actuated flow generators. In Table 23 an overview is given of

different air pumps.

Table 23: Overview of gas flow pumps [149, 150, 157, 164-168]

Company Type Year Availability

(Com./Lab.)

Pump technique, actuator

Volume

(mm3)

Max gas Flow (μL min-1)

SPR

(min-1)

Power ratio

(mW min L-1)

Kaemper - 1998 L diaphragm, piezo 504 3,500 6.9 -

Cabuz - 2001 L diaphragm, piezo 225 30,000 133.3 2.67E-4

Bartels MP5 2006 C diaphragm, piezo 686 15,000 21.9 0.013

MP6 2008 C diaphragm, piezo 1,710 20,000 11.7 0.010

thinXXS MDP1304 2007 C diaphragm, piezo 1,383 22,000 15.9 0.046

Schabmueller

2002 L diaphragm, piezo 120 690 5.8 -

Richter standard 2007 L diaphragm, piezo 54 10,000 185 -

high flow 2009 L diaphragm, piezo 54 40,000 740 -

high press. 2009 L diaphragm, piezo 54 1,000 18.5 -

Micro 2009 L diaphragm, piezo 18 4,000 222 -

high perform2009 L diaphragm, piezo 2,827 35,000 124 -

Bohm - 1999 L diaphragm, EM 1,000 40,000 40.0 -

Xavitech V200-GAS 2008 C diaphragm, EM 16,300 450,000 27.6 0.005

Thomas BL-G 085 M 2006 C rotary vane, EM 50,000 8,500,000 172.2 2.8

For passive airflow the output characteristics will fluctuate, meaning the

desired airflow is not always met. Because of inconstant airflow this option is

left out and an actuator is used to force the air in the cells.

Based on Table 23 electromagnetic and electrostatic diaphragm pumps seem

to be less bulky for the function of delivering airflow, shown by the Self

Pumping Ratio. Thus, the two Bartels MP5 pumps are replaced by an

electromagnetic diaphragm pump with dimensions based on fsp =34. The

pump dimensions will be almost 40% smaller, 883mm3, than the two Bartels

MP5 pumps together. Electromagnetic actuated pumps are mostly not flat

146

designs but more long and cylindrical, making a cylindrical pump with a

length of 31mm and a diameter of 6mm feasible. Based on the pumps

developed by Fraunhofer Institut IZM the SPR could even be an extra factor

10 higher. Again it has to be taken into account that all data of the

laboratory pumps is without casing (~60% of total volume).

8.3.6 Other components: tubing and electronics

Besides the previously mentioned components air and fuel has to be

delivered to the fuel cells. Flexible tubing is used with an outer diameter of

∅2mm. The tubes have a limited bending ratio which has to be taken into

account. Two tanks are needed to fuel the DMFC system, one filled with

water and one with methanol. A small passive mixer will mix the water and

methanol to the right concentration (3%w/w). The mixture will be controlled

by two valves. A small methanol micro sensor will measure the concentration

after the mixing tank, controlling the valves. The micro sensor is under

development by ISSYS and KEM, and an estimate of the dimensions is made

based on published data [151].

8.3.7 Assembly of the design

In Figure 55 the final assembly of the fuel-cell power-system is shown. All

components are modeled excluding the electronic interconnections. In this

design it is assumed the controllers are integrated on the PCB of the MP3

player. The specifications of the main components are summarized in Table

24.

147

Figure 55: The fuel cell power system assembly.

Table 24: Specifications of the main components used in the redesigned fuel-cell power system.

Component Type Specs Dimensions [mm3]

Fuel cell membrane 3x Nafion 117 0.53VOC 3x 14x14x2

Capacitor Super capacitor 10.6mWh 13.6x10.8x4.8

Methanol tank blow molded t=0.25mm 3,570

methanol 3.17Wh 2,790

Water tank blow molded - 3,570

Fuel pump Piëzo diaphr. - 7.9x7.9x2.0

Air pump Electromagn. - 31x∅6

148

8.4 Discussion of the design, redesign and the model

8.4.1 Evaluation of the designs

A conventional and scaled design of a 150mW fuel cell power system is made.

Both fuel cell hybrid systems can power a flash-drive MP3 player. The

conventional design is based on commercially available components and the

second design is based on scaled components, where every component

specific is scaled to fulfill the performance needs for that component.

Components are not available but the feasibility of the system, when

components are available, is hereby tested.

For both designs the choice of components and the efficient placing of the

components strongly depend on the qualities of the designer. He or she is

responsible for choosing the right components and an efficient architecture

by for instance structural variants. Both these jobs can be automated,

especially during concept design when the designer is not known with the

availability of different components and its performance specifications.

For both designs the CAD model shows the power system cannot fit in the

space available, 8,700mm3 (66x33x4mm3). The power systems size is for the

conventional design and the scaled design respectively 18mL (83x36x6mm3)

and 24,400mm3 (86x35x8mm3). Table 25 shows the volumetric specifications

of both designs compared to the lithium polymer battery present in the MP3

player.

In the conventional design the fuel tanks were designed as a box. In the

scaled version the tanks are formed fitting in between the tubing and acting

as a platform for fitting the other components. This gives two advantages

over the conventional design: (i) more empty space is effectively used, and (ii)

during assembly the components can be fitted to the tanks, making the

subassembly easier to handle. Table 25 compares the lithium polymer

battery used in the MP3 player with the conventional design and the scaled

design. The scaled design has been a great improvement compared to the

conventional design, but still does not fit into the compartment available.

149

Just like energy density, power density lags behind the lithium polymer

battery.

Table 25: comparison between the lithium polymer battery, the conventional design and

the scalded design as described above.

lithium polymer battery

conventional design redesigned DMFC hybrid

Volume (mm3) 8,700 24,400 18,000

Component (mm3) - 16,300 14,000

Dimensions (mm3) 66x33x4 86x35x8 83x36x6

Energy [Wh] 3.1 3.3 3.17

E.density [Wh L-1] 348 135 / 202 176 / 226

P.density [W L-1] 103 37 / 55 50 / 64

Because it was not feasible to extract weight data from the CAD data the

weight characteristics of the CAD models are not presented.

8.4.2 Evaluation of the model

The volume characteristics of the conventional design and the second design

can be found in Table 26 and 27 respectively. The model used in Chapter 5

models a stand-alone fuel-cell power system and not a hybrid system, as

designed for the MP3 player. The predicted volume of the fuel cell system is

equal to 69,000mm3, and the total volume of the conventional design and

the redesign is equal to 24,400mm3 and 18,000mm3 respectively, making

the model described in Chapter 5 not very useful for a DMFC hybrid.

In the model the volume of the fuel cell stack is based on the maximum

power output. In the design-case the maximum power output of the fuel-cell

is equal to the nominal power 150mW, less than the maximum power output

of the whole system, 868mW. Using the nominal power of 150mW in the

model a volume of 15,000mm3 is derived, lower than the outcome from the

design.

150

Table 26: Estimates of the volume for the peak and nominal load profiles (at η=21%) for the conventional design.

Conventional design

Based on Ppeak

Based on Pmean

Mean/peak power (W) 150 867 150

Fuel cell (mm3) 4,600 11,200 1,900

Fuel tank (mm3) 7,800 4,000 4,200

BOP (mm3) 3,700 53,900 9,300

Empty space (mm3) 8,300 53,900 9,300

Total volume (mm3) 24,400 69,300 15,500

Table 27: Estimates of the volume for the nominal load profiles (at η=29%) for the scaled design.

Scaled design Based on Pmean

Mean/peak power (W) 150 150

Fuel cell (mm3) 3,600 1,900

Fuel tank (mm3) 7,200 3,100

BOP (mm3) 3,200 9,300

Empty space (mm3) 4,000 9,300

Total volume (mm3) 18,000 14,300

The difference between the fuel-cell system designed by Motorola and the

design proposals of Section 8.2 and 8.3, is the extra added accumulator for

short power burst. The design of the fuel-cell stack strongly depends on the

nominal power, and the selection of the intermediate accumulator depends

strongly on peak power. Because in the Motorola case no intermediate

accumulator is used, an extra accumulator volume should be added for the

fuel-cell hybrid model. The model proposed is a simple approximation of the

volume based on the power density of the accumulator:

( )accu

peakaccu p

PV

ρ⋅= (18)

Besides the power difference the designed system has two tanks, one

methanol fuel tank and one water tank. The Motorola design case is with

only one tank containing 100% methanol and no extra water. All water

produced by the cell is re-cycled and used to dilute the methanol to a 3%wt

151

mixture. The effective energy-density of the fuel tank decreases in this

redesign case with a factor two. An extra concentration parameter cmeoh is

introduced in the model. This parameter indicates the amount of methanol

over the total liquid amount. In the Motorola case this parameter is equal to

1.0, in the MP3 design case this parameter is equal to 0.5.

The volume of the fuel-cell flat-pack in the design case has a power density

equal to almost 1/2th, namely 51.5W L-1, of the Motorola case (100W L-1)11.

The Motorola fuel cell is a 4-cell stacked version using bi-polar plates which

are thinner than the end-plates. Making use of a flat-pack design, linking

several independent cells to each other in series, increases the volume of

different cells, but also results in an overall thinner power/energy source. An

extra parameter is added to the final analytical model, the flat-pack constant

cfp, equal to 1 for stacked design and 1.5 for a flat-pack design.

System efficiency ηsys is difficult to predict. On the other hand the fuel cells

efficiency (ηfc) can be derived from the specification sheets of the fuel cell,

based on load. In the Motorola case the fuel cell voltage efficiency is 29% and

the total systems voltage efficiency 22%12. The systems voltage efficiency can

be modeled as:

bopfcsys ηηη ⋅= (19)

Where ηbop is the BOP efficiency [29]:

au

DCDCbop r+=

1

1/ηη (20)

With rau the ratio of auxiliary power to net power (~18.75%) and ηDC/DC the

efficiency of the DC/DC convertor (~90%) makes the BOP efficiency equal to

76%. Both the ratio and the efficiency of the DC/DC convertor are

11 This is the direct power output over the stacks volume. The power density is based

on the systems power-output over the fuel cell stacks volume (77.3W L-1) 12 It is assumed the overall system efficiency equals the systems voltage efficiency.

The overall system efficiency (Eout/Ein) of the Motorola fuel-cell system is 20%

152

approximations based on literature [29] and are outside the design-

boundaries of the fuel-cell system. If power conditioning circuitry is used

inside the electronic device, the device can be powered directly from the fuel

cell without conditioning, meaning without a DC/DC convertor, and achieve

a BOP efficiency of 84%.

The efficiency of the fuel cell also influences the volume of the fuel cell stack.

Higher efficiency means higher power density. For now the power density is

set at (p ρ)fc = 100W L-1. Higher power densities up to 500W L-1 have been

reported by PolyFuel [169]. The volume of the fuel-cell can now be modeled

as follows:

( ) fcfc

nomfpfc p

PcV

ρη ⋅⋅⋅

= (21)

The volume for BOP is based on the volume of wiring, tubing, air and fuel

pumps, other auxiliaries and empty space. A large chunk of this volume is

‘empty space’. In the standard design case almost 32% of the total volume

was taken by empty space. In the redesign this amount was decreased to

almost 20%. For the function proposed in Chapter 5 the BOP volume only

depends on the mean power Pmean and power density of the components

(including empty space) (p ρ)bop. The value for the power density of the BOP

components has to be corrected for, because empty space is diminished. The

empty space will be excluded from the BOP volume and and extra volume is

added, the empty space volume Vempty:

( )*

nombop

bop

PV

p ρ=

⋅ (22)

And

empty es totalV c V= ⋅ (23)

Where: (p ρ)*bop = 70W L-1

ces

= 0.2-0.32

153

8.5 Modified heuristic models for a DMFC hybrid power system

Taking all modification of the previous section into account a new heuristic

model is produced estimating the volumes of the different parts in a scaled

DMFC hybrid power source. The model can be used as a general model for a

first approximation of a DMFC system and a DMFC hybrid system. The

proposed model of Chapter 5 consisted only of three parts, the fuel cell, the

fuel tank and BOP. The modified model is extended with two other parts like

the empty space and the intermediate accumulator.

8.5.1 Model for DMFC system

The following model for an actively fueled DMFC system resulted from the

design case as described in the previous section. This model is a general

model to predict the volume characteristics of a DMFC (in 103mm3) system

less the intermediate accumulator:

( )1

1total fc fuel bopes

V V V Vc

= ⋅ + +−

(24)

Where:

( )

( )( )

( )*

can

fp meanfc

bop fc

runfuel

meoh v bop fc feul

meanbop

bop

empty es total

c PV

p

EV

c c u

PV

p

V c V

η ρ

η η ρ

ρ

⋅=

⋅ ⋅

=⋅ ⋅ ⋅ ⋅

=⋅

= ⋅

As can be seen in the model the following variables are inputted by the user:

• Mean power Pmean

• Runtime on one charge described as the total amount of energy

needed: mean

run

Prun tE Δ=

• Total run time of the system on one charge Δtrun

154

• Single cycle run Δtcycle

• And peak power Ppeak

The normalized variables, like the energy density of the methanol (u ρ)fuel, the

power density of the fuel cells (p ρ)fc [125] and the power density of the

intermediate accumulator (p ρ)accu, in this case a capacitor [69], are generally

available in literature or specification sheets. The power density for the

balance of plant (p ρ)bop is based on scaled down components derived in the

previous section. This data will be stored in the database which should be

updated for new developments in the field:

(uρ)fuel = 4373Wh L-1

(pρ)fc = 100-500W L-1

(pρ)bop = 16-70W L-1

The efficiencies ηx can be found in literature [120, 125] or calculated:

( )

/

1DC DC

bopaur

ηη =+

(20)

Where: ηfc = 21-29%

ηDC/DC = 90-100%

rau = 18.75%

The constants cx are “rule-of-thumbs” developed in the previous section:

cfp = 1.0 or 1.5

ces = 0.2-0.32

cmeoh =

2

meoh

meoh H O

V

V V+

cv can = 0.85

8.5.2 Model for a rechargeable battery

The volume for a rechargeable battery can be modeled with two different

performance parameters, the energy needed for the total run cycle Erun or the

155

maximum current draw for which the battery is still safe to use, defined as C.

A simple approximation model for the rechargeable battery is:

( )

( )

if

if

peakrun

runaccuaccu

peak peak

runaccu

PEC

u p EV

P PC

p p E

≤ ⋅=

> ⋅

(25)

Where (p ρ)accu and (u ρ)accu are the respectively the specific power and energy

of the accumulator (either a battery or capacitor) and maximum rate of

discharge C can be found in specification sheets and have to be stored in a

database. General numbers for different types of rechargeable accumulators

are described in Table 28.

Table 28: Overview of the mean power and energy density, and the maximum rate of discharge for different types of rechargeable batteries. Power density data is extracted

from [170] and recalculated with the mean density available from own database. Energy densities of the different battery types are taken from [13].

Battery type

max specific

power

p

(W kg-1)

max power

density

(p ρ)

(W L-1)

mean energy

density

(u ρ)

(Wh L-1)

max rate of

discharge

C

(A/Ah)

Lead-Acid 180 470 80 6C

NiCd 50-1000 125-2500 86 30C

NiMh - - 147 -

Li-ion 500-2000 800-3500 212 15C

Li-polymer 50-250 70-350 189 2C

Supercapacitor 10,000 15,000 0.3 50,000C

8.5.3 Model for a DMFC hybrid

The previous sections describe both the model for a DMFC and for a

rechargeable accumulator, either a battery or a super capacitor. A

combination of both can be used as the model for a DMFC-battery hybrid

system:

156

( )*1

1total fc fuel bop accues

V V V V Vc

= ⋅ + + +−

(26)

The battery in a DMFC hybrid is used as a peak-power generator containing

enough energy for one single cycle. The volume for this intermediate

accumulator is thus either based on peak power draw (mainly for battery

type of accumulators) or energy needed for a single cycle (mainly for

capacitor type of accumulators). The battery can be modeled as described in

the Equation 25, with the energy per run replaced by the energy needed for

one cycle:

( )

( )

*

if

if

cycle peak

cycleaccuaccu

peak peak

cycleaccu

E PC

u EV

P PC

p E

ρ

ρ

≤ ⋅=

> ⋅

(27)

Where C is the maximum discharge rate (h-1) of the accumulator

8.6 Evaluation of the newly developed model The newly developed model was based on numbers from literature and the

rules of thumb. These rules of thumb were produces by means of the

designs produced in the previous sections, making a validation with this

design not feasible. In Table 29 the newly developed model was tested for the

scaled design, when the following efficiencies, constants and power and

energy densities are used:

ηfc = 29%

ηDC/DC = 90%

cfp = 1.5

ces = 0.2

cmeoh = 1

cvcan = 0.85

(pρ)fc = 100W L-1

(uρ)bop = 70W L-1

157

A new metric, the volume metric, is introduced which gives an instant

insight in the feasibility of the DMFC system compared to the benchmarked

battery. The metric is a factor of the value for the power source’s volume over

the benchmarked batteries’ volume.

totalvol

li ion

VM

V −

= (28)

Where the volume of the benchmarked battery Vli-ion is in our case equal to

8,700mm3.

To validate the model Smart Fuel Cell (SFC) was contacted for more abstract

data on their DMFC systems [141, 142]. SFC produces low power DMFC

generators in the range of 25W to 90W (Efoy series), containing an amount of

energy ranging from 600-2200Wh. Data received from SFC is listed in Table

30 for the highly miniaturized Jenny (25W, 400Wh) and in Table 31 and the

Efoy 2200 (90W, 5.5kWh). Both the Jenny and the Efoy 2200 have a stacked

fuel-cell design (cfp = 1), and a 100% methanol fuel tank (cmeoh = 1.0). It is

assumed that the DC/DC convertor has an efficiency of 90%.

