Studies in Engineering and Technology, Vol. 1, No. 1, February 2014
Transcript of Studies in Engineering and Technology, Vol. 1, No. 1, February 2014
Editorial Team
Editorial Assistant
Andrew Walter, Redfame Publishing, United States
Editorial Board Members
Abdulrahman F Almarshoud, College of Engineering, Qassim University, Saudi Arabia
Ahmet Can ALTUNIŞIK, Karadeniz Technical University, Civil Engineering Department, Turkey
Alexander Medvedev, Transport and Telecommunication Institute, Latvia
Alexander Pisarevskiy, Bauman Moscow State Technical University, Russian Federation
Alexander Russell, Otto von Guericke University of Magdeburg, Germany
Andrea Amicarelli, RSE, Italy
Giovanni Angrisani, Università degli Studi del Sannio, Italy
Girish Upreti, University of Tennessee, United States
Haiming Wen, University of California, Davis Northwestern University, United States
Hala Abd El Megeed, National Institute for Standards, Egypt
Halil KARAHAN, Civil Engineering Department, Pamukkale University, Turkey
Hassan Ibrahim Shaaban, Prof of Metallurgy, Atomic Energy Authority of Egypt, Egypt
Hossam Adel Zaqoot, Visiting assistant professor at Al-Azhar University-Gaza, Palestinian Territory, Occupied
Hossein Lavvafi, Case Western Reserve University, United States
Isaac Atuahene, University of Tennessee, United States
Kaan Yetilmezsoy, Yildiz Technical University, Turkey
KaiLong Hsiao, Taiwan Shoufu University, Taiwan, Province of China
Kalyanmoy Michigan State University, United States
Lei Kang, University of California at Berkeley, United States
Mahsa Seyyedian Choobi, University of Applied Science and Technology, Iran, Islamic Republic Of
Maibritt Pedersen Zari, Victoria University, New Zealand
Marco A Ruano, Economics Department Universidad Carlos III de Madrid, Spain
Martin Jaeger, Australian College of Kuwait, Kuwait
Pau Redon, Fundación Hospital General de Valencia, Spain
Paul Steskens, Belgian Building Research Institute, Belgium
Quamrul H. Mazumder, University of Michigan-Flint, United States
Roohollah Kalatehjari, Universiti Teknologi Malaysia 81310 Johor, Malaysia
Shahab Bahrami, University of British Columbia, Canada
Simona Rainis, Cirmont-International Research Center for Mountain, Italy
Sivasubramanian V, National Institute of Technology Calicut, India
Sonia Maria Gomez Puente, Eindhoven University of Technology (TU/e) the Netherlands, Netherlands
Sudip Chakraborty, Department of Chemical Engineering and Materials, University of Calabria, Italy
Tangming Yuan, University of York, United Kingdom
Wael A. Salah, Faculty of Engineering, Multimedia University Jalan Multimedia, Selangor, Malaysia
Xiuyu Gao, Wood Group Kenny, United States
Yao Liu, University Malaysia Pahang, Malaysia
Studies in Engineering and Technology
Vol. 1, No. 1; February 2014
ISSN 2330-2038 E-ISSN 2330-2046
Published by Redfame Publishing
URL: http://set.redfame.com
i
Contents
Sewer Networks Optimization Using Cellular Automata
Maryam Rohani, Mohammad Hadi Afshar 1-12
Finding a Common Weight Vector of Data Envelopment Analysis Based upon Bargaining Game
Manabu Sugiyama, Toshiyuki Sueyoshi 13-21
Examples of Mental Mistakes Made by Systems Engineers While Creating Tradeoff Studies
James Bohlman, A. Terry Bahill 22-43
The Valve Timing Optimization of the Diesel Engine Based on Response Surface Methodology
Jun Li, Lei Ji, Yangjiao Xu, Jinli Xie 44-49
Network-based Management on Repairing Tool Kits of Civil Aviation Engineering Maintenance
Xiaoxu Tian, Xinlei Zheng, Ting Wang, Na Li, Haifeng Wang, Fuqing Huang 50-55
Times Semi-Passive RFID Tags with Double Loop Antennas Arranged as a Shifted Gate Stability
Optimization
Ofer Aluf 56-73
Reviewer Acknowledgements
Andrew Walter 74
Studies in Engineering and Technology
Vol. 1, No. 1; February 2014
ISSN 2330-2038 E-ISSN 2330-2046
Published by Redfame Publishing
URL: http://set.redfame.com
1
Sewer Networks Optimization Using Cellular Automata
Maryam Rohani1, Mohammad Hadi Afshar
1
1School of Civil Engineering, Iran Univ. of Science and Tech., Tehran, Iran
Correspondence: Maryam Rohani, PhD Student, School of Civil Engineering, Iran Univ. of Science and Tech., Tehran,
Iran. Tel: 98-9123162480. E-mail: [email protected]
Received: October 15, 2013 Accepted: October 27, 2013 Available online: November 7, 2013
doi:10.11114/set.v1i1.237 URL: http://dx.doi.org/10.11114/set.v1i1.237
Abstract
The Hybrid Cellular Automata (HCA) method is used in this paper for the optimal design of sewer network problems
with the fixed layout. The HCA method decomposes the problem into two sub-problems with considering the pipe
diameters and nodal cover depths as decision variables. Two stages are solved iteratively for determining the decision
variables in a manner to minimize the total cost of the sewer network subject to the operational constraints. The HCA
method is used to optimally solve three benchmark examples with different sizes and the results are presented and
compared to those of the existing methods. The results indicate that the HCA method is more efficient and effective
than the alternative methods.
Keywords: Sewer Network, Optimization methods, Cellular Automata, design problem.
1. Introduction
Sewer network systems as a necessary urban infrastructure play an important role in the urban areas. The main objective
of optimal sewer network design problem is minimization of the capital investment on infrastructure whereas ensuring
good performance under specified design criteria. This topic has received considerable attention and different numerical
optimization approaches have been introduced and applied to the optimal design of sewer networks (Afshar, Shahidi,
Rohani, & Sargolzaei, 2011, Afshar & Rohani, 2012). These include the Linear Programming (LP), Non-Linear
Programming (NLP), Dynamic Programming (DP), and Evolutionary Algorithms.
There have been some attempts using the Linear Programming method to solve the problem of sewer network design,
such as Deininger (1970), Dajani and Gemmell (1971), Froise and Burges (1978), and Walters and Templeman (1979).
Gupta, Agarwal, & Khanna (1976), Lemieux, Zech, & Delarue (1976), and Price (1978) applied NLP and Swamee
(2001) used the Lagrange multiplier method to yield optimal sewer network design.
Among these methods, DP is the mostly used method for the optimal design of the sewer networks due to serial features
of these networks. Merrit and Bogan (1973), Mays and Wenzel (1976), Robinson and Labadie (1981), Yen, Cheng, Jun,
Voohees, & Wenzel (1984), and Kulkarni and Khanna (1985) applied DP for the optimal design of wastewater and/or
storm water networks. Although, DP methods are theoretically capable of finding the global optimum solution, but they
suffer from the curse of dimensionality limitation and therefore are not appropriate method for the large scale real-world
sewer networks.
Recently, Evolutionary approaches are being used for the problem due to their simplicity and flexibility for both
continuous and discrete problems without any assumption about the optimization objectives and good results have been
reported using these methods. Heaney, Wright, Sample, Field, & Fan (1999) had used Genetic Algorithm (GA) on
spreadsheet templates to get near-optimal solutions for the optimal design of sewer networks. Liang, Thompson, &
Young (2004) applied GA and Tabu Search (TS) algorithm for the optimal design of sewer networks. Afshar (2007)
proposed the sequential feature of solution construction in the Ant Colony Optimization Algorithm (ACOA) to develop
two partially constrained ACO algorithms for the solution of storm sewer network design problems. Pan and Kao (2009)
had integrated Quadratic Programming (QP) with GA to solve the sewer network optimization problem. The
applicability and efficiency of the GA-QP model were tested and the results indicated that the GA-QP model could
obtain various near optimum design alternatives within an acceptable computational time. Wang and Zhou (2009)
analyzed and compared the performances of GA, Particle Swarm Optimization (PSO) and Ant Colony Algorithms
(ACA) from three aspects of convergence, speed and complexity. It was shown that ACA is superior to other methods
Studies in Engineering and Technology Vol. 1, No. 1; 2014
2
for better convergence and satisfactory speed. Afshar (2010) applied the Continuous Ant Colony Optimization
Algorithm (CACOA) for the optimal design of sewer networks. Yeh, Chu, Chang, & Lin (2011) applied TS and
Simulated Annealing (SA) for the optimization of sewer network problems.
Since the basic theories of Evolutionary Algorithms, emulating the natural optimization process of evolution, are similar
to the natural evolutionary process, they need a big search space in spite of intelligent process, and so they involve high
computational costs and need a large number of iteration and computational efforts to achieve the optimum solutions.
Moreover, they require some free parameters that should be sensitive analyzed for obtaining the optimal solution.
Cellular Automata (CA), a novel optimization algorithm, has recently introduced and attracted much attention and has
been widely applied to some engineering problems.
In this paper, Hybrid CA (HCA) methods are used for the design of sewer network with the fixed layout. In the HCA
methods, the problem is decomposed in two stages solved iteratively to get the results. In the first stage, nodal cover
depths of the network are determined with the fixed values for pipe diameters using a CA method with the nodes
considered as the CA cells and nodal cover depths as cell states. In the second stage, the obtained nodal cover depths in
the first stage are used to calculate the pipe diameters with another CA method. In this stage, the pipes considered as the
CA cells and their diameters as the cell states. Two different updating rules, Continuous and Discrete approaches, are
used for CA updating rule of the second stage depending on the treatment of the pipe diameters. The CA updating rule is
derived by requiring that the network cost is minimized in the neighborhood of each cell. The HCA methods are used to
design three benchmark examples and comparison the results with the existing ones show the efficiency and
effectiveness of the methods to solve the sewer design optimization problems.
2. Cellular Automata
Cellular Automata (CA), a model of self-reproducing system, was conceived by Ulam (1960) and Von Neumann (1966)
and later completed and improved with the work of other researchers like Thatcher (1964), Codd (1968), and Burks
(1972).
CA has a set of identical elements, called cells with finite possible value called cell state. The new states of all cells are
defined simultaneously using an updating rule, which is a function of the previous state of the cell itself and its
neighborhoods.
CA had been used as a simulator in various fields such as computer science, (Wolfram 1988), chemistry (Packard, 1986),
and medical profession (Sentos & Coutinho, 2001). Recently, some research showed that CA can be used as a practical
and efficient optimization engine, which relies on the key properties of: locality of the neighborhoods interactions,
homogeneity of the evolving mechanism, parallelism of the computation, and simplicity of the model structure. CA has
been extensively used as a viable and efficient optimization algorithm for the structural design (Kita & Toyoda, 2000,
Missoum, Gürdal, & Setoodeh, 2005, and Setoodeh, Gürdalb, & Watson, 2006), estimating shortest path (Adamatzky
1996) or trip distribution problems (El Dessouki, Fathi, & Rouphail, 2001), and computer networks (Shuai & Zhao,
2004).
In the early applications in water resource problems, CA was used to produce good initial populations for a GA leading
to improved performance of the GA (Keedwell & khu, 2005, Guo, Keedwell, Walters, & Khu, 2007a). The first use of
CA as a stand-alone optimizer was demonstrated by Guo, Walters, Khu, & Keedwell (2007b) for optimal design of
storm sewer networks based on the simplifying assumption of known slopes. Afshar and Shahidi (2009) were the first to
propose CA with mathematically derived transition rules for the optimal water supply and hydropower operation of a
single reservoir. Later, Afshar et al. (2011) proposed a single stage CA for the optimal design of sewer networks with
fixed layout using the nodal excavation depth as the decision variables of the problem. Afshar and Rohani (2012)
extended the single stage CA method of Afshar et al. (2011) to two stage CA and proposed Hybrid CA, in which nodal
cover depths and pipes diameter were considered as decision variables.
3. Sewer Network Size Optimization
Sewer network system is one of the urban infrastructure systems with huge construction and operation cost and any
attempt to reduce these costs result in considerable saving. A sewer network is an underground system built to convey
waste water to one or more collection points (outfalls).
Optimal sewer network design with a fixed layout aims to find a cost-effective solution by determining the pipes
diameters and slopes which minimizes the capital investment whilst ensuring a good system performance under specific
design criteria. The problem of sewer network design for a fixed network layout, in the absence of pumps and drops,
can be formulated as:
Min Cnetwork =∑Cpl+∑Cmi=∑LlKp(Dl,Xi,Xj)+∑Km(hmk) (1)
Studies in Engineering and Technology Vol. 1, No. 1; 2014
3
Subject to:
Vmin≤ Vl≤ Vmax l=1,...,NL (2)
βmin≤ βl≤ βmax l=1,...,NL (3)
Smin≤ Sl≤ Smax l=1,...,NL (4)
Xmin≤ Xk≤ Xmax k=1,...,NN (5)
Dlє D l=1,...,NL (6)
Dl≤ Dl' l=1,...,NL (7)
Where, Cnetwork is the total cost of the network, Cpl is the installation cost of lth
pipe, Cmi is the cost of ith
manhole, NL is
the number of pipes in the network, NN is the number of nodes in the network, Ll is the length of lth
pipe, Kp is the unit
cost of lth
pipe defined as a function of its diameter (Dl) and upstream and downstream nodal cover depth (Xi,Xj), and
Km is the cost of manhole construction as a function of manhole depth (hm).
Equations (2) to (7) represents the constraints of velocity, water-depth ratio, pipe slope, nodal cover depth,
commercially available pipe diameter, and progressive diameter for the sewer network problem, respectively, where, Vl
is the velocity of lth
pipe, βl = yl/Dl, yl is the flow depth of lth
pipe, Sl is the slope of lth
pipe, Xk is the cover depth of kth
node, D is the set of commercially available pipe diameters, l' refers to the set of pipe located downstream of pipe l, and
min, max are the allowable minimum and maximum parameters, respectively.
4. Hybrid Cellular Automata (HCA)
Application of CA to any problem requires that four basic components of the CA method, cell, cell state, neighborhood,
and updating rule, are properly defined. In this paper, pipe diameters and nodal cover depths are chosen as the decision
variables.
The Hybrid Cellular Automata (HCA) formulation requires decomposing the problem into two sub-optimization
problems which are solved iteratively in two stage manners. In the first stage, each node of the sewer network is
regarded as a cell and nodal cover depths are considered as the cell states, which are determined with the fixed values of
pipe diameters. In the second stage, the pipe diameters are calculated by solving a second nonlinear sub-optimization
problem with considering the calculated nodal cover depths from the first stage as fixed values. Two different updating
rules, continuous and discrete, are derived and used depending on the treatment of pipe diameters. In the continuous
approach, the pipe diameters are considered as continuous variables and the corresponding updating rule is derived
mathematically from the original objective function of the problem and followed by a rounding process in which the
continuous pipe diameters calculated are rounded, if required, to find the discrete optimal pipe diameters, while in the
discrete approach, an ad-hoc updating rule is derived based on the discrete nature of pipe diameters. The described two
stage process is iterated until convergence is achieved.
Considering the nodal cover depths, Xk;k=1,…,NN, and pipe diameters, Dl;l=1,…,NL, as the decision variables, these
constraints can be easily applied as box constraints. Using a penalty method for satisfaction of remaining constraints,
the total penalized objective function of the sewer network optimization problem can be defined as follows:
Min C=∑Cpl+∑Cmi+∑(αCSVvl+ αCSVsl+ αCSVβl) (8)
Where, α is the penalty parameters with large enough positive value, and CSVv, CSVs, CSVβ represent the violation
from the constraints of velocity, slope, and water-depth ratio for each pipe, respectively, CSVvl=(1-Vl/Vmin)2+(Vl/Vmax-1)
2,
CSVβl=(1-βl/βmin)2+(βl/βmax-1)
2, CSVsl=(1-sl/smin)
2+(sl/smax-1)
2.
Subject to:
Xmin≤ Xk≤ Xmax k=1,...,NN (9)
Dlє D l=1,...,NL (10)
The process is started with arbitrary sets of pipe diameters and nodal cover depths satisfying the constraints of (9) and
(10).
4.1 First Stage
In the first stage, the nodal cover depths are calculated with the minimization of the following local objective function
over the cell neighborhood (k ) considering the fixed values for the pipe diameters:
Ck=∑(Cpl+αCSVvl+ αCSVsl+ αCSVβl)+Cmk (11)
Studies in Engineering and Technology Vol. 1, No. 1; 2014
4
Subject to the box constraints of Xmin≤ Xk≤ Xmax.
Minimization of the local objective function of Equation (11) with respect to the nodal cover depth (Xk) leads to the
nonlinear equations to be solved with the Newton-Raphson method which results in the updated nodal cover depths
defined as:
F(Xk)=∂Ck/∂Xk=0 ∆Xk=-Fk/(∂F/∂X)k│kk
∆Xk=∆Xkkk+1
-∆Xkkk
(12)
Where, Fkkk
=F(Xkkk
), kk is the nonlinear iteration index, and ∆Xk is the change in the value of the cell state. Fk and
(∂F/∂X)k are both implicit functions of the Xk which can be calculated using the chain rule of differentiation and
Manning equation. This procedure is repeated for the cell under consideration until the convergence is met and the
process of updating is repeated for all cells of the network at the end of which the first stage is terminated.
4.2 Second Stage
In the second stage, the values of nodal cover depths obtained in the first stage are used to get the optimal pipe
diameters. Two CA approaches are applied to solve this problem, in which the pipes and corresponding diameters are
considered as the CA cell and cell state, respectively. The neighborhood of the CA is simply considered as the cell itself
without any neighboring cells.
4.2.1 Discrete Approach
In discrete method, the pipe diameters are treated as discrete values leading to an ad-hoc CA updating rule derived
based on engineering judgment. The diameter of each pipe is changed such that the pipe cost is minimized in a manner
to satisfy the constraints of velocity and water depth ratio. The following three engineering based ad-hoc updating rule
is used to update the cell state:
1) If constraints of velocity and water-depth ratio are all satisfied, pipe diameter is decreased to the lower diameter
available to minimize the objective function.
2) If one or both of the maximum velocity and water-depth ratio constraints are violated, the pipe diameter is increased
to the upper diameter available.
3) If one or both of the minimum velocity and water-depth ratio constraints are violated, the pipe diameter is decreased
to the lower diameter available.
4.2.2 Continuous Approach
In continuous method, the updating rule is derived mathematically assuming pipe diameter as continuous variable
followed by a rounding procedure to convert the continuous solutions to discrete available diameters such that the
following localized objective function is minimized:
Minimize Cl=Cpl+αCSVvl+αCSVβl+Cmi+Cmj (13)
Subject to the box constraints of:
Dmin≤ Dl≤ Dmax (14)
Where i and j refer to the upstream and downstream node of lth
pipe, Dmin and Dmax represent the minimum and
maximum components of available commercial diameters, respectively. The updating rule is derived mathematically by
requiring that the objective function of Equation (13) is stationary with respect to the cell state with applying the
Newton-Raphson linearization:
Gl=∂Cl/∂Dl=0 ∆Dl=-Gl/(∂G/∂D)l ∆Dl=∆Dlkk+1
-∆Dlkk
(15)
This procedure is repeated for the cell under consideration until the convergence is met. Once the process of updating of
all cell states are completed, the continuous diameters so calculated are rounded to the upper discrete diameters
available before returning to the first stage. More details of the method can be found in the work of Afshar and Rohani
(2012).
Studies in Engineering and Technology Vol. 1, No. 1; 2014
5
Figure 1. Schematic description of the HCA method.
A schematic description of the HCA method is illustrated in Figure 1 for both discrete and continuous versions in which
the cells, cell states, neighborhood and the updating rules are briefly described for more clarification.
5. Test Examples
The performance of the HCA method is investigated in this section by applying the model to three design problems with
different sizes in the literature. These hypothetical test examples were previously proposed and used by Moeini and
Afshar (2012 a,b) for the simultaneous layout and size optimization of sewer network using ACOA based methods.
First Stage:
Cells: Network nodes
Cell state: nodal cover depths
Neighborhood: Neighboring pipes
i=1
kk=1 kk=kk+1
i=i+1
If k=NN
Yes
Update the nodal cover
depths, Xk, using Eq. 12
∆Xk<ε No
Yes
No
Second Stage
Cells: Network pipes
Cell state: Diameter
Neighborhood: Null
l=1 l=l+1
Update the pipe diameter,
Dl, using steps 1-3 in
section 4.2.1
Discrete CA Continuous CA
Update the pipe diameter, Dl
∆Dl<ε
Finish
Convergence is met?
Yes
If l=NL
kk=kk+1
Yes
No
Yes
kk=1
No
No
Start with random nodal cover depths, Xk;k=1,…,NN,
and pipe diameters, Dl;l=1,…,NL
Studies in Engineering and Technology Vol. 1, No. 1; 2014
6
Figure 2. Network layouts for three examples: a) small scale sewer network, b) medium scale sewer network, c) large
scale sewer network.
The optimal layouts obtained by Moeini and Afshar (2012 b), shown in Figure 2, is used here to assess the efficiency
and effectiveness of the CA and HCA method. The small scale sewer network consists of 9 nodes and 12 pipes, the
medium scale sewer network has 25 nodes and 40 pipes and the large scale sewer network includes 81 nodes and 144
pipes. All the networks have two treatment plants with fixed elevation located at the bottom corner of the sewer network.
(c)
[144] [143] [142]
67
[33]
17
[16] [9] [10] [11] [12] [15] [17]
[32]
[49]
24
[136]
[137] 73 74 [138] 75 [139] 76 [140] [141]
[132]
77
[133]
78 79
[134] [135]
80 81
72 71 [127] [126] 70
[128] [129] [130] [131]
69 [125] 68 [124] [123] 66 [122] 65 [121] 64 [120]
[117] [118] [119] [116] [115] [114] [113] [112] [111]
[103] 55
0
56 [104] 57 [105] 58 [106] 59 [107] 60 [108] 61 [109] 62 [110] 63
[102]
54 [93] 53
[101] [100]
[92] 52
[94] [95] [96] [97] [98] [99]
46 [86] 47 [87] 48 [88] 49 [89] 50 [90] 51 [91]
[85]
[76]
[84] [83] [82] [81] [80] [79] [78] [77]
37 [69] 38 [70] 39 [71] 40 [72] 41 [73] 42 [74] 43 [75] 44 45
[68] [67] [66]
W.T.P
8 [8] 9 7 [7] 6
[23] [24]
[65] [64] [60]
28 [52]
[61] [62] [63]
29 [53] 30 [54] 31
[55] 32 [56] 33
[57] 34 [58] 35 [59] 36
26
[50]
[42] 27
[51] [47] [48]
25 [41] [40] 23 [39]
[43] [44] [45] [46]
[38] 22 [37] 21 19 [35] 20 [36]
[26]
10 [18]
[27] [28] [29] [30]
[25] 18
[34]
16 15
[31]
14 [22] 12 [20] 11 [19]
[6]
[14] [13]
13 [21]
1 2 3 [2] [1] [3] [5] 5 4 [4]
(a)
W.T.P W.T.P
[2] [1]
[3] [4] [5]
[7] [6]
[12] [11]
[8] [9] [10]
1
2
3
4 6 5
7 9 8
(b) W.T.P W.T.P
[2] [1] [4] [3]
[5] [6] [7] [8] [9]
[11] [10] [13] [12]
[20] [19] [22] [21]
[29] [28] [31] [30]
[14] [15] [16] [17] [18]
[23] [24] [25] [26] [27]
[32] [33] [34] [35] [36]
1 3 2 4 5
6 8 7 9 10
11 13 12 14 15
16 18 19 20
21 23 22 24 25
17
[40] [39] [38] [37]
W.T.P
Studies in Engineering and Technology Vol. 1, No. 1; 2014
7
The ground elevation of a reference point located at the middle of the upper edge of the area denoted by node 8, 23 and
77 in small, medium and large scale networks, respectively, is considered 1000 metres. The area is considered to have a
constant slope of 2% from the reference point to the left and right and toward the bottom edge. The lengths of pipes in
three networks are considered as 100 meters and the set of commercially available pipe diameters for all the pipes is
included in the range of 100 mm to 1500 mm with an interval of 50 mm from 100 mm to 1000 mm and an interval of
100 mm from 1000 mm to 1500 mm. The Manning coefficient is considered as 0.015 and the problem is constrained to
a maximum and minimum velocity of 6 m/s and 0.75 m/s, maximum and minimum cover depth of 10 m and 2.5 m, and
maximum and minimum relative flow depth of 0.83 and 0.1, respectively.
The terms of pipe and manhole construction costs are defined as (Moeini and Afshar, 2012 a,b):
Kp=10.93e3.43D
+0.012X1.53
+0.437X1.47
D (16)
Kh=41.46hm
Where, D is the pipe diameter (m), X is the buried depth (m), and hm is the depth of manhole (m).
These test problems are here solved using CA method of Afshar et al. (2011) and HCA methods and the results are
presented and compared with other existing methods.
Table 1. Optimal network cost obtained by different methods
Test Example Model Cost Time (milli second)
small scale network
ACOA-TGA
(Moeini and Afshar, 2012 a) 23467 -
CACOA-TGA
(Moeini and Afshar, 2012 b) 23467 -
CA 23811 1.6
HCA-Discrete 23460 4.7
HCA-Continuous 23513 3.1
medium scale network
ACOA-TGA
(Moeini and Afshar, 2012 a) 86204 -
CACOA-TGA
(Moeini and Afshar, 2012 b) 85990 -
CA 88096 14.1
HCA-Discrete 85873 46.9
HCA- Continuous 86678 15.6
large scale network
ACOA-TGA
(Moeini and Afshar, 2012 a) 365600 -
CACOA-TGA
(Moeini and Afshar, 2012 b) 363922 -
CA 370486 103.1
HCA-Discrete 361685 200.0
HCA- Continuous 367436 54.6
Table 1 compares the optimal network costs and the CPU time required by the CA methods introduced here and those of
ACOA-TGA (Moeini and Afshar, 2012 a) and CACOA-TGA (Moeini and Afshar, 2012 b) methods using a 2 MHz
Pentium 4. It can be seen that both HCA methods produce better solutions than CA method in three sewer networks.
Moreover, the HCA methods result in comparable solutions to the ACOA in small scale network, while with increasing
the scale of the problem, the HCA methods produce superior solutions than ACOA. Furthermore, the HCA-discrete
method results in better solution than the existing methods while requiring less than one second CPU time to achieve
the solutions. It should be noted that ACOA needs much more time than CA methods, because of its mechanism and it is
one the population based methods. Since the HCA method requires an initial guess for the decision variables of the
problem, pipe diameters and the nodal cover depths, to start off the solution procedure, a sensitivity analysis is carried
out here to assess the sensitivity of the final solution to the initial guess.
Studies in Engineering and Technology Vol. 1, No. 1; 2014
8
Table 2. Maximum, Minimum and Average solution costs over 10 runs
Cost
Test Example Model Minimum Maximum Average Scaled Standard
Deviation
small scale network
ACOA-TGA
(Moeini and Afshar, 2012 a) 23467 23467 23467 0.0000
CACOA-TGA
(Moeini and Afshar, 2012 b) 23467 23467 23467 0.0000
CA 23811 23811 23811 0.0000
HCA-Discrete 23460 23747 23546 0.0059
HCA-Continuous 23513 34064 25798 0.1683
medium scale network
ACOA-TGA
(Moeini and Afshar, 2012 a) 86204 87127 86642 0.0037
CACOA-TGA
(Moeini and Afshar, 2012 b) 85990 86591 86187 0.0020
CA 88096 88096 88096 0.0000
HCA-Discrete 85873 86953 86410 0.0052
HCA- Continuous 86678 87786 87397 0.0038
large scale network
ACOA-TGA
(Moeini and Afshar, 2012 a) 365600 381484 372605 0.0127
CACOA-TGA
(Moeini and Afshar, 2012 b) 363922 367174 365606 0.0030
CA 370486 370489 370488 0.0000
HCA-Discrete 361685 367131 363894 0.0040
HCA- Continuous 367436 371350 369661 0.0031
Table 2 represents the maximum, minimum and average solution costs over 10 runs using different initial designs along
with the scaled standard deviation of the solutions defined as the ratio of the standard deviation to the average solution.
This table emphasizes on the insensitivity of the CA methods to the initial population.
Table 3. Results obtained from HCA method (discrete) for the small scale sewer network
Cover Depth (m)
Pipe D (mm) Upstream Downstream V (m/s) y/d
1 100 2.50 4.50 1.39 0.57
2 150 2.50 4.50 1.83 0.58
3 200 3.40 4.50 2.04 0.67
4 150 2.50 2.50 1.27 0.56
5 200 2.50 4.50 2.08 0.50
6 150 2.50 3.40 1.67 0.82
7 150 2.50 2.50 1.27 0.56
8 150 2.50 3.40 1.47 0.50
9 150 2.50 2.50 1.27 0.56
10 150 2.50 2.50 1.27 0.56
11 100 2.50 2.50 1.05 0.73
12 100 2.50 2.50 1.05 0.73
Studies in Engineering and Technology Vol. 1, No. 1; 2014
9
Table 4. Results obtained from HCA method (discrete) for the medium scale sewer network
Cover Depth (m)
Cover Depth (m)
Pipe D(mm) Upstream Downstream V (m/s) y/d Pipe D(mm) Upstream Downstream V (m/s) y/d
1 200 2.50 4.50 2.18 0.55 21 150 2.50 2.50 1.27 0.56
2 150 2.50 2.50 1.38 0.74 22 150 2.50 2.50 1.27 0.56
3 100 2.50 2.50 1.05 0.73 23 200 2.50 2.50 1.60 0.62
4 150 2.50 4.50 1.94 0.71 24 200 2.50 2.80 1.79 0.75
5 350 3.45 4.50 2.92 0.65 25 150 2.50 3.40 1.67 0.82
6 150 2.50 2.50 1.27 0.56 26 200 2.50 2.50 1.68 0.80
7 150 2.50 2.50 1.27 0.56 27 200 2.50 2.50 1.60 0.62
8 150 2.50 2.50 1.27 0.56 28 150 2.50 2.50 1.27 0.56
9 300 3.17 4.50 2.82 0.73 29 150 2.50 2.50 1.27 0.56
10 300 3.47 3.45 2.19 0.82 30 150 2.50 2.50 1.27 0.56
11 150 2.50 3.47 1.69 0.81 31 150 2.50 2.50 1.27 0.56
12 150 2.50 2.50 1.27 0.56 32 150 2.50 2.50 1.27 0.56
13 250 2.50 3.17 2.26 0.82 33 150 2.50 2.50 1.38 0.74
14 200 2.50 3.45 2.03 0.76 34 150 2.50 2.50 1.27 0.56
15 250 2.80 3.47 2.26 0.82 35 150 2.50 2.50 1.38 0.74
16 150 2.50 2.50 1.27 0.56 36 150 2.50 2.50 1.27 0.56
17 250 2.50 2.50 1.92 0.71 37 100 2.50 2.50 1.05 0.73
18 200 2.50 3.17 1.94 0.79 38 100 2.50 2.50 1.05 0.73
19 150 2.50 2.50 1.27 0.56 39 100 2.50 2.50 1.05 0.73
20 200 3.40 2.80 1.41 0.82 40 100 2.50 2.50 1.05 0.73
Table 5. Results obtained from HCA method (discrete) for the large scale sewer network
Cover Depth (m)
Cover Depth (m)
Pipe D(mm) Upstream Downstream V (m/s) y/d Pipe D(mm) Upstream Downstream V (m/s) y/d
1 350 4.14 4.50 2.59 0.68 73 200 2.50 2.50 1.52 0.53
2 300 3.34 4.14 2.61 0.82
74 200 2.50 3.00 1.88 0.82
3 200 2.50 3.34 1.90 0.62 75 400 3.00 3.17 2.77 0.76
4 150 2.50 2.50 1.38 0.74 76 150 2.50 2.50 1.27 0.56
5 100 2.50 2.50 1.05 0.73
77 250 3.00 2.50 1.69 0.80
6 200 2.50 2.50 1.52 0.53 78 400 4.02 4.37 2.84 0.70
7 200 2.50 2.50 1.68 0.80 79 150 2.50 2.50 1.27 0.56
8 200 2.50 4.50 2.38 0.81 80 150 2.50 2.50 1.27 0.56
9 500 3.93 4.50 3.52 0.82 81 150 2.50 2.50 1.27 0.56
10 150 2.50 4.14 1.60 0.47 82 150 2.50 2.50 1.27 0.56
11 300 3.88 3.34 1.88 0.82 83 350 2.71 3.00 2.62 0.82
12 150 2.50 2.50 1.27 0.56 84 250 2.50 3.17 2.26 0.82
13 150 2.50 2.50 1.27 0.56 85 250 3.00 2.50 1.69 0.80
14 150 2.50 2.50 1.27 0.56 86 150 2.50 3.00 1.39 0.52
15 150 2.50 2.50 1.27 0.56 87 300 2.50 4.02 2.93 0.82
16 150 2.50 2.50 1.27 0.56 88 200 2.50 2.50 1.65 0.70
17 550 3.78 4.50 3.86 0.82 89 150 2.50 2.50 1.27 0.56
18 200 2.50 3.93 2.05 0.58 90 200 2.50 3.00 1.88 0.82
19 150 2.50 2.50 1.27 0.56 91 250 3.00 2.71 1.81 0.82
20 200 2.50 3.88 2.04 0.58 92 150 2.50 2.50 1.27 0.56
21 150 2.50 2.50 1.27 0.56 93 150 2.50 3.00 1.39 0.52
22 150 2.50 3.40 1.67 0.82 94 200 2.50 3.00 1.88 0.82
23 200 3.40 3.89 1.88 0.82 95 250 2.50 4.02 2.41 0.58
24 250 3.89 3.60 1.81 0.82 96 300 2.50 2.50 2.20 0.78
25 550 3.60 3.78 3.43 0.76 97 150 2.50 2.50 1.27 0.56
26 500 3.71 3.93 3.26 0.82 98 200 2.50 2.50 1.65 0.70
27 150 2.50 2.50 1.27 0.56 99 150 2.50 3.00 1.39 0.52
28 200 2.50 3.88 1.77 0.78 100 300 2.5 2.71 2.30 0.74
29 150 2.50 2.50 1.27 0.56 101 250 2.50 2.50 1.92 0.71
Studies in Engineering and Technology Vol. 1, No. 1; 2014
10
30 150 2.50 2.50 1.27 0.56 102 200 2.50 3.00 1.88 0.82
31 150 2.50 3.40 1.47 0.50 103 150 2.50 2.50 1.27 0.56
32 150 2.50 3.89 1.56 0.48 104 150 2.50 2.50 1.27 0.56
33 550 4.00 3.60 2.95 0.77 105 200 2.50 2.50 1.65 0.70
34 300 2.50 3.78 2.71 0.64 106 150 2.50 2.50 1.27 0.56
35 500 4.32 3.71 2.59 0.82 107 150 2.50 2.50 1.27 0.56
36 150 2.50 4.32 1.63 0.46 108 250 2.50 2.50 1.92 0.71
37 200 2.50 2.50 1.65 0.70 109 150 2.50 2.50 1.27 0.56
38 150 2.50 2.50 1.27 0.56 110 150 2.50 2.50 1.27 0.56
39 200 2.50 2.50 1.52 0.53 111 200 2.50 2.50 1.60 0.62
40 200 2.50 3.00 1.88 0.82 112 200 2.50 2.50 1.68 0.80
41 250 3.00 4.00 2.31 0.65 113 250 2.50 2.50 1.94 0.76
42 150 2.50 2.50 1.27 0.56 114 150 2.50 2.50 1.27 0.56
43 300 2.50 3.71 2.60 0.58 115 200 2.50 2.50 1.52 0.53
44 500 5.08 4.32 2.44 0.82 116 200 2.50 2.50 1.68 0.80
45 150 2.50 2.50 1.27 0.56 117 200 2.50 2.50 1.68 0.80
46 150 2.50 2.50 1.27 0.56 118 200 2.50 2.50 1.68 0.80
47 150 2.50 2.50 1.27 0.56
119 200 2.50 2.50 1.60 0.62
48 150 2.50 2.50 1.27 0.56 120 150 2.50 2.50 1.27 0.56
49 150 2.50 3.00 1.39 0.52 121 150 2.50 2.50 1.27 0.56
50 450 3.09 4.00 3.48 0.82 122 200 2.50 2.50 1.68 0.80
51 300 2.83 2.50 2.00 0.73 123 150 2.50 2.50 1.27 0.56
52 150 2.50 2.50 1.27 0.56 124 150 2.50 2.50 1.27 0.56
53 200 2.50 5.08 2.54 0.76 125 150 2.50 2.50 1.27 0.56
54 200 2.50 2.50 1.65 0.70 126 150 2.50 2.50 1.27 0.56
55 150 2.50 2.50 1.27 0.56 127 150 2.50 2.50 1.27 0.56
56 150 2.50 3.40 1.67 0.82
128 150 2.50 2.50 1.27 0.56
57 250 3.40 2.50 1.42 0.70 129 150 2.50 2.50 1.38 0.74
58 250 2.50 3.09 2.17 0.68 130 150 2.50 2.50 1.38 0.74
59 150 2.50 2.83 1.35 0.53 131 150 2.50 2.50 1.38 0.74
60 300 2.50 2.50 2.07 0.60 132 150 2.50 2.50 1.27 0.56
61 400 4.37 5.08 3.11 0.82 133 150 2.50 2.50 1.38 0.74
62 150 2.50 2.50 1.27 0.56 134 150 2.50 2.50 1.38 0.74
63 150 2.50 2.50 1.27 0.56 135 150 2.50 2.50 1.38 0.74
64 150 2.50 2.50 1.27 0.56 136 150 2.50 2.50 1.27 0.56
65 150 2.50 3.40 1.47 0.50 137 100 2.50 2.50 1.05 0.73
66 150 2.50 2.50 1.27 0.56 138 100 2.50 2.50 1.05 0.73
67 450 3.17 3.09 2.83 0.82 139 100 2.50 2.50 1.05 0.73
68 250 2.50 2.83 2.11 0.82 140 100 2.50 2.50 1.05 0.73
69 150 2.50 2.50 1.27 0.56 141 100 2.50 2.50 1.05 0.73
70 200 2.50 4.37 2.34 0.82 142 100 2.50 2.50 1.05 0.73
71 200 2.50 2.50 1.65 0.70 143 100 2.50 2.50 1.05 0.73
72 150 2.50 2.50 1.27 0.56 144 100 2.50 2.50 1.05 0.73
Details of the optimal solution obtained by the HCA-discrete method for three sewer networks are also shown in Table
3, 4, and 5, respectively.
6. Concluding Remarks
In this paper, Hybrid Cellular Automata approach was used for the optimal solution of sewer network design problems.
The problem was decomposed to two stages with considering the nodal cover depths and pipe diameters as decision
variables. In the first stage, nodal cover depths were calculated assuming fixed values for the pipe diameters, while in
the second stage, the pipe diameters were determined in two approaches of continuous and discrete with nodal cover
depths of the first stage. Two stages were solved iteratively until the convergence was achieved. The HCA methods
were used to solve three benchmark examples in the literature and the comparison of results with two versions of Ant
Colony Optimization Algorithm indicated the ability and efficiency of the HCA methods to produce better results
comparable to those of heuristic search methods with much higher efficiency.
Studies in Engineering and Technology Vol. 1, No. 1; 2014
11
References
Adamatzky, A. I. (1996). Computation of shortest path in cellular automata. Mathematical and Computer Modelling,
23(4), 105-113.
Afshar, M. H. (2007). Partially Constrained Ant Colony Optimization Algorithm for the solution of constrained
optimization problems: application to storm water network design. Adv Water Resour, 30, 954–965.
Afshar, M. H., & Shahidi, M. (2009). Optimal solution of large scale reservoir operation problems: Cellular Automata
versus heuristic-search methods. Engineering Optimization, 41(3), 275-293.
Afshar, M. H. (2010). A parameter free Continuous Ant Colony Optimization Algorithm for the optimal design of storm
sewer networks: Constrained and unconstrained approach. Advances in Engineering Software, 41(2), 188–195.
Afshar, M. H., Shahidi, M., Rohani, M., & Sargolzaei, M. (2011). Application of cellular automata to sewer network
optimization problems. Scientia Iranica A, 18(3), 304–312.
Afshar, M. H., & Rohani, M. (2012). Optimal design of sewer networks using cellular automata-based hybrid methods:
Discrete and continuous approaches. Engineering Optimization, 44(1), 1-22.
Burks, E. (1972). Essays on Cellular Automat, University of Illinois Press, Champaign, IL.
Codd, E F. (1968). Cellular Automata, Academic Press, New York.
Dajani, J. S., & Gemmel, R. S. (1971). Economics of wastewater collection networks, Civil Eng. Res. Rep. 43, Water
Resour. Center, Univ. of Ill. At Urbana-Champaign, Urbana.
Deininger, R. A. (1970). Systems analysis for water supply and pollution control. Natural Resource Systems Models in
decision Making, edited by G.H. Toebes, Water Resources Center, Purdue Univ. Lafayette, Ind., 45-65.
El Dessouki, W. M., Fathi, Y., & Rouphail, N. (2001). Meta-optimization using cellular automata with application to the
combined trip distribution and assignment system optimal problem. Computer-Aided Civil and Infrastructure
Engineering, 16(6), 384-396.
Froise, S., & Burges, S. J. (1978). Least-cost design of urban drainage networks. Water Resources Planning and
Management Division, ASCE, 104(1), 75-92.
Guo, Y., Walters, G. A., Khu, S. T., & Keedwell, E. (2007a). A novel cellular automata based approach to storm sewer
design. Engineering Optimization, 39(3), 345-364.
Guo, Y., Keedwell, E. C., Walters, G. A., & Khu, S. T. (2007b). Hybridizing Cellular Automata Principles and NSGAII
for Multi-objective Design of Urban Water Networks. Evolutionary Multi-Criterion Optimization, 546-559.
Gupta, J. M., Agarwal, S. K., & Khanna, P. (1976). Optimal design of wastewater collection systems. The
Environmental Engineering Division, 102(EE5), 1029–1041.
Heaney, J. P., Wright, L. T., Sample, D., Field, R., & Fan, C. Y. (1999). Innovative methods for the optimization of
gravity storm sewer design. Proceedings the 8th international conference on urban storm drainage, Sydney,
Australia, 1896–1903.
Keedwell, E., & Khu, S. T. (2005). Using cellular automata to seed genetic algorithms for water distribution network
design problems. Engineering Applications of Artificial Intelligence, 18(4), 461-472.
Kita, E., & Toyoda, T. (2000). Structural design using cellular automata. Structural and Multidisciplinary Optimization,
19, 64–73.
Kulkarni, V. S., & Khanna, P. (1985). Pumped wastewater collection systems optimization. Environ Eng, ASCE, 111(5),
589–601.
Lemieux, P. F., Zech, Y., & Delarue, R. (1976). Design of a stormwater sewer by nonlinear programming. Canadian
Journal of Civil Engineering, 3(1), 83-89.
Liang, L.Y., Thompson, R. G., & Young, D.M. (2004). Optimizing the design of sewer networks using genetic
algorithms and tabu search. Engineering, Construction and Architectural Management, 11(2), 101-112.
Mays, L. W., & Wenzel, H. G. (1976). Optimal design of multi-level branching sewer systems. Water Resources
Research, 12(5), 913-917.
Merrit, L. B., & Bogan, R. H. (1973). Computerbase Optimal Design of Sewer System. The Environmental Engineering
Division, ASCE, 99(EE1), 35-53.
Missoum, S., Gürdal, Z., & Setoodeh, S. (2005). Study of a new local update scheme for cellular automata in structural
design. Struct. Multidisciplinary Optimization, 29, 103–112.