Table 29: Output from the CAD model of the scaled design and the modified analytical

model. CAD scaled

design

Modified

analytical

model

Error Volume

metric

Mvol

Volume FC (mm3) 3,600 3,661 +2%

Volume fuel tank (mm3) 7,200 7,589 +5%

Volume BOP (mm3) 2,700 2,643 -2%

Volume empty space (mm3) 3,900 1,751 -

Volume accu13 (mm3) 710 710 -6%

Total volume (mm3) 18,010 18,254 +1% 2.10

The data from these tables show that the values produced by the model find

itself in a wide spectrum. This is because the ranges of the power and energy

13 The volume of the intermediate accumulator is based on the supercapacitor from

Fuji with an energy density of 14.93Wh L-1 and a power density of 1025 W L-1. The volume is related to the energy needed for one cycle (10.6mWh).

158

densities of the different components are taken widely. It can be seen that

the Jenny finds itself in the left part of the volume spectrum and the Efoy

2200 finds itself in the middle part. This is not really a surprise because the

Jenny is designed to be as compact as possible and the Efoy 2200 is a

conventionally designed DMFC system.

Table 30: Data for the SFC Jenny [141, 142]. An intermediate battery is not available. Jenny Model

Volume FC (106mm3) 0.08 0.07 - 0.33

Volume fuel tank (106mm3) 0.40 0.49 – 0.68

Volume BOP (106mm3) 0.70 0.36 – 1.56

Volume empty

space

(106mm3) 0.20 0.23 – 1.21

0.0 0.5 1.0 1.5 2.0

Total volume (106mm3) 1.38 1.14 – 3.780.0 1.0 2.0 3.0 4.0

Table 31: Data for the SFC Efoy 2200, 5.5kWh [141, 142]. An intermediate battery is not available.

Efoy Model

Volume FC (106mm3) 1 0.24 – 1.19

Volume fuel tank (106mm3) 5.5 6.73 – 9.30

Volume BOP (106mm3) 3 1.29 – 5.62

Volume empty

space

(106mm3) 5 2.06 – 7.58

0 2.5 5 7.5 10

Total volume (106mm3) 14.5 10.32 – 23.69

0 5 10 15 20

The Jenny fuel cell system is a system designed for minimized volume. The

stack is not constructed as a standard bi-polar stack, but a new mono-polar

configuration is used, making the design very thin. The thickness of two cells

is 0.8mm for the mono-polar configuration compared to 2 to 4 mm for two

bi-polar stacked cells [141]. This new innovative design makes the Jenny

fuel cell stack multiple times smaller than regular designs, with an improved

power density of more than 400W L-1. For the Efoy fuel cell system, the

model predicts the volume of the fuel cell stack quite well. The power-density

of the stack is 120W L-1. For conventional designs values in the range of 100

to 150 W L-1 are a good estimate.

159

The lower heating value (LHV) of methanol is fairly constant (4.4kWh L-1),

but the model predictions are for both systems lower than the real values.

This either means the efficiency of the fuel cell system ηfc (21% - 29%) or the

BOP ηbop (76%) or the canisters’ volume coefficient cv can (0.85) is taken to

conservative. An increase in the efficiency to 35% will result in good

predictions of the fuel-tank volume.

The real value of the BOPs’ volume is within the range of both fuel-cell

systems. The power density of the BOP for the Jenny is around 35W L-1, and

for the Efoy this is eqaul to 30W L-1.

The model uses an empty space coefficient ees in between 0.2 and 0.32. The

real empty space coefficient for the Jenny system is 0.15, which is a very

high, resulting in a prediction on the left side of the graph. On the other

hand the real empty space coefficient for the Efoy system is 0.35, which is a

lot higher than the Jenny. Both systems are at the outside limits of the

models’ empty space coefficient.

8.7 Conclusions This chapter describes the development of the first-order model. The model

is based on the zero-order model produced in Chapter 5. The zero-order

model, models the DMFC system broken down in three parts, the fuel cell,

the tank and BOP. The new first-order model extends and improves the

preliminary model by breaking down the system in five parts, the fuel cell,

tank, BOP, empty space and an intermediate battery. The intermediate

battery is introduced to deliver peak powers, while the fuel cell only has to

produce a constant mean power. By introducing a hybrid system the product

designer also has to input the load-profile of the application in the model.

This data is not always at hand during concept-development, making this

model applicable in later phases of the design process.

In general the proposed model approximates the volume of different parts

quite well. The results show that the energy and power-dense Jenny system

moves in the left limit of the model and the more conventional Efoy system

more in the middle. By changing the fuel-cell stack configuration from a bi-

160

polar to a mono-polar system, the power density of the fuel-cell stacks

improves with a factor 4. The influence of this component on total volume of

the system is still low compared to the influence of the BOP and the fuel

tank. When volume is an issue the packing efficiency could be higher than

assumed in the model. The Jenny system shows that an empty space factor

of 0.15 is feasible.

The predictions for the fuel tank are outside the left limits of the model. This

can be explained either by a to conservative efficiency of the fuel cell system

ηfc or the BOP ηbop, or the canisters’ volume coefficient cv can is taken to

conservative.

The first-order model at hand only delivers volumetric evaluation of a DMFC

power system for a specific application. The results from the model deliver a

wide outcome, where innovative design with high packing-efficiency moves

on the left side of the predicted outcome, and conventional designs can be

found in the middle of the predicted outcome. More commercially available

DMFC systems have to tested to evaluate the model more exhaustively, and

define more strict ranges in variables.

Besides volume it does not predict its form, meaning dimensions are not an

outcome. By introducing a more accurate model which places commercially

available components in a 3D space the dimensions of the architecture could

be predicted and give more accurate outcome to volume and dimensions.

161

9 Second order model: database-driven metrics-based design

The results from Chapter 8 shows that the simple analytical approach to

approximate volume and other characteristics are too basic and an approach

with more accuracy is desired. In this Chapter a different approach is

proposed to test the feasibility of a fuel cell system in a specific application.

The approximation does not consist of a simple formula but evaluates

multiple architectures, or structural variants. The evaluation is based on a

multi-parametric optimization algorithm, where volume, weight and costs

are the three basic properties.

In Section 9.1 an introduction is given on preference-based optimization,

followed by an explanation of all metrics which will be optimized for. To

facilitate understanding, comparison and weighing of the effect of different

properties, all properties are made dimensionless. The dimensionless

properties are called metrics M appropriately and are not all equally

important. The importance of the metric is defined by the preference factor λ,

and the values are based on the values found in Chapter 5. Multiplying the

dimensionless factors with the specific preference factors gives the function

F(D) to optimize for, will be described in the last part of Section 9.1. The

optimization function is applied in the design algorithm proposed in Section

9.2. The optimization function is used to find the most optimized

architecture of a DMFC system applied in a specific application. The

algorithm presented is evaluated in Sections 9.3 to 9.6 by applying it to the

design of a DMFC system for the Samsung YP-Z5F MP3 player, as described

in Chapter 8. This Chapter will finish with a discussion, conclusions and

recommendations in Sections 9.7, 9.8 and 9.9.

162

9.1 “Automated design” approach The approach proposed in this chapter is a numerical evaluation of multiple

structural variants optimized for a single objective consisting of different

dimensionless metrics. The computer is used to do more than merely

analyze engineering designs, but also makes design decisions and lead the

design to an improved solution, in short an “automated design” approach. In

this automated design process an optimization algorithm is used to optimize

the system for the basic metrics. In literature the approach for optimized

layout design is used in aeronautical environments, where the 3D packing

problem with performance constraints is generally impossible to solve only

by engineers experience and intuition [183, 184], but also in the design of

circuit boards and IC chip layout [171] and loading of ships, trucks and

trains [172].

First the design variables and the construction of the objective function is

explained in Sections 9.1.1 to 9.1.3. The metrics used and the relative

preference of those metrics are explained in Section 9.1.4 and 9.1.5. Section

9.1.7 will introduce the different types of algorithms which can be used to

find the minimized objective.

9.1.1 Design variables

The numerical quantities for which values are to be chosen in producing the

design will be called “design variables” [173]. For the DMFC system the

position in space and rotation of the different components are design

variables. Amongst other properties we want to minimize volume of the total

design by changing these variables. All design variables can be combined in

a Design vector D, which simply is a list containing all the design variables

for a particular problem. For the case study the only design variables used to

define a solution is the placing of components in space:

( ) ( )( )1

, , , , , ,t r t r iD r r r rθ θ= , for i = 1,2, …, n (29)

Where rt , rr and θ are the translation vector, the rotation vector around

which the object will rotate about, and the angle to rotate respectively for all

n objects in the design. The objects in the design are components, amongst

163

others the fuel cell, fuel tank and pumps. Every combination of these

variables refers to a specific architecture or placing of objects related to each

other in space (Figure 56).

rt rr <0,0,0>

Component i

θ

Figure 56: Definition of the translation vector rt, the rotation vector rr and the rotation

angle θ for a specific component i.

9.1.2 Design constraints

At this point a design is now simply a set of values of design variables

defined by the design vector D. It must be noted that the size of the

components are taken from a database of components which are chosen

based on there performance specifications (see Chapter 8 and Section 9.3 of

this chapter). To produce an acceptable design restrictions are introduced,

further called the “design constraints”. There are three categories of

constraints: side constraints, behavior constraints and constraints arising

from a discrete-valued design variable [173].

A constraint restricting the range of design variables for other than the direct

consideration of performance is called a side constraint. In our design

example a side constraint is amongst others the systems length that should

be in between a minimum and maximum:

( ) 0.5 1.5battery system batteryg D l l l≡ ≤ ≤ (30)

164

In the above example the minimum and maximums systems length is related

to the benchmarked value used in the application, in our case the length of

the battery

A constraint derived from the performance or behavioral requirements is

called behavioral constraint. A behavior related constraint is in our case

amongst others the design of the fuel tank which is related to the energy E

needed in the application:

( ) ( )t t t fuelh D l h w E uρ≡ = (31)

A discrete-value constraint is a constraint which arises when the design

variable is not selected from a continuous range of values but is permitted to

take only one of discrete values. In our case the rotation angle is constrained

to rotations in quantities of ½π:

( ) 1

2j D kθ π≡ = , with k ∈ Z (32)

9.1.3 The objective function

In general the goal of multi-objective optimization is to find a set of solutions

as close as possible to the Pareto optimal front, Figure 57 [174]. This means

there is more than one optimal solution (which is always the case) to the

multi-objective optimization problem. In Figure 57 an optimized Pareto front

is created by changing the preference factor λi for both the optimized

functions, or the objectives, f1 and f2, where fi=λi Mi. The gradient of the lines

in the figure describe the importance of the different optimized functions. By

changing the preference factors the Pareto optimal front is created. A change

in the preference factors will result in a different optimal solution. The

factors should therefore not be arbitrarily chosen. To find the best trade-off

the results from the conjoint analyses are used to produce the best set of

preferences.

Preference bases multi-objective optimization is often used to simplify a

problem to a single objective optimization problem, resulting in a single

solution, the combined optimization function. The method used in this work is

called the Weighted Sum method [174], which scalarizes a set of metrics into

165

a single objective function F(D) by pre-multiplying each metric Mi(D) with a

user-supplied weight λi:

( ) ( )1

minm

i ii

F D M Dλ=

= (33)

All metrics used have to be handled and converted in the same type, a

maximizing or minimizing metric. The object metrics in our case are all of

the minimizing type: minimizing volume, minimizing weight and minimizing

costs. Dimensional unit effects are taken out of the picture by making all

metrics dimensionless, resulting in normalized metrics Mi, see Section 9.1.4.

Feasible region

λ1

λ2

Decreasing F(D)

f1

f2

Figure 57: The weighted sum approach for a convex objective space.

9.1.4 Explanation of metrics

Following this procedure of the weighted sum method, the metrics used have

to be defined. As described in Chapter 7 three basic properties are proposed

for use in the second-order model: volume, weight and costs. To sum the

minimizing properties to each other we have to make them similar in

dimension. To do this the properties are normalized to the specifications of

the benchmarked power source used in the application. This means the

166

calculated costs for the DMFC power system is divided by the costs of the

benchmarked lithium polymer battery:

fc system

Caccu

CM

C−= (34)

The same thing can be done with the other two basic properties:

fc system

maccu

mM

m−= (35)

And:

fc system

Vaccu

VM

V−= (36)

Weight and dimensions are taken from specifications sheets for all

components. To make an assumption for the “purchase price” the costs for

the benchmarked power source and all components used in the fuel cell

system are set at a price per piece when purchasing 100 pieces.

9.1.5 Relative preference factor λ

The optimization function depends on three metrics defined in the previous

section. It is impossible to find the optimum in which all metrics are

minimized, making a trade-off necessary. Because some metrics are more

important than the others, the preference factor λi is introduced. The

preference factor gives the importance of the specified metric compared to

the other metrics. All preference factors summed together must be equal to 1:

1 21

1m

i mi

λ λ λ λ=

= + + + = (37)

Based on higher-level information, the preference factor is first chosen.

167

9.1.6 Presentation of the objective function

The objective function is defined and can be used in an algorithm, in search

for the optimal design with the following minimizing function (Equation 33):

( ) ( ) ( ) ( ) ( )min C C m m V VF D M D M D M D p Dλ λ λ= + + +

(38)

Subject to:

( )( )( )( )( )( )( )( ) ( )

( )

1

2

3

4

5

6

7

1

2

0.5 1.5

0.5 1.5

0.5 1.5

1

0 1

0 1

0 1

battery system battery



C m V

C

m

V

t t t fuel

nomfc fc

fc cel

g D l l l

g D h h h

g D w w w

g D

g D

g D

g D

h D l h w E u

Ph D l w

n V

λ λ λ

λ

λ

λ

ρ

≡ ≤ ≤

≡ ≤ ≤

≡ ≤ ≤

≡ + + =

≡ ≤ ≤

≡ ≤ ≤

≡ ≤ ≤

≡ =

≡ = =⋅

( )( )

12

intersections

1

2

, with

0 if

other

fc

l cell

i j

n

i

j D k k

Obj Objp D

n

θ π

⋅ ⋅

≡ = ∈ Ζ

∉=

Where g1(D) to g3(D) define the search field in which the optimization may

take place, and g4(D) to g7(D) define the field for the preference factors λ. The

behavioral constraints h1(D) and h2(D) define the two flexible components as

will be described in Section 9.3.5 and 9.3.6. To decrease computation time

the values for rotation of the objects is restricted to discrete values by j(D).

The penalty factor p(D) is introduced to the function when objects from the

component set intersect with each other.

168

9.1.7 Optimization methods

The geometrical packing problem, those involving spheres, cubes and

parallelepipeds can be solved by using computers and heuristics. First the

problem is written with the design vector D such that F(D) min, where D is

the vector with all design variables. The methods to optimize this kind of

problems are numerous, varying from classical methods like Grid. Random

Walk and Gradient-based Algorithms [173] to non-classical optimization

algorithms like Evolutionary Algorithms (EA), which mimics nature’s

evolutionary principles to drive its search towards an optimal solution [174].

In Figure 58 an overview is given of different types of optimization algorithms.

Every variable used in the Design Vector defines one degree of freedom in the

design space. The grid based algorithm samples the whole field of solutions

in search for the minimized objective function F(D). With five components

this means 25 design variables14 and thus a design space with 25 Degrees Of

Freedom. This will result in multiple optimal solutions but also a large

number of numerical evaluations. Even with state-of-the-art computers this

is not desired. A deterministic approach also uses the gradient of the

objective function to direct the solution to a local optimum. Using this

method will not always result in the best solution of the optimization

because there are probably multiple local optimums.

Gradient descent

Evolution strategy

Increasing numerical evaluations

Random Deterministic

Grid based

Random walk

Increasing algorithm complexity

Grid

Increasing number of (local) optimums

Heuristic

Figure 58: Breakdown of types of optimization algorithms.

14 Three variables in the translation vector, one variable in the rotation vector, and

one in the rotationangle of the object, equals 5 DOF per object.

169

A less calculation-intensive method is the random search method generating

a random cloud of design vectors D’s within the design space. Every

improved trial to the previous trial is stored in a table. This method is

analogous to shooting with hail. A more sophisticated random-based method

is the Random Walk. This version is based on a sequence of improved

approximations to the minimum, each derived from the preceding

approximation. The sequence is determined from the prescription [173]:

1n n rD D eρ+ = +

(39)

Where Dn is the “old” approximation to the minimum and Dn+1 is the “new”

approximation, ρ is the scalar step length and er is a unit random vector.

The random walk method guides the solution to a minimum. Depending on

the starting point of the algorithm the solution converges to a local minimum

and not always to the overall minimum.

Within this thesis a variant on the classical method of Random Walk is

chosen to be used in the first approach towards an automated design and

evaluation of multiple structural variants (Section 9.4 and 9.5). To decrease

calculation time and increase accuracy the algorithm is extended with an

evolutionary algorithm in Section 9.6. The solution is directed by a Genetic

Algorithm (GA) [171] where a set of design vectors is generated, which all

represent a unique point in the design space. The objective function

evaluates the designs and decides which solutions are more fit over the

others. A new generation of solutions is generated based on mating the

solutions taking the fitness of each solution into account.

9.2 Presentation of the optimization algorithm Figure 59 shows the flowchart of the algorithm which not only minimizes the

objective function but also chooses the components based on simple

performance specifications, explained in Section 9.3.

The algorithm consists of three parts: the performance input and component

selection (Section 9.2.2), the multi-parametric optimization (Section 9.2.3)

170

and the part presenting the results graphically and in data files (Section

9.2.4). Before we are going to deal with these three parts the preference

factors have to be defined.