Studies in Engineering and Technology Vol. 1, No. 1; 2014
12
Moeini, R., & Afshar, M. H. (2012a). Layout and size optimization of sanitary sewer network using intelligent ants.
Advances in Engineering Software, 51, 49–61.
Moeini, R., & Afshar, M. H. (2012b). Constrained Ant Colony Optimisation Algorithm for the layout and size
optimisation of sanitary sewer networks. Urban Water Journal, 1–20.
Packard, N. H. (1986). Lattice Models for Solidification and Aggregation. In First International Symosium for Science
on Form.
Pan, T. C., & Kao, J. J. (2009). GA-QP model to optimize sewer system design. Environmental Engineering, 135(1),
17–24.
Price, R. K. (1978). Design of storm water sewers for minimum construction cost. In Proc. 1st International Conference
on Urban Strom Drainage, Southampton, United Kingdom, 636-647.
Robinson, D. K., & Labadie, J. W. (1981). Optimal design of urban storm water drainage system. International
symposium on urban hydrology, hydraulics, and sediment control, University of Kentucky Lexington, 145–56.
Sentos, R. M. Z. D., & Coutinho S. (2001). Dynamics of HIV Approach: A Cellular Automata Approach. Phys. Rev.
Lett., 87(16), 102-104.
Setoodeh, S., Gürdalb, Z., & Watson, L. T. (2006). Design of variable-stiffness composite layers using cellular automata.
Computer Methods in Applied Mechanics and Engineering, 195(9-12), 836-851.
Shuai, D., & Zhao, H. (2004). A new generalized cellular automata approach to optimization of fast packet switching.
Computer Networks, 45, 399-419.
Swamee, P. K. (2001). Design of sewer line. Environmental Engineering, 127(9), 776–781.
Thatcher, J. (1964). Universality in the von Neumann cellular model. Technical Report 03105-30-T, University of
Michigan.
Ulam, S. (1960). A collection of mathematical problems. NewYork, Interscience publishers.
Von Neumann, J. (1966). Theory of self-reproducing automata. Champaign, IL, University of Illinois Press.
Walters, G. A., & Templeman, A. B. (1979). Non-optimal dynamic programming algorithms in the design of minimum
cost drainage systems. Engineering Optimization, 4, 139-148.
Wang, L., Zhou, Y. (2009). Comparative Study on Bionic Optimization Algorithms for Sewer Optimal Design. 2009
Fifth International Conference on Natural Computation, 24-29.
Wolfram, S. (1988). High Speed Computing: Scientific Application and Algorithm Design. University of Illinois Press.
Yeh, S. F., Chu, C. W., Chang, Y. J., & Lin, M. D. (2011). Applying tabu search and simulated annealing to the optimal
design of sewer networks. Engineering Optimization, 43(2), 159–174.
Yen, B. C., Cheng, S. T., Jun, B. H., Voohees, M. L., & Wenzel, H. G. (1984). Illinois least cost sewer system design
model. User’s guide, Department of Civil Engineering, University of Texas at Austin.
This work is licensed under a Creative Commons Attribution 3.0 License.
Studies in Engineering and Technology
Vol. 1, No. 1; February 2014
ISSN 2330-2038 E-ISSN 2330-2046
Published by Redfame Publishing
URL: http://set.redfame.com
13
Finding a Common Weight Vector of Data Envelopment Analysis Based
upon Bargaining Game
Manabu Sugiyama1 & Toshiyuki Sueyoshi
2
1Faculty of Social and Information Studies, Gunma University, Maebashi-City, Gunma, Japan
2Department of Management, New Mexico Institute of Mining & Technology, Socorro, NM, USA
Correspondence: Manabu Sugiyama, Faculty of Social and Information Studies, Gunma University, 4-2 Aramaki-machi,
Maebashi-City, Gunma 371-8510, Japan. Tel: 81-27-220-7522. E-mail: [email protected]
Received: October 20, 2013 Accepted: November 6, 2013 Available online: November 11, 2013
doi:10.11114/set.v1i1.277 URL: http://dx.doi.org/10.11114/set.v1i1.277
Abstract
Data Envelopment Analysis (DEA) is a mathematical programming method for measuring the relative efficiency of
Decision Making Units (DMUs) by evaluating their outputs and inputs. In the history of DEA, the cross-efficiency of
jth DMU is widely used as an efficiency measure of a given DMUo among researchers. The approach always utilizes
weights related to inputs and outputs in the assessment. Unfortunately, the weights are not always uniquely determined
in the cross-efficiency measurement because DEA always suffers from an occurrence of multiple solutions, so
indicating an occurrence of multiple weights. To overcome such a difficulty, this paper proposes a new approach for
determining a common weight vector of DEA based on bargaining game.
Keywords: Bargaining Game, Kalai-Smorodinsky Bargaining Solution, Data Envelopment Analysis, Common Weight
Vector
1. Introduction
A large number of studies on Data Envelopment Analysis (DEA) have developed after DEA was first proposed by
Charnes, Cooper and Rhodes (1978), as confirmed by Glover and Sueyoshi (2009). DEA is a mathematical
programming approach to assess relative efficiencies within a group of Decision Making Units (DMUs). An important
result of such an analysis is a set of virtual multipliers, or weights, accorded to production factors (i.e., inputs or
outputs). The set of weights are often different for each of the participating DMUs.
Sexton, Silkman and Hogan (1986) have defined the cross-efficiency of jth DMU (DMUj) as a measure of DMUo that is
the ratio of weighted outputs to weighted inputs obtained when we use both input and output levels of DMU j. There
were several research efforts (e.g. Kao & Hung, 2005, Sugiyama & Yamada, 2001) that applied the cross-efficiency.
Moreover, there were other articles (e.g. Kao & Hung, 2005, Boussofiane & Dyson, 1991, Cook, Kress & Seiford, 1992,
and Roll, Cook & Golany, 1991) that discussed about a use of common weights. However, it is widely known that the
weights are not always uniquely determined. Therefore, the cross-efficiency method is not uniquely determined in DEA.
The DEA measurement process regarding efficiency of each DMU can be considered as playing a bargaining game
(Peters, 1992, and Thomson, 1994). In some cases, the measurement of relative efficiency by using a scheme of the
bargaining game is useful for group decision making. Thus, this study proposes a new approach for determining a
common weight vector based on the bargaining game. Furthermore, this research uses an example on Japanese electric
power industries in order to document the practicality of the proposed approach.
The remainder of this paper is organized as follows: Section 2 introduces a basic DEA model for making the proposed
analysis that incorporates cross-efficiency and its related total efficiency measures. Section 3 defines a feasible set from
the bargaining game for DEA. The section also proposes the calculation on Kalai-Smorodinsky bargaining solution by a
feasible set. Section 4 applies the proposed approach to measure the productivity analysis of Japanese electric power
industry. Section 5 summarizes conclusions and future extensions.
2. DEA Model and the Cross-efficiency
There are various descriptions about DEA. To uniform symbols and expressions, this study follows a description of
Cooper, Seiford and Tone (2006).
Studies in Engineering and Technology Vol. 1, No. 1; 2014
14
2.1 DEA Model
DEA presupposes total “n” DMUs as research objects of its analysis. It is also assumed that each DMUj (j=1,2,…,n) has
common inputs and outputs which contain m inputs xij>0 (i=1,2,…,m) and s outputs yrj>0 (r=1,2,…,s). Input-oriented
radial model under constant Returns to Scale (RTS), which is a basic DEA model on DMUo (o=1,2,…,n) to be
examined, has the following mathematical formulation:
.,...,2,1 0
,,...,2,1 0
,,...,2,1 0
,,...,2,1
,,...,2,1 0 subject to
, min.
1
1
11
mis
srs
nj
srysy
misxx
ss
io
ro
jo
roro
n
jjorj
io
n
jjoijioo
s
rro
m
iioo
(1)
The dual form of Model (1) are expressed by
.,...,2,1
,,...,2,1
,,...,2,1 0
,1 subject to
, max.
11
1
1
miv
sru
njyuxv
xv
yu
io
ro
s
rrjro
m
iijio
m
iioio
s
rroroo
(2)
The DMUo has an efficiency measure o that is obtained by a relative comparison with total n of DMUs’ production
activities. The optimal value *o obtained from Model (1) is often called “technical efficiency” in production
economics. The status of DEA-efficiency needs both 1* o and all slack variables are zero, i.e., 0* ros for all
,...,s,r 21 and 0* ios for all mi ,...,2,1 . If 1* o and at least one or more slack variables are 0*
ros for some
sr ,...,2,1 and 0* ios for some mi ,...,2,1 or 1* o , DMUo is defined as inefficient. Hence, Model (1) can be
usually solved by two steps of optimization without providing a specific value to ε, which is a non-Archimedean
infinitesimal (Cooper et al., 2006). This study extends the discussion by using this radial model under constant RTS.
The radial model is expressed as:
.,...,2,1
,,...,2,1
,,...,2,1 1 subject to
, max.
1
1
1
1
miv
sru
nj
xv
yu
xv
yu
io
ro
m
iijio
s
rrjro
m
iioio
s
rroro
o
(3)
Here, let *rou be an optimal value of rou , and, let *
iov be an optimal value of iov in the Model (2). This study refers
to these *rou and *
iov as DEA-solutions.
2.2 The Cross-efficiency and the Accommodated Total Efficiency
The cross-efficiency was first proposed by Sexton et al. (1986). They have defined the cross-efficiency of DMUj as
measured by DMUo as the ratio of weighted outputs to weighted inputs obtained when we use the input and output
levels of DMUj along with these weights derived for DMUo, as discussed previously. Mathematically, the
Studies in Engineering and Technology Vol. 1, No. 1; 2014
15
cross-efficiency is the ratio of the sums on the left side of constraint j in Model (3) for DMUo:
.,...,2,1 ; ,...,2,1
1
*
1
*
njno
xv
yu
Em
iijio
s
rrjro
oj
(4)
The cross-efficiencies are simply the ratios in the constraints of Model (3). The cross-efficiencies are easily summarized
by an nn matrix, whose (o,j) component is ojE . Sexton et al. (1986) called it as cross-efficiency matrix. The
conventional efficiency measures exist on the diagonal of the cross-efficiency matrix.
By examining the row o of the cross-efficiency matrix, this study can identify how DMUo rates each of the other DMUs,
that is, how efficient each of the other DMUs is when an optimal weights generated by DMUo are used for its
measurement. The mean efficiency in the row o (including the diagonal) is called EROW(o)
nj ojn
E1
1 . The measure
indicates the average efficiency of all DMUs according to DMUo. In a similar manner, this study can examine the
column j of the cross-efficiency matrix to identify how DMUj is rated by each of the other DMUs when it is evaluated
by means of the optimal weights that they are generated. The mean efficiency in the column j (including the diagonal) is
referred to as ECOL(j)
no ojn
E1
1 . The measure indicates the average efficiency of DMUj according to all other
DMUs. They can compute the average of all the cross-efficiency values, or EBAR. However, there may be no common
weights of the cross-efficiency because DEA always suffers from an occurrence of multiple solutions.
The accommodated total efficiency was proposed by Sugiyama and Yamada (2001). They showed that the
accommodated total efficiency is a general form for the cross-efficiency. The accommodated total efficiencies of DMUs
are calculated from the following three steps. (a) In the first step, they evaluate the relative efficiencies of DMUs as
group members. (b) In the second step, the mutual evaluation information of DMUs can be defined and calculated by
using their weights. They indicate the mutual evaluation information by a form of matrix which they called it as
“Accommodation Efficiency Matrix.” It is widely known that the weights are not always uniquely determined. The
mutual evaluation information is not uniquely determined. Here, they have proposed a method for determining the
weights uniquely by minimizing square of the weights differences. (c) In the third step, they calculate the
accommodated total efficiency which is the group efficiencies of DMUs by using the maximum eigenvalue of
“Accommodation Efficiency Matrix.”
3. A Common Weight Vector by the Bargaining Game Approach
This section defines some feasible sets of bargaining game on DEA and proposes the Kalai-Smorodinsky bargaining
solution by using those feasible sets. There are various descriptions about the bargaining game, and many articles. See,
for example, Peters (1992) and Thomson (1994). The study of DEA with the Game theory can be found in Banker,
Charnes, Cooper and Clarke (1989), Semple (1996), Hao, Wei and Yan (2000), Nakabayashi and Tone (2006), and etc.
On the other hand, Du, Liang, Chen, Cook and Zhu (2011) described DEA on the bargaining game. A Nash bargaining
game has also been proposed for measuring the performance of a two-stage network DEA system.
3.1 The Bargaining Game
By nN ,...,2,1 , this study denotes the set of players. The bargaining game (Peters, 1992, and Thomson, 1994) is
defined by a pair of d,S . The players in N try to reach a unanimous agreement on some outcome Sη , yielding
utility k for player k. If they fail, the disagreement outcome or disagreement point d occurs in the game. The set S is
referred to as a feasible set of the bargaining game. The set S needs to be convex, bounded and closed. There is at least
one point of S strictly dominating d.
This study chooses one of the bargaining solutions by applying an axiomatic approach. In the axiomatic approach, the
typical solution on the bargaining game is the Nash bargaining solution (Thomson, 1994). The Nash bargaining solution
is a single solution on the feasible set S satisfying Pareto-optimality, symmetry, scale invariance and independence of
irrelevant alternatives (IIA). Here, a solution of the proposed bargaining game on DEA satisfies Pareto-optimality,
symmetry and scale invariance. However, the solution of the bargaining game on DEA may not satisfy independence of
irrelevant alternatives (IIA). Therefore, the Nash bargaining solution is not appropriate. The rationale is because a
DEA-efficiency score is a relative evaluation score.
Meanwhile, a DEA solution of the proposed bargaining game fully satisfies individual monotonicity. The
Kalai-Smorodinsky bargaining solution (Thomson, 1994) is the only solution on a feasible set S satisfying
Pareto-optimality, symmetry, scale invariance and individual monotonicity. Consequently, this study selects the
Kalai-Smorodinsky bargaining solution.
These axioms are described in research efforts (i.e., Peters, 1992, and Thomson, 1994). The definition of the
Kalai-Smorodinsky bargaining solution (Thomson, 1994) is specified as follows.
Studies in Engineering and Technology Vol. 1, No. 1; 2014
16
[The Kalai-Smorodinsky bargaining solution] SK :
SK
is the maximal point of S on a segment connecting the origin to Sa , the ideal point of S, defined by
SSa kk ηmax for all k.
[In another definition, d,SK is the maximal point of S on a segment connecting d to d,Sa where
dηηd ,max, SSa kk for all k.]
3.2 The Feasible Set S of the Bargaining Game on DEA
Let the players be DMUs. Consequently, the number of the players is n. In addition, game situations assume the
bargaining game. Then, this study generalizes a feasible set of bargaining game on DEA. The feasible set in a correlated
pure strategy is expressed as follows:
,...,m,iv
,...,s,ru
Nj
xv
yu
Nj
xv
yu
S
i
r
m
iiji
s
rrjr
m
iiji
s
rrjr
j
n
21 0
,21 0
, 1
,
R
1
1
1
1
P
η . (5)
The feasible set in correlated pure strategy PS is the feasible set of a common weight vector of DEA. Moreover, the
comprehensive hull of PS is expressed as follows:
,...,m,iv
,...,s,ru
Nj
xv
yu
Nj
xv
yu
S
i
r
m
iiji
s
rrjr
m
iiji
s
rrjr
j
n
21 0
,21 0
, 1
,
Rcom
1
1
1
1
P
η. (6)
Here, it is expected that the feasible set in correlated pure strategy PS is a convex set. However, it is difficult to prove
whether PS is a convex set because PS includes the fractional equations. Let the input and output data for DMUj be
T21 ,...,, mjjjj xxxx and T21 ,...,, mjjjj yyyy , respectively. The symbol “T” denotes vector transpose. In addition, let
the weight vector be suuu ,...,, 21u and mvvv ,...,, 21v . Therefore, this study sets Model (6) by relaxing the equality
constraints jjj vxuy into inequality ones. This relaxation of Pcom S is naturally accepted in the bargaining
game. Then, this study calculates the Kalai-Smorodinsky bargaining solution by these feasible sets in the proceeding
section. By comparing results obtained by the Kalai-Smorodinsky bargaining solution that obtains from the two feasible
sets, it is possible to confirm that these feasible solutions are convex sets near the Kalai-Smorodinsky bargaining
solution.
3.3 The Computational Mode for Bargaining Solution
Let the ideal point of S be each DMU's efficiency score *j . The proposed bargaining game can establish various
kinds of points as the disagreement point d. Therefore, this study sets the disagreement point d as the origin. However,
this study cannot accept the bargaining solution from the origin. Because the DEA-solutions become 0u * or 0v * .
The weight of the DEA problem (3) is a non-zero vector. Since PS η0 and 0u are equivalent, a
Kalai-Smorodinsky bargaining solution η such as a common weight vector of DEA is a non-zero vector. In addition,
since 0y j for all Nj , 0u and 0u , and 0η are equivalent. Therefore, this study refers to as a
Kalai-Smorodinsky bargaining solution 0η as a positive Kalai-Smorodinsky bargaining solution. A
Kalai-Smorodinsky bargaining solution corresponds to a desirable common DEA weights is a positive
Kalai-Smorodinsky bargaining solution.
Studies in Engineering and Technology Vol. 1, No. 1; 2014
17
Since **2
*1
P ,...,, nSa and a line segment between the origin 0 and PSa is NjηSa j ,10Pη , finding a positive
Kalai-Smorodinsky bargaining solution PSK of 0,PS is the same of solving the positive optimal value *j of
.21 0
,21 0
, 01
, 1
, subject to
, max.
1
1
1
1
1
*
,...,m,iv
,...,s,ru
Nj
Nj
xv
yu
Nj
xv
yu
i
r
j
m
iiji
s
rrjr
jm
iiji
s
rrjr
n
jjj
(7)
The maximization of Model (7) is an unbounded problem. Thus, the maximization problem form, modified by adding
the equation 1jvx , becomes:
.21 0
,21 0
, 0
, 1 subject to
, max.
11
1
1
1
1*
,...,m,iv
,...,s,ru
Njyuxv
Njxv
xv
yu
i
r
s
rrjr
m
iiji
m
iiji
n
jm
iiji
s
rrjr
j
(8)
If Pcom S is a feasible set of the bargaining game, then finding a Kalai-Smorodinsky bargaining solution Pcom SK
of 0,com PS is the same of solving the positive optimal value *j of
.21 0
,21 0
, ,01
, 1
, subject to
, max.
1
1
1
1
1
*
,...,m,iv
,...,s,ru
Nj
Nj
xv
yu
Nj
xv
yu
i
r
j
m
iiji
s
rrjr
jm
iiji
s
rrjr
n
jjj
(9)
The maximization of Model (9) is an unbounded problem, as well. Thus, the maximization problem by adding the
equation 1jvx becomes as follows:
Studies in Engineering and Technology Vol. 1, No. 1; 2014
18
.21 0
,21 0
, ,01
, 0
, 0
, 1 subject to
, max.
11
11
1
1
*
,...,m,iv
,...,s,ru
Nj
Njyuxv
Njyuxv
Njxv
i
r
j
s
rrjr
m
iiji
s
rrjr
m
iijij
m
iiji
n
jjj
(10)
That is, P*P comcom SaSK η . Here, this study can set the ideal point of S be an each DMU's efficiency score.
Furthermore, it is possible to set the disagreement point d as the origin.
4. Numerical Example
This section documents the productivity analysis of Japanese electric power industry by applying the proposed approach.
This example was given in Sugiyama and Yamada (2001).
4.1 Data
The subjects of analysis are nine electric power companies in Japan. This study utilizes the management indexes, given
below, as input/output data of each company in the fiscal year 1991. The data source is “Hand Book of Electric Power
Industry '91” (Statistics Committee of Electric Utilities Association (Ed.), 1992), in the form of Table 1.
[Input/Output Items]
Inputs: x1j “Number of Employees”, x2j “Maximum Generation Capacity” and x3j “Total Assets”.
Outputs : y1j “Electricity Sales” and y2j “Number of Customers”.
Table 1. Observed Inputs and Outputs
Number of
Employees
Maximum
Gen. Cap.
Total
Assets
Electricity
Sales
Number of
Customers
x1j x2j x3j y1j y2j
Hokkaido 6,457 5,315 1,320,938 21,389 3,256
Tohoku 13,557 10,150 2,657,112 55,227 6,445
Tokyo 40,063 46,905 11,627,131 227,631 23,221
Chubu 20,285 22,799 4,896,313 103,140 8,711
Hokuriku 5,338 4,453 1,252,893 21,711 1,712
Kansai 25,166 33,158 5,931,094 122,749 11,331
Chugoku 10,898 9,433 2,148,717 44,498 4,578
Shikoku 6,603 5,423 1,214,685 20,548 2,490
Kyushu 13,669 14,063 3,305,687 57,272 7,007
Scale 100 103(kW) 106(YEN) 106(kWh) 103
Table 2. Modified Input/Output Data by Average
Number of
Employees
Maximum
Gen. Cap.
Total
Assets
Electricity
Sales
Number of
Customers
x1j x2j x3j y1j y2j
Hokkaido 0.4091 0.3153 0.3461 0.2855 0.4262
Tohoku 0.8590 0.6022 0.6961 0.7373 0.8437
Tokyo 2.5386 2.7828 3.0460 3.0388 3.0398
Chubu 1.2853 1.3526 1.2827 1.3769 1.1403
Hokuriku 0.3382 0.2642 0.3282 0.2898 0.2241
Kansai 1.5946 1.9672 1.5538 1.6387 1.4833
Chugoku 0.6905 0.5596 0.5629 0.5940 0.5993
Shikoku 0.4184 0.3217 0.3182 0.2743 0.3260
Kyushu 0.8661 0.8343 0.8660 0.7646 0.9173
Studies in Engineering and Technology Vol. 1, No. 1; 2014
19
4.2 Analysis and Evaluation
First, DEA is appiled on the nine electric power companies as DMUs. Table 3 summarizes the computational results
obtained from the proposed analysis, i.e., DEA-efficiency scores * and its reference set, for each DMU. As a result,
Hokkaido, Tohoku, Tokyo and Chubu are determined DEA-efficient. In contrast, Hokuriku, Kansai, Chugoku, Shikoku
and Kyushu are determined DEA-inefficient.
Table 3. DEA-efficiency of Electric Power Companies
DEA-efficiency reference set
*o
Hokkaido 1.0000 ---
Tohoku 1.0000 ---
Tokyo 1.0000 ---
Chubu 1.0000 ---
Hokuriku 0.9245 Tohoku, Tokyo
Kansai 0.9928 Tohoku, Tokyo, Chubu
Chugoku 0.9906 Tohoku, Chubu
Shikoku 0.8430 Hokkaido, Tohoku
Kyushu 0.9560 Hokkaido, Tohoku, Tokyo
Second, the study calculates the efficiency of each electric power company from a DEA common weight vector based
on the Kalai-Smorodinsky bargaining solution. As discussed in Section 3.3, let the ideal point of S be an each DMU's
efficiency score. The proposed bargaining game can establish various kinds of points as the disagreement point d. This
study sets the disagreement point d as the origin. To calculate the Kalai-Smorodinsky bargaining solution, this study
uses the maximization problems (8) and (10), modeled by each feasible set: Sa and d. Tables 4 and 5 indicate the
results on efficiencies for each DMU and a common weight vector obtained from the Kalai-Smorodinsky bargaining
solution.
Table 4. The Kalai-Smorodinsky Bargaining Solution of Electric Power Companies
The Kalai-Smorodinsky Bargaining Solution DEA-efficiency
the feasible set PS the feasible set Pcom S *o
Hokkaido 0.8411 0.8411 1.0000
Tohoku 1.0000 1.0000 1.0000
Tokyo 1.0000 1.0000 1.0000
Chubu 1.0000 1.0000 1.0000
Hokuriku 0.8082 0.8082 0.9245
Kansai 0.9928 0.9928 0.9928
Chugoku 0.9716 0.9716 0.9906
Shikoku 0.8071 0.8071 0.8430
Kyushu 0.8863 0.8863 0.9560
Table 5. A Common Weight Vector
1v 2v 1u 2u 3u
the feasible set PS 0.6779 0.0000 2.3778 2.3584 0.5912
the feasible set Pcom S 0.6779 0.0000 2.3778 2.3584 0.5912
In Tables 4 and 5, the results based on the feasible set PS and the feasible set Pcom S are same in each company.
Consequently, this study assumes that the Kalai-Smorodinsky bargaining solution, the feasible set PS and Pcom S are
a convex set. Therefore, this study determines that Tohoku, Tokyo and Chubu are efficient. Hokkaido, Hokuriku, Kansai,
Chugoku, Shikoku and Kyushu are inefficient. Thus, this study is able to identify that Tohoku, Tokyo and Chubu attain
a desirable performance level. On the other hand, Hokkaido, Hokuriku, Kansai, Chugoku, Shikoku and Kyushu do not
attain the desirable performance level. In this study, the calculated common weight 2v score is 0000.02 v . In other
words, the input item jx2 “Maximum Generation Capacity” may not be necessary for measuring the relative
efficiencies of DMUs. The details of this analysis will become very important information at selecting the input/output
items on DEA.
Studies in Engineering and Technology Vol. 1, No. 1; 2014
20
Table 6. Efficiency of Nine Electric Power Companies
The Kalai-Smorodinsky
Bargaining Solution
The Accommodated
Total Efficiency
The Cross-Efficiency
(by the weights
determined uniquely)
Hokkaido 0.8411 0.9155 0.8884
Tohoku 1.0000 1.0000 0.9757
Tokyo 1.0000 0.9658 0.9419
Chubu 1.0000 0.8643 0.8476
Hokuriku 0.8082 0.7244 0.7120
Kansai 0.9928 0.8500 0.8321
Chugoku 0.9716 0.9018 0.8824
Shikoku 0.8071 0.7879 0.7689
Kyushu 0.8863 0.8893 0.8655
Third, Table 6 indicates the results obtained from the Kalai-Smorodinsky bargaining solution and conventional
approaches. The accommodated total efficiencies of each DMU were given in Sugiyama and Yamada (2001), as well.
Furthermore, the cross-efficiencies (ECOL(j)) of each DMU were calculated by using the weights determined uniquely
in Sugiyama and Yamada (2001). The Kalai-Smorodinsky bargaining solution satisfies Pareto-optimality, symmetry,
scale invariance and individual monotonicity, thus the result obtained from the Kalai-Smorodinsky bargaining solution
may be more useful for a group decision making than other efficiencies. Therefore, each DMU may be able to accept
the solution more easily.
Fourth, Table 7 indicates a common weight vector obtained from each approach. Here, the weights of the accommodated
total efficiency were given in Sugiyama and Yamada (2001), as well. There were not the weights satisfying the
cross-efficiency in the feasible set of DEA-solutions. Hence, these weights were not calculated in this study.
Table 7. A Common Weight Vector of Each Approach
1v 2v 1u 2u 3u
The Kalai-Smorodinsky Bargaining Solution 0.6779 0.0000 2.3778 2.3584 0.5912
The Accommodated Total Efficiency 0.3739 0.3199 0.6984 0.5755 0.6824
The Cross-Efficiency --- --- --- --- ---
5. Conclusions and Future Extensions
This research discussed a methodology for determining a common weight vector of DEA based on the
Kalai-Smorodinsky bargaining solution. The calculated common weight vector was uniquely determined. This study
applied the proposed approach to examine the productivity analysis of Japanese electric power industry. The application
indicates the practicality of the proposed approach.
This research is the first effort for applying a scheme of bargaining game to determine a common weight vector of DEA.
Furthermore, this paper contributed to the progress of the study of DEA with the Game theory.
In conclusion, it is hoped that this study makes a contribution in DEA. We would like to anxiously wait for future
extensions that are originated from this research effort.
Acknowledgement
This work is supported by JSPS Grant-in-Aid for Scientific Research (C) 23510159.
References
Banker, R. D., Charnes, A., Cooper, W. W., & Clarke, R. (1989). Constrained Game Formulations and Interpretations
for Data Envelopment Analysis, European Journal of Operational Research, 40, 299-308.
http://dx.doi.org/10.1016/0377-2217(89)90422-0
Boussofiane, A., Dyson, R. G., & Thanassoulis, E. (1991). Applied Data Envelopment Analysis, European Journal of
Operational Research, 52, 1-15. http://dx.doi.org/10.1016/0377-2217(91)90331-O
Charnes, A., Cooper, W.W., & Rhodes, E. (1978). Measuring the Efficiency of Decision Making Units, European
Journal of Operational Research, 2, 429-444. http://dx.doi.org/10.1016/0377-2217(78)90138-8
Cook, W. D., Kress, M., & Seiford, L. M. (1992). Prioritization Models for Frontier Decision Making Units in DEA,
European Journal of Operational Research, 59, 319-323. http://dx.doi.org/10.1016/0377-2217(92)90148-3
Cooper, W. W., Seiford, L.M., & Tone, K. (2006). Data Envelopment Analysis : A Comprehensive Text with Models,
Applications, References and DEA-Solver Software (2nd ed.), New York, Springer Science.
Studies in Engineering and Technology Vol. 1, No. 1; 2014
21
Du, J., Liang, L., Chen, Y., Cook, W. D., & Zhu, J. (2011). A bargaining game model for measuring performance of
two-stage network structures, European Journal of Operational Research, 210, 390-397.
http://dx.doi.org/10.1016/j.ejor.2010.08.025
Glover, F., & Sueyoshi, T. (2009). Contributions of Professor William W. Cooper in Operations Research and Management
Science, European Journal of Operational Research, 197, 1-16. http://dx.doi.org/10.1016/j.ejor.2008.08.011
Hao, G., Wei, Q., & Yan, H. (2000). The Generalized DEA Model and the Convex Cone Constrained Game, European
Journal of Operational Research, 126, 515-525. http://dx.doi.org/10.1016/S0377-2217(99)00306-9
Kao, C., & Hung, H. T. (2005). Data envelopment analysis with common weights: the compromise solution approach,
Journal of the Operational Research Society, 56, 1196–1203. http://dx.doi.org/10.1057/palgrave.jors.2601924
Nakabayashi, K., & Tone, K. (2006). Egoist's Dilemma: a DEA Game, Omega, 34, 135-148.
http://dx.doi.org/10.1016/j.omega.2004.08.003
Peters, H. J. M. (1992). Axiomatic Bargaining Game Theory, Kluwer Academic Publishers.
http://dx.doi.org/10.1007/978-94-015-8022-9
Roll, Y., Cook, W. D., & Golany, B. (1991). Controlling Factor Weights in Data Envelopment Analysis, IIE
Transactions, 23, 2-9. http://dx.doi.org/10.1080/07408179108963835
Semple, J. (1996). Constrained Games for Evaluating Organizational Performance, European Journal of Operational
Research, 96, 103-112. http://dx.doi.org/10.1016/S0377-2217(96)00068-9
Sexton, T. R., Silkman, R. H., & Hogan, A. J. (1986). Data Envelopment Analysis: Critique and Extensions, Silkman, R.
H. (Ed.), Measuring Efficiency: Assessment of Data Envelopment Analysis (pp. 73-105), San Francisco, Jossey Bass.
Statistics Committee of Electric Utilities Association (Ed.). (1992). Hand Book of Electric Power Industry '91 (in
Japanese), Tokyo, Japan Electric Association.
Sugiyama, M., & Yamada, Y. (2000). Data Envelopment Analysis Using Virtual DMU as Intermediates: An Application
to Business Analysis of Japan's Automobile Manufatures, Journal of Japan Industrial Management Association, 50,
341-354. http://ci.nii.ac.jp/naid/110003945554/en
Sugiyama, M., & Yamada, Y. (2001). Group-Consensus Biilding on DEA (in Japanese), Communications of the
Operations Research Society of Japan, 46, 284-289. http://ci.nii.ac.jp/naid/110001184921/en
Thomson, W. (1994). Cooperative Models of Bargaining, Aumann, R. J. & Hart, S. (Eds.), Handbook of Game Theory
with Economic Applications (pp. 1237-1284), Elsevier Science B.V..
von Neumann, J., & Morgenstern, O. (1953). Theory of Games and Economic Behavior (3rd ed.), New Jersey, Princeton
University Press.
Yamada, Y., Matsui, T., & Sugiyama, M. (1994). An Inefficiency Measurement Method for Management Systems (in
Japanese), Journal of the Operations Research Society of Japan, 37, 158-168.
http://ci.nii.ac.jp/naid/110001184395/en
This work is licensed under a Creative Commons Attribution 3.0 License.
Studies in Engineering and Technology
Vol. 1, No. 1; February 2014
ISSN 2330-2038 E-ISSN 2330-2046
Published by Redfame Publishing
URL: http://set.redfame.com
22
Examples of Mental Mistakes Made by Systems Engineers While Creating Tradeoff Studies
James Bohlman1 & A. Terry Bahill
1
1Systems and Industrial Engineering, University of Arizona, USA
Correspondence: Terry Bahill, Systems and Industrial Engineering, University of Arizona, 1622 W. Montenegro, Tucson
AZ 85704-1822, USA. E-mail: [email protected]
Received: October 21, 2013 Accepted: November 4, 2013 Available online: November 21, 2013
doi:10.11114/set.v1i1.239 URL: http://dx.doi.org/10.11114/set.v1i1.239
Abstract
Problem statement: Humans often make poor decisions. To help them make better decisions, engineers are taught to
create tradeoff studies. However, these engineers are usually unaware of mental mistakes that they make while creating
their tradeoff studies. We need to increase awareness of a dozen specific mental mistakes that engineers commonly
make while creating tradeoff studies.
Aims of the research: To prove that engineers actually do make mental mistakes while creating tradeoff studies. To
identify which mental mistakes can be detected in tradeoff study documentation.
Methodology: Over the past two decades, teams of students and practicing engineers in Bahill’s Systems Engineering
courses wrote the system design documents for an assigned system. On average, each of these document sets took 100
man-hours to create and comprised 75 pages. We used 110 of these projects, two dozen government contractor tradeoff
studies and three publicly accessible tradeoff studies. We scoured these document sets looking for examples of 28
specific mental mistakes that might affect a tradeoff study. We found instances of a dozen of these mental mistakes.
Results: Often evidence of some of these mistakes cannot be found in the final documentation. To find evidence for
such mistakes, the experimenters would have had to be a part of the data collection and decision making process. That
is why, in this paper, we present only 12 of the original 28 mental mistakes. We found hundreds of examples of such
mistakes. We provide suggestions to help people avoid making these mental mistakes while doing tradeoff studies.
Conclusions: This paper shows evidence of a dozen common mental mistakes that are continually being repeated by
engineers while creating tradeoff studies. When engineers are taught about these mistakes, they can minimize their
occurrence in the future.
Keywords: validation, decision making, mistakes, design documentation
1. Introduction
Humans often make poor decisions. To help them be better decision-makers, engineering professors teach their students
to create tradeoff studies. Tradeoff studies are broadly recognized as the method for simultaneously considering
multiple alternatives with many criteria, and as such are recommended and mandated in the Capability Maturity Model
Integration (CMMI®) (CMMI, 2010; Chrissis, Konrad and Shrum, 2003) Decision Analysis and Resolution (DAR)
process. The decision-making fields of Judgment and Decision Making, Cognitive Science and Experimental
Economics have built up a large body of research on human biases and errors in considering numerical and
criteria-based choices. Relationships between experiments in these fields and the elements of tradeoff studies show that
tradeoff studies are susceptible to human mental mistakes. Smith, Son, Piattelli-Palmarini and Bahill (2007) postulated
28 specific mental mistakes that could affect the ten specific components of a tradeoff study.
However, Smith et al. (2007) did not have examples from the system engineering literature. So we sought to validate
their model by finding and documenting specific instances of these mental mistakes in tradeoff studies. This present
paper validates the proposals of the Smith, et al. (2007). It presents real-world examples of a dozen of these mistakes. It
would have been a daunting challenge to find examples of such mistakes in the open systems engineering literature,
because the literature just does not publish mistakes.
Studies in Engineering and Technology Vol. 1, No. 1; 2014
23
1.1 What Are Tradeoff Studies?
Tradeoff studies are broadly recognized and mandated as the method for simultaneously considering multiple
alternatives with many criteria (Daniels, Werner & Bahill, 2001; Smith et al., 2007). Tradeoff studies provide an ideal,
rational method for choosing among alternatives. Tradeoff studies involve a mathematical consideration of many
evaluation criteria for many alternatives simultaneously, in parallel.
Tradeoff studies are performed at the beginning of a project to help state the problem, select the desired system
architecture and make major purchases. However, throughout a project tradeoffs are continually being made: creating
team communication methods, selecting tools and vendors, selecting components, choosing implementation techniques,
designing test programs, and maintaining schedule. Many of these tradeoffs should be formally documented.
The components of a tradeoff study are the (1) problem statement, (2) evaluation criteria, (3) weights of importance, (4)
alternate solutions, (5) evaluation data, (6) scoring functions, (7) normalized scores, (8) combining functions, (9)
preferred alternatives and (10) sensitivity analysis.
The following is a cartoon of a simple tradeoff study. The analyst is designing a system to help a baseball umpire to call
balls and strikes. He is trying to select the preferred alternative. His two alternatives are a complex instrument named
the Umpire’s Assistant and a Seeing Eye dog. His evaluation criteria are Accuracy of the Call and Silence of
Communication between the alternative systems and the umpire. (We do not want it to intrude on the classic game.) The
very bottom row shows that the Umpire’s Assistant was the recommended alternative.
Table 1. A Generic Tradeoff Study
Criteria
Weights of
Importance of the
Criteria
Alternative-1 Scores Alternative-2 Scores
Criterion-1 wt1 s11 s12
Criterion-2 wt2 s21 s22
Alternative
Rating 1 1 11 2 21sum wt s wt s
2 1 12 2 22sum wt s wt s
Table 2. Tradeoff Study Numerical Example
Alternatives
Evaluation Criteria Weights of
Importance
Umpire’s
Assistant
Seeing
Eye Dog
Accuracy of the Call 0.75 0.67 0.33
Silence of Communication 0.25 0.83 0.17
Sum of weight times score 0.71
The winner 0.29
The numerical values go from 0 to 1, with 1 being the best.
2. Research Methodology
Over the last quarter of a century, Bahill has sought out and collected examples of tradeoff studies. He collected 110
university projects, two dozen government contractor tradeoff studies (these reports are proprietary and cannot be
published) and three publicly accessible tradeoff studies such as the San Diego Airport site selection study.
Then the authors read these reports looking for evidence of the 28 specific mental mistakes postulated by Smith et al
(2007). Bohlman read all 8000 pages of the university projects and filled a database with over 800 examples of such
mistakes. These examples were put into categories according to the 28 specific mental mistakes that we were looking
for. After several iterations, we decided that our examples fit into only 12 of these categories. Bahill then rated each
example for the perceived heuristic value, meaning the goodness of illustrating each type of mental mistake. The
examples with the least perceived value were set aside. The remaining examples were compared with the original
documents checking for context and accuracy. They were then re-rated based on the perceived heuristic value, uniform
coverage of the dozen mental mistakes and reducing the number of different projects that would have to be explained.
We created the 20 development cases that are in this paper, and another 50 cases that were used for testing. Finally,
Studies in Engineering and Technology Vol. 1, No. 1; 2014
24
Bahill read the government contractor tradeoff studies and the three identified publicly accessible tradeoff studies and
similarly created case studies.
Often evidence of some of these mental mistakes cannot be found in the final documentation. To find evidence for such
mistakes, the experimenters would have to have been a part of the data collection and decision making process. That is
why, in this paper, we present only 12 of the original 28 mental mistakes. This paper presents examples of the following
mental mistakes.
1. Using Dependent Criteria
2. Not Stating the Problem in Terms of Stakeholder Needs
3. Vague Problem Statement
4. Substituting a Related Attribute
5. Sensitivity Analysis Mistakes
6. Forer Effect
7. Weight of Importance Mistakes
8. Anchoring and the Status Quo
9. Treating Gains and Losses Equally
10. Not Using Scoring Functions
11. Implying False Precision
12. Obviating Expert Opinion
This paper is organized as follows: for each of the above (1) the section heading announces the type of mental mistake,
(2) Smith et al’s (2007) explanation of the mistake is given, (3) Smith et al’s (2007) recommendations for avoiding that
mistake are stated, (4) the problem that the teams were modeling is explained (if it is the first time the problem has been
presented), (5) an excerpt of this type of mistake is presented in the Garamond font, (6) it is explained why the
particular example contains a mental mistake in the Times New Roman font, (7) a suggested rewrite is given in the
Arial font. The mistake excerpts have not been edited: they are direct quotes: however, they might have been
reformatted. The sections from Smith, et al. (2007) have been edited.
In the decision making literature there are hundreds of names for such mistakes, names such as attribute substitution,
dominated criteria, cognitive illusions, emotions, fallacies, simplifying heuristics, fear of regret, paradoxes, herding,
group think, psychological traps (Marquard & Robinson, 2008) and cognitive biases (Sage, 1981: Smith,
Piattelli-Palmarini & Bahill, 2008; Smith & Bahill, 2010). However, in this paper we will simply call them mental
mistakes.
The purpose of this paper is to explain and give examples of certain types of mental mistakes that humans often make
while doing tradeoff studies. Many of these mistakes are subtle and it is hard to avoid making them. However, it is
hoped that systems analysts who read this paper will be able to recognize these mistakes in tradeoff studies created by
others and to avoid these mistakes in their own tradeoff studies.
3. Examples of Mental Mistakes in Tradeoff Studies
This section shows examples of a dozen types of common mental mistakes that were made while doing tradeoff studies.
We believe these mistakes are continually being repeated by systems engineers worldwide. We hope that reading this
paper will help systems engineers to avoid such mental mistakes in the future.
3.1 Using Dependent Criteria
Evaluation criteria should be independent. In a tradeoff study, alternatives should be evaluated based on independent
evaluation criteria. However, analysts often choose dependent criteria. When scoring these criteria for the different
alternatives, having multiple dependent criteria can magnify or diminish the final scores of the alternatives, thus
recommending the wrong alternative. Here are some simple examples of evaluation criteria. For evaluating humans,
Height and Weight are not independent: Sex (male versus female) and Intelligence Quotient are independent. In
selecting a car, the following criteria are dependent: Maximum Horse Power, Peak Torque, Top Speed, Time for the
Standing Quarter Mile, Engine Size (in liters), Number of Cylinders and Time to Accelerate 0 to 60 mph (Smith et al.,
2007).
Recommendation: Dependent criteria should be grouped together as subcriteria. The seven subcriteria for the car given
in the previous paragraph could all be grouped into the criteria Power.
Studies in Engineering and Technology Vol. 1, No. 1; 2014
25
For each year, the class project was designed to be unprecedented, with no possible optimal solution, realistic, solved
by teams of three or four engineers who would have to analyze, synthesize and evaluate alternative designs. The
following example does not depend on the particular project: it could have occurred in any project.
Team Excerpt
The maximum current drawn by the system shall not exceed 15 amperes.
The system shall operate on 120 volt, 60 hertz electricity.
The system shall not consume more than 1.8 kilowatts.
Why is this a mental mistake?
If you know the voltage and the current, then you know the power; Power = Voltage x Current.
Suggested rewrite
The system shall operate on 120 volt, 60 hertz AC electricity.
The system shall draw a maximum of 15 amperes.
3.2 Not Stating the Problem in Terms of Stakeholder Needs
Committing to a class of preconceived solutions (instead of stating the true stakeholder needs) causes a lack of
flexibility. Identifying the true stakeholder needs can be difficult because stakeholders often refer to both problem
domains and solution domains – whichever comes most naturally. In systems engineering, the initial problem statement
must be written before looking for solutions (Wymore, 1993; Smith et al., 2007).
Recommendation: Communicate with and question the stakeholders in order to determine their values and needs. State
the problem in terms of customer requirements (Bahill & Dean, 1999, 2009; Hooks & Farry, 2001; Daniels & Bahill,
2004; Hull, Jackson & Dick, 2005). Later, after a better understanding of evaluation criteria and weights of importance
has been gained, one must find alternative solutions that provide a good match to the requirements.