Input load cycle & requirements (design brief)

Specs non-flexible

components

Calculate Pnom, Ppeak, tcycle, Ecycle, Etotal

Calculate specs of flexible

components

Candidates flexible

components

Candidates non-flexible components

Generate initial layout sets Lij for

j=1..m

Component set Ci for i=1..n

Objects intersect

Generate new layout set

N

Calculate bounding box

dimensions, …

Y

All candidate components

Calculate Fij(D)

Choose min Fij(D)

Data of layout set Lij

Max # generations or convergence

reached

NY

Generate new layout sets Lij

where aR and a<1

i=n

N

End

Y

i=i+1

Figure 59: Flowchart of the automated design procedure.

171

9.2.1 Preference factors

In Chapter 6 a conjoint analysis has been performed to give more insight in

the factors influencing the user’s choice when buying a cell-phone and a

laptop computer. Five properties are investigated and in Table 32 the

average importance score for these properties are described. We can use the

average importance score as the preference factor for the five properties

investigated.

Table 32: Average importance score of the five properties derived from the conjoint analysis in Chapter 6, and the used values for the algorithm (right two columns).

Conjoint analysis Numbers used in algorithm

Type of product Small handheld

Large portable

Small handheld

Large portable

Example Cell phone Laptop computer

Cell phone Laptop computer

λC Costs 0.22 0.24 0.34 0.37

λt,charge Charge time 0.17 0.18 - -

λt,use Time of use 0.18 0.17 - -

λm Weight 0.09 0.15 0.14 0.23

λV Volume 0.34 0.25 0.52 0.40

Σλ 1 1 1 1

In Chapter 7 only three basic properties for optimization were defined:

volume, weight and costs. To define the preference for these basic properties

the five properties investigated in Chapter 6 have to be reduced to three.

“Charge-time” and “time of use” is left out of the equation, and the resulting

distribution is presented in the latter two columns of Table 32. For smaller

products, like cell phones, volume is more important than for larger portable

products, like the laptop computer.

9.2.2 Part 1: Performance input and component selection

In part 1 the user has to input performance data of the application he/she

wants to analyze. The performance data consists of a load-profile for one

cycle and data of the benchmarked power source used in the device, like size,

weight, costs and energy specifics. Based on these parameters the algorithm

will calculate the required physical characteristics of all flexible components

172

(as the fuel cells and fuel container) and design them, and make the choice

for non-flexible components out of a database by matching the required

performance with the actual characteristics of commercially available

components (fuel and air pumps, and the intermediate accumulator). For

every component the algorithm will choose multiple candidates, which could

fulfill the performance needs. All candidate components are listed and a

component set is made which will be optimized in part 2. The equations used

to calculate the physical and performance specifications of all flexible and

non-flexible components are explained in Section 9.3 by means of the

example of the DMFC power system for a MP3 player.

9.2.3 Part 2: Multi parametric optimization

A combination of components is arbitrarily chosen and listed in a component

set. For this set n initial solutions are defined. One solution is described in

the form of a Design vector D, describing the placing of all components in

space. The solutions are constrained by the design space defined by the

different design constraints gi(D), hi(D) and ji(D). An initial design vector is

defined D[0,0], which is tested on possible intersection of objects [175, 176].

If the objects do not intersect a feasible solution is generated. The

architecture of this feasible solution is analyzed by calculating the size of the

bounding box, its volume, the costs of the system based on the components

costs and the weight of the total system. The objective function F(D) is

calculated and stored in a list.

A new design vector D[0,j] is generated, based on the initial design vector

D[0,0]. The new design vector is generated in the neighborhood of the initial

design vector and is called neighboring design vector, see Figure 60. The

algorithm generates neighboring design vectors randomly within a

predefined field defined by the radius vector α rt (Figure 60). The newly-

constructed neighboring design-vectors are defined as:

[ ]( )( )*[ , ] [ ,0] , ,0 ,rt rj

D i j D i r r i eα θ= + − + Δ

for j =1…m (40)

Where: i = the primary design vector number, with i = 1 … n

173

j = the neighboring design vector number, with j = 1… m

α = defining the maximum field (step) in which the

neighboring design vectors are constructed

rt = a random generated vector for the translation of all objects

depending on the dimensions of the benchmarked power

source: ( ), ,t datum datum datumr R l w h=

, with R∈ [0,100]/100

rr[i,0] = vector for the initial radius angles of the i-th primary

design vector D[i,0]

er = a randomly generated unit vector for the new radius angle

(1,0,0), (0,1,0) or (0,0,1)

Δθ* = a randomly generated vector of angles ∈ [-π,- ½ π, 0, ½ π]

x

y Primary design vector D[i,0]

Neighboring design vectors D[i,j]

Search field

α rt

Figure 60: A 2D representation of the primary design vector D[i,0] for one object and the

generated neighboring design vectors D[i,j].

The algorithm generates m neighboring design vectors and all values for the

objective function are noted. After m random generated neighboring points

the solution with the lowest value of the objective function, and with no

intersecting objects, is chosen to be the new primary design vector D[i+1,0].

The loop is rerun, until the maximum number of generations or convergence

is reached after n iterations: D[n,0]. All best design vectors D[i,0], with i =

1…n, are stored separately in a list for further presentation.

174

9.2.4 Part 3: Presentation of the results

Based on a single component set the optimal layout is defined by the final

design vector D[n,0]. The architecture of this design is presented in a 3D

graphic, together with all generations of best options made on the road

towards this optimal. The next step is to choose a new component set from

the candidate components lists, and rerun the whole process for this new set.

After a couple of runs the best component-set layout is found optimized for

the minimizing objective function F(D).

9.3 Component selection: MP3 player To test the algorithm as described in the previous section a program is

written to design an optimized DMFC system for a MP3 player. The program

is written in Mathematica [177] and the code can be found in Appendix B.1.

This Section describes the algorithm by means of the case-study: the design

of a DMFC power system powering the Samsung YP-Z5F MP3 player

(Chapter 6).

Figure 61: Load profile for one WMA and MP3 file.

175

9.3.1 Performance input data

The data inputted by the user is in the form of load profile of one cycle. In

the case of the MP3 player this is the coarsened load-profile of the WMA file

(- - -) as pictured in Figure 61. The power data is imported as a list of power

in mW versus the time in seconds. The benchmarked battery specifications

are inputted as a list of the capacity (mAh), the working voltage (V), the

weight (g), price (€) and its dimensions (mm).

9.3.2 Introduction into component selection

With the data imported the algorithm will select non-flexible components

from a database of commercially available components and designs flexible

components matching the performance requirements of the device. In basis

the DMFC system consists of five main components:

1. air pump (non-flexible)

2. fuel pump (non-flexible)

3. intermediate accumulator (non-flexible)

4. fuel cell stack or flat pack (flexible)

5. fuel tank (flexible)

Other components like the fuel-water mixer and mixing chamber, water

condenser, air filters and other minor components are left out of this

optimization because of lack of information about these components or the

low impact on volume or weight. It is also assumed that the controller will be

integrated on the PCB of the device.

Data from non-flexible components is extracted from a table (CSV file)

imported in the code. A component is represented by its outer bound

dimensions (length, width and height), its weight, its price, and performance

specific variables. A component is represented as a parallelepiped (or a

Cuboid in Mathematica) by only its height, width and length parameters.

Figure 62 shows the real form dimensions of a fuel pump and its

representation in the program, and in Table 33 an excerpt for the fuel-pump

table is shown. A full overview can be found in Appendix B.3. The choice for

176

a specific component is based on its performance characteristics. For the air

pump and the fuel pump this is the maximum flow (ml min-1) and for the

intermediate accumulator this is the voltage (V) and the capacity (mAh). The

ordering is done on basis of the minimized value of these performance

characteristics.

Representation of the pump

Real form of the component

length[pump]

height[pump]

width[pump]

Figure 62: Real form and the representation of the Schwartzer Precision SP100 EC fuel

pump within the program.

Table 33: An excerpt of the fuel-pump table in the component database. For a full overview see Appendix B.3.

Brand Type V (V)

P (mW)

max flow(ml min-1)

l (mm)

w (mm)

h (mm)

m (g)

Price (€ 100+)

Bartels MP5 Diaphragm, piezo 250 200 5 14 14 3.5 1 0.1

Bartels MP6 Diaphragm, piezo 250 200 6 30 15 3.8 2 0.1

thinXXS MDP1304 Diaphragm, piezo 5 250 10 25.4 26.2 7.5 3 0.1

thinXXS MDP2205 Diaphragm, piezo 5 250 11.8 26.2 25.4 9 3 0.1

HNP MZR2521Annular gear, electromagnetic

18 3000 9 68.8 13 13 56 0.1

… … … … … … … … … … …

9.3.3 Air and fuel pump selection

The air and fuel pump are selected on there fuel flow performance. For every

pump in the table the maximum air or fuel flow is entered. On basis of the

load profile, the air and fuel flow needed is calculated. Methanol crossover

177

over the cell is hereby neglected15 . The following equations are used to

calculate the mass (g min-1) and volume flow (ml min-1) [112, 178]:

60meanstoch

cell e

Pm xM

V n Fλ= (41)

And:

m

Vρ

= (42)

Where: x = amount of moles used/produced in the reaction

Pmean = required mean power [mW]

Vcell = one cells voltage = 0.32V

ne = the amount of electrons per mole = 6

F = Faradays constant = 96485 Coulomb

ρ = density of the medium [g ml-1]

M = molar mass [kg mole-1]

λstoch = the stochiometric ratio at the anode and cathode

In the reaction of the fuel cell three media will react, water (H2O), methanol

(CH3OH) and oxygen (O2), in our case as part of air. In Table 34 the values

needed to calculate the fuel flows for these media are listed. Besides these

values the mass and volume flows are calculated needed to match with the

nominal power of the MP3 case (Pnom=150mW).

Table 34: Specific fuel characteristics for the used media in the DMFC. Water

H2O

Methanol

CH3OH

Oxygen

O2

x - 1 1 1.5

M kg mole-1 0.018 0.032 0.032

λ - 10 10 2

ρ g ml-1 1.0 0.79 0.00143

m g min-1 0.000146 0.01325 0.00398

V ml min-1 0.000146 0.01677 2.7798

15 In reality methanol crossing over the membrane is still a problem and almost 20%

of the methanol will cross over, resulting in a fuel loss, but also a voltage difference loss.

178

Besides for reaction water is also used as a carrier for the fuel to the

membrane. Methanol is normally diluted in water at a percentage equal to

3%m (2.3%v), resulting in an extra water flow needed:

3

2 2.3%CH OH

H O mix

VV =

And thus the total fuel flow at the anode is equal to:

3

3 2 2.3%CH OH

liquid CH OH H O

VV V V= + +

= 0.75 ml min-1

At the cathode the fuel flow consists of oxygen. Because the design does not

include an oxygen tank but the oxygen from air is used, and thus air has to

be pumped around. The oxygen percentage in air is equal to 21%, resulting

in a fuel flow at the cathode equal to:

2

21%O

air

VV =

= 14.6 ml min-1

Based on the calculated fuel flows at the anode and cathode the pumps can

be selected, from a table consisting the dimensions, weight and, when

available, price (for 100 items).

9.3.4 Intermediate accumulator selection

The intermediate accumulator is selected by matching the capacity of the

battery with the required capacity when the battery has to be discharged and

charged within one cycle, when the battery is discharged at a depth of

discharge (DOD) equal to 80%:

0

1

80%

cyclet

nomt

C P P t=

= ⋅ − ⋅ (43)

The dimensions, weight and, when available, price (for 100 items) is taken

from the table.

179

9.3.5 Dimensioning the fuel cell

The fuel cell is the first flexible component in our component set. The fuel

cells are designed as if they were custom build for the application. The

designer has to indicate if the fuel cell is stacked or is designed with flat-

pack architecture. The latter design is more voluminous but has the

advantage of being thin, and thus fitting a thin product design.

The design of the fuel cell stack is based on the design from [120] working at

a cell voltage of Vcell = 0.32 V at a current density of icell = 140 mA cm-2. The

number of cells needed is based on the maximum battery voltage:

battery

fccell

Vn

V

=

(44)

When assuming a flat pack design the required active area of the membrane

can be calculated with:

1

2fcnom

cellfc cell cell

nPA

n V i= ⋅ ⋅

⋅ (45)

And thus the length and the width of all of the cells, when assuming a

squared fuel cell flat pack architecture:

cell cell celll w A= = (46)

The thickness of the fuel cell flat pack is taken to be equal to the thickness

of two endplates making use of the innovative mono-polar architecture with

a thickness of 0.8mm for two cells [141].

Weight is calculated based on density numbers taken from the Motorola case

[29] equal to ρfc = 2.078 kg L-1. Prices are based on the price for one hundred

square centimeters of Membrane Electrode Assembly (MEA) taken from [132]

(may 2010), equaling $356.36, and thus $c3.56 or €c2.97 per mm2. The

MEA used in the design has an active area of 1,420mm2, making the price

for only this component equal to €42.17.

180

9.3.6 Dimensioning the fuel tank

The second flexible component is the fuel tank, which has to match with the

energy need for a single charge. The dimensions of the fuel tank are flexible

and can be used to fill up empty space. In our algorithm we have assumed

the length and width of the tank be equal to the length and the width of the

benchmarked power source. For the MP3 player this is equal to 66 and

33mm. The thickness of the fuel tank is a result from the matching volume

needed to store the required amount of energy (820mAh at 3.7V) and the

amount of water described by cmeoh, as described in Chapter 7, which is in

our case equal to ½ (the amount of water equals the amount of methanol).

The total thickness of the fuel tank is now equal to:

tan

1

meoh

batteryk

meoh E tank tank

Et

c l wρ= ⋅

⋅ ⋅ (47)

The weight of the fuel tank can be calculated with the density of water

(1 kg L-1) and methanol (0.79 kg L-1). Price is assumed to be equal to the

price for a fill-pen refill, where a five-pack of 1.45 mL cartridges costs €1.90

[179], resulting in a price-density of: €1.31x10-3 mm-3.

9.3.7 Ordering of components

A list of components is selected based on the performance requirements

imported by the user. A list of compliant components is chosen and sorted

on its performance, where the best performing version is at number one and

the worst is the latter in the list. For the pumps the ordering is based on

volume flow with the lowest value as the most interesting, and for the

intermediate accumulator the lowest value of the energy capacity satisfying

the requirements is the most interesting accumulator. Both flexible

components, the fuel cell and fuel tank designs match the performance

requirements and thus no ordering is needed.

9.4 First test run: initial algorithm To test the algorithm the requirements for a power source applied in the

Samsung YP-Z5F MP3 player are entered into the program. The optimization

181

is in first instance only based on optimization of the volume for only one

component-set, thus λC = λw = 0 and λV = 1.

9.4.1 Test run results

Based on the performance requirements the following non-flexible

components are chosen:

• Air pump: Bartels MP5

• Methanol/water pump: ThinXXS MDP2205

• Intermediate accumulator: Varta CR1/2AA

The dimensions of all non-flexible and flexible components are listed in Table

35. The total volume of all components summed individually is exceeding the

volume of the benchmark (8,720mm3), which means a minimum packing

ratio of 2.16 is feasible.

With the chosen component-set the algorithm first chooses a primary design

vector D[0,0] and 20 random generated neighboring design vectors D[0,j]

with j = 1…20 using a radius factor α = 0.1. After every run the best solution

is chosen to be the next primary design vector D[1,0]. This process has been

repeated for n = 4,800 times.

Table 35: overview of selected and designed components for the DMFC system applied in

the Samsung YP5 MP3 player.

Component Type Length(mm)

Width(mm)

Height(mm)

Volume (mm3)

Weight (g)

Price (€)

Air pump Bartels MP5 14 14 3.5 689 0.80 n.a.

Fuel pump ThinXXS MDP2205 26.2 25.4 9 5,986 3.00 n.a.

Accumulator Varta CR1/2AA 14.75 14.75 25.1 5,460 11.50 n.a.

Fuel cell flat pack Nafion 117 [120] 37.7 37.7 0.8 1,141 2.37 42.17

Fuel tank CH3OH:H2O = 1:1 66 33 2.53 5,516 4.94 7.23

Total 18,793 22.60 >49.40

In Figure 63 all best values of F(D) are plotted for every primary design vector

D[i,0] with i =0…4,800. The run took 11 hours to converge the objective

function equaling 11.87. This value is, in this case, the volume packing ratio

making a total DMFC system volume of 103,400mm3. The numbers recorded

182

in Figure 63 show the evolution of the designs produced by the algorithm. In

Figure 64 all components are named in the set of parallelepipeds, and in

Figure 65 the generated designs are plotted as a set of parallelepipeds in a

defined space.

0

5

10

15

20

25

0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000

number of iterations

Ob

ject

ive

Fu

nct

ino

F(D

)

0,1,2,3

4

5 6

7 8 9 10

Figure 63: First optimization of the objective function F(D), for 4,800 primary design

vectors D[n,0] with 20 neighboring design vectors.

Air pump

Battery

Fuel cell

Fuel pump

Fuel tank

Figure 64: Definition of all component names to the parallelepiped.

183

Figure 65: Evolution of design 0 to design 10 produced by the algorithm.

184

Evaluating the results from this first run, it shows that all flat objects are

directed in the longitudinal direction for minimized volume. The algorithm

guides the overall volume towards a flat design. To improve calculation time

the direction of the components is set to zero and only the α rt is randomly

changed. A new evaluation run is started up with the lat design from the

previous run as starting point. In this following optimization run the

neighboring design-vector is redefined as follows:

( )( )[ , ] [ ,0] ,0,0tj

D i j D i rα= +

for j =1…m (48)

In Figure 66 the follow-up optimization shows an improvement in the

minimizing objective function from 11.87 to 7.91. The second part of this

run took 4 hours on a Dell PC (Intel Core 2Duo CPU E8400@3Ghz, 1.97GHz,

3.25GB RAM) before converging.

0

5

10

15

20

25

0 200 400 600 800 1000 1200 1400 1600


Ob

ject

ive

Fu

nct

ion

F(D

)

4

0

5,6

9

12…14

15

18

2326

27

Figure 66: Second part of the optimization of the objective function F(D), for an extra

1,600 primary design vectors D[n,0] with 20 neighboring design vectors.

185

Figure 67: Evolution of design 3, 6, 9, 12, 15, 18, 21, 24, and 27 produced by the

algorithm when only changing the radius vector. The design procedure started from the results of the first test run.

186

9.4.2 Evaluation

The results from the first test run show convergence towards a smaller total

volume. The minimizing function is improved to 7.91. The lowest value of the

objective function feasible with this component set is 2.16. The improvement

steps show a slow convergence to a high end-value of the objective function,

not nearly nearing this minimum value feasible. Three problems arise from

the first test run:

1. Observing the evolution of the structural variants in Figure 67, it

can be noted that the fuel-cell (largest squared volume) does not

place itself besides the fuel tank (largest rectangular volume).

2. The lowest achievable end-value of the objective function is very high

for this component set. This value has to be equal or lower than 1

before it will be a feasible solution to the problem.

3. The convergence takes up a long time (15 hours for a test-run of

128,000 iterations). Besides good solutions the algorithm produces a

large amount of solutions which do not contribute to finding the

optimum.

Problem 1 is investigated by conducting a parametric analysis where the

radius vector α rt is varied. Increasing the α stepwise from 0.2 to 1.0 will

increase the search field for neighboring design vectors as defined in Figure

60. This increase of the design field did unfortunately not result in an

improved solution.

Problem 2 is about ordering of components. To improve the test for feasibility

the ordering of the components will be based on the objective function F(D),

instead of the performance parameters as proposed in Section 9.3.7. In our

case the smallest component will be chosen to be the best option. A welcome

circumstance is that this adjustment in the algorithm makes the primary

loop, of evaluating multiple component-sets, redundant (see Figure 59). In

Section 9.5 the algorithm is changed and evaluated with the improvements

as proposed above.

187

In the second part of the first test run the number of design variables were

decreased to only varying the translation vector, as described by Equation 48.

To decrease time to converge even more an evolutionary algorithm is

proposed which guides itself to an optimized solution (Section 9.6).

9.5 Second test run: improved algorithm As described in the evaluation of Section 9.4 the component selection is

based on F(D) instead of performance characteristics. A second test run is

conducted with m=20 neighboring design vectors with a radius factor α=0.2.

This process has been repeated for n=1,600 times.


Based on the objective function F(D) the following non-flexible components

are selected from the available components:

• Air pump: Bartels MP5

• Methanol/water pump: Bartels MP5

• Intermediate accumulator: Maxell ML2016T25 button cell

The dimensions of all non-flexible and flexible components are listed in Table

36. The total volume of all components summed individually is in this case

almost equal to that of the benchmark (8,720mm3), which means a

minimum packing ratio of 1 is feasible.

Table 36: overview of selected and designed components for the DMFC system applied in the Samsung YP5 MP3 player.

Component Type Length

(mm)

Width

(mm)

Height

(mm)

Volume

(mm3)

Weight

(g)

Price

(€)

Air pump Bartels MP5 14 14 3.5 686 0.80 n.a.

Fuel pump Bartels MP5 14 14 3.5 686 0.80 n.a.

Accumulator Maxell ML2016T25 20 20 1.6 640 1.80 3.75

Fuel cell flat packNafion 117 [120] 37.7 37.7 0.8 1,137 2.37 42.17

Fuel tank CH3OH:H2O = 1:1 66 33 2.53 5,510 4.94 7.23

Total 8,659 10.71 >53.15

In Figure 68 the evolution of the primary design vectors D[i,0] are plotted. It

shows the algorithm converges quickly to a low value of the objective

188

function from 31.05 to afinal 2.98. In Figure 69 all best values of F(D) are

plotted for every primary design vectors D[i,0] with i = 0…1,600. The run

took 2 hours and 45 minutes.

0

5

10

15

20

25

30

35

1 201 401 601 801 1001 1201 1401 160


Ob

jec

tiv

e F

un

cti

on

F(D

)

Figure 68: The values of the objective function for 1,600 primary vectors generated with

the second test run.

9.5.2 Evaluation

The algorithm converges to a value for the objective function equal to 2.98,

where the minimum attainable value would be 1. The plots in Figure 69

show that, again, the algorithm produces flat compositions of the objects.

Thus a flat architecture results in the lowest value of the objective function.

Guiding the objects at the start of the optimization will result in a decrease

in calculation time because only compositions inherently leading to an

optimized objective function are evaluated during the search and others are

not. It is proposed to use an algorithm to direct the objects based on initial

form, where all objects are typed as a cuboid, beam, flat square or flat

rectangular object. First the base-object is chosen which is leading for the

longitudinal direction of all other objects. The objects used in the evaluation

are, with the exception of the fuel tank, all flat square objects.

189

Figure 69: Evolution of the second test run. The algorithm produced 11 succeeding designs: 1, 2, 3, 4, 5, 38, 107, 271, 457, 701 and 840.

190

9.6 Third test run: applying the evolutionary algorithm

To increase time to converging an evolutionary algorithm approach is

proposed. For this thesis an extension to the random walk algorithm is

proposed, which takes into account the gradient sign of the objective

function for every value in the design vector D. This method guides the

design vector more quickly to a (local) optimized solution. Figure 70 is a copy

of Figure 58 showing the new algorithm in its context of optimization

algorithms.

Gradient descent

Evolution strategy

Increasing numerical evaluations

Random Deterministic

Grid based

Random walk

Increasing algorithm complexity

Grid

Increasing number of (local) optimums

Heuristic

Figure 70: Breakdown of types of optimization algorithms.

The newly developed program made in Mathematica is presented in

Appendix B.2. The following procedure is followed:

1. Components are selected based on the objective function.

2. The direction of the components is based on its intrinsic form and

fixed at start.

3. The largest component is set as the base at start.

4. The primary design vector D[0,0] is evaluated and a random cloud of

neighboring design vectors D[0,j] is generated.

191

5. The fitness of every neighboring design vector is calculated by means

of the objective function Fj(D) plus a penalty function when

intersection occurs p(D).

6. The “gravitational” mean of the cloud is calculated by:

( )

( )

( )

( )

1 1 ( ) ( )

1 ( )

20 20

0 0

20 20

0 0

, ,obj t obj t

obj t

t j t j

j jcloud

j j

j j

r F D r F D

rF D F D

= =

= =

⋅ ⋅ =

for obj(t) =1…5 objects

(49)

Where: j = 0 is the fittest solution from the previous iteration and

j=1…20 are the solutions belonging to the new generated

cloud.

7. The cloud vector is subtracted from the primary translation vector rt ,

consisting of all translation vectors of all objects, and the sign of

every object in the design vector can be determined (Figure 71).

8. A new primary design vector is generated with the new translation

vector rt equal to:

[ ] [ ] ( ) 01,0 ,0t t t cloudr i r i Sign r r r+ = − − (50)

Where: 0r Rα= a randomly generated length within the search field

limited by the maximum radius R.

9. Repeat until the maximum number of generations is reached or

convergence is reached.

192

x

y

D[i,0]

D[i,j] Directed search field

Figure 71: A 2D representation of the primary design vector D[i,0] for one object and the

generated neighboring design vectors D[i,j].


The component selection is again based on the objective function and listed

in Table 36. The algorithm is tested by conducting several runs, where the

starting translation vector is changed. Three of five runs are plotted in Figure

72, where all best values of F(D) are plotted for every primary design vectors

D[i,0] with i = 0…300-500 and a cloud of n = 20 neighboring design

vectors. Figure 73 shows the design at start and finish of the Run 1 to 3. The

300 to 500 iteration runs take up 15 to 30 minutes on a Dell PC (Pentium

Intel Core 2Duo CPU E8400@3Ghz, 1.97GHz, 3.25GB RAM).

The algorithm is evaluated several times and the objective functions value is

converging generally nearing 2 for run 1 and run 2. Run 3 converges to a

higher value nearing 3. To test if this starting point will converge to the

minimum value in new evaluations, this run is repeated. In Run 4 the

number of evaluations n is increased from 500 to 1,000 (Figure 74 a). In Run

5 the number of neighboring design vectors m is increased from 20 to 40

(Figure 74 b). The convergence of these test are also shown in Figure 72, and

do not see a convergence nearing 2.

193

0

1

2

3

4

5

6

7

8

9

10

1 201 401 601

Number of iterations

Ob

jec

tiv

e F

un

cti

on

F(D

)

Run 1 (n=300, m=20)

Run 2 (n=500, m=20)

Run 3 (n=500, m=20)

Run 4 (n=1000, m=20)

Run 5 (n=500, m=40)

Figure 72: Optimization of the objective function F(D), for n primary design vectors

D[n,0] and m=20-40 neighboring design vectors.

194

a) Run 1 where n=300 evaluations and m=20, starting with F(D)=13.51 and finishing

at 2.28 (16 minutes).

b) Run 2 where n= 500 evaluations and m=20, starting with F(D)=17.56 and finishing


c) Run 3 where n=500 evaluations and m=20, starting with F(D)=25.61 and finishing


Figure 73: Convergence of the algorithm for three algorithm evaluations with thee different starting points (a, b and c), as shown at the left side. The results after the

optimization are shown on the right side.

195

a) Run 3 again, where n=1,000 evaluations with m=20, starting with F(D)=25.61 and

finishing at 2.84.

b) Run 3 again, where n=500 evaluations with m=40, starting with F(D)=25.61 and

finishing at 3.16.

Figure 74: Evaluation of Run 3 but with a) an increase of the number of iterations, n=500 1,000 (at m=20) and an increased number of points in the cloud from m= 20

100 (n=500).

196

9.6.2 Evaluation

The convergence of the new evolutionary algorithm is improved in speed but

also in the minimizing the value of the objective function. The algorithm

produces designs with a higher packing ratio than the algorithm proposed in

the Sections 9.4 and 9.5. The test is evaluated three times with different

starting points. This test shows that the starting point influences the end-

result, even after increasing the number of evaluations or the number of

neighboring design vectors. Thus random generated start-points can

converge to different local minima, which is not always the best case. To test

if the DMFC system is a good alternative to the benchmarked battery, the

optimization run has to be repeated several times.

Viewing the end results in Figure 73 and Figure 74 shows that the form of

the fuel cell, which is the large squared form, is inefficiently chosen. If the

fuel-cell has the form of a rectangular object, the packing ratio will decrease.

The automated design procedure of this flexible object has to be redefined.

9.7 Discussion on the results In the following sections the optimization algorithm, including the

component selection, and the results from these optimizations are discussed

based on costs, weight and volume.

9.7.1 Algorithm

In the case study presented in this chapter multiple structural variants are

generated for DMFC power systems powering the Samsung YP-Z5F MP3-

player. The structural variants are evaluated and optimized for the minimum

value of F(D). Three properties for optimization are introduced: volume,

weight and costs, which are reduced to corresponding metrics M. To test the

working of the algorithm the preference factor λ for volume is set to 1 and

the rest (weight and costs) are set to 0.

In Section 9.4 an initial test run was executed. Based on these results the

initial component-selection method has changed. Instead of selection based

on performance specifications the selection of the component-set is based on

197

the objective function. The new algorithm with an improved selection method

is evaluated in Section 9.5 and results are presented. The optimizations were

all performed on a Dell PC (Intel Core 2 Duo CPU E8400@3Ghz, 1.97GHz,

3.25GB RAM), to compare the time-to-converge for the algorithms with each

other. The optimization method used in Sections 9.4 and 9.5 are both very

time-consuming and a new approach is proposed, making use of an

evolutionary algorithm, presented in Section 9.6. The evolutionary algorithm

is less time-consuming (from 8 hours to 15 minutes) and a lower number of

calculations is needed before the algorithm converges to an improved

objective function F(D). A factor 5 to 10 lower number of calculations are

used. Besides improved time-to-convergence the evolutionary algorithm

produces a design with a lower value of the objective function (from 3 to

almost 2). The algorithm starts with a randomly generated starting point,

which not always results in the smallest value of the objective function. It is

thus proposed to do several runs to test if the DMFC system is a good

alternative to the benchmarked battery

The algorithm is useable when the load-curve is available. During the

conceptual design process this is not always the case, and the designer

should make a guess about the power over time for one cycle and over one

charge. A small change in the load profile could mean a great difference in

the selected component set, making the use of accurate load-curves a

necessity when the designer wants to test the feasibility of a DMFC system.

This makes the algorithm interesting for products already developed, where

the load profile is known, and less for new to develop products, where the

load-profile is based data from similar products or a previous version of the

device.

9.7.2 Costs

In the last column of Table 35 and Table 36 the prices for all components are

listed individually. The prices for non-flexible components are not always

available, which makes it difficult to draw conclusions on the feasibility of

the system compared to benchmarked battery. One thing can be noticed is

the price for the fuel cell flat pack. The price is based on 100cm2 MEA [132]

and is in this case a large contributor to the total price of the system.

198

Compared to the price of the benchmarked battery, the price for only the

MEA is almost 84% of the total price of the battery (€50).

According to Darnell [130] the MEA accounts for 40% of the total DMFC

costs and 60% can be carried back to assembly and other components.

Using this number the total price for the DMFC system would than be

€105.43, making the initial price for the DMFC system a factor two higher

than the lithium-ion battery used, at same specifications. This makes the

DMFC system, at the moment, economically not competing with the

benchmark.

Besides the high price for the MEA and thus the fuel cell, prices for different

components are not always available and liable for change. Because of the

lack of price data for non-flexible components, the optimization with costs as

part of the optimization function is not feasible within the algorithm at the

moment. The component database should be extended with more optional

components, and with more data, specifically on price.

9.7.3 Weight

In Figure 75 the weight break down is plotted for the DMFC system. The fuel

tank is the heaviest component of the system. The total weight of the

component-set selected is 10.7g making this system half the weight of the

benchmark (21g), and thus in this respect very feasible. It must be taken

into account that weight for wiring and plumbing is not yet added to the

weight, which will result in a small increase of the total system weight.

199

Air pump7%

Fuel Pump7%

Accumulator17%

Fuel Cell flat pack22%

Fuel tank47%

Figure 75: Weight breakdown of the DMFC system.

9.7.4 Volume

The algorithm is used to optimize merely for volume. The results should

present a final solution with a low amount of empty space, and a high

packing ratio. The final design after the third run shows that this is not the

case with a packing ratio nearing 2. The empty space in the generated DMFC

system is more than 50% of the total volume. The design of the fuel-cell

greatly influences this packing ratio. If the automated design of this flexible

component, described in Section 9.3.5, takes the maximum width into

account the packing ratio could decrease.

When comparing the total volume of only the components (8,659mm3) with

the volume of the benchmarked battery (8,720mm3), it shows the selected

component-set can lead to solution nearing that of the benchmark. If we look

at the volume breakdown in Figure 76 of all components than we notice that

the fuel tank takes up the most space. It must be noted that the tanks

consists of a water and a methanol tank (50/50). If a 100% methanol tank is

feasible than the value of this component will be halved. In the algorithm the

interconnections like wiring and tubing is not taken into account, thus extra

volume will be added.

200

Air pump8%

Fuel Pump8%

Accumulator7%

Fuel Cell flat pack13%Fuel tank

64%

Figure 76: Volume breakdown of the DMFC system.

The DMFC system is only feasible with the chosen component set when

empty space equals zero and no extra components or interconnections have

to be added. An existing DMFC systems like the SFC Jenny [142] has an

empty-space ratio of 15%. This product is optimized for volume and the

amount of empty space is minimized to the maximum. Adding 15% to the

current chosen component-set volume will result in a minimum feasible

packing ratio of 1.15. Concluding from this the volume is a bottleneck in the

current selected component-set, and thus will be a bottleneck for the

feasibility of DMFC systems as an alternative for the lithium-ion battery

when applied in the MP3 player case.

Note that costs and weight are not taken into account in the optimization

algorithm, thus it is unknown if the selected components were selected as

the primary component set when those factors were included. The algorithm

is a good tool supporting the designer to place the components in the most

opportune way, minimizing systems’ bounding-box volume.

At the moment the algorithm makes use of only commercially available

components. Besides the limited amount of alternatives, often candidates are

over-dimensioned. The list should be updated and scalability of components

should be an option added to the algorithm. When scalable components are

201

introduced in the algorithm, it will produce designs which are feasible on the

long term.

9.8 Conclusions In this Chapter an algorithm is proposed, and implemented in a

Mathematica program, which consists of three parts:

1. Selecting of non-flexible components from a database and designing

of flexible components. The selection and design are based on

performance input in the form of a single-cycle load-profile and the

performance characteristics of the current power source used in the

portable electronic device. The different feasible components are

combined in feasible component sets.

2. One of the component sets is selected and used to generate multiple

structural variants. The structural variants are evaluated with the

objective function F(D), in our case only for volume and optimized by

minimizing this value.

3. Presentation of the results in the form of 3D layouts, tables

consisting of the id of the design, value for the objective function and

the design vector.

During part 1 the feasible non-flexible components are selected and multiple

component sets are produced. The first component-set evaluated by the first

test run is based on the best performing components which satisfy the

minimum requirements. The ordering of the list of feasible components

should not be based on performance but on the objective function. In the

second test run a new component-set is selected based on the objective

function and tested. The selection based on the objective function makes the

primary loop redundant. This results in the overall run-time of the program,

because only one component-set is evaluated instead of multiple.

In part 2 several structural variants are produced, which are evaluated with

the objective function. To test the program the optimization is only based on

volume, and not on all three defined basic properties. The results from

second test run show that the algorithm converges to a value 2.98 after

1,600 iterations (~8 hours). The time-to-convergence is very high in time and

202

in number of iterations (1,600x20 design vectors have been evaluated). To

increase convergence and time needed to achieve the minimum value of the

objective function, an evolutionary algorithm is proposed and tested in the

third test run. The third test run shows a convergence to a lower value of the

objective function than test run two (nearing 2), in a much lower number of

iterations (factor 5 to 10). The convergence depends strongly on the starting

point of the optimization, and thus several starting points have to be

evaluated to be sure it will converge to the lowest value of several local

minima.