The project for 2006 was the SpinCoach™. When a spinning object (like a baseball) is put in a moving fluid (like air),
it will experience a force that pushes it sideways (Bahill, Baldwin, & Venkateswaran, 2005; Baldwin, Bahill & Nathan,
2007; Bahill & Baldwin, 2008; Bahill, Baldwin & Ramberg, 2009.) Some highly successful baseball players have said
and written that they see this spin of the ball and use it to track the ball with saccadic and smooth pursuit eye
movements (Bahill & Stark, 1975; Bahill & Stark, 1977; Bahill & LaRitz, 1984). But at present, there is no system that
can teach high school and college baseball and softball players to predict this spin-induced deflection of the pitch.
Therefore, this project was to design and document the design of a system that would help train baseball and softball
players to pick up the spin on the ball and predict the spin's effect on the ball's deflection. The system would be capable
of displaying images of spinning balls, allowing the subject to predict the spin induced deflection and providing
feedback to facilitate learning. The key architectural decision in this project is whether to design the first version for
baseball or softball.
Team Excerpt
The batter believes he can predict the trajectory until the ball and bat connect and therefore swing accordingly.
Because of the deflection in the last milliseconds of the balls flight however, the batter is not able to adjust his swing
in time to compensate and therefore the likelihood for there to be a strike called is increased because of the batter
swing. In essence the pitcher is trying to fake out the batter and get him to swing at the ball or not swing at the ball
based on a seemingly predictable trajectory. It appears to the batters, however that the ball seems to change direction
in the last moments of its trajectory and is cause for batters to swing without connecting with the ball.
Why is this a mental mistake?
This problem statement does not mention the batter’s needs. It is certainly stating correct things about
what the batter and the pitcher are doing. But it fails to describe the batter’s needs in terms of the
system being designed. If they were designing a different system, then the excerpt might be useful.
But they were supposed to be designing a system to help the batter learn the spin-induced deflection
of the ball.
Suggested rewrite
Baseball and softball players need to learn how to recognize the spin of a pitched ball
and use that information to predict the spin-induced deflection of the ball.
Second Team Excerpt
A video game trainer could also be used, in which a batter tries to gauge the spin on a video ball and predict where it
would end up. This option could be implemented with a CD-ROM. The CD would go through the process of how to
see the spin based on video taken from a laboratory. Dr. Terry Bahill, a professor in the Systems and Industrial
Engineering (SIE) department at the University of Arizona, has set up a laboratory with equipment to simulate
Studies in Engineering and Technology Vol. 1, No. 1; 2014
26
pitches with different types of spins. This laboratory would be a valuable resource in designing experiments to train a
player to pick up the spin on a ball.
Why is this a mental mistake?
This problem statement does not mention the batter’s needs.
Suggested rewrite
The top-level system function is to teach batters to recognize the spin of a pitched
ball and then use this information to predict the spin-induced movement of the ball.
3.3 Vague Problem Statement
If a problem statement is vague, proposed solutions could vary greatly, and derive support for very different reasons
and in different ways. If a problem statement is poorly written or ambiguous, dissimilar alternative solutions could
remain in the solution pool, obfuscating their rational consideration, especially if the rationale for the different
psychologically attractive values of the alternative solutions are not well understood (Keeney, 1992). Failing to mention
customer needs makes the problem statement seem vague (Smith et al., 2007).
Recommendation: Stating the problem is the most important and possibly the most difficult aspect of a tradeoff study.
The problem should be stated in terms of the stakeholder’s needs. But be sure to state the problem so that it is
independent of preconceived solutions. State the problem so that it can be satisfied by a large number of alternative
solutions.
The project for 2007 was the PopupCoach™. Even professional baseball players occasionally find it difficult to
gracefully approach seemingly routine pop-ups. McBeath, Nathan, Bahill & Baldwin (2008) describe a set of towering
pop-ups with trajectories that exhibit cusps and loops near the apex. For a normal fly ball, the horizontal velocity is
continuously decreasing due to drag caused by air resistance. But for pop-ups, the Magnus force (the force due to the
ball spinning in a moving airflow) is larger than the drag force: therefore, the horizontal velocity decreases in the
beginning, like a normal fly ball, but after the apex, the Magnus force accelerates the horizontal motion (Bahill &
Baldwin, 2007). We refer to this class of pop-ups as paradoxical because they appear to misinform the typically robust
optical control strategies used by fielders and lead to systematic vacillation in running paths, especially when a
trajectory terminates near the fielder. In short, some of the dancing around when infielders pursue pop-ups can be
explained as a combination of bizarre trajectories and misguidance by the normally reliable optical control strategy,
rather than apparent fielder error. Former major league infielders confirm that our model agrees with their experiences.
But at present, there is no methodological system that can teach high school and college baseball and softball players
(more specifically catchers and infielders) to track pop-ups. Batting practice and ball games offer few opportunities for
a player to learn this skill. Our customer needs a system that will provide frequent convenient opportunities to learn to
field pop-ups. The goal of this project is to design and document the design of such a system.
Team Excerpt for the PopupCoach
The occurrence of pop-ups in major league games is an average of five times per game. This frequency provides
reason for effective practice to be devoted to train players in order to hone their ability to catch these routine
pop-ups.
Why is this a mental mistake?
This problem statement is vague.
Suggested rewrite
For baseball and softball fielders who need to catch pop-ups, the PopupCoach is a
training system that teaches fielders to track and catch pop-ups: unlike present
coaches and books, the PopupCoach explains the trajectory, offers convenient
practice opportunities and complements players’ improvement.
Team Excerpt for the SpinCoach
The differences in (baseball) spin result in visually detectable differences in the appearance of the spinning ball as it
approaches the batter. Currently, players practice the assessment of spin and trajectory prediction during actual game
play and team practice sessions. This process requires coordination between multiple players, and the use of a pitcher
capable of delivering a repeatable baseball pitch to the batter.
Why is this a mental mistake?
This problem statement is vague.
Suggested rewrite
For the baseball batter who needs to predict the trajectory of the pitch, the
SpinCoach is a training system that helps him to recognize the spin on the pitch and
predict the ball’s spin-induced movement; unlike present coaches and books, the
SpinCoach shows the batter how each pitch spins and helps him to recognize this
spin.
Studies in Engineering and Technology Vol. 1, No. 1; 2014
27
3.4 Substituting a Related Attribute
Attribute substitution occurs when a subject is assessing an attribute and substitutes a related attribute that comes
more readily to mind. In effect, people who are confronted with a difficult question sometimes answer an easier one
instead (Kahneman, 2003). In a similar vein, when confronted with a choice among alternatives that should properly be
decided by a full tradeoff study, there is a strong tendency to substitute a seemingly equivalent yet much simpler
decision question in place of the tradeoff study process (Smith et al., 2007).
Recommendation: Sponsors of tradeoff studies should realize that a premature reduction of a tradeoff study process to
a simpler decision question is a common heuristic that prevents consideration of the original multi-objective decision.
Team Excerpt for the SpinCoach
Use Case 2.
Name: Learn Spin-induced Deflections
Iteration: 2.3
Derived from: Concept of operations
Brief description: Player uses the SpinCoach and learns to predict the spin-induced deflection of a ball.
Added value: Player will be better able to predict the trajectory of the ball and consequently should have a higher
batting average.
Second Team Excerpt for the SpinCoach
5.2.1.2 Effectiveness
The measurement of effectiveness determines the percent increase on the user’s batting average over time.
The purpose of the SpinCoach is to teach batters to predict the spin-induced deflection of the baseball, but as a
measure of success, these teams proposed to record the player’s batting average.
Why is this a mental mistake?
What we really want to know is “Does training with the SpinCoach teach batters to predict the
spin-induced deflection of the baseball?” But this is too hard to measure; therefore, our engineers
substituted the player’s batting average as a measure of success.
Suggested rewrite
We want to teach baseball batters to predict the spin-induced deflection of the
baseball. Therefore, we want to know how well they predict the spin-induced
deflection, but that is hard to measure, therefore we substitute the player’s batting
average as a measure of success.
In this particular design what we really want to know is “Does training with the SpinCoach teach batters to predict the
spin-induced deflection of the baseball?” But this is too hard to measure; therefore, we substitute the player’s batting
average as a measure of success. But we do tell our readers that we are doing this.
After he received the Nobel Prize for developing Prospect Theory (RSAS, 2002; Kahneman, 2002), Kahneman spent
most of his time trying to unify mental mistakes. He suggested that many or most could be explained by attribute
substitution. This is a very difficult mistake to avoid. People do it all the time. The point of this section is that analysts
should understand what attribute substitution is and avoid making it as a mistake. However, if the analyst really wants
to use a substitute attribute, then he or she should go ahead and use it. But be sure to tell that reader that he or she is
using attribute substitution.
Team Excerpt for the PopupCoach
The ability of a defending baseball team to catch pop-up balls can provide a key advantage needed to win a baseball
game. That pop-ups occurs in major league games an average of five times underscores the need for players to
improve their ability to catch them. Pop-ups are difficult to catch because their trajectories are irregular and it is not
readily predictable where players should stand and catch them. There is no existing training system to improve
players’ catching percentage.
Second Team Excerpt for the PopupCoach
Thus, the need for a system to teach fielders about and offer practice scenarios involving trajectory and spin is
necessary if it can help increase fielding percentages and improve the win percentages for players and teams.
Why is this a mental mistake?
What we really want to know is “Does training with the PopupCoach teach fielders to use the optimal
running path and running speed while fielding fly balls. But this is too hard to measure; therefore, our
engineers substituted the player’s catching percentage and fielding percentages.
Suggested rewrite for the PopupCoach
Studies in Engineering and Technology Vol. 1, No. 1; 2014
28
For each pop-up, we will compute the player’s optimal running path and then
compare it to the player’s actual running path. We will also compute the optimal
running speed at the time of the catch and compare this to the player’s actual running
speed. These two metrics will be combined to indicate how well the fielder tracked
the pop-up.
Attribute substitution is a tricky mistake, because everyone does it. In this design what we really want to know is “Does
training with the PopupCoach teach fielders to use the Optical Acceleration Cancellation algorithm (McBeath, Nathan,
Bahill, & Baldwin, 2008) to catch Pop-ups?” But we cannot use as a metric the probability of catching a pop-up,
because 9 of our 12 alternative designs do not use real pop-ups. For these 9 designs we think it would be too expensive
(if not impossible) to gather enough data for every player to make valid inferences. Furthermore we have no method of
measuring the spin of pop-us. Therefore, we think variability in the speed and spin rate of real pop-ups would obscure
any evidence that the fielder used the Optical Acceleration Cancellation algorithm to catch pop-ups. Because we could
not measure the desired attribute, we substituted a simpler attribute, the running speed and path. Attribute substitution is
not always a mistake: often it is deliberate and stated.
3.5 Sensitivity Analysis Mistakes
Most people are not well trained in the machinery and methods of sensitivity analysis. They often fail to compute
second- and higher-order partial derivatives. When estimating partial derivatives, they often use too large a step size.
When estimating partial derivatives of functions of two parameters, they often use the wrong formula; they use a
formula with two instead of four numerator terms. Smith, Szidarovszky, Karnavas and Bahill (2008) has shown that
interactions among parameters in tradeoff studies can be important, step sizes for the approximation of effects should be
small, and second-order derivatives should be calculated accurately. It is expected that only the best-trained personnel
will know of such results, illustrating the gap between truth and training (Smith et al., 2007).
Recommendation: Investments in sensitivity analysis training must be made. Perhaps enabling software can substitute
for much sensitivity analysis knowledge. (Hsu, Bahill & Stark, (1976); Karnavas, Sanchez & Bahill, (1993); Smith,
Szidarovszky, Karnavas & Bahill, (2008) describe the development and use of sensitivity analyses.
First Team Excerpt
Since the training methods are independent of any resources and a tradeoff analysis was not required, no sensitivity
analysis will be done for it.
Second Team Excerpt
The I/O performance weights more than the utilization of resources in the trade-off analysis. TW1P0 and TW2P0
values will be switched to determine the sensitivity of the design concepts to weighting and the results are shown
below:
Third Team Excerpt
The current trade study has an emphasis placed on the performance requirements of the SlugMaster to determine
the sensitivity of results to the weighting of the requirements, a plot was constructed as a function of the two
weights: I/O Performance and utilization of resources.
Why is this a mental mistake?
These Teams considered at most one parameter, the relative tradeoff weight between performance and
cost.
Suggested rewrite, derived from Smith, Szidarovszky, Karnavas & Bahill, (2008).
Concept
Weights are Weights are
0.70/0.30 0.30/0.70
Score Score
1 0.3811 0.7347
2 0.5172 0.3665
3 0.5191 0.3061
4 0.568 0.3416
5 0.563 0.3249
6 0.6275 0.3362
7 0.5739 0.3132
Studies in Engineering and Technology Vol. 1, No. 1; 2014
29
Table 1. A Generic Tradeoff Study
Criteria Weights of
Importance Alternative 1 Alternative 2
Criterion-1 Wt1 S11 S12
Criterion-2 Wt2 S21 S22
Alternative
Rating
1
1 11
2 21
Sum
Wt S
Wt S
2
1 12
2 22
Sum
Wt S
Wt S
Table 2 gives numerical values for one particular tradeoff study, The Umpire’s
Assistant (http://www.sie.arizona.edu/sysengr/sie577/UmpiresAssistant.doc).
Table 2. Tradeoff Study Numerical Example
Alternatives
Criteria Weights of
Importance
Umpire’s
Assistant
Seeing
Eye
Dog
Accuracy
of the call 0.75 0.67 0.33
Silence of
Signaling 0.25 0.83 0.17
Sum of
weight
times
score
0.71
The
winner
0.29
Definition of the semirelative sensitivity function:
0
NOP
F FS
%
These are the semirelative sensitivity functions for Tables 1 and 2.
1
1
1
2
1
11
1
21
1
12
1
22
11 12 1
21 22 2
1 11
2 21
1 12
2 22
0.26
0.16
0.50
0.21
-0.25
-0.04
PI
Wt
PI
Wt
PI
S
PI
S
PI
S
PI
S
S S S Wt
S S S Wt
S Wt S
S Wt S
S Wt S
S Wt S
%
%
%
%
%
%
These functions show that the most important parameter is the score for alternative-1
on criteria-1. Sensitivity analyses need mathematical detail. Failure to do the
mathematics right produces erroneous results. Smith, Szidarovszky, Karnavas and
Bahill, (2008) present the correct mathematical equations.
For more comments about sensitivity analyses, see these online documents:
http://www.sie.arizona.edu/sysengr/sie554/SpinCoach/JA2/index.html
http://www.sie.arizona.edu/sysengr/sie554/PopUpCoach/index.html
Studies in Engineering and Technology Vol. 1, No. 1; 2014
30
3.6 Forer Effect
According to the Forer Effect, previously existing criteria will be adopted if (1) the analyst believes that the criteria
apply to the present problem, (2) the criteria are well presented and (3) the analyst believes in the authority of the
previous criteria writer. For example, the analyst might fail to question or re-write criteria from a legacy tradeoff study
that originated from a perceived authority and is now seemingly adaptable to the tradeoff at hand. This is called the
Forer effect. Forer (1949) gave a personality test to his students. He then asked them to evaluate his personality analyses
of them, supposedly based on their test's results. Students rated their analyses on a scale of 0 (very poor) to 5 (excellent)
as to how well it applied to them. The average score was 4.26. Actually, Forer had given the same analysis to all the
students. He had assembled this analysis of a generally likeable person from horoscopes. Variables that contribute to
this fallacy in judgment are that the subject believes the analysis only applies to them, the subject believes in the
authority of the evaluator, and the analysis contains mainly positive traits (Smith et al., 2007).
Recommendation: Spend time considering and formulating criteria from scratch, before consulting and possibly
reusing previously written criteria.
The textbook for the course (Chapman, Bahill & Wymore, 1992) had a primitive sensitivity analysis that only
considered one parameter, the tradeoff weight between cost and performance. In Section 5.5.4 it states, “The system is
sensitive to tradeoff weightings. For example changing the weights of the Trade-Off Requirement can easily sway the
answer. The current trade-off puts heavy emphasis on the I/O performance of the system (0.90) and not on the
utilization of resources (0.10). Changing the degree of emphasis can change the results…” In the course, we had a
lecture and a homework that described how to take partial derivatives and form the semirelative sensitivity functions for
each parameter in the tradeoff study (Smith, Szidarovszky, Karnavas & Bahill, 2008). However, the students repeatedly
copied the method of conducting a sensitivity analysis from a legacy tradeoff study published in the course textbook.
Despite warnings about the inadequacy of that sensitivity analysis, the students conducted their analyses in that very
same way. This was an example of the Forer effect (Forer, 1949) Students failed to question a sensitivity analysis that
was presented by a perceived authority and was seemingly adaptable to their own tradeoff study.
Team Excerpt
Figure 1. Team summary of a sensitivity analysis. It shows how the overall tradeoff scores vary for a dozen
alternatives as the cost becomes less important and the performance becomes more important.
cost performance 1wt wt
Why is this a mental mistake?
Studies in Engineering and Technology Vol. 1, No. 1; 2014
31
The only parameter being varied here is the relative weight of performance versus cost. In this figure,
the team just put their new data into a figure from their textbook (Chapman, Bahill & Wymore, 1992).
They ignored all sensitivity analyses that we developed later. Students failed to question a sensitivity
analysis that was presented by a perceived authority and was seemingly adaptable to their own
tradeoff study.
3.7 Weight of Importance Mistakes
When a group of people is asked to assign a weight of importance for an evaluation criterion, each person might
produce a different value. Different weights arise not only from different preferences, but also from irrational severity
amplifiers (Bahill & Karnavas, 2000; Bahill & Smith, 2009). These include the factors of lack of control, lack of choice,
lack of trust, lack of warning, lack of understanding, manmade, newness, dreadfulness, personalization, recallability and
immediacy. Excessive disparities occur when a person assesses a weight of importance after framing the problem
differently. An evaluation may depend on how the criterion affects that person, how well that person understands the
alternative technologies, the dreadfulness of the results, etc. As a result, each person might assign a different weight of
importance to any criterion. The decision analyst should assign weights to the criteria so that the more important ones
will have more effect on the outcome. Weights are often given as numbers between 0 and 10, but are usually
normalized so that in each category they sum to 1.0. These methods can be used by individuals or teams. If pair-wise
comparisons of preferences between the criteria can be elicited from experts, then the weights of importance can be
determined through the Analytic Hierarchy Process (AHP). However, performing pair-wise comparisons can lead to
intransitive preferences: Therefore, the AHP computes an inconsistency index to warn if the domain expert is giving
intransitive responses (Smith et al., 2007).
Recommendation: Interpersonal variability can be reduced with education, peer review of the assigned weights, and
group discussions. But be aware that people are like lemmings: if you reveal how other people are voting, then they are
likely to respond with the most popular answers. It is also important to keep a broad view of the whole organization, so
that criteria in one area are considered in light of all other areas. A sensitivity analysis can show how important each
weight is. For unimportant weights, move on. For important weights, spend more time and money trying to get
consensus: this might include showing the recommended alternatives for several different sets of weights.
The project for 2005 was the Umpire’s Assistant. For the baseball umpire who needs to call balls and strikes, the
Umpire’s Assistant is an intelligent decision aiding system that helps him or her to call balls and strikes accurately,
consistently and in real-time. Unlike unassisted human umpires, the Umpire’s Assistant uses the same strike-zone
standards for all leagues, parks, umpires, batters and pitchers.
Table 3. Umpire’s Assistant Team Excerpt
Utilization of Resources Figures
of Merit Requirements
Weight
Value
Normalized
weight
1. Available Money 2 0.02326
2. Available Time 2 0.02326
2.1 System design &
prototyping by 12/31/05
2 0.02326
2.2 System verification testing
by 2/06
2 0.02326
3. Technological Restrictions 10 0.11628
3.1 to not significantly alter
the dynamics of baseball
9 0.10465
3.2 to comply with local,
regional, state, federal laws
10 0.11628
3.3 to comply with FCC rules 10 0.11628
4. Adaptability 8 0.09302
4.1 to comply with Standards
& Specifications of MLB
8 0.09302
4.2 to comply with Standards
& Specifications of NCAA
8 0.09302
Why is this a mental mistake?
The normalized weights add up to 0.826. They should add up to 1.0 in each category and each
subcategory.
Table 4. Suggested rewrite for the Umpire’s Assistant
Studies in Engineering and Technology Vol. 1, No. 1; 2014
32
Utilization of Resources Evaluation Criteria
We
ights
of
Import
ance
Crite
ria
Norm
aliz
ed
We
ights
Subcrite
ria
Norm
aliz
ed
We
ights
1. Available Money 2 0.09
2. Available Time 2 0.09
2.1 System design & prototyping by
12/31/05
2 0.5
2.2 System verification testing by 2/14/06 2 0.5
3. Technological Restrictions 10 0.45
3.1 to not significantly alter baseball
dynamics
9 0.31
3.2 to comply with local, state & federal laws 10 0.35
3.3 to comply with FCC rules 10 0.35
4. Adaptability 8 0.36
4.1 to comply with MLB rules 8 0.5
4.2 to comply with NCAA rules 8 0.5
Of course, there would be a paragraph explaining each of these short evaluation
criteria tags. The abbreviations would be explained in these paragraphs.
3.8 Anchoring and the Status Quo
The order in which the alternatives are listed has a big effect on the values that humans give for the evaluation data
(Piattelli-Palmarini, 1994; Tversky & Shafir, 1992). Therefore, tradeoff study matrix should be filled out row by row
and the status quo should be the alternative in the first column. This will make the evaluation data for the status quo the
anchors for estimating the evaluation data for the other alternatives. This is good because the anchoring alternative is
known, is consistent, and you have control over it (Smith et al., 2007).
Note: The status quo will probably have low evaluation data values (inputs for scoring functions) for performance, cost,
schedule and risk. If the status quo had high performance values, then you probably would not be trying to replace it.