The results from the second and third test run show good convergence of the

algorithm and result in a design proposal for placement of the different

components in space. Viewing those results it can be concluded that the

feasibility of the DMFC system strongly depends on the components volume

and costs. The total weight of the summed components, in de component-set,

is 50% of the weight of the benchmarked battery, and thus no issue for the

feasibility. The third test run, using the evolutionary algorithm, converges to

the lowest value of the objective function, in this case volume, nearing 2.

This means the DMFC system will be twice as voluminous as the

benchmarked battery. This can be decreased by a change in the design of

the fuel-cell from a squared to rectangular form.

When only the component volumes of all components’ individually are

summed together the minimum value of the objective function is 0.99. With

the selected component-set it is thus theoretically feasible to produce a

DMFC system equaling the volume of the benchmarked battery. Because we

did not take other components into account, like wiring, tubing and the

mixing chamber, the total volume will be higher than 0.99, and thus lead to

a more voluminous design.

Besides components, the system will in practice always have a certain

amount of empty space. For the optimized SFC Jenny the empty space is

15% of total volume. If this number is used it can be concluded that DMFC

systems in the MP3 player case, using only commercially available

components, is not feasible at the moment because of volume. The

203

improvement in weight is apparent and will not be an issue for the feasibility

of DMFC systems in the MP3 player case.

The feasibility of the third property, price, is difficult to test. For non-flexible

components prices are not always available, and a small amount of different

options are available. The prices for the components are based on

commercially available sale prices when 100 pieces are bought. The price of

the MEA is also based on sale prices from the Fuelcellstore [132], and the

MEA only will cost €42.17. This price for the MEA alone is 84% of the total

price for the benchmarked battery. The total sale price of the component set

selected by the algorithm is higher than €53.15. An estimate based on the

figures given by Darnell [130] result in a total price for the DMFC system

equaling €105.43. At same specifications this price is twice the price of the

lithium-ion battery used, making the DMFC system not a commercially

attractive alternative.

9.9 Recommendations The amount of commercially available components which can be used in 0 to

100W DMFC systems is low. Smaller components matching low-power

DMFC systems have to be developed. To address this lack of components the

algorithm could be introduced with scalable non-flexible components. These

components are based on existing components but scaled up or down to

match the required performance specifications. Scalable components give

the designer an indication of the long-term feasibility and the constraints of

having to develop new components.

The algorithm can evaluate multiple structural variants with the objective

function as optimization variable. The use of tables and computerized

producing structural variants is a clear way of making quick conceptual

feasibility tests. In our case the feasibility of the DMFC system is tested as

an alternative to the benchmarked rechargeable battery. The tables for non-

flexible components are easy to update to current available components. The

algorithm uses only the three basic properties for optimization, and

extending the number of properties is proposed for the third-order model.

Furthermore the algorithm evaluates only structural variants where

components are represented by simple parallelepipeds. This representation

204

does not follow the form of the actual component very accurate. To improve

the accuracy of the model the objects should be represented by other forms

like cylinders, but also by combination of forms. Interconnections are not

taken into account in the second-order model and the third-order model

should include these in the algorithm. A fourth order model can also include

the influence of heat, fluid flow and other multi-physical parameters.

205

10 Conclusions

This chapter discusses the initial and main Research Questions and gives

recommendations for further research. First the main question will be

answered in Section 10.1. The technological and scientific relevance of the

research, and thus the thesis, will be discussed in Section 10.2, and

recommendations are given in Section 10.4.

10.1 Answer to the main question

10.1.1 Initial Research Question

This thesis started out by asking the following initial Research Question:

“Which power systems can compete with the commonly used rechargeable

lithium based battery in the application field of portable electronic devices,

and how can a systematic approach help product designers select

appropriate power systems during the preliminary design phase?”

The first part of this question is answered by doing a thorough literature

search. First a portable electronic device was defined as a ‘device which uses

electric energy, with an integrated power source and carriable in once bag.

Size is limited to a large laptop computer (6.2kg and 4.5liters), and the

device should be intended to be carried around’. Power ranges from 0.01W to

100W systems with energy specifications in between 0.1 and 200Wh. A large

amount of data was gathered about the physical, economical and

technological properties for a large variety of power systems applied within

the defined application field. Within this search a power system is defined as

a power generator in combination with an energy carrier. The data gathered

was thus broken up in two parts and compared to commonly used

rechargeable batteries on three properties: power, energy and costs. This

206

exploratory research concluded by pinpointing two power systems which

could compete and even outperform the lithium based rechargeable battery,

namely (i) the internal combustion engine (ICE) in combination with a

carbon-based fuel, and (ii) the fuel cell system fueled with hydrogen or

carbon based fuels. The first combination consists of a power-dense

generator with a high energy-dense fuel. Biggest issues with this power

system are its high noise and toxics output, thus makes this system

interesting when applied in outside-the-home environments. The second

power-system of interest is the low-temperature fuel cell systems, the direct

methanol fuel cell (DMFC) or the polymer exchange membrane fuel cell (PEM

FC) system. The PEM FC system makes use of hydrogen as a fuel, stored in

high pressure vessels of up to 600bars, what makes this fuel is less

attractive for application in portable electronic devices. Methanol on the

other hand has a high energy density. The potential of this fuel is storing two

to seven times the amount of energy as contained in current lithium-ion

rechargeable batteries. Within this thesis the research focused will be on

fuel-cell systems fueled with methanol.

The comparison described above is based on existing power systems and

general numbers available at the time. To make a fair comparison between

the different power systems, the systems ought to be designed in different

applications. The second part of the initial Research Question focuses on the

systematic approach of selecting or designing power systems. The literature

review in the field of tools and methods for selecting or sizing a power system

shows that there is a limited amount of tools available. During the

specification phase, and even before, indentifying tools like Powerquest, CES

method and Ragone plots can be used. T-max evaluates the consequences

for the power system when specifications are changed, and is especially

useful when conceptualizing a product. System modeling is generally used

during the concept/embodiment phase, where different concepts are

evaluated and choices are made based on different objectives. Finally

enumerative tools, like POWER, are found which can be used to evaluate

different concepts using single, hybrid or more batteries. These tools are

generally used during the embodiment phase where the devices specification

is already laid out.

207

The tools to be used during the specification phase of the design process

merely identify the opportunities of alternative power systems. It is

recommended to improve the tools used during these phases of the design

process by presenting the feasibility instead of only the visibility.

10.1.2 Main Research Question

Based on the conclusions from the preliminary research the main Research

Question is defined as followed:



design process?”

To answer this question five sub-questions have been formulated which will

be answered underneath. The answers to these questions are given in

Chapters 5 to 9. The first four Research Questions are answered in Chapters

5, 6, 7.1 and 7.2 respectively (Figure 77). Research Question 5 is answered

within Chapters 8 and 9 using an explorative design research method.

RQ

4 R

Q3

RQ

2 R

Q5

RQ

1 5. DFMC as an alternative for lithium ion batteries

6. The user and DFMC systems

7.1 Orders or modeling

8. First order model, a heuristic approach

9. Second order model, database driven parametric d i

7.2 Differentiating properties

Figure 77: Systematic approach to answering the five sub-questions.

208

RQ1: Which technological and economical properties differentiate the

DMFC power system from the lithium-based rechargeable battery in the

field of portable electronic devices? To answer the first Research Question

a comparison is made between the two power systems in Chapter 5. Both

power systems are introduced by explaining the technological details, based

on theoretical but also practical values. For portable electronic devices the

power system takes up 20 to 25% of its volume and 10 to 25% of its weight.

This makes the power source an important factor giving form to the device.

The DMFC is compared to the lithium ion battery with a zero-order model,

which estimates volume and weight specifications of the power system based

on power and energy densities of the three main contributors, the fuel cell,

the tank and the bill of products (BOP). The model is based on a case

prototype from literature [29, 125], using a stand-alone fuel cell that follows

the load profile of the application, a cell-phone recharger. No intermediate

battery was used to level peak powers. With this model a Ragone plot is

produced which shows the main physical differences between the DMFC and

one of the best performing lithium-ion batteries at the time [126]. Compared

to the lithium-ion battery the DMFC is a power system with a high energy

and a low power density. In general the DMFC outperforms the battery in

application areas that need high energy and low peak power. Based on the

Ragone plot the DMFC could be a factor 2 to 3 smaller and six times more

lightweight. The potential improvement in volume is not as large as that of

weight and thus it can be postulated that volume is the main bottleneck for

DMFC systems applied in portable electronics, if volume specifics improve,

this will always have a positive effect on weight.

The zero-order model is used to make rough estimates of the physical

properties of the power systems when applied in different electronic devices.

The physical improvements are mainly found in applications with low-power

requirements and high-energy demands, like sensors, smoke detectors and

flash-drive MP3 players with no display. In the application field of laptop

computers, cell phones and displayed devices, only the combination of a

DMFC and battery, a DMFC hybrid or DMFC system, is feasible.

209

Besides physical properties the costs and life-cycle costs are discussed in

Chapter 5. Based on the result presented in this chapter the initial costs of

the DMFC power system should be cut back with at least a factor five before

it becomes as cost-effective as the lithium-ion battery. The life-cycle costs

(LCC) are the main bottleneck in terms of economical feasibility, and the

price of methanol has to be cut down with a factor 4 before its becomes

competitive with grid-price, and thus with rechargeable lithium-ion batteries.

It is common that the consumer is willing to pay more for better performing

power systems, which is the case in the transition from nickel-metal hydride

batteries to lithium-ion based rechargeables. The initial price of lithium ion

batteries is higher compared to the NiMH battery, but the time-in-between-

charges is improved greatly. Besides the LCC the improvement in time-in-

between charges/replacements is calculated. The DMFC power system has

to be replaced 2.7 times less than an alkaline battery, and 3.9 times less

than the lithium-ion battery has to be recharged. The endurance

improvement for the DMFC system is evident and might be an argument for

the consumer to make the switch to DMFC systems, even at a higher price.

RQ2: Which demands with respect to DMFC power systems in portable

electronic devices are most significant from a user point of view, and

how can we make use of this in a design method? To test the preference

of future consumers on their willingness to buy a fuel cell powered cell

phone or laptop computer a conjoint analysis is executed. Five differentiating

properties are used for this studies, weight, volume, purchase price, charge

or cartridge (at extra costs), and time of use. From these five differentiating

properties volume is for both products the most important property of the

product. Weight is of the lowest issue for the participants. The differences

are larger for the cell phone than for the laptop computer. For the cell phone,

the participants had a strong preference for quick charging over an instant

recharge by means of a cartridge. For the laptop computer this was vice

versa. In general it can be stated that cartridges, and thus fuel-cell power

systems, are more appreciated when use-time is an issue, when the

cartridge can provide a longer time-in-between-charges. Use-time is an issue

for devices with runtimes of around 3 hours or less. The preferences of the

user for the different properties are quantified and are to be used in following

preference-based design models.

210

RQ3: How can we identify the opportunities of DMFC power systems for

portable electronic devices, during early phases of the design process?

Different orders of modeling are proposed which range from simple rule-of-

thumbs (zero-order model) to enumerative algorithms (second and third-

order model), ranging from low to high accuracy. The opportunity for a

DMFC power system can be identified, with the zero and first order model.

These models are based on heuristics, and the accuracy of both models is

low. Therefore a second-order model is proposed which is a table-driven

preference-based model using a computer to design and evaluate multiple

structural variants. The model optimizes the process using an objective

function based on the main differentiating properties described in the

previous paragraph.

RQ4: What are the basic properties that help the designer to model a

DMFC power system from a techno-economical perspective? To define

the differentiating properties to compare the DMFC power system with other

power systems the list of considerations influencing the selection of a battery

is evaluated. Three properties, volume, weight and initial costs, are defined

as the three main differentiating properties. These properties are used in the

zero, first and second-order model, for optimizing systems’ specifications.

The three properties give sufficient information on the technological and

economical feasibility of a DMFC power system during the early phase of the

design process. More properties can be introduced in a third-order model,

making the model more advanced and accurate but also more complex,

needing more input data from the tables and from the product designer. The

third-order model is thus proposed to be used in a later phase of the design

process, e.g. the embodiment phase, where engineering of the system is

more important.

RQ5: Which algorithms can be formulated to give the product designer

not only insight in the opportunity of DMFC power systems, but also

help the designer in an appropriate systematic approach? In Chapter 7

four orders of modeling are proposed, the zero to third-order model. The

models all have there own function during the design process. Both the zero

and first-order model can be used to identify opportunities or make a DMFC

system more visible for the designer during the specification and

211

conceptualization phase. The first-order model is an optimized version of the

zero-order model, where the DMFC system is broken down in five parts

instead of three: the fuel cell, its tank, BOP, empty space and the

intermediate battery. The zero-order model is adapted to the first-order

model by means of heuristics obtained by designing a DMFC system for a

MP3 player. The model consists of simple equations which can be used as

rule-of-thumb during preliminary design phases. The input for the model is

minimized to only mean power, peak power and energy requirements for the

specific application. Only volume is worked out within this thesis, because

this is the main driving force in DMFC system design for portable electronic

devices and can be evaluated properly by means of CAD modeling. The

model was validated with two commercially available DMFC systems,

produced by SFC energy, the Jenny and Efoy 2200 fuel cell system [141].

The Jenny fuel cell system is designed and engineered with volume

compactness in mind, and this fuel-cell system is represented by the best-

case estimates from the first-order model (left side of the estimates). The Efoy

2200 fuel cell system is a conventionally designed and engineered product

and finds itself in the middle of the approximations. Based on these two

validations the model estimates the volume of the DMFC system for highly

packed systems (left side estimates), via conventional designs (middle) to bad

designs (right side). The estimate volume of the fuel-tank is for both cases to

high. Either system efficiency, BOP efficiency or the canisters volume is

taken to conservative. More validation tests are needed to fine-tune the

defined coefficients, densities, constants and efficiencies, and make the

model more accurate. The first-order model can be used during the

preliminary phases of the design process and needs a low amount of input

information (Figure 78).

A second-order model is proposed to get a more accurate estimate of the

total volume, but also to get the physical dimensions of the power system.

The second-order model makes use of enumerative methods done on a

computer. Different structural variants are evaluated by means of a

minimizing objective function F(D). The input for the algorithm needs more

information than just the systems’ energy, and mean and peak power, used

in the zero and first-order model. It also needs the load-profile and the

212

specifications of the benchmarked power system, in our case study the

lithium polymer battery.

nth order model

concurrent design rooms

[several days - weeks]

0 to 1st order model

pen & paper

[1 - 2 minutes]

1st to 2nd order model

calculator & spreadsheet

[5 - 10 minutes]

2nd to 3th order model

desktop computer

[1 hour - 1 day]

Figure 78: Tools to be used during the design engineering process.

Based on the inputted specifications of the benchmarked battery and the

load profile of the device, the algorithm selects commercially available

components from a database and designs custom made components like the

fuel cell and fuel tank. A set of components is chosen which satisfies the

requirements of the device and which result in the minimum value of the

objective function. Multiple structural variants are produced and evaluated

on the preference-based objective function. The objective function is

minimized by means of an evolutionary algorithm, and convergence to a local

minimum is reached after a low number of iteration steps (500 iterations, 15

to 30 minutes). The convergence depends strongly on the starting point of

the optimization, and thus several starting points have to be evaluated

assure convergence is reached to the lowest value of several local minima.

The algorithm is tested on the case-study of the DMFC power system for the

MP3 player. The algorithm converges to a volume which is twice as large as

the benchmarked battery. Weight is halved compared to the benchmark and

the sales price is probably twice as high. Based on these results the DMFC

system is, when applied in the case-study assessed, not feasible because of

volume problems. Other products have to be evaluated before a verdict can

be given about DMFC systems over the whole range of portable electronic

devices.

213

The improvement of the second-order model over the first-order model is its

accuracy, and introduction of size besides volume. A third-order model is

proposed where the component set is extended with other components like

the mixer and sensors, but also with interconnections like wiring and

plumbing. Furthermore the components in the second order model are

represented by simple parallelepipeds. To improve accuracy of the

automated design the objects should be represented by other forms like

cylinders, but also by combination of forms. This type of models is

specifically intended to be used in concurrent-design environments, where

the design engineer has to make quick evaluations when specifications or

product requirements change (Figure 78).

10.2 Scientific and technological relevance This thesis covers research in the field of design engineering. One of the

focuses of this thesis is how to transfer the scientific knowledge to design. In

this thesis this is done by exploring the field of alternative power systems for

portable electronic devices, and making these alternatives visible for the

designer. The scientific relevance aims at the methods used and the

technological relevance aims on the practical applicability of the methods

used and results presented. In order to transfer knowledge from theory to

practice a combination of reviewing scientific literature, user testing,

mathematical modeling and research design has been used.

10.2.1 Part I: explorative research

In the first chapters of this thesis an explorative research has been

conducted to power systems. By acquiring large amounts of data on physical,

economical and technological properties of different power generators and

energy carriers, a state-of-the-art and even a historical overview is generated.

This overview proved to be a good starting point to compare power systems

with each other and find differentiating properties. It must be noted that

data acquired is temporarily useable because of fast improvements in the

field. Furthermore for certain technologies, like the DMFC, a low number of

data point are found and presented, which can result in a misrepresented

picture. In order to optimize the reliability of the study all data used in

Chapter 2 were verified by peers by presenting interim results of the study at

214

the International Power Sources Symposium (2004) and by publishing in the

Journal of Power Sources in 2006.