The status quo already exists, so (1) it will not be expensive, which gives it a low (good) value for cost, (2) it should not
have schedule problems, which gives it a low (good) value for schedule and (3) it should also be low risk.
Recommendation: Put the status quo alternative in the first column. In the first iteration, evaluate the scores left to
right and in the next iteration evaluate them right to left. The more alternatives that exist and the more complicated the
decision, the more the status quo will be favored. Do not needlessly increase the number of alternatives in a tradeoff
study. More alternatives increase the difficulty of the decision. However, in the very beginning of a project it is good to
have many alternatives in order to better understand the problem and the requirements. View the problem from different
perspectives. Use different starting points. When estimating values for parameters of scoring functions, think about the
whole range of expected values for the parameters.
Table 5. Team Excerpt for ranking process alternatives
Criterion 1 2 3 4 5 6 7 Max
Score
Metric
Value 1.8 0.7 2.3 1.5 0.7 1 2 100
Raw Wtd Raw Wtd Raw Wtd Raw Wtd Raw Wtd Raw Wtd Raw Wtd Solution
Alternative Scr Scr Scr Scr Scr Scr Scr Scr Scr Scr Scr Scr Scr Scr Score
Solution 1 2 7.2 10 6.3 9 14 3 9 3 2.1 8 2 2 4 41.8
Solution 2 9 3.6 7 7 9 21 9 4.5 2 1.4 2 8 10 20 76.6
Solution 3 3 16 2 4.9 4 9.2 8 14 3 2.1 8 9 1 2 67.3
Solution 4 9 5.4 10 1.4 5 12 9 12 9 6.3 9 8 10 20 76.6
Solution 5 3 16 3 7 5 12 8 14 8 5.6 7 9 2 4 67.3
Solution 6 9 16 3 2.1 5 12 9 11 9 6.3 5 7 10 20 76.6
Solution 7 3 5.4 8 5.6 9 21 7 11 3 2.1 5 5 9 18 67.3
Studies in Engineering and Technology Vol. 1, No. 1; 2014
33
Do
Nothing 10 18 10 7 10 23 10 15 10 7 10 10 10 20 100
Why is this a mental mistake?
In a tradeoff study matrix the alternatives should be in columns, not rows, because humans find it
easier to compare across rather than down. The alternatives and evaluation criteria should be
identified by names, not numbers. The alternatives and evaluation criteria should be explained with
sentences and paragraphs. Finally, the do nothing alternative should be in the first column.
Why is this a mental mistake?
Table 5 also shows a mistake at the implementation level: the engineers intended each weighted score
to be its raw score times the metric value, and the solution score to be the summation of the weighted
scores in that row. But their Excel worksheet had mistakes: only the Do Nothing row was correct.
Table 6. Suggested rewrite of the Tradeoff matrix
for alternative architectures of the SpinCoach
Alternatives
Criteria
Do N
oth
ing
Com
pute
r
Sim
ula
tion
CD
RO
M
DV
D
Web
Pa
ge
Vid
eo G
am
e
Fidelity of Images
Feedback Time
Product Production Cost
Shipping Cost
Updatability
Each evaluation criterion and each alternative architecture must have a paragraph of
explanation, as in the following paragraphs.
Previously, we have written that the tables and figures of a tradeoff study do not contain the evaluation criteria and the
design alternatives: they merely contain tags. The actual criteria and alternatives must be explained elsewhere using
sentences and paragraphs. The following is an example for the SpinCoach.
Alternative architectures for the SpinCoach
1. The Status Quo. Some batters can recognize the spin on the ball and predict its spin induced movement. But they
have difficulty verbalizing this capability and teaching it to others. In batting practice, we can have the pitcher announce
to the batter “curve” and then throw a curveball: announce “slider” and then throw a slider. Etc. This could be done with
a human pitcher or a pitching machine.
2. Computer Simulations. Images of spinning balls can be simulated and presented on a computer monitor. This system
is described in SpinTeacherGray.doc.
4. Spinning Balls. Holes are drilled into baseballs or softballs and the balls are skewered on bolts. These bolts are
chucked into drills. The drills are spun at controlled speeds. Videos of this setup are on my web site.
http://www.sie.arizona.edu/sysengr/baseball/index.html. Do not let someone’s tie or scarf get tangled in the drills.
4. CD-ROM or DVD. Balls spinning on drills can be photographed and their images stored on CD-ROM or DVD disks.
Such videos are on my web site. http://www.sie.arizona.edu/sysengr/baseball/index.html. These images along with the
software program will be transferred from the CD-ROM to the user’s hard disk using a license key provided by BICS.
There after the user runs the SpinCoach from his or her hard disk. The user must login for each session. The information
gathered at login is used to track user performance history. System upgrades will be provided with new CD-ROMs.
5. Web-based Application. Balls spinning on drills can be photographed and their images stored on the BICS web
server. Such videos are on my web site. http://www.sie.arizona.edu/sysengr/baseball/index.html. These images along
with the software program will be on an Internet accessible web site. This system will be based on the Apache web
server with web pages written in PHP or HTML and video clips in AVI format. Access to the system will be granted by
monthly subscription and login based authentication control. The user will have a profile in the system and can access
this profile from any terminal connected to the Internet. The system will store user information in a database. This
Studies in Engineering and Technology Vol. 1, No. 1; 2014
34
information is used to track user performance history. System upgrades can be made on the web site at any time and
will be transparent to the user.
6. Make it into a video game and sell it to Nintendo, Sony or Microsoft.
Evaluation criteria for the SpinCoach
Fidelity of Images. How realistic are the images? Are they two or three-dimensional? What is the resolution? What is
the color depth? What is the update rate? Will the presentation vary depending on the processor speed or the
communications bandwidth? For example, would the system degrade with a dialup telephone connection to the
Internet? This criterion traces to the Operational Concept Description (OCD). Importance weight is 6.
Feedback Promptness. The system shall provide positive or negative feedback to the player after each prediction. The
system shall provide this feedback to the player within 500 milliseconds of the player's response. This will be a Boolean
(yes or no) function. This traces to customer requirement 10. Importance weight is 10.
Product Production Cost is a measure of how much it will cost in U. S. dollars for BICS to produce one unit of the
product. A monotonic decreasing scoring function shall be used (L=0, B=10, S=-0.1, U=500). Input range is 0 to 500
dollars, baseline is 10 dollars and slope is -0.1. This traces to customer requirement 1. Importance weight is 6.
Figure 2. A scoring function for the Product Production Cost evaluation criteria
The Shipping Cost evaluation criterion is composed of Shipping Weight, Shipping Expenses, Shipping Effort and
Billing Cost per unit. A scoring function is not necessary if the subcriteria have scoring functions and normalized
weights. Our target value is ten dollars. This traces to customer requirement 1. Importance weight is 5.
Updatability. This criterion evaluates how easy and convenient updates are expected to be. The system shall be
continually improved and updated throughout the system life cycle. Corrective maintenance such as bug fixes should be
performed within weeks. Adaptive maintenance, which includes revisions necessary to allow the system to run on new
or improved hardware and software, should be accomplished in a monthly time frame. Performance and functional
updates will be performed yearly. This should trace to the business plan. Importance weight is 5.
3.9 Treating Gains and Losses Equally
People do not treat gains and losses equally. Kahneman earned the Nobel Prize for explaining the fact that people
prefer to avoid losses rather than to acquire gains. Prospect Theory (Kahneman & Tversky, 1979) suggests that
psychologically losses are twice as powerful as gains. Would you rather get a 5% discount, or avoid a 5% penalty?
Most people would rather avoid the penalty. In a tradeoff study, you will get a different result if the scoring function
expresses losses rather than gains (Abdellaaoui, 2000) (Smith et al., 2007).
Recommendation: Human unequal treatment of gains and losses suggests that scoring functions in a tradeoff study
should uniformly express either gains or losses. Principles of linguistic comprehensibility suggest that criteria should be
worded in a positive manner, so that more is better. For example, use Uptime rather than Downtime, Mean Time
Between Failures rather than Failure Rate, and Probability of Success rather than Probability of Failure. Finally, when
using scoring functions, make sure that more output is better.
Team Excerpt
2.2 Number of Complaints
2.3 Number of Problems with the System
3.1 Number of Accidents per visit.
2.5.3. Number of Curses per day
5.2.6. Injury -- Is it possible for the design to inflict bodily injury on the batter? This rated by the players on a scale of
1 – 10 (1 being no bodily harm, 10 being serious injury requiring hospitalization).
Why is this a mental mistake?
These criteria are phrased negatively.
Suggested rewrite
Studies in Engineering and Technology Vol. 1, No. 1; 2014
35
2.2 Customer Approval Rating (%)
2.3 Mean Time to Failure (MTTF)
3.1 Number of accident-free visits
2.5.3 Time without cursing
5.2.6. Safety – Mean time between injuries.
Second Team Excerpt
3.5.2.6 Availability of system
This figure of merit measures the availability of the system, in terms of the number of hours per year for which the
system is unavailable during working hours due to failure or unscheduled maintenance.
3.5.2.7 Safety of system
This figure of merit measures the safety of the system in terms of injuries caused by the system.
Why is this a mental mistake?
Evaluation criteria should be phrased so that more is better. Availability of System and Safety of
System contradict themselves. The titles are phrased positively, but the descriptions contain the
negative words unavailable and injuries.
Suggested rewrite
3.5.2.6 System Availability
This criterion is the percent of time that the system is available during working hours,
average per week.
3.5.2.7 System Safety
The units for this criterion are the number of days that the system has operated
without an injury.
3.10 Not Using Scoring Functions
Evaluation data are transformed into normalized scores by scoring functions (utility curves) or qualitative scales (fuzzy
sets). The shape of scoring functions should ideally be determined objectively, but often, subjective expert opinion is
involved in their preparation. Creating scoring function packages takes time and effort (Bahill, 2008). A scoring
function package should be created by a team of analysts, and be reevaluated with the stakeholders with each use. Most
tradeoff studies that we have observed in industry did not use scoring functions. In some cases, scoring functions were
explained in the company process, but they were not convenient, hence they were not used (Smith et al., 2007).
Recommendation: Wymorian standard scoring functions (or similar scoring functions, fuzzy sets or utility functions)
should be used in tradeoff studies. Easy-to-use scoring functions should be referenced in company processes.
Team excerpt:
Many teams just did not use scoring functions.
Evaluation data are transformed into normalized scores by scoring functions (Wymore, 1993, pp. 385-398; Daniels,
Werner & Bahill, 2001). The shape of scoring functions should ideally be determined objectively, but often, subjective
expert opinion is involved in their preparation. Scoring functions are also called utility functions, utility curves, value
functions, normalization functions and mappings. Evaluation criteria should always have scoring functions so that the
preferred alternatives do not depend on the units used. For example, see what would happen if you were to add values
for something that cost about one hundred dollars and lasted about a millisecond.
Alt-1 cost a hundred dollars and lasts one millisecond, Sum = 100.001.
Alt-2 only cost ninety-nine dollars but it lasts two millisecond, Sum = 99.002.
The duration does not have any effect on the decision. A simple program that creates scoring functions is available free
at http://www.sie.arizona.edu/sysengr/slides. It is called the Wymorian Scoring Function tool. An example of a scoring
function was given in Figure 2 for the Product Production Cost.
Scoring functions must state the units for the input: for example, actual dollar values will be used as input to a cost
scoring function. Without scoring functions, the preferred alternative would depend on the units used, for example,
whether the costs were given in U. S. Dollars or British pounds. With scoring functions, this will not happen.
3.11 Implying False Precision
The most common mistake that we have seen in tradeoff studies is implying false precision. For example, a tradeoff
analyst might ask a subject matter expert to estimate values for two criteria. The expert might say something like, “The
first criterion is about 2 and the second is around 3.” The analyst puts these numbers into a calculator and computes the
ratio as 0.666666667. This is nonsense, but nine digits might be used throughout the tradeoff study. The Forer Effect
might explain this. The analyst believes that the calculator is an impeccable authority in calculating numbers. Therefore,
what the calculator says must be true (Smith et al., 2007).
Studies in Engineering and Technology Vol. 1, No. 1; 2014
36
Recommendation: Use significant figures methodology. Furthermore, in numerical tables, print only a sufficient
number of digits after the decimal place as is necessary to show a difference between the preferred alternatives.
Team Excerpt
5.3.1.1 Trade-off scores
Concept 1: Customer chooses the bat 0.6 * 0.22500 + 0.4 * 0.41435 = 0.30074
Concept 2: Store owner chooses the bat 0.6 * 0.25125 + 0.4 * 0.41435 = 0.31649
Concept 3: BatChooser chooses the bat 0.6 * 0.67500 + 0.4 * 0.96840 = 0.79236
Concept 4: BatSelect Chooses with the help of the BatChooser 0.6 * 0.84280 + 0.4 * 0.96840 = 0.89304
Presenting five digits after the decimal point obfuscates the equations and does not help to
differentiate between the alternatives.
Suggested rewrite
5.3.1.1 Trade-off scores
Concept 1: Customer chooses the bat 0.6 * 0.23 + 0.4 * 0.41 = 0.30
Concept 2: Storeowner chooses the bat 0.6 * 0.25 + 0.4 * 0.41 = 0.32
Concept 3: BatChooser chooses the bat 0.6 * 0.68 + 0.4 * 0.97 = 0.79
Concept 4: BatSelect chooses the bat with the help of the BatChooser
0.6 * 0.84 + 0.4 * 0.97 = 0.89
Table 7. Second Team Excerpt for Bat
Chooser
Concept Weighted
Score Ranking
Concept 1 Score 0.4414 7
Concept 2 Score 0.5420 4
Concept 3 Score 0.2196 8
Concept 4 Score 0.6728 2
Concept 5 Score 0.6558 3
Concept 6 Score 0.4652 6
Concept 7 Score 0.6802 1
Concept 8 Score 0.4860 5
Table 8. Suggested rewrite for the Bat Chooser
Alternative Weighted
Score Ranking
Concept 1 0.44 7
Concept 2 0.54 4
Concept 3 0.21 8
Concept 4 0.67 2
Concept 5 0.65 3
Concept 6 0.46 6
Concept 7 0.68 1
Concept 8 0.48 5
Table 9. Team Excerpt for the SpinCoach
Performance
Requirements Value
Normalized
weight
Studies in Engineering and Technology Vol. 1, No. 1; 2014
37
1. Accuracy 8 0.235294
1.1 Spin Rate 10 0.384615
1.2 Launch Angle 8 0.307692
1.3 Launch Speed 8 0.307692
2. Consistency 7 0.205882
3. Ease of Use 6 0.176471
3.1 Portability 6 0.260870
3.2 Location 7 0.304348
3.3 # of Operators 10 0.434783
4. Opportunity 8 0.235294
5. Feedback 5 0.147059
Why is this a mental mistake?
The original Value data have one significant digit. Therefore, the normalized weights certainly should
not have six digits after the decimal point.
Table 10. Suggested rewrite for the SpinCoach
Evaluation Criteria
We
ights
of
Import
ance
Crite
ria
Norm
aliz
ed
We
ights
*
Subcrite
ria
Norm
aliz
ed
We
ights
*
1. Accuracy 8 0.24
1.1 Spin Rate 10 0.38
1.2 Launch Angle 8 0.31
1.3 Launch Speed 8 0.31
2. Consistency 7 0.21
3. Ease of Use 6 0.18
3.1 Portability 6 0.26
3.2 Location 7 0.30
3.3 Number of
Operators
10 0.43
4. Opportunities per Hour 8 0.24
5. Feedback Response Time 5 0.15
*Significant figures methodology suggests that the
normalized weights should only have one significant
digit. But here we have used two to make the
calculations obvious.
Table 10 has also put the criteria and subcriteria weights in separate columns. All of
the subcriteria weights for a particular criteria sum to 1.0. All of the criteria weights
sum to 1.0.
On the other hand, Table 5 shows an example not of false precision, but of inconsistent precision. In some columns,
Excel is set to display integers while in others it is set to one decimal place.
When determining how many digits should be printed consider (1) how many digits are necessary to differentiate
between the preferred alternatives, (2) the sensitivity of the final recommendations to the parameters, the most sensitive
parameters should be given extra resources and therefore perhaps more significant digits (Karnavas, Sanchez & Bahill,
1993; Smith, Szidarovszky, Karnavas & Bahill, 2008) and (3) no parameters need to be more exact than the least
accurate parameter (presuming of course that the tradeoff matrix is as described in this paper and it does not have a
multi-step process for estimating any parameters).
As an example, when humans state preferences between risky prospects, their judgments are not linear in probability.
Humans overweight small probabilities and underweight high probabilities. This has been modeled with several
different equations (Abdellaoui, 2000; Bleichrodt & Pinto, 2000).
Studies in Engineering and Technology Vol. 1, No. 1; 2014
38
1( )
(1 )
pw p
p p
( )(1 )
pw p
p p
and
( ) exp( ( ln ) )w p p .
In these equations w is the probability weighting function, p is the probability of a particular prospect, , , and are parameters fit to experimental data of individual humans. Now the question becomes, Should a tradeoff analyst use
equations like these when eliciting information for a tradeoff study? The answer depends on the accuracy of the other
parameters. In all of the design documents that we examined, the weights of importance had only one significant digit.
Therefore correcting for each human’s incorrect estimation and use of probabilities is not warranted. In a tradeoff study,
the number of significant figures should be determined for each parameter. Then resources should not be committed to
increasing the number of significant figures for any except the most important and the least precise parameters.
3.12 Obviating Expert Opinion
An analyst could hold a circular belief that expert opinion or review is not necessary, because no evidence for the need
of expert opinion is present. This is especially true if no expert has ever been asked to comment on the tradeoff study
(Smith et al., 2007).
Recommendation: Experts should be sought, formally or informally, to evaluate the results of tradeoff studies.
The most common mistake that we have found in design projects over the last 35 years is failing to talk with
stakeholders and failing to consult experts and experienced advisors. The university and local industry is full of experts
in the fields of every project that we have done. In this time, very few teams have sought advice from domain experts.
Why do people fail to seek out the advice of experts and experienced advisors? The students rated the following
possible reasons. In each category, the reasons are arranged from the most frequent to the least.
It was common for our teams to not seek outside advice or guidance in the course of performing their tradeoff studies. If
they had sought this guidance, expert review or opinion, they might have avoided the errors we detected in their tradeoff
studies. This would most likely be the case if the guidance concerned the tradeoff study itself (not just the technical
matters) and elicited high-quality examination of all tradeoff study components.
Table 11. Possible reasons for failing to talk with stakeholders, experts and advisors.
Timidity
Perhaps they do not want to inconvenience the wise men or waste their time. However, people are not reluctant
to seek the advice of physicians, tax accountants and lawyers. To overcome timidity, before you talk to an
expert, you should formulate your questions and explain your problem in a way that the expert can quickly
understand. Tailoring a message and formulating the right questions is hard and must be done iteratively.
Before you leave your meeting, you should state what you think the expert said, to make sure that you
understood and to prevent having to go back later for clarification.
Perhaps they are shy or intimidated by experts.
Perhaps they fear that the incompleteness of their project will be interpreted as incompetence.
Perhaps they think that a face-to-face meeting would display their naïveté. This is not a problem with e-mails,
because most people do not expect e-mails to be thoughtful, coherent and grammatically correct: most
students do not edit their e-mails or use a spelling and grammar checker on them.
Perhaps they think that seeking advice reveals their ignorance, and that ignorance is shameful.
Perhaps they think that consulting experts shows weakness, whereas going it alone shows strength.
Perhaps they feel that, because they do not have a charge number, they cannot ask experts in their company for
advice.
Perceived Value
Studies in Engineering and Technology Vol. 1, No. 1; 2014
39
Perhaps they do not realize the usefulness of face-to-face meetings with experts.
Perhaps it is a matter of return on investment. Consulting experts takes time and effort. Perhaps these teams
thought the improvement in their tradeoff studies would not be worth the effort of consulting experts.
Perhaps the smart people thought, “We can get an A without wasting our time talking to our advisor.”
Perhaps the new technology generation thinks that they can just Google the web and get all the information that
experts might provide.
Perhaps they noticed that other courses at the university do not provide world-class experts to meet with them, so
it must not be important.
Perhaps they do not see a direct correlation between their grade and meetings with their advisor.
Perhaps they do not perceive added value.
Time (obviously time and perceived value will be traded off)
Perhaps they thought that they were too busy; meeting with their advisor would take time and effort; it would be
hard to schedule meetings with their advisor. Maybe they were just lazy.
Communication
Perhaps they have had no experience initiating a meaningful conversation with a stranger and are therefore
reluctant to do so.
Most of our students communicate with cell phones, twitter, the internet or e-mail. So they are multitaskers,
jumping from task to task. Therefore, they are good at multitasking, but their attention spans are perhaps
short. So they do not know how to talk face-to-face with an expert.
Other
Perhaps they have been taught that engineers work alone: after all, cooperating on exams is frowned upon.
However, in the modern industrial environment, engineering is done by teams and when success is
important consultants are also hired.
Perhaps they are reluctant to change or they don’t want to do it someone else’s way. If you ask for advice, then
you should use the advice you are given.
Foreign students said, “It’s embarrassing to show weakness in the English language” and “Our culture teaches us
to not approach an advisor or mentor.”
Similarly, the medical profession does not practice extensive consultation with experts. In one study of over 300 breast
cancer surgeons (Katz et al., 2010), only one-fourth typically consulted medical oncologists, radiation oncologists or
plastic surgeons prior to surgery. About two-thirds of the surgeons reported that almost none of their patients
participated in patient decision-support activities, such as attending a practice-based presentation, viewing web-based
materials or participating in peer-support programs.
“He who trusts in himself is a fool, but he who walks in wisdom is kept safe” (Proverbs 28; 26)
4. Statistical Summary of Mental Mistakes
We examined 110 project reports composed of over 8000 pages of text that had been submitted over the last two
decades and we compiled the following statistics.
Type of Mental Mistake Number of
mistakes found
Using Dependent Criteria 75
Not Stating the Problem in Terms of
Stakeholder Needs 62
Vague Problem Statement 65
Substituting a Related Attribute 24
Sensitivity Analysis Mistakes 91
Forer Effect 61
Studies in Engineering and Technology Vol. 1, No. 1; 2014
40
Weight of Importance Mistakes 11
Anchoring and the Status Quo 69
Treating Gains and Losses Equally 46
Not Using Scoring Functions 32
Implying False Precision 59
Obviating Expert Opinion 85
Other Mental Mistakes 128
In these 110 tradeoff study reports, we found 808 mental mistakes. We put these in a spreadsheet. Then for 100 of these
mental mistakes, we made example cases using (1) direct quotes from the original reports, (2) an explanation of the
mistake, (3) related paragraphs from Smith et al. (2007) and (4) recommended revisions. We used a format similar to
that used in this paper. Based on the perceived heuristic value, uniform coverage of the 12 mental mistakes and
minimization of the number of projects that would have to be explained, we selected the two dozen development cases
(team excerpts) that are in this paper, and another 50 cases that were used for testing. Then 20 Raytheon engineers and
50 University of Arizona students tried to identify the mental mistakes in the 50 excerpts of the test set. The average
agreement was about 80%.
In general, we found no correlation between the number of mistakes we detected and the students’ grades on the reports.
There are several reasons for this. First, the reports with the highest grades were usually written better and they were
therefore easier to understand. The better we understand something the easier it is to find and identify mental mistakes
in it. Second, the better reports tended to be longer and more complete, and thus there was more opportunity for mental
mistakes. On the other hand, the poor reports showed more instances of mistakes of omission, such as failure to use
scoring functions, incomplete sensitivity analyses and failure to get advice from experts.
5. Who Cares?
Who cares about mistakes in doing tradeoff studies? Perhaps everyone should. If a tradeoff study is not performed or is
done badly it could cost a company a lot of money. As an example will now consider the San Diego Airport Site
Selection Tradeoff Study. This is a large, expensive, publicly accessible tradeoff study that contains mental mistakes
like those we present in this paper. This tradeoff study took six years and cost 17 million dollars. When its results were
presented to the voters in November of 2006, the voters turned the proposal down and the $17M was wasted. Some
mistakes might have been made in conducting this tradeoff study.
They did a tradeoff study, but only four of the ten tradeoff study components were utilized: Problem Statement,
Alternate Solutions, Evaluation Criteria and Preferred Alternatives.
They used five evaluation criteria: Aeronautical, Environmental, Market, Military and Financial. The criteria were
arranged hierarchally with subcriteria and subsubcriteria. However, the criteria did not have weights of importance or
scoring functions.
They had a dozen alternative sites, including the Do Nothing alternative. They often added and deleted alternatives. For
example, the floating platform in the Pacific Ocean was dismissed early. The Campo and Borrego Springs sites were
added late, so these sites had greater visibility in the public mind. However, the Campo and Borrego Springs sites were
similar so, because of distinctiveness by addition of alternatives, they faded away.
They did a rudimentary sensitivity analysis looking at changes in their planning parameters at two different demand
levels. They also did a small sensitivity analysis showing changes in total cost as a function of available funding
(without issuing bonds or increasing taxes).
The interim results of the study were continually being reported in the press. So they certainly received a lot of expert
opinions. However, in the end, the voters did not trust the study. The objectivity of the Regional Airport Authority that
conducted the study was questioned. It appeared over time that the Authority was more interested in supporting a
particular airport site than in explaining the various options for the voters. The Authority was perceived as being
pro-business and anti-military. The difficulties of military and civilian joint use were not ameliorated. San Diegans were
happy with the status quo: Lindberg Field was good for its community and the Marine Corps Air Station was good for
its community. The Authority did not show a burning platform or a compelling reason for change. It seemed that they
only considered future business growth.
The ballot proposal asked, “Should Airport Authority and government officials work toward obtaining 3,000 acres at
MCAS Miramar by 2020 for a commercial airport, providing certain conditions are met?” It was turned down 38% to
Studies in Engineering and Technology Vol. 1, No. 1; 2014
41
62%. We are not saying that if they had done a more thorough tradeoff study, then the ballot proposal would have
passed. We are only saying that they could have done a better tradeoff study.
6. Summary
Good industry practices for improving the probability of success of tradeoff studies include having teams evaluate the
data, evaluating the data in many iterations and expert review of the results and recommendations. It is important that
the review teams have a substantial number of reviewers that are external to the project and that the reviewers consider
design decisions as well as simple checking to ensure that tasks were done. Reviews are often hampered by failure to
allow external reviewers access to proprietary or classified data. To improve your tradeoff study process you should
inform your decision makers about how mental mistakes affect tradeoff studies (forewarned is forearmed), encourage a
long-term institutional decision horizon, use a team approach with frequent iterations and institute both expert and
public reviews. Finally try to reduce mental mistakes by using the recommendations of this paper.
The literature on Decision Making, Cognitive Science and Experimental Economics contains hundreds of experiments
showing persistent human mistakes of judgment. Smith et al. (2007) postulated 28 specific types of mental mistakes that
could affect the ten specific components of a tradeoff study. This paper has shown examples in a dozen categories of
mental mistakes that systems engineers have actually made while creating tradeoff studies. The research found over
eight hundred examples of such mistakes: these examples prove that systems engineers really do make such mistakes.
References
Abdellaaoui, M. (2000). Parameter-free elicitation of utility and probability weighting functions, Management Science,
46(11), 1497-1512. http://dx.doi.org/10.1287/mnsc.46.11.1497.12080
Bahill, A. T. (2008). Decision Making and Tradeoff Studies,
http://www.sie.arizona.edu/sysengr/slides/decisionMaking.ppt, last accessed March 2010.
Bahill, A. T., & Baldwin, D. G., (2007). Describing baseball pitch movement with right-hand rules, Computers in
Biology and Medicine, 37, 1001-1008. http://dx.doi.org/10.1016/j.compbiomed.2006.06.007
Bahill, A. T., & Baldwin, D. G. (2008). Mechanics of baseball pitching and batting, Chapter 16 Applied Biomedical
Engineering Mechanics, Dhanjoo Ghista, CRC Press and Taylor & Francis Asia Pacific, pp. 445-488.
Bahill, A. T., Baldwin, D. G., & Ramberg, J. S. (2009). Effects of altitude and atmospheric conditions on the flight of a
baseball, International Journal of Sports Science and Engineering, 3(2), 109-128. ISSN 1750-9823 (print).
http://www.worldacademicunion.com/journal/SSCI/online.htm
Bahill, A. T., Baldwin, D. G., & Venkateswaran, J. (2005). Predicting a baseball's path, American Scientist, 93(3),
218-225.
Bahill, A. T., & Dean, F. F. (1999, 2009). Discovering system requirements, Chapter 4 in Handbook of Systems
Engineering and Management, A.P. Sage & W.B. Rouse (Eds), John Wiley & Sons, first edition 175-220, 1999,
second edition, 205-266.
Bahill, A. T., & Karnavas, W. J. (2000). Risk analysis of a pinewood derby: A case study, Systems Engineering, 3(3),
143-155. http://dx.doi.org/10.1002/1520-6858(200033)3:3<143::AID-SYS3>3.0.CO;2-0
Bahill, A. T., & LaRitz, T. (1984). Why can't batters keep their eyes on the ball, American Scientist, 72, 249-253.
Bahill, A. T., & Smith, E. (2009). An industry standard risk analysis technique, Engineering Management Journal,
21(4), 16-29.
Bahill, A. T., & Stark, L. (1975). Neurological control of horizontal and vertical components of oblique saccadic eye
movements, Mathematical Biosciences, 27, 287-298. http://dx.doi.org/10.1016/0025-5564(75)90107-8
Bahill, A. T., & Stark, L. (1977). Oblique saccadic eye movements: Independence of horizontal and vertical channels,
Archives of Ophthalmology, 95, 1258-1261. http://dx.doi.org/10.1001/archopht.1977.04450070156016
Baldwin, D. G., Bahill, A. T., & Nathan, A. (2007). Nickel and dime pitches, Baseball Research Journal, 35, 25-29.
Bleichrodt, H., & Pinto, J. L. (2000). A parameter-free elicitation of the probability weighting function in medical
decision analysis, Management Science, 46(11), 1485-1496. http://dx.doi.org/10.1287/mnsc.46.11.1485.12086
Chapman, W. L., Bahill, A. T., & Wymore, A. W. (1992). Engineering modeling and design, CRC Press Inc., Boca
Raton, FL.
Chrissis, M. B., Konrad, M., & Shrum, S. (2003). CMMI: Guidelines for Process Integration and Product Improvement,
Pearson Education Inc., Boston.
Studies in Engineering and Technology Vol. 1, No. 1; 2014
42
CMMI for Development, ver. 1.3, (2010). Software Engineering Institute, Pittsburgh, PA,
http://cmmiinstitute.com/resource/cmmi-for-development-version-1-3-microsoft-word/
Retrieved November 2013
Daniels, J. P., & Bahill, A. T. (2004). Hybrid process combines traditional requirements and use cases, Systems
Engineering, 7(4), 303-319. http://dx.doi.org/10.1002/sys.20013
Daniels, J., Werner, P., & Bahill, A. T. (2001). Quantitative methods for tradeoff analyses, Systems Engineering, 4(3)
190-212. http://dx.doi.org/10.1002/sys.1016
Forer, B. R. (1949). The fallacy of personal validation: A classroom demonstration of gullibility, Journal of Abnormal
Psychology, 44, 118-121. http://dx.doi.org/10.1037/h0059240
Hanselmann, M., & Tanner, C. (2008). Taboos and conflicts in decision making: sacred values, decision difficulty, and
emotions, Judgment and Decision Making, 3(1), 51-63.
Hooks, F., & Farry, K. A. (2001). Customer-centered products: Creating successful products through smart
requirements management, AMACOM, New York.
Hsu, F. K., Bahill, A. T., & Stark, L. (1976). Parametric sensitivity of a homeomorphic model for saccadic and
vergence eye movements, Computer Programs in Biomedicine, 6, 108-116.
http://dx.doi.org/10.1016/0010-468X(76)90032-5
Hull, E., Jackson, K., & Dick, J. (2005). Requirements engineering, Springer, London.
INCOSE Systems Engineering Handbook v3.2.2, (2011). Retrieved November 2013
http://www.incose.org/ProductsPubs/incosestore.aspx
Kahneman, D. (2002). Maps of bounded rationality: a perspective on intuitive judgment and choice, Nobel Prize
Lecture, December 8, 2002, this site also has a video of the speech,
http://nobelprize.org/nobel_prizes/economics/laureates/2002/kahneman-lecture.html
Kahneman, D. (2003). A perspective on judgment and choice: Mapping bounded rationality, American Psychologist,
58(9), 697-720. http://dx.doi.org/10.1037/0003-066X.58.9.697
Kahneman, D., & Tversky, A. (1979).Prospect theory: An analysis of decision under risk, Econometrica, 46(2),
171-185.
Karnavas, W. J., Sanchez, P. J., & Bahill, A. T., (1993). Sensitivity analyses of continuous and discrete systems in the
time and frequency domains, IEEE Transactions on Systems, Man and Cybernetics, SMC-23, 488-501.
http://dx.doi.org/10.1109/21.229461
Katz, S. J., Hawley, S. T., Morrow, M., Griggs, J. J., Jagsi, R.., Hamilton, A. S., Graff, J. J., Friese, C. R., & Hofer, T. P.
(2010). Coordinating cancer care: patient and practice management processes among surgeons who treat breast
cancer, Medical Care, 48(1), 45-51. http://dx.doi.org/10.1097/MLR.0b013e3181bd49ca
Keeney, R. L. (1992). Value-focused thinking: A path to creative decision making, Harvard University Press,
Cambridge, MA.
Marquard, L., & Robinson, S. M. (2008). Reducing perceptual and cognitive challenges in making decisions with
models, in Decision Modeling and Behavior in Complex and Uncertain Environments, Eds. T. Kugler, J. C. Smith,
T. Connolly & Y. J. Son., Springer, Science+Business Media, NY, NY, pp. 33-55.
McBeath, M. K., Nathan, A. M., Bahill, A. T., & Baldwin, D. G. (2008). Paradoxical pop-ups: Why are they hard to
catch? American Journal of Physics, 76(8), 723-729. http://dx.doi.org/10.1119/1.2937899
Piattelli-Palmarini, M. (1994). Inevitable illusions: How mistakes of reason rule our minds, John Wiley & Sons, New
York.
RSAS, Press release: Nobel Prize in Economics 2002, Retrieved September 2008 from The Royal Swedish Academy of
Sciences: http://nobelprize.org/nobel_prizes/economics/laureates/2002/press.html.
San Diego Airport Site Selection,
http://www.san.org/documents/sdcraa/archives/Airport%20Site%20Selection%20Program/Technical%20Analysis/
Financial%20Comparative%20Analysis%20Presentation.pdf Accessed November 2013.
Sage, A. P. (1981). Behavioral and organizational considerations in the design of information systems and processes for
planning and decision support, IEEE Transactions on Systems, Man and Cybernetics, SMC-11(9) 640-680.
http://dx.doi.org/10.1109/TSMC.1981.4308761
Studies in Engineering and Technology Vol. 1, No. 1; 2014
43
Smith, E. D., Piattelli-Palmarini, M., & Bahill, A. T., (2008). Cognitive biases affect the acceptance of tradeoff studies,
in Decision Modeling and Behavior in Complex and Uncertain Environments, Eds. T. Kugler, J. C. Smith, T.
Connolly & Y. J. Son., Springer, Science+Business Media, NY, NY, pp. 227-249.
Smith, E. D., Son, Y. J., Piattelli-Palmarini, M., & Bahill, A. T., (2007). Ameliorating the effects of cognitive biases on
tradeoff studies, Systems Engineering, 10(3), 222-240. http://dx.doi.org/10.1002/sys.20072
Smith, E. D., Szidarovszky, F., Karnavas , W. J., & Bahill, A. T. (2008). Sensitivity analysis, a powerful system
validation technique, The Open Cybernetics and Systemics Journal, 2, 39-56.
http://dx.doi.org/10.2174/1874110X00802010039.
Tversky, A., & Shafir, E. (1992). The disjunction effect in choice under uncertainty, Psychological Science, 3(5),
305-309. http://dx.doi.org/10.1111/j.1467-9280.1992.tb00678.x
Wymore, A. W. (1993). Model-based systems engineering, CRC Press Inc., Boca Raton, FL.
Zhu, A., Aurum, I., Gorton. & Jeffery, R. (2005). Tradeoff and sensitivity analysis in software architecture evaluation
using analytic hierarchy process, Software Quality Journal, 13(4), 357-375.
http://dx.doi.org/10.1007/s11219-005-4251-0
Biographical Sketches
James Bohlman works for Intex, a small startup making HF IR emitters for gas detection technology. He is working to
specify and obtain equipment for testing and burn-in. He joined STRC a small startup in May 2008 as the senior
engineer and has been involved almost from the beginning in the construction of the lab space. STRC is a research
group dedicated to the purification of metallurgical-grade silicon into solar-cell useable silicon by metallurgical
refination and hydrometallurgy. He attended University of Wisconsin at Madison where receiving a BS in Metallurgical
and Mining Engineering. He worked at Texas Instruments in Dallas Texas in the pilot line wafer fab metallization area
running sputter deposition PVD equipment. He then worked in a production wafer fab called DMOS IV processing 6"
silicon wafers into one megabit DRAM memory circuits later working on application specific integrated circuits (ASICs)
as the memory business went to Japan. James also worked at Analog Devices in Norwood, Massachusetts as a process
engineer in metallization and implant primarily on high-end bipolar circuits customizable with a trimable SiCr (Silicon
Chrome) resistor. Then at Burr Brown Corp., (which subsequently became Texas Instruments in 2000) he was a CVD
and PECVD process engineer producing high-performance bipolar A-to-D circuits. He also worked on implementation
of a special resistor process to make the parts more reliable, reducing quality issues and customer concerns. He has an
MBA from the University of Phoenix and a Masters of Engineering in Systems Engineering from the University of
Arizona; two degrees which complement each other in their approach to projects, systems and business.
Terry Bahill is an Emeritus Professor of Systems Engineering and of Biomedical Engineering at the University of
Arizona in Tucson. He received his Ph.D. in electrical engineering and computer science from the University of
California, Berkeley, in 1975. Bahill has worked with dozens of high-tech companies presenting seminars on Systems
Engineering, working on system development teams and helping them to describe their Systems Engineering processes.
He holds a U.S. patent for the Bat Chooser, a system that computes the Ideal Bat Weight for individual baseball and
softball batters. He was elected to the Omega Alpha Association: the Systems Engineering Honor Society. He received
the Sandia National Laboratories Gold President's Quality Award. He is a Fellow of the Institute of Electrical and
Electronics Engineers (IEEE), of Raytheon Missile Systems, of the International Council on Systems Engineering
(INCOSE) and of the American Association for the Advancement of Science (AAAS). He is the Founding Chair
Emeritus of the INCOSE Fellows Selection Committee. His picture is in the Baseball Hall of Fame's exhibition
"Baseball as America." You can view this picture at http://www.sie.arizona.edu/sysengr/.
This work is licensed under a Creative Commons Attribution 3.0 License.
Studies in Engineering and Technology
Vol. 1, No. 1; February 2014
ISSN 2330-2038 E-ISSN 2330-2046
Published by Redfame Publishing
URL: http://set.redfame.com
44
The Valve Timing Optimization of the Diesel Engine Based on Response
Surface Methodology
Jun Li, Lei Ji, Yangjiao Xu & Jinli Xie
Chongqing Jiaotong University, Chongqing, China
Correspondence: Jun Li, Mechantronics and Automotive Engineering, Chongqing Jiaotong University, 400074,
Chongqing, China. Tel: 86-138-8390-1379. E-mail: [email protected]
Received: October 30, 2013 Accepted: November 11, 2013 Available online: December 5, 2013
doi:10.11114/set.v1i1.262 URL: http://dx.doi.org/10.11114/set.v1i1.262
Abstract
To study the effect of valve timing on the diesel engine performance, the simulation model of diesel engine was
established with AVL BOOST and its accuracy was proved. The volumetric efficiency is one of the important indicators
to evaluate engine performance. The volumetric efficiency as optimization objective and valve timing were optimized
and discussed by using Box-Behnken test method and the response surface methodology. Optimization result shows that
volumetric efficiency of the diesel engine can been increased by 6.42% under rated speed.
Keywords: response surface methodology, valve timing, volumetric Efficiency
1. Introduction
The simulation of the engine’s working process has become an important method to the research and development (Lei
et al., 2011; Rakopoulos et al., 2004; Razmjooei et al., 2010). It is established the one-dimensional simulation model of
the working process of diesel engine by AVL BOOST. And the simulation model is validated by speed characteristic.
Then the influence of valve timing on the volumetric efficiency of diesel engine is analyzed and optimized in using the
response surface methodology to obtain the optimal valve timing and the volumetric efficiency of engine. The
volumetric efficiency can be increased 6.42% under the rated speed.
2. Model Establishment
2.1 The Model Establishment
The 4-cylinder, 4-stroke and turbocharged diesel engine was used in bench test. The basic parameter of the diesel engine
is shown in Tab.1. The simulation model was established, shown in Fig.1. The Vibe 2 Zone heat release model and
Woschni1978 heat transfer model was used (Liu, 2011).
Table 1. The main parameters of diesel engine
Basic parameters Parameter values
Compression Ratio 18
Bore×Stoke ( mm ) 75×80
Rated Power( kW ) 65
Rated Speed( min/r ) 4500
Intake Valve Close(IVO)(ºCA) 63
Intake Valve Open(IVC)(ºCA) 101
Exhaust Valve Open(EVO)(ºCA) 105
Exhaust Valve Close(EVC)(ºCA) 63
2.2 Validation of the Simulation Model
The simulation model of diesel engine was verified from calculated results and experimental results, shown in Fig.2.
Studies in Engineering and Technology Vol. 1, No. 1; 2014
45
Figure 1. Simulation model of diesel engine
Figure 2. The calculated results compared with the experimental results
The comparative result shows that the relative deviation between simulation and experiment are both below
5%.Therefor the simulation model of diesel engine established in this paper is accurate and reliable.
3. Valve Timing Optimization Based on Response Surface Methodology
3.1 Response Surface Methodology
Response Surface Methodology is a product with the development of statistics, mathematics and computer science.
Experimenting, modeling, analyzing data, using graphics technology are used in order to show up the relationship of
response system and we can know and select the optimized response of experiment design directly (Montgomery, 2007;
Chen et al, 2009; Simate et al., 2009; Li et al., 2007; Liu, S. S. et al., 2012; Liu, C. et al., 2012). Response surface
analysis of experiment includes Central Composite Design, Box-Behnken Design, Quadratic saturation D-optimal
Design, Uniform Design, etc. Box-Behnken Design which can be called the efficiency design method, can be estimated
in the Linear and Quadratic polynomial with the Linear interaction of polynomial model by fewer tests (Chen et al,
2009). The influence of valve timing on the volumetric efficiency of diesel engine under the rated speed is discussed
and the response surface experiment is designed by using Box-Behnken Design. The test points and test data are
showed in Tab.2.