10.2.2 Part II: Differentiating properties

Besides data acquisition this thesis also introduces the consumer as part of

the research and design process in the second part of the thesis. For product

developers the human factors are often difficult to quantify and often limited

to ergonomic data. The method of conjoint analysis (presented in Chapter 6)

showed that this method is a good tool to present the order of preference for

different properties which are not equally dimensioned. This method can

broaden the mind of product and engineering designers not only to focus on

technological optimization but also use preferences of the user. It must be

noted that the number of participants was low (n=21) and thus the results

from this studies can not be generalized for the whole user group. The

results are merely seen as indicative.

10.2.3 Part III: Modeling of DMFC power systems

In the third part of this thesis a research design method has been used to

explore the feasibility of DMFC power systems in portable products. To

validate the models all models are presented at different conferences and

discussed with scientists and developers in the field. The first-order model is

validated by two existing case-studies, which showed a fit with the estimated

values.

To improve this model the second-order model is proposed. This model is

intrinsically validated, because it uses existing components of which

dimensions, weight and other properties are all found in original

specification sheets of the manufacturer. The input data from the tables is

limited because all components are represented by simple forms

(parallelepipeds). This makes this model easy to update, but the results do

not show an accurate result. Extended models, like the third order model,

have to include more physical specifications like connection points and

different sub-sizes but also other constraining properties like temperature

requirements, Voltage, safety and reliability, environmental conditions,

maintenance, etc. This makes the third-order model more accurate, but also

215

labor-intensive in acquiring the data and will increase maintenance for the

developer of the tool.

The method of computerized generation of structural variants and evaluation

of the variants by an objective function is not commonly used in the field of

consumer products design. The proposed approach is relevant for the future

product designers who want to make use of more advanced computer tools

during concept design. In general computers, and automated design during

the concept phase, can contribute to shorten the heuresis phase, and thus

the design cycle. It introduces possible ways to a “first time right” design

process. Other computerized tools could be developed to help the product

designer in making choices during the concept phase of design, and opens a

new direction in product development research.

10.3 Generalization of the models In this thesis the feasibility of DMFC power systems in portable electronic

devices is researched. The research methods used in this study encompass

data acquisition and data mining, desktop literature research, case-study

reviews, user tests, and design research, using heuristic and numerical

modeling. To test the feasibility four models have been proposed, of which

three have been presented. The preliminary model of the zero-th order and

the first-order model are models that can be used during the preliminary

identification phase of the design process. The results give a first glance on

the feasibility of this power system. The model is based on energy and power

densities of main parts of the power system, and thus generally applicable to

other power generators and energy carriers.

For more accuracy and getting more insight in the size of the fuel-cell system,

the second-order model is introduced, making use of the strengths of the

computer to automatically generate a large amount of structural variants,

using the components offered by the tables. Besides generating the

structural variants the second-order model also evaluates every variant. This

type of modeling can be generalized and used by product designers letting

the computer take over the dull part of generating and evaluating structural

variants of a wide range of products. Selection of components and logical

choices can be made by the computer, following a preference based objective

216

function. The designer is in this way forced to think about this the objectives

of its power system in early phases of the design process.

The second and third order models are based on knowledge acquired by

specifications sheets, in-house tests and expert interview. This type of

modeling is generally useable for concurrent design groups where engineers

from different field of expertise work together to generate a single solution

(product) to a problem initiated. Quick decisions have to be made based on

low amount of information. The type of models proposed to make quick but

accurate decisions are fundamental for decision-room design. It must be

noted that the development of these tools should be cost-effective for the

application or technology field the engineer is working in.

The models used in this thesis are specifically produced for estimating the

feasibility of DMFC power systems for portable electronic devices. This

makes the model not applicable to other products smaller or larger scale

than the scope defined in this thesis. The results of the tests described in

Chapter 8 and 9 are only applicable to the device designed, the MP3 player.

Because the power and energy specifications of other applications within this

application field do not differ a great deal from the investigated case study, it

is plausible the results also will apply to these cases. It is recommended to

explore this.

10.4 Recommendations for future research The research executed in this thesis consisted of two parts: to test the

feasibility of DMFC power system as an alternative for the lithium-based

rechargeable battery, and developing a systematic approach for the product

designer, supporting him/her in his decision making during the heuresis

stage of the conceptual design phase.

10.4.1 Direct Methanol Fuel cell systems in portable electronics

Within this thesis only one case is tested on its feasibility. The second-order

model can be used to test a wider range of portable electronic devices, and

give more insight in fields of opportunity for the DMFC system.

217

To increase more potential component-sets for the second-order model, the

component tables should be extended with more commercially available

components. In the performance range of the research conducted the

number of available components is still very low, and thus to test long-term

feasibility of DMFC systems scalability of components could be introduced.

The results will not only show a new design but also pinpoints the research

and development before application is feasible. This could trigger component

developers to produce more miniature components in the required field.

The initial sales price of the MEA is still very high and thus will result in a

high initial price of the fuel cell system (twice as high as the benchmarked

battery). This high initial price of the power source could be of subordinate

concern, when the improvement in convenience is there. Based on results

from Chapter 5 the fuel-cartridge is time-in-between-replacements for fuel

cartridges is 2.7 times longer than for alkaline batteries and even 3.9 times

longer than the lithium-ions’ time-in-between-recharges. If this longer

runtime weighs up to the extra initial price of the power system is certainly a

thing to find out.

Besides initial price the methanol pricing, and thus the life-cycle costs, is

very high compared to recharging or buying alkaline batteries. Again the

improved performance could weigh up to the extra costs, and finding the

tipping point is of interest.

10.4.2 Model development

With regard to the first-order model, the estimates represent a wide range of

possible designs. The left side estimates represent well-engineered DMFC

systems with a high packing ratio. The right side estimates represent bad

engineered DMFC systems, and right in the middle the conventional designs

are situated. To validate the model more commercially available DMFC

systems have to be evaluated by the model, and coefficients, densities,

constants and efficiencies have to be fine tuned. This type of modeling is of

great interest for the conceptual product designers, because a feasibility test

can be made with low amount of input data from the designer. Besides

making use of commercially available systems the second or third order

218

model can be used to produce several DMFC systems powering applications

over the whole range of interest. The automatically produced designs can

than be used to improve the analytical models for the different parts of the

fuel cell system.

Cost estimates are implemented but due to lack of data not tested within the

second order model. This property is one of the main important issues and

the data tables containing commercially components data should therefore

be available, accurate and up to date.

The second-order model is limited in accuracy because the selected

components are represented by only parallelepipeds. Besides the simple

representations no interconnection is taken into account. For the third-order

model it is recommended to improve accuracy by (i) more advanced

representation of the components, (ii) extending the basic set of components

with amongst other the mixer and sensors, and by (iii) introducing

interconnections like wiring and plumbing. Besides more accurate designs

the objective function could be extended by taking more properties in to

account, like the environmental impact but also with thermal requirements

and systems’ endurance.

At the moment the second-order model is prototyped in Mathematica. To

decrease calculation time, especially needed in concurrent engineering

design groups, the model should be written in a lower-level coding

environment.

Besides the DMFC power system other alternative power systems could be

modeled according to the data acquired in Chapter 2. The equations can

than be used to compare the power systems with each other in the early

phases of the design process.

10.4.3 Concept development support by using computers

The models developed within this thesis are the first steps towards

conceptual modeling supporting the product designer in making funded

decision during early phases of the design process. In the systematic

219

approach of design engineering its common to develop an analytical model to

simulate different design-concepts. Mostly these models are prototyped in

numerical programs as Matlab/Simulink and spreadsheet programs as MS

Excel. The access to computer programs is increasing and it is encouraging

to see that new generation designers and scientists make more use of these

tools. As a university and as the school of Industrial Design Engineering we

are obligated to introduce the next generation designers with conceptual

modeling, making use of the support of the computer. The author would like

to recommend research in the field of analytical modeling during the concept

phase of the design process. This thesis is a first step towards generalized

models supporting the designer in conceptual decision making. The

computer can be used to, not only generate an unlimited number of

structural variants, but also for instance in the field of quantitative

morphological charts.

221

References

1. Sternblitz-Rubenstein, M.S. Number of Smartphones Worldwide Exceeded 180 Million Units in 2009. 2010 [cited 2010 September 13]; Available from: http://www.marketwire.com/press-release/Number-of-Smartphones-Worldwide-Exceeded-180-Million-Units-in-2009-1167459.htm.

2. John, J. and C. Riddle. Smart phone power. in Proceedings - Design Automation Conference. 2010. Anaheim, CA.

3. Kundu, A., J.H. Jang, J.H. Gil, C.R. Jung, H.R. Lee, S.H. Kim, B. Ku, and Y.S. Oh, Micro-fuel cells--Current development and applications. Journal of Power Sources, 2007. 170(1): p. 67-78.

4. Hallmark, J. End User Perspective - Industrial Consumer Electronics Power (< 20-50W). in DOE Fuel Cell Portable Power Workshop. 2002. Solid State Research Center, Phoenix, Arizona, USA.

5. Flipsen, S.F.J. and J.I.v. Hout, Haalbaarheidsonderzoek naar een opwindbare tandenborstel (for MOY Design). 2000, TNO Industrie, Delft: Delft.

6. Flipsen, B. and M. van der Vlugt, Case study durable energy sources for portables and wearables: walkman case (in Dutch). 1999, TNO Institute of Industrial technologies: Delft, the Netherlands.

7. Flipsen, S.F.J., V. Meeussen, H. Hakkesteegt, and J.W. Raadschelders, PDO eindrapportage: brandstofcellen voor draagbare elektronica. 2002: TNO Industrie, Delft.

8. Flipsen, B., G. Timmers, R. van den Hoed, and I. Leeflang. In search for application fields of fuel cells. in Proceedings of Electronics Goes Green 2000+. 2000: Fraunhofer institut.

9. Flipsen, S.F.J. and V.M.A. Meeussen, Marketscan fuel-powered products (for Swedish Match). 2000, TNO Industrie: Delft.

10. Torenbeek, E., Synthesis of subsonic airplane design. 1982, Delft: Springer.

11. Fu, L., T.J. Lu., and J.E. Huber, Database and selection method for portable power sources. Advanced Engineering Materials, 2005. 7(8): p. 755-765.

12. Cook, K.A. and A.M. Sastry, An algorithm for selection and design of hybrid power supplies for MEMS with a case study of a micro-gas chromatograph system. Journal of Power Sources, 2005. 140(1): p. 181 - 202.

13. Flipsen, B., Power sources compared: The ultimate truth? Journal of Power Sources, 2006. 162(2): p. 927 - 934.

14. van der Voort, E., Design of a methanol fuel cell powered PDA. 2005. 15. Carroll, A. and G. Heiser, An analysis of power consumption in a

smartphone, in 2010 USENIX Annual Technical Conference. 2010: Boston, MA, USA.

16. Mahesri, A. and V. Vardhan, Power consumption breakdown on a modern laptop, in Lecture Notes in Computer Science (including

222

subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). 2005. p. 165-180.

17. Mourik, F.v., Laagvermogen product op DMFC (in Dutch). 2001, TNO, TU Delft: Delft.

18. Juliussen, E. Worldwide PDA & smartphone forecast 2000-2010. 2003; Available from: http://www.etforecasts.com/products/ES_pdas2003.htm.

19. Onino. Onino website: compare prices and research thousands of products. 2010 [cited 2010 October]; Available from: http://www.onino.co.uk/.

20. Samsung. Specification sheet Samsung Galaxy S. 2010 [cited 2010 October]; Available from: http://www.samsung.com.

21. Apple. Specifications of the Iphone 4. 2010 [cited 2010 October]; Available from: http://www.apple.com/iphone/specs.html.

22. Engelsman, E. and R. Melchior, Trends in batteries for portable computers. 2005, Delft University of Technology: Delft, the Netherlands.

23. Starner, T., Human-powered wearable computing. IBM Systems Journal, 1996. 35(3-4): p. 618-629.

24. Diehl, J.C., P.v. Gennip, and A. Mestre, Design guidelines for renewable energy technologies in mobile products, in Int. design conference. 2006, Design society: Glasgow. p. 1375-1381.

25. Dowling, K., Power Sources for Small Robots. 1997, Robotics Institute: Pittsburgh, PA.

26. Johnson, M.P., Physical limits of portable power storage. 1998, MIT Media Lab: Cambridge, MA, USA.

27. Limmeren, R.v. and M.v.d. Berg, Trends in energy and power supply of mobile phones. 2004, Delft University of Technology: Delft, the Netherlands.

28. Miyake, Y., Y. Akiyama, A. Hamada, Y. Itoh, K. Oda, S. Sumi, K. Nishio, and N. Nishizawa, Status of fuel cells R&D activities at Sanyo. Journal of Power Sources, 1996. 61(1-2): p. 155--160.

29. Xie, C., J. Bostaph, and J. Pavio, Development of a 2 W direct methanol fuel cell power source. Journal of Power Sources, 2004. 136(1): p. 55 - 65.

30. Pavio, J., J. Hallmark, J. Bostaph, A. Fisher, B. Mylan, and C.G. Xie, Developing micro-fuel cells for wireless communications. Fuel Cells Bulletin, 2002. 2002(4): p. 8 - 11.

31. Seiko, C. Seiko Thermic 6C12. 1999 [cited 1999; Available from: www.seiko-corp.co.jp.

32. Elsner, N.B., J.C. Bass, S. Ghamaty, D. Krommenhoek, A. Kushch, J. Mccoy, V. Javonovic, and J. Elsner. Performance specifications of the HZ-2, HZ-9, HZ-14, HZ-20 and the demonstration generator. Power Generating Modules 2004 [cited 2004; Available from: http://www.hi-z.com/.

33. ThermoElectric, G. Performance specifications of: model 5015, 5030, 5060, 5120, 5220, 8550, 1120 and 1500. GlobalTEGs 2004 [cited 2004; Available from: http://www.globalte.com.

34. Piezo Systems, I. Catalogus: performance specifications of T215-A4 and H4 series, T220-A4 and H4 series, and the T226-A4 and H4 series. Piezo Bending Generators 2004 [cited 2004; Available from: www.piezo.com.

223

35. Mabuchi, M., Performance specification sheet: RF-500TB, in Precious metal-brush motors, M. Motor, Editor. 2004: Japan.

36. Juden, C. Dynotest of the Soubitez 42 and AXA-HR dynamo. 2004 [cited 2004; Available from: http://www.myra-simon.com/myra/.

37. Kinetron. Kinetron Micro Generator performance specifications of: MG861, 205, 204, 401, 026 and 901. 2004 [cited 2004; Available from: http://www.kinetron.nl.

38. Toshiba, Toshiba Announces World's Smallest Direct Methanol Fuel Cell with Energy Output of 100 Milliwatts. 2004, Press release.

39. HelioCentris. Performance specifications of: Science education kit. 2004 [cited 2004.

40. SFC. Performance specifications of: A25. 2004 [cited 2004; Available from: http://www.sfc.com.

41. Fuelcellstore. Webshop of fuel cell products and components related, performance specification of: Cute FC series, PEMFC kit, Fuel cell eco, Cozy FC series, Fuel cell Mini, CoolFC series, and others. 2004 [cited 2004; Available from: http://www.fuelcellstore.com.

42. Xie, C., J. Pavio, J. Hallmark, J. Bostaph, and A. Fisher. Key requirements of micro fuel cell system for portable electronics. in Proceedings of the Intersociety Energy Conversion Engineering Conference. 2002.

43. Toshiba, Toshiba Announces World's First Small Form Factor Direct Methanol Fuel Cell for Portable PCs. 2003, Press release.

44. NEC, NEC Develops Smaller Fuel Cell for Notebook PCs. 2003, Press release by NEC.

45. Hepperle, M., Cox Tee-Dee .010. Aeromodeller Oktober, 1979. 46. Webra. Performance specifications of: Speed 50 Aero/Heli 1025 series,

Speed 40 Aero 1034GT, 32GT Aero 1028GT, 61F Aero and 120 Aero. 2004.

47. Fu, K., A.J. Knobloch, F.C. Martinez, D.C. Walther, C. Fernancez-Pello, A.P. Pisano, D. Liepmann, and K. Miyaska. Design and experimental results of small-scale rotary engines. in 2001 ASME International Mechanical Engineering Congress and Exposition. 2001. New York (NY).

48. SuperTigre, Performance specifications of: G34 Aero, GS40 Rin, GS45 Abc, GS45 H ABC, G51 Ring, G61K Ring, G61H Abc-Ring, G75 Ring, G90 Ring, S2000/25, G3250, G4500, Twin 60CC and G20/23, in Supre Tigre, superior quality and performance. 2000: Champaign, IL.

49. A.S.P. Performance specifications of: 108A, 91A, 75A, 46A, 40A, 32A-S, 28, 25A, 15A, 12A AP090, AP15A and four-stroke series. 2004 [cited 2004; Available from: http://www.conrad.nl.

50. X-race. Performance specifications of: 25SZ, M21SZ, 21KS, pro 21SX, pro 21SR, pro 15ST, pro 15SX, M15ES, M15SZ, 17ES, 17SZ, 15ES and 15SZ. 2004.

51. Honda. Performance specifications of: GX-22, GX-31, GX120, GX160, GX200, GX240, GX270 and GX340. 2004 [cited 2004; Available from: http://www.hondashop.com.

52. O.S.Engines, Performance specification of the OS.30 Wankel Rotary 49-PI, O.S. Engines, Editor. 2004: Champaign, Il, USA.

53. Jansen, A.J. and P. Slob. Human power; comfortable one-hand cranking. in Proceedings of International Conference on Engineering Design. 2003. Stockholm, Sweden.

54. Andriesse, C.D., Energieconversies. 1991: University of Utrecht.

224

55. Kawai, K. and M. Harada, Man as Locomotive Engine. 1997, Kyoto University: Kyoto, Japan.

56. Varta, Performance specifications of different primary batteries from Varta: Alkaline, Lithium cilinder, Zinc chloride cilinder; and secondary batteries: NiMH cilinder and button cells, . 2004, Varta.

57. Panasonic, Performance specifications of different secondary batteries: Lead-acid, NiCd cilinder and power packs, NiMh 3000 sub C packs, Li-ion camcorder cells; and goldcaps super capacitors., Panasonic, Editor. 2004.

58. Voltcraft, Performance specifications of Voltcraft Sealed Lead Acid batteries, Voltcraft, Editor. 2004.

59. SAFT, Performance specifications of SAFT Li-ion batteries series, SAFT, Editor. 2004.

60. Rayovac, Performance specifications for Rayovac Alkaline rechargebale batteries, Rayovac, Editor. 2004.

61. Emmerich, Performance specifications of Emerich NiCD button cells, Emmerich, Editor. 2004.

62. Canon, Performance specifications fo different Canon Camcorder batteries, Canon, Editor. 2004.

63. Samsung, Performance specification of Samsung li-ion batteries, Samsung, Editor. 2004.

64. GP, Performance specifications of GP NiMh cilindrical and prismatic cells, GP, Editor. 2004.

65. JVC, Performance specifications of JVC li-ion camcorder batteries. 2004.

66. Sony, Performance specifications of different Sony camcorder batteries, S. Batteries, Editor. 2004.

67. Epcos, Performance specifications of Epcos UltraCaps, Epcos, Editor. 2004.

68. Rutronik, Performance specification of Rutronik Electrolytic condensors, R. EYF, Editor. 2004.

69. Maxwell, Electric Double Layer Capacitor: BOOSTCAP (R) Ultracapacitor. 2004, Maxwell Technologies.

70. Schneuwly, A. and R. Gallay, Properties and applications of supercapacitor, from the state-of-the-art to future trends, in International Conference on Power Electronics, Intelligent Motion and Power Quality. 2000, Mesago PCIM GmbH: Nuernberg, Germany.

71. Zuettel, A., Materials for hydrogen storage. Materials Today, 2003. 6(9): p. 24-33.

72. C.E.S., Cambridge Engineering Selector. 2004. 73. Hockaday, R. Energy Related Devices. 2010 [cited 2010; Available

from: http://www.energyrelatedevices.com. 74. Tsuchiya, H. and O. Kobayashi, Mass production cost of PEM fuel cell

by learning curve. International Journal of Hydrogen Energy, 2004. 29(10): p. 985--990.

75. Dyer, C.K., Fuel cells for Portable Applications. Journal of Power Sources, 2002. 44(2): p. 109--115.

76. Advitek. Brandstofcel. 2010 [cited 2010 November 12]; Available from: http://www.advitekmobilehome.nl.

77. Instant-Power, Performance specification of the Instant power Zinc Air battery, I. Power, Editor. 2004.

78. Mars, P., Power delivery for notebooks and wireless PDA's with supercapacitors. 2001.

225

79. Flipsen, B. Selection of power sources for portable applications. in Portable Energy 2009. 2009: Knowledge Foundation.

80. Roozenburg, N.F.M. and J. Eekels, Productontwerpen, structuur en methoden. 2nd ed. 2003, Utrecht: Lemma.

81. Pahl, G., W. Beitz, and K. Wallace, Engineering Design: a systematic approach. 2nd ed. 1996: Springer.

82. Spitas, C., Analysis of systematic engineering design paradigms in industrial practice: A survey. Journal of Engineering Design, 2010.

83. Eder, W.E., Design Modeling - A design science approach (and why does industry not use it?). Journal of Engineering Design, 1998. 9(4): p. 355-371.

84. Scopus. Scopus scientific search engine. 2009 [cited 2009; Available from: http://www.scopus.com.

85. ScienceDirect. ScienceDirect scientific search engine. 2009 [cited 2009; Available from: http://www.sciencedirect.com.

86. N.A.P., Appendix C: Measures of Performance, in Meeting the energy needs of future warriors. 2004, The National Academies press: Washington D.C., USA. p. 80-87.

87. Christen, T. and M.W. Carlen, Theory of ragone plots. Journal of Power Sources, 2000. 91(2): p. 210-216.

88. Ragone, D.V., Review of battery systems for electrically powered vehicles, in Mid-year meeting of the Society of Automotive Engineers. 1968: Detroit, Mi.

89. Rincón-Mora, G.A. and M. Chen, Squeezing operational life out of a shrinking energy capsule. Power Management Design Line, 2006.

90. Ashby, M.F., Y.J.M. Brechet, D. Cebon, and L. Salvo, Selection strategies for materials and processes. Materials & Design, 2004. 25(1): p. 51 - 67.

91. van Gennip, P., S.Y. Kan, S. Silvester, and B. Flipsen. Power quest, an online software tool to find mobile product-energy-system matches. in Proceedings of the TMCE 2006. 2006. Ljubljana, Slovenia.

92. Cook, K.A., F. Albano, P.E. Nevius, and A.M. Sastry, POWER (power optimization for wireless energy requirements): A MATLAB based algorithm for design of hybrid energy systems. Journal of Power Sources, 2006. 159(1): p. 758 - 780.

93. Wu, Q., Q. Qiu, and M. Pedram. An interleaved Dual-Battery Power supply for Battery-operated Electronics. in Proceedings of Asia and South Pacific Design Automation Conference. 2000: ACM.

94. del Real, A.J., A. Arce, and C. Bordons, Optimization strategy for element sizing in hybrid power systems. Journal of Power Sources, 2009. 193(1): p. 315 - 321.

95. Lilienthal, P. and T. Lambert, HOMER, the optimization model for distributed power. 2009, website.

96. Bernal-Agustin, J.L. and R. Dufo-Lopez, Multi-objective design and control of hybrid systems minimizing costs and unmet load. Electric Power Systems Research, 2009. 79(1): p. 170 - 180.

97. Bernal-Agustin, J.L. and R. Dufo-Lopez, Efficient design of hybrid renewable energy systems using evolutionary algorithms. Energy Conversion and Management, 2009. 50(3): p. 479 - 489.

98. Flipsen, B., Personal interview with Thomas Valdez, senior member engineering staff. 2009: Pasadena, CA, USA.

99. Flipsen, B., Personal interview with Paul Timmerman, senior power system engineer at JPL. 2009: Pasadena, CA, USA.

226

100. Icardi, U.A., S. Specchia, G.J.R. Fontana, G. Saracco, and V. Specchia, Compact direct methanol fuel cells for portable application. Journal of Power Sources, 2008. 176(2): p. 460 - 467.

101. Lin, B., Conceptual design and modeling of a fuel cell scooter for urban Asia, in School of Engineering and Applied Sciences, Department of Mechanical and Aerospace Engineering. 1999, Princeton University: Princeton NJ, USA.

102. Rohrschneider, R.R., Olds J.R. Braun R.D. Hutchinson Jr. V. Kuhl C.A. Steffes S.R., Flight system options for a long-duration Mars airplane. Collection of Technical Papers - AIAA 3rd "Unmanned-Unlimited" Technical Conference, Workshop, and Exhibit, 2004. 2: p. 833-846.

103. Rohrschneider, R.R., Olds J.R. Braun R.D. Hutchinson Jr. V. Kuhl C.A. Steffes S.R., Modeling approach for analysis and optimization of a long-duration Mars airplane. Collection of Technical Papers - 10th AIAA/ISSMO Multidisciplinary Analysis and Optimization Conference, 2004. 1: p. 134-147.

104. Wikipedia. Design Structure Matrix - Wikipedia, The Free Encyclopedia. 2009; Available from: http://en.wikipedia.org/wiki/Design_structure_matrix.

105. Malone, B. and M. Papay, ModelCenter (TM): an integration environment for simulation based design. 1999.

106. Spitas, C., Analysis of systematic engineering design paradigms in industrial practice: Scaled experiments. Journal of Engineering Design, 2010.

107. Creswell, J.W., Research Design: Qualitative and Quantitative Approaches. 2nd. ed. 2003: Sage.

108. Baarde, D.P. and M.P.M. de Goede, Basisboek Kwalitatief Onderzoek. 2001, Houten: Stenfert Kroese.

109. Linden, D. and T.B. Reddy, Handbook of Batteries. 3rd ed. 2001: McGraw-Hill Inc.

110. Buchmann, I. Battery University. 2005 [cited 2010; Available from: www.batteryuniversity.com.

111. DeMuro, D., Whither Fuel Cells. 2003, Motorola Energy Systems Group, Battery Power Products & Technology Magazine.

112. Larminie, J. and A. Dicks, Fuel cell systems explained. 2003: Chichester: Wiley.

113. Shukla, A.K., C.L. Jackson, K. Scott, and R.K. Raman, An improved-performance liquid-feed solid-polymer-electrolyte direct methanol fuel cell operating at near-ambient conditions. Electrochimica Acta, 2002. 47(21): p. 3401--3407.

114. Qi, Z. and A. Kaufman, Open circuit voltage and methanol crossover in DMFCs. Journal of Power Sources, 2002. 110(1): p. 177 - 185.

115. Nakagawa, N. and Y. Xiu, Performance of a direct methanol fuel cell operated at atmospheric pressure. Journal of Power Sources, 2003. 118(1-2): p. 248--255.

116. Okajima, k., M. Suetaka, K. Furukawa, and M. Sudoh, Evaluation of Liquid-feed direct methanol fuel cell at near-ambient temperature for mobile applications. Electrochemistry, 2002. 70(12): p. 928--930.

117. Ramani, V., H.R. Kunz, and J.M. Fenton, Effect of particle size reduction on the conductivity of Nafion®/phosphotungstic acid composite membranes. Journal of Membrane Science, 2005. 266(1-2): p. 110-114.

227

118. Wang, Z.H. and C.Y. Wang, Mathematical Modeling of Liquid-Feed Direct Methanol Fuel Cells. Journal of The Electrochemical Society, 2003. 150(4): p. A508--A519.

119. Oedegaard, A. and C. Hentschel, Characterisation of a portable DMFC stack and a methanol-feeding concept. Journal of Power Sources, 2006. 158(1): p. 177 - 187.

120. Padhy, B.R. and R.G. Reddy, Performance of DMFC with SS 316 bipolar/end plates. Journal of Power Sources, 2006. 153(1): p. 125 - 129.

121. Chen, C.Y., D.H. Liu, C.L. Huang, and C.L. Chang, Portable DMFC system with methanol sensor-less control. Journal of Power Sources, 2007. 167(2): p. 442 - 449.

122. Peiwen Li, N.D., Anthony Dumay, Development and Fabrication of Passive DMFCs, in Seventh International Fuel Cell Science, Engineering and Technology Conference. 2009, ASME: Newport Beach, CA, USA.

123. Qian, W., D.P. Wilkinson, J. Shen, H. Wang, and J. Zhang, Architecture for portable direct liquid fuel cells. Journal of Power Sources, 2006. 154(1): p. 202--213.

124. Kim, D., E.A. Cho, S.A. Hong, I.H. Oh, and H.Y. Ha, Recent progress in passive direct methanol fuel cells at KIST. Journal of Power Sources, 2004. 130(1-2): p. 172--177.

125. Bostaph, J., C. Xie, J. Pavio, A.M. Fisher, B. Mylan, and J. Hallmark, 1W direct methanol fuel cell system as a desktop charger. 2003, internet.

126. Sony, Announcement of New Lithium Ion Batteries, Realizing Industry's Highest Level of Energy Density (A8 series, 383450A8). 2004, Press release.

127. Fisher, A., J. Pavio, J. Bostaph, R. Koripella, B. Mylan, J. Hallmark, S. Rogers, S. Samms, and C. Xie. Technology Development in Methanol-Based fuel cells. in Proceedings of the Third International Micro and Nanotechnology Meeting. 2003. Grenoble, France.

128. Ryynaenen, M. and S. Tasa, Optimized Energy for Future Mobile Multimedia Devices. 2005, Website: www.h2foresight.info/Nokia.ppt.

129. Broussely, M. and G. Archdale, Li-ion batteries and portable power source prospects for the next 5-10 years. Journal of Power Sources, 2004. 136(2): p. 386--394.

130. Wichert, R., Fuel Cells for Portable Power: Markets, Manufacture and Cost. 2003, Darnell Group Inc.: Corona, California, U.S.A.

131. Schuttenhelm, R. Prijs electriciteit fors gestegen (in Dutch), based on: Bron: Nederlands huishouden duur uit met elektriciteit. CBS, mei 2006. 2006 [cited 2010 November]; Available from: http://www.hoesnel.nl/energie_ontwikkeling/elektriciteitsprijs-nederland.html.

132. Fuelcellstore. Webshop for fuel cell and materials. 2010 [cited 2010 november].

133. Kitco. Precious Metal Store: price for Platinum. 2010 [cited 2010 November 2010]; Available from: http://www.kitco.com.

134. Crompton, T.R., Battery Reference Book. 2000: Reed Educational and Professional publishing Ltd.

135. Toshiba's New Rechargeable Lithium-Ion Battery Recharges in Only One Minute. 2005, Press release.

228

136. Knights, S.D., K.M. Colbow, J. St Pierre, and D.P. Wilkinson, Aging mechanisms and lifetime of PEFC and DMFC. Journal of Power Sources, 2004. 127(1-2): p. 127--134.

137. Schoormans, J. and C. de Bont, Consumentenonderzoek in de produktontwikkeling. 1995: Lemma.

138. SPSS, SPSS: data mining, statistische analyse software. 2004, IBM. 139. Mangione-Smith, W.H., "Beauty Depends on Size"-Aristotle. IEEE

Concurrency, 1998. 6(2): p. 8--10. 140. Chapin, T., Density of consumer eelctronics and user perceptionsof

quality, acceptability of weight, level of technology, durability and fragility. 1999, TUFTS University: Medford, USA.

141. Mueller, J., Personal contact with Jens Mueller, Smart Fuel Cell. 2010, Personal contact.

142. SFC. The JENNY 600S fuel cell by SFC. 2010 [cited 2010; Available from: http://www.sfc.com/en/man-portable-jenny.html.

143. Geyer, R. and V. Doctori Blass, The economics of cell phone reuse and recycling. The International Journal of Advanced Manufacturing Technology, 2010. 47(5): p. 515-525.

144. Andrae, A. and O. Andersen, Life cycle assessments of consumer electronics — are they consistent? The International Journal of Life Cycle Assessment, 2010. 15(8): p. 827-836.

145. Flipsen, B., Designing micro fuel cells for portable products. Journal of Fuel Cell Science, Engineering, and Technology, 2010. 7.

146. Samsung. Samsung YP-Z5F MP3 player, specification sheet 2007 [cited 2007; Available from: http://www.samsungmobile.com.

147. Flipsen, B. Design Challenges for a Fuel Cell Powered MP3 Player. 2007. Bath (Great Britain): International Power Sources Symposium.

148. Micro Annular Gear Pump mzr(R)-2521 in Dosing pump for analytical instrumentation. 2006, HNP Mikrosysteme GmbH.: www.hnp-mikrosysteme.de.

149. ThinXXS, Datasheet MDP1304, micro diaphragm pump. 2006, ThinXXS Microtechnology AG.

150. Bartels. Datasheet of the MP5 and MP6 micropump. 2008. 151. Sparks, D., C. Laroche, N. Tran, D. Goetzinger, N. Najafi, K.

Kawaguchi, and M. Yasuda. A new Methanol Concentration Microsensor for Improved DMFC Performance. 2005: Fuel Cell Summit 2005.

152. Burke, A., R&D considerations for the performance and application of electrochemical capacitors. Electrochimica Acta, 2007. 53(3): p. 1083 - 1091.

153. Yogmogita, H., Electrical Storage Technology Combines Li-ion Capacitors. Nikkei Electronics Asia Volume, 2005.

154. Vishal, S., V.G. Suresh, and R. Arvind, Microscale pumping technologies for microchannel cooling systems. 2004: p. 191-221.

155. Laser, D.J. and J.G. Santiago, A review of micropumps. Journal of Micromechanics and Microengineering, 2004. 14(6): p. R35-R64.

156. SP V 125 PZ-L datasheet, The revolutionary Generation of Piezo-Pumps. 2009, Schwartzer GmbH: www.schwartzer.com.

157. Bohm, S., W. Olthuis, and P. Bergveld, A plastic micropump constructed with conventional techniques and materials. Sensors and Actuators A: Physical, 1999. 77: p. 223-228.

229

158. Richter, A., A. Plettner, K.A. Hofmann, and H. Sandmaier, A micromachined electrohydrodynamica (EHD) pump. Sensors and Actuators A, 1991. 29: p. 9.

159. Chen, C.H., S. Zeng, J.C. Mikkelsen, and J.G. Santiago, Development of a planar electrokinetic micropump. 2000.

160. Richter, M., Micropumps - From the Lab to the Fab. Actuator 2006, 2006.

161. Richter, M., ErsterPreis fur die kleinste Siliziumpumpe der Welt. 2007. 162. Zhu, M., P. Kirby, M. Wacklerle, M. Herz, and M. Richter,

Optimization design of multi-material micropump using finite element method. Sensors and Actuators A: Physical, 2009. 149(1): p. 130 - 135.

163. Flipsen, B., Personal communication with M. Richter (Fraunhofer Institut) about Fraunhofers micropump. 2009.

164. Kaemper, K.P., J. Doepper, W. Erhrfeld, and S. Oberbeck. A self-filling low-cost membrane micropump. 1998: IEEE Micro electro Mechanical Systems.

165. Cabuz, C., W.R. Herb, E.I. Cabuz, and S.L. Lu. The dual diaphragm pump. 2001: IEEE.

166. Schabmueller, C.G.J., Self-alignement gas/liquid micropump. Journal of Micromechanical Microengineering, 2002. 12: p. 420-424.

167. Datasheet: V200 and P200 GAS. 2007, Xavitech AB: www.xavitech.com.

168. Rotary Vane Pump BL-G 085 M. 2006, Gardner Denver Thomas: www.rtpumps.com.

169. Polyfuel, Membranes push boundary of power density for DMFC stacks. Membrane Technology, 2007. 2007(11): p. 8 - 9.

170. Hadjipaschalis, I., A. Poullikkas, and V. Efthimiou, Overview of current and future energy storage technologies for electric power applications. Renewable and Sustainable Energy Reviews, 2009. 13(6-7): p. 1513 - 1522.

171. Cagan, J., K. Shimada, and S. Yin, A survey of computational approaches to three-dimensional layout problems. Computer-Aided Design, 2002. 34(8): p. 597-611.

172. Grignon, P.M. and G.M. Fadel, A GA Based Configuration Design Optimization Method. Journal of Mechanical Design, Transactions of the ASME, 2004. 126(1): p. 6-15.

173. Fox, R.L., Optimization methods for engineering design. Mechanics and thermodynamics, ed. H.W.E.a.B. budiansky. 1971: Addison-Wesley Publishing Company. 270.

174. Deb, K., Multi-objective Optimization using Evolutionary Algorithms, ed. S. Ross and R. Weber. 2001: John Wiley & Sons, LTD.

175. Wikipedia. Line-plane intersection - Wikipedia, The Free Encyclopedia. 2010; Available from: http://en.wikipedia.org/w/index.php?title=Line-plane_intersection&oldid=346414975.

176. Wikipedia. Collision detection - Wikipedia, The Free Encyclopedia. 2010; Available from: http://en.wikipedia.org/w/index.php?title=Collision_detection&oldid=349217034.

177. Wolfram, Mathematica (R) 7. 2010, Wolfram Research, Inc.: Champaign, IL, USA.

230

178. Strijk, R., Development and future feasibility of a direct methanol fuel cell system with a power of 50Watts, in Industrial Design Engineering. 2002, Delft University of Technology: Delft.

179. Penworld. Prices for a 5-pack 1.45mL fill-pen refills. 2010 May 2010]; Available from: http://www.penworld.be.

231

A Modeling of the open cell Voltage for a DMFC

The theoretical reversible open cell voltage E0 can be calculated according to

[112]:

0 fg

EzF

−Δ= = 1.206V (51)

In practice this theoretical open cell Voltage is never reached and in our

model the practical open cell Voltage VOC is used equal to [1]:

0

0log( )OCV E A i= + (52)

The working Voltage of a DMFC cell is equal to:

( ) ( )00log( ) log( ) expcV E i r A i i A i m n i= − ⋅ − ⋅ + − + ⋅ ⋅ (53)

Combining Equation 52 with 53 makes:

( )log( ) expOC cV V i r A i i m n i= − ⋅ − ⋅ + + ⋅ ⋅

The practical open cell Voltage is in this equation a great unknown. By

means of multiple regression analysis using 4 case-studies [125-128] and 47

measured points the value of this Voltage is derived in a formula. First a list

of parameters is derived which influence the performance of the cell:

232

• The active surface area of the cell A (cm2)

• The methanol concentration N (mol dm-3)

• The fuel flow vmeoh (ml min-1)

• The cells temperature T (degrees C)

• The air flow vair (ml min-1)

• The platinum loading on the anode (mg cm-2)

• The platinum loading on the cathode (mg cm-2)

In Table 37 the results of the first analysis is shown, and the accuracy of the

predicting equation is presented in Figure 79.

Table 37: Multiple regression analysis on all variables.

Regression Coefficient t(N-P-1;0,05) 2,022691

Estimate SE t(cal)

P

(T<=t(cal))

Lower

95%

Upper

95%

VIF (Variance

Inflation

Factor)

b0 606,213

7

60,3328

6

10,0478

2

***

(P<=0.001) 2,24E-12 484,179 728,2484

b1:deg C 1,10224

1

0,47642

1

2,31358

7 * (P<=0.05) 0,026051

0,13858

9 2,065894 1,576477

b2:cm2 3,82070

5

3,09008

1

1,23644

2

N.S.

(P>0.05) 0,223689 -2,42957 10,07098 10,76056

b3:mg/cm2 (a) -12,941711,1603

7 -1,15961

N.S.

(P>0.05) 0,253256 -35,5157 9,632251 2,776644

b4:ml/min (a) -6,124951,87012

8 -3,27515 ** (P<=0.01) 0,00222 -9,90764 -2,34226 10,0846

b5:mol/dm3

(a) -43,9511

18,1835

3 -2,41708 * (P<=0.05) 0,020422 -80,7308 -7,17146 2,324562

b6:mg/cm2 (k) 26,4237

1

9,86341

3

2,67896

2 * (P<=0.05) 0,010757

6,47307

5 46,37435 4,439991

b7:ml/min (k) -0,010450,01949

4 -0,53603

N.S.

(P>0.05) 0,594979 -0,04988 0,028982 7,493789

233

0

100

200

300

400

500

600

700

800

900

0 200 400 600 800 1000

mV (OC) observed

mV

(O

C)

pre

dic

ted

Figure 79: Observed and predicted value of the open cell voltage for the first estimate.

The significance of the variable is shown by the value P. The non-significant

variables are removed (P>0,05), like the active surface-area (b2), the

platinum loading on the anode (b3) and the air flow (b7), and a new multiple

regression analysis is executed, shown in Table 38 and the accuracy of the

prediction is shown in Figure 80.

Table 38: Multiple regression analysis on a selected number of variables. Regression Coefficient t(N-P-1;0,05) 2,018082

Estimate SE t(cal) P(T<=t(cal))Lower 95%Upper 95% VIF (Variance

Inflation Factor)

b0 574,833558,787049,778235*** (P<=0.001) 2,18E-12 456,1964 693,4705

b1:deg C 0,9929670,422966 2,34763 * (P<=0.05) 0,02368 0,139388 1,846547 1,141642

b2:ml/min (a) -5,20097 0,815925 -6,37433*** (P<=0.001) 1,15E-07 -6,84758 -3,55437 1,763722

b3:mol/dm3 (a) -47,3301 17,43898 -2,71404 ** (P<=0.01) 0,0096 -82,5234 -12,1368 1,964447

b4:mg/cm2 (k) 34,464486,7660115,093767*** (P<=0.001) 7,85E-06 20,81012 48,11885 1,919582

234

0

100

200

300

400

500

600

700

800

900

0 200 400 600 800 1000

mV (OC) observed

mV

(O

C)

pre

dic

ted

Figure 80: the observed and predicted value of the open cell voltage for the second

estimate.

With a significance of 99% and a R2 = 0.65 (n=47) the following equation

predicting the practical open cell Voltage (in mV) can be defined:

574.8 0.9930 5.2010 47.3301 34.4645OC MeOH cV T v N m= + − − + (54)

In this equation it can be seen that the influence of methanol concentration

is negatively correlated to the open cell Voltage. In [2] the dependency of the

fuel flow is researched experimentally. This research confirms the negative

correlation of the methanol concentration with the open cell Voltage. Besides

this negative correlation the research showed the neglecting influence of fuel

flow on the cell Voltage when it was in between 1 and 4.5 ml min-1. Because

the fuel flows in small portable fuel cells is low, this variable is excluded and

a new multiple regression analysis is executed which will result in a relation

between the open cell Voltage and the temperature T, the platinum loading

on the cathode mc and the methanol concentration N.

Table 39 shows the results from the multiple regression analysis when only

taking the previous variables into account. Again the accuracy of the

prediction is shown in Figure 81.

235

Table 39: Final multiple regression analysis on the most significant variables, temperature, fuel concentration and the platinum loading on the cathode.

Regression Coefficient t(N-P-1;0,05) 2,016692

Estimate SE t(cal) P(T<=t(cal)) Lower 95% Upper 95% VIF

b0 456,757277,34189 5,90569 *** (P<=0.001) 5,01E-07 300,7824 612,732

b1:deg C 1,5809110,5722232,762754 ** (P<=0.01) 0,0084 0,426914 2,734909 1,087351

b2:mol/dm3 (a) -10,7873 22,83112 -0,47248 N.S. (P>0.05) 0,638973 -56,8306 35,25608 1,752156

b3:mg/cm2 (k) 18,59993 8,72175 2,132592 * (P<=0.05) 0,038709 1,010849 36,18902 1,659848

0

100

200

300

400

500

600

700

800

900

0 200 400 600 800 1000

mV (OC) observed

mV

(O

C)

pre

dic

ted

Figure 81: the observed and predicted value of the open cell voltage for the third estimate.

With a significance of 99% and a R2 =0.31 (n=47) the following equation can

be stated:

456.8 1.5809 10.7873 18.5999OC kathV T N m= + − + (55)

236

Table 40: Overview of alle measured points used for the VOC model. Cel/Stack Anode Cathode Ppp ipp Vpp VOC Tcell nr.cells Atotal ma N Va mc vc Ref. mW/cm2 mA/cm2 mV mV deg C cm2 mg/cm2 mol/dm3 ml/min mg/cm2 ml/min

175 650 270 820 90 1 25 1.00 1.0 15.00 4.6 1000 [113] 90 450 210 805 90 1 25 1.00 1.0 15.00 4.6 1000 105 315 310 750 60 1 25 1.00 1.0 15.00 4.6 1000 65 240 260 800 60 1 25 1.00 1.0 15.00 4.6 1000 30 138 220 617 60 1 25 4.80 0.5 40.00 4.8 920 [114] 30 138 220 615 60 1 25 4.80 0.5 40.00 4.8 643 30 138 220 613 60 1 25 4.80 0.5 40.00 4.8 397 28 125 220 601 60 1 25 4.80 0.5 40.00 4.8 180 33 143 232 590 60 1 25 4.80 1.0 40.00 4.8 920 29 107 270 580 60 1 25 4.80 1.0 40.00 4.8 643 22 90 242 570 60 1 25 4.80 1.0 40.00 4.8 397 17 66 170 455 60 1 25 4.80 1.0 40.00 4.8 180 36 160 226 368 80 1 25 4.80 1.0 40.00 4.8 920 31 143 216 495 80 1 25 4.80 1.0 40.00 4.8 643 17 81 211 526 80 1 25 4.80 1.0 40.00 4.8 397 7 50 148 530 80 1 25 4.80 1.0 40.00 4.8 180 20 129 155 550 80 1 5 1.00 2.0 2.75 1.0 910 [115] 44 300 147 600 80 1 5 2.70 2.0 2.75 1.0 910 94 525 179 615 80 1 5 3.75 2.0 2.75 1.0 910 87 425 205 625 80 1 5 5.00 2.0 2.75 1.0 910 95 400 238 650 80 1 5 5.70 2.0 2.75 1.0 910 16 110 145 500 30 1 5 3.75 2.0 3.00 1.0 1930 53 300 177 515 60 1 5 3.75 2.0 3.00 1.0 1930 64 325 197 530 70 1 5 3.75 2.0 3.00 1.0 1930 72 350 206 540 80 1 5 3.75 2.0 3.00 1.0 1930 59 300 197 550 90 1 5 3.75 2.0 3.00 1.0 1930 20 120 170 500 30 1 5 3.75 2.0 3.00 1.0 1930 74 400 185 580 60 1 5 3.75 2.0 3.00 1.0 1930 87 440 197 600 70 1 5 3.75 2.0 3.00 1.0 1930 97 475 204 620 80 1 5 3.75 2.0 3.00 1.0 1930 100 500 199 640 90 1 5 3.75 2.0 3.00 1.0 1930 120 600 200 640 100 1 5 3.75 2.0 3.00 1.0 1930 35 200 173 550 90 1 5 3.75 2.0 3.00 1.0 310 53 260 205 550 90 1 5 3.75 2.0 3.00 1.0 910 303 673 450 800 120 1 25 5.40 1.0 4.00 6.3 4000 251 561 447 780 110 1 25 5.40 1.0 4.00 6.3 4000 158 321 490 770 90 1 25 5.40 1.0 4.00 6.3 4000 99 276 359 760 70 1 25 5.40 1.0 4.00 6.3 4000 35 100 350 750 25 1 25 5.40 1.0 4.00 6.3 4000 304 680 447 750 110 1 25 5.40 1.0 4.00 4.0 4000 217 486 447 690 110 1 25 5.40 1.0 4.00 2.0 4000 52 201 261 515 110 1 25 5.40 1.0 4.00 1.0 4000 205 500 411 750 110 1 25 5.40 1.0 4.00 1.0 4000 1 5 250 575 27 1 5 2.00 0.5 0.10 2.0 150 [116] 8 40 194 550 27 1 5 2.00 1.5 0.10 2.0 150 6 45 135 495 27 1 5 2.00 3.0 0.10 2.0 150 4 30 126 460 27 1 5 2.00 5 0.10 2.0 150

237

B

Mathematica files

To test the algorithm proposed in chapter 9 this is implemented in two

Mathematica [177] programs, one used for the first and second level run

(Section B.1) and the third-level run, using an evolutionary strategy (Section

B.2). The inputtables including the commercially available pumps and

intermediate accumulator can be found in Section B.3.

B.1 Initial algorithm On the following pages the initial program is presented which is used for the

second test run.

248

B.2 Evolutionary algorithm On the following pages the program is presented which is used in the third

test run.

254

B.3 Input tables (dbase) All non-flexible components specifications are stored in CSV file format. In

this Section the tables are presented for the gas pump, the liquid pump and

the intermediate accumulator (either a rechargeable battery or

supercapacitor).

The following tables are included in this appendix:

• ComponentA_gaspumps.csv

• ComponentB_liquidpumps.csv

• ComponentC_accu.csv (2 parts)

255

Tab

le B

.1: A

ll o

pti

onal

gas

pu

mps

incl

uded

in

th

e al

gori

thm

256

Tab

le B

.2: A

ll o

pti

onal

liq

uid

pu

mps

incl

uded

in

th

e al

gori

thm

257

Tab

le B

.3: A

ll o

pti

onal

in

term

edia

te a

ccu

mu

lato

rs in

clu

ded

in

th

e al

gori

thm

(par

t 1).

258

Tab

le B

.3: A

ll o

pti

onal

in

term

edia

te a

ccu

mu

lato

rs in

clu

ded

in

th

e al

gori

thm

(par

t 2).

259

Acknowledgements

This bookwork is the result of a long period of research which wouldn’t be

possible without the help of a few people. First of all I want to thank my

promotors and co-promotor, Han Brezet, Christos Spitas and Joris Vergeest.

Han, thank you for your support during the whole period, you were the only

constant factor making research a fun thing. Christos, for making me finish

this PhD and support during the last year of work, we finished it together.

Joris, for being a support during the last period of draft reading and giving

me constructive feedback.

I would like to thank Kas Hemmes for our talks about fuel cells and you

being available when I needed information. Thank you, professor Hans de

Deugd and professor Flip Doorschot for introducing me to the research field

of conceptual modeling and for being supportive in the intermediate phase of

this research.

In 2009 I visited the Jet Propulsion Lab, and my visit was received openly by

Thomas Valdez and Paul Timmerman. I would like to thank you for your

openness and cooperation in my search for information. I have learned a lot

from you guys.

Arjen, for being my roommate, our talks, us cracking jokes and of course

being a good friend. I would like to express my appreciation to all my

colleagues at the Design Engineering/Product Engineering (at random) and

especially: Bert, Maarten, Sander, Gerard, Erik and Erik, JC, Marco, Martin,

Herman and Fred. During the last period of my PhD, you guys were there to

reset my mind and take over a lot of my regular work. Dave for proofreading

and helping me out with my Dunglish (sometimes leaving it like it is). Katrijn

for proofreading the thesis and the mental text-messages during the last

couple of months.

260

This work wouldn’t be feasible without the contribution of several students.

Especially I would like to thank Maarten for your excellent CAD models, Eric

for your down-to-earth approach to fuel cell design, and Anne for your help

in the user assessment.

During the writing process I listened a lot to DJ Sven, DJ Sylverius and Ben

Liebrand (Veronica radio). Hereby my thanks for being in my headphones

during the long periods of programming, writing and thinking. In

anticipation of tonight I would like to thank DJ Fiesto for keeping me up on

the dance floor tonight.

During the PhD I am accompanied by two excellent paranymphs. Ruben

thanks for being an excellent (ex)colleague, roommate and thank you for

introducing me jazz music. Xander, thanks for being a friend and your

support during the defense.

I would like to thank my own family, my parents, brother and sister, and the

family of Petra for being supportive and interested, even during the last

couple of rough months, where it became clear what is really important. I

would like to show my gratitude to Petra, for being supportive and loving

during the whole period. Especially during the last couple of months taking

care of so many things in our lives. Lotte, Karsten and Evie, for making me

smile when I come home and showing me that learning something new every

day is a blessing.

261

Curriculum Vitae

Bas Flipsen was born in Vlierden, the Netherlands on July 13, 1971. He

obtained his HAVO and VWO diplomas in 1988 and 1990, respectively. As a

child he builds aircrafts, objects and repaired his go-cart using dad’s yellow

construction tape and cardboard. In continuation with his early interests he

went to Delft in August 1990, to study Aeronautical Engineering where he

graduated in 1997, on a preliminary design of a ‘Quiet Short Take Off and

Landing Aircraft’. During this period he learned about mathematical

modeling in de conceptualization phase and wanted to explore the creativity

part more. He continued his studies at the school of Industrial Design

Engineering, and graduated in 1999 on the design of a ‘low-power close-in

water heater using a heat pump’. He became an engineering product-

developer at the Delft Product Centre (TNO Industrial Technologies), where

he made first contact with sustainable power systems for portable

electronics. As a mathematical designer he worked on different projects at

the Sustainable Product Innovation group, and in 2002 he made the switch

back to extend his academic career at the Delft University of Technology,

where he started out as an assistant professor. In 2004 the PhD was set to

go and the result is what you have in front of you. Bas loves cycling, eighties

and nineties disco music and likes to draw coloring pages for his kids.

For my full CV:

http://nl.linkedin.com/in/basflipsen

Direct Methanol Fuel Cell systems in portable electronics A ...

Documents

Transcript of Direct Methanol Fuel Cell systems in portable electronics A ...