Studies in Engineering and Technology Vol. 1, No. 1; 2014
46
Table 2. Testing and test data of Box-Behnken design
Std IVO IVC EVO EVC volumetric efficiency
1 73 101 105 73 0.8231
2 63 91 105 73 0.8066
3 63 111 105 53 0.8466
4 63 101 105 63 0.8400
5 63 91 105 53 0.7903
6 63 91 105 63 0.8400
7 53 101 105 73 0.8496
8 73 91 105 63 0.7896
9 73 101 115 63 0.8264
10 53 101 105 53 0.8268
11 63 101 105 63 0.8400
12 63 91 95 63 0.8034
13 63 111 105 73 0.8606
14 63 101 115 53 0.8216
15 63 101 105 63 0.8400
16 53 101 115 63 0.8481
17 73 101 95 63 0.8252
18 63 101 115 73 0.8441
19 63 101 105 63 0.8400
20 63 91 115 63 0.8090
21 63 101 95 73 0.8432
22 73 111 105 63 0.8469
23 53 101 95 63 0.8444
24 63 101 95 53 0.8258
25 63 111 95 63 0.8576
26 63 111 115 63 0.8628
27 53 91 105 63 0.8120
28 53 111 105 63 0.8632
29 73 101 105 53 0.8115
3.2 The Analysis of Response Surface
Design-Expert is a software which is used in optimized experiment by response surface widely. The equation of test
response surface is some supposed. The hypothesis is: H0:β1=β2=β3=…=βi=0, H1:β1, β2, β3, …, βi
Where: at least, one parameter is not equal to zero, and the significance level is α=0.05.The "P value" is a concept
which is the judgment instead of rejection region in statistics. The Box-Behnken Design of optimized analysis is used in
this paper. The Tab.3 shows that the P values of AC, BC, BD, CD and C2 are more than 0.05, which indicates the terms
in model are not notable, and need to be re-optimized.
Table 3. ANOVA for response surface quadratic model
Source Sun of Squares df Mean Square F value P-value
Prob>F
Model 0.012 14 0.00083 344.14 <0.0001
A-IVO 0.00122 1 0.00122 504.82 <0.0001
B-IVC 0.00878 1 0.00878 3609.10 <0.0001
C-EVO 0.000012 1 0.000012 5.10 0.0038
D-EVC 0.000945 1 0.000945 389.24 <0.0001
AB 0.000009 1 0.000009 3.82 0.0453
AC 0.000001 1 0.000001 0.64 0.4363
AD 0.000036 1 0.000036 2.44 0.00140
BC 0.000001 1 0.000001 0.037 0.8502
BD 0.000001 1 0.000001 0.0092 0.9247
CD 0.000009 1 0.000009 2.67 0.1244
A2 0.000151 1 0.000151 62.07 <0.0001
B2 0.000352 1 0.000352 145.02 <0.0001
C2 0.000002 1 0.000002 1.50 0.2409
D2 0.000336 1 0.000336 138.22 <0.0001
Residual 0.000034 14 0.000002
Lack of Fit 0.000034 10 0.000002
Pure Error 0.000 4 0.000003
Cor Total 0.012 28 R2=0.9971
Studies in Engineering and Technology Vol. 1, No. 1; 2014
47
The Tab.4 shows that the P values of all are less than 0.05, which indicates the terms in model are significant; then the
response surface equation can be used. The response surface model equation is:
22
2
007342.0*007517.0
004967.0**0028.0**001525.0*008883.0
*001017.0*027.0*010.0814.0efficiencyvolumetric
DB
ADABAD
CBA
(1)
Table 4. ANOVA for the optimized response surface quadratic model
Source Sum of Squares df Mean Square F Value p-value Prob>F
Model 0.012 9 0.00130 538.74 <0.0001
A-IVO 0.00122 1 0.00122 508.82 <0.0001
B-IVC 0.00878 1 0.00878 3635.70 <0.0001
C-EVO 0.000012 1 0.000012 5.14 0.0038
D-EVC 0.000945 1 0.000945 392.11 <0.0001
AB 0.000009 1 0.000009 3.85 0.0453
AD 0.000031 1 0.000031 12.99 0.0019
A2 0.000166 1 0.000166 68.72 <0.0001
B2 0.000380 1 0.000380 157.39 <0.0001
D2 0.000362 1 0.000362 150.15 <0.0001
Residual 0.000034 19 0.000002
Lack of Fit 0.000034 15 0.000002
Pure Error 0.000 4 0.000
Cor Total 0.012 28 R2=0.9971
Figure 5(a). Contour plot of response surface methodology
Figure 5(b). Contour plot of response surface methodology
Studies in Engineering and Technology Vol. 1, No. 1; 2014
48
Figure 6(a). 3D surface of response surface methodology
Figure 6(b). 3D surface of response surface methodology
The Fig.5 and Fig.6 shows that: IVO is 53ºCA, IVC is 111ºCA, EVO is 115ºCA, EVC is 71ºCA.
3.3 Verifying the Optimized Results of the Response Surface
The optimal valve timing which is got from the response surface methodology is used to verify the simulation model.
The Tab.5 shows that the volumetric efficiency is increased by 6.42% after optimized.
Table 5. Volumetric efficiency contrast of before and after optimization
Type IVO ºCA IVC ºCA EVO ºCA EVC ºCA volumetric efficiency
Original 63 101 105 63 0.841
Optimization 53 111 115 71 0.895
4. Conclusions
1) Based on the response surface methodology, we can optimize the volumetric efficiency and valve timing of diesel
engine, establish the response surface quadratic model, draw the 3D surface of response surface methodology and then
can carry out the optimization and interaction between volumetric efficiency and valve timing.
2) Getting the valve timing of the diesel engine from the response surface methodology, analyzed the simulation model
of the diesel engine. The results show that the volumetric efficiency can be optimal value under the rated speed.
Studies in Engineering and Technology Vol. 1, No. 1; 2014
49
Acknowledgements
This project is supported by National Natural Science Foundation of China (No. 51305472) and Education and
Teaching Reform Project of Chongqing CSTC, China (KJ090408; No.0903070)
References
Chen, X. K., Li, B. G., & Lin, Y. (2009). Application of Improved RSM in the Optimization Design of Automotive
Frontal Crashworthiness. Transaction of Beijing Institute of Technology, 1076-1077.
http://lib.cqvip.com/read/detail.aspx?ID=32654508
Lei, J. L., Huang, Z. P., Shen, L. H., et al. (2011). Working Process Simulation and Performance Optimization of
High-pressure Common Rail Diesel Engine. Vehicle Engine, 53-57.
http://202.202.244.3/kns50/detail.aspx?QueryID=27&CurRec=1
Li, J., & Liang, X. G. (2007). Scramjet Inlet Multi-Objective Optimization Based on response surface methodology.
Transactions of Nanjing University of Aeronautics & Astronautics, 205-209.
http://en.cnki.com.cn/Article_en/CJFDTOTAL-NJHY200703006.htm
Liu, C., Pan, X., Yan, Q. D., et al. (2012). Effect of Blade Number on Performance of Torque Converter and Its
Optimization Based on DOE and Response Surface Methodology. Transaction of Beijing Institute of Technology,
689-693. http://lib.cqvip.com/qk/95548X/201207/43098212.html
Liu, F. F., Jiang, C. H., & Xia, X. L. (2011). Effect on Performance of Gasoline Engine about Intake Valve Closing
Angle. Journal of Nanchang University (Engineering & Technology), 66-68.
http://lib.cqvip.com/read/detail.aspx?ID=37154422
Liu, S. S., Gu, Z. Q., Wu, W. G., et al. (2012). Multi-objective Collaborative Optimization of Vehicle Based on
Response Surface Methodology. Journal of Central South University (Science and Technology), 2586-2591.
http://lib.cqvip.com/qk/90745B/201207/42974900.html
Montgomery, D. C. (2007). Design and Analysis of Experiments, 6th Ed. New York: 404-406.
Rakopoulos, C. D., Giakoumis, E. G., Hountalas, D. T., et al. (2004). The Effect of Various Dynamic, Thermodynamic
and Design Parameters on the Performance of a Turbocharged Diesel Engine Operating under Transient Load
Conditions. SAE Paper. http://papers.sae.org/2004-01-0926/
Razmjooei, K., & Nagarajan, G. (2010). Optimization of DI Diesel Engine Operating Parameters Using a Response
Surface Method. SAE Paper. http://papers.sae.org/2010-01-1262/
Simate, G. S., Ndlovu, S., & Gericke, M. (2009). Bacterial leaching of nickel laterite using chemolithotrophic
microorganisms: Process optimization using response surface methodology and central composite rotatable design.
Hydrometallurgy, 4(3), 241-243. http://www.sciencedirect.com/science/article/pii/S0304386X09001236
This work is licensed under a Creative Commons Attribution 3.0 License.
Studies in Engineering and Technology
Vol. 1, No. 1; February 2014
ISSN 2330-2038 E-ISSN 2330-2046
Published by Redfame Publishing
URL: http://set.redfame.com
50
Network-based Management on Repairing Tool Kits of Civil Aviation
Engineering Maintenance
Xiaoxu Tian1, Xinlei Zheng
2, Ting Wang
2, Na Li
2, Haifeng Wang
2, Fuqing Huang
3
1Person-in-charge of project, Civil Aviation University of China, Tianjin, China
2Members of project, Civil Aviation University of China, Tianjin, China
3Instructor, Civil Aviation University of China, Tianjin, China
Correspondence: Xiaoxu Tian, Person-in-charge of project, Civil Aviation University of China, Tianjin 300300, China.
E-mail: [email protected]. Fuqing Huang, Instructor, Civil Aviation University of China, Tianjin 300300, China.
E-mail: [email protected]
Received: December 5, 2013 Accepted: December 23, 2013 Available online: January 16, 2014
doi:10.11114/set.v1i1.319 URL: http://dx.doi.org/10.11114/set.v1i1.319
Abstract
Based on features of high speed of network transmission and easy operations, this thesis covers two aspects to realize
network-based management on repairing tool kits of civil aviation engineering maintenance. Firstly, develop a network
inquiry system, which can help employees and administrators inquire tool borrowing information. Secondly, a module
is designed that has functions to transmit text messages to tool kits borrowers prompting them to return tools at times
when the returning date approaches.
Keywords: network, management, inquiry system, text messages
1. Introduction
As the civil aviation industry is developing very fast in China, working loads of engineering maintenance is increasing
with each passing day. The traditional registration method for repairing tool kits can no longer meet requirement by
modern management and hence. It is extremely urgent to develop a highly efficient management platform for
engineering maintenance tool kits. This thesis solves the problem from two aspects. Firstly, using ASP (Active Server
Page) technology and network database theory, we design a network inquiry system for repairing tool kits based on the
B/S (Browser/Server) mode. This system contains two parts, which are employee inquiry and administrator inquiry. For
employees, they can inquire their own real-time tool borrowing information and submit tool booking requests on line
according to their needs. For administrators, they are able to remotely inquire tool borrowing information of all the
employees and approve or reject applications by employees. Secondly, design a module that has text message
transmission functions based on GSM (Global System For Mobile Communication) network to tool borrowers
prompting them to return tools at times when the returning date is approaching.
2. Methodology
This system using the B/S model , the Dreamweaver homepage manufacture software, the ASP dynamic homepage
technology and the SQL Server(Structured Query Language Server) backstage data server, realizes network-based
management on repairing tool kits of civil aviation engineering maintenance.
In the mode of B/S, the client does not need any special software except the browser. The browser exchanging
information with database through the web (World Wide Web) server, can work in different platforms expediently. This
system achieves the function that employees and administrators inquire their borrowing information through the
browser based on the B/S model. Dreamweaver MX 2004 is used to design or develop web pages and web applications.
It is a strong software, which contains the visual layout tools, application development function and code editor support
and it has a strong function of multimedia webpage design.
ASP (Yue, et al., 2002) is a server-side scripting environment and it can create and run dynamic web pages or web
applications. When the browser of client sends a request to server, the ASP interpretive program of server carries out
the ASP program at the server-side and delivers the result to the browser in HTML (Hyper Text Markup Language)
format. In the production of web pages, the name of the object which saves and takes database is ADO (ActiveX Data
Studies in Engineering and Technology Vol. 1, No. 1; 2014
51
Objects). The main target of ADO is saving, taking or altering the data of source or adding data to the specified data
source.
This system uses SQL Server2000 (Shaosong, et al., 2006) as the data server. SQL Server2000 has many advantages,
such as convenient operation, good scalability and high related software integration degree and it can be used in many
platforms. In this system, the name of database is SCHOOL and many tabulations in the SCHOOL have been used,
such as EMP (tabulation of employee), ADM (tabulation of administrator), BOR (tabulation of borrowing tools),
ABOR (tabulation of available borrowing tools), BOO (tabulation of booking tools) and APP (tabulation of approval).
Furthermore, this system uses a lot of scripting language, which is an explanatory language. Not only can scripting
language be executed by browser directly, but also it can access to the various objects in the browser. With the help of
scripting language, this system completes the task of data validation before the tabulation is submitted and achieves the
function of opening or closing the window.
The above methodology is for the network inquiry system, the methodology of text message prompt modules will be
introduced in the fourth part of the thesis.
3. Network Inquiry System
3.1 Employee Inquiry
The network inquiry system (Chenzhi, et al., 2013; Iahad, et al., 2012) offers employee a platform, where they can
inquire their real-time tool borrowing information in order to avoid forgetting owing to too long a time. Figure 1
illustrates flowchart of the employee inquiry. Employees only input their account number in the log-in page to enter
their own tool borrowing information interface, where information about tool kits they have borrowed and tools that can
be borrowed can be searched. Based on the above two basic functions, we have perfected the network platform so as to
realize tool online booking functions. If employees apply online, tool kits can be borrowed, thus simplifying borrowing
procedures and increasing efficiency. Figure 2 shows the interface of employee inquiry.
Figure.1 Flowchart of employee inquiry
Figure 2. Interface of employee inquiry
Figure 1. Flowchart of employee inquiry
Studies in Engineering and Technology Vol. 1, No. 1; 2014
52
3.2 Administrator Inquiry
How to manage tools in an overall and unified way is hard to administrators in maintenance practices and in order to
solve this problem, we have developed the administrator inquiry, flowchart of which is shown in Figure 3. In the
flowchart, you can find four functions of the administrator inquiry.
There is no difference in appearances between employee and administrator inquiry interface. However, there is essential
difference in power and permission of them. First, for employee inquiry, only tool borrowing information of the
employee himself can be inquired. While for administrator inquiry, tool borrowing information of all employees can be
inquired. Second, employees only have right to apply for tool borrowing, while administrators have right to approve or
reject applications by employees.
4. Text Message Prompt Modules
4.1 Basic Targets of the Modules
In maintenance practices of employees, it is hard to avoid forgetting tool borrowing owing to too long a time, thus
decreasing tool utilization rates. We have developed text message transmission modules based on GSM network to deal
with this problem. When tool returning date is approaching, the module will transmit text messages to borrowers
prompting them to return the tools they have borrowed on time.
4.2 Fundamental Principles of the Text Message Transmission Modules
Flowchart of the module is shown in Figure 4.The upper computer transmits instructions to a single-chip
microcontroller for processing and AT (Attention) instructions are then transmitted to the GSM modules through a
serial port, completing control on the GSM modules and transmitting text messages.
4.3 Hardware Development of the Text Message Modules
The system hardware comprises two modules, i.e. controller and communication. As the communication module will
generate strong current during starting and text message transmission and reception . In order to enhance resistance to
interferences and expandability, we made the two modules in two separate boards. Figure 5 is a picture of the text
message transmission module.
Figure 3. Flowchart of administrator inquiry
Figure 4. Flowchart of text message transmission module
Studies in Engineering and Technology Vol. 1, No. 1; 2014
53
The controller mainly controls and protects the single-chip microcontroller, and C8051F020 (Ling, et al., 2006) is used
in this design as the MCU (Micro Control Unit), where pipeline architecture is used in its internal core at a speed as
high as 25MIPS (Million Instructions Per Second), ten times faster than common ones. The controller circuits comprise
eight sub-modules. They are C8051F020 minimum system, NOKIA5110 liquid crystal display, keyboard, SD card,
power, RS232 (Recommend Standard 232) communication interfaces, MC52 control interface and external memory
interface, totaling eight. Two modules, i.e. SD card and RS232 are described in detail as the following.
The reason for designing a SD memory module in the controller circuits is mainly for real-time records of messages
transmitted for future access. There are two bus modes, one being SD and one being SPI(Single Program Initiation), the
more popular SPI mode is used in this system. Figure 6 shows actual SD card driving circuit, where R31 - R35 are
pull-up resistors. The RS232 interface circuits are mainly for realization of data transmission between the single-chip
microcontroller and GSM module. Because the main controlling board is supplied with power 3.3V, so, the level
translation chip used is MAX3232, actual circuit of which is shown in Figure 7.
Core of the communication module is a GSM module (Wei, et al., 2008), which is a communication module similar to
the cell phone, having functions of sending text messages, voice communications and so on. Although being small, this
module has been utilized widely in many fields. The GSM module makes control by means of AT instructions and can
be connected to computer RS232 serial ports. Furthermore, it can be controlled by single-chip microcontrollers.
Compared with other kinds of remote transmission networks, GSM has higher security and confidentiality features,
which makes it suitable for applications in sensitive industrial sites. Besides, GSM networks covers wide areas and is
basically not limited by transmission distances and there is no need to establish special network and maintenance
Figure 5. A picture of text message transmission module
Figure 7. RS232 Interface circuits Figure 6. SD card driving circuits
Studies in Engineering and Technology Vol. 1, No. 1; 2014
54
networks, enjoying great superiority in communication costs. Hence, GSM modules are selected for this text message
prompting module.
4.4 Software Development of the Text Message Transmission Modules
The software developed this time is based on the platform Keil uVision 4, using language C as shown in Figure 8
program flowchart. After the device is powered, all the modules will be initialized and text messages can be sent after
success. Information transmitted can be stored in the SD card.
The AT instructions general apply in the connection and communication between PC and terminal equipment. As
mentioned in the previous paragraphs, the GSM (Biancucci, et al., 2013) modules perform control through AT
instructions. In other words, control processes on the GSM modules are, in reality, processes of sending AT instructions.
Sending a piece of text message is completed through the following steps:
1) Sending AT+CMGF, setting text message format;
2) Sending AT+CSCA, setting text message center number;
3) Sending AT+CMGS, setting user’s cell phone number;
4) Inputting characters to be sent.
The SD card memory (Huan, 2009) is another highlight in this design. In order to read text messages on different
platforms, FAT32 (File Allocation Table) file system is transplanted into the SD card read-write operations so as to
write all communication messages into SD card in format TXT, realizing data records.
5. Conclusions
We complete the network inquiry system and the text message prompt modules through intensive research and
development, realizing network management and efficient management on civil aviation engineering maintenance tool
kits. This system solves practically the hard-to-solve problem facing the civil aviation engineering maintenance sector
and it meets the modern requirement of the development. We believe that once this project is promoted and used widely,
efficient management on civil aviation engineering maintenance tool kits is expected.
Based on the system, we will continue to develop the management of tool kits to improve the speed of borrowing and
returning tools. We will establish a POS (Point Of Sales) machine terminal, which can replace the way of manual input
with scan mode. If the terminal is complete, the whole system will be a powerful tool management system.
Acknowledgement
This work was supported by National Training Programs of Innovation and Entrepreneurship for Undergraduates.
Figure 8. Flowchart of programs
Studies in Engineering and Technology Vol. 1, No. 1; 2014
55
References
Anonymous. Record management system based on the B/S model. Retrieved from
http://wenku.baidu.com/view/d307220c6c85ec3a87c2c53d.html
Biancucci, G., Claudi, A., & Dragoni, A. F. (2013). Secure data and voice transmission over GSM voice channel:
Applications for secure communications. Proceedings - 4th International Conference on Intelligent Systems,
Modelling and Simulation, ISMS 2013, 230-233. http://dx.doi.org/ 10.1109/ISMS.2013.10
Chenzhi, G., & Zhenya, L. (2013). An inquiry-based blended learning system for computer network curriculum.
Proceedings of the 8th International Conference on Computer Science and Education, ICCSE 2013, 1340-1345.
http://dx.doi.org/10.1109/ICCSE.2013.6554130
Huan, H. (2009). Development of network monitoring system based on the CAN buses. Retrieved from
http://d.wanfangdata.com.cn/Thesis_D066827.aspx
Iahad, N. A, Mirabolghasemi, M., & Huspi, S. H. (2012). A blended community of inquiry approach: The usage of
social network as a support for Course Management System. 2012 International Conference on Computer and
Information Science, ICCIS 2012 - A Conference of World Engineering, Science and Technology Congress,
ESTCON 2012 - Conference Proceedings,180-183. http://dx.doi.org/ 10.1109/ICCISci.2012.6297235
Ling, X., & Xiaoyan, C. (2006). Single-chip microcontroller C8051F020, its features and applications in
telecommunication systems. Journal of Henan Mechanical and Electrical Engineering College, 14(6), 15-16.
http://dx.doi.org/10.3969/j.issn.1008-2093.2006.06.007
Shaosong, Y., Xiangfa, R., & Yan, Z. (2006). Design and research of storehouse manage system based on SQL Server
2000 database. MACHINERY DESIGN & MANUFACTURE. http://dx.doi.org/
10.3969/j.issn.1001-3997.2006.02.078
Wei, X., & Jianqing, Z. (2008). GSM modules – A New prominent in realization of remote control. Radio, 10, 54-56.
http://lib.cqvip.com/qk/94496X/200810/28294784.html
Yue, Z., & Yushun, F. (2002). The Design and Implementation of Workfiow Management System Based on COM and
ASP. COMPUTER ENGINEERING AND APPLICATIONS, 38(1),
http://dx.doi.org/10.3321/j.issn:1002-8331.2002.01.080
This work is licensed under a Creative Commons Attribution 3.0 License.
Studies in Engineering and Technology
Vol. 1, No. 1; February 2014
ISSN 2330-2038 E-ISSN 2330-2046
Published by Redfame Publishing
URL: http://set.redfame.com
56
Times Semi-Passive RFID Tags with Double Loop Antennas Arranged as
a Shifted Gate Stability Optimization
Ofer Aluf
Correspondence: Physical Electronics Dept., Tel-Aviv University, Ramat-Aviv, 69978, Israel
Received: December 13, 2013 Accepted: December 29, 2013 Available online: January 17, 2014
doi:10.11114/set.v1i1.320 URL: http://dx.doi.org/10.11114/set.v1i1.320
Abstract
In this article, Very Crucial subject discussed in Semi-Passive RFID TAGs system stability. Semi-Passive TAGs with
double loop antennas arranged as a shifted gate system stability optimization under delayed electromagnetic
interferences. The double loop antenna is employed due to the fact that this antenna consists of two parallel loops; i.e.,
primary and secondary loops. We define Vi1(t) and Vi2(t) as the voltages in time on double loop antennas. Vi1(t) is the
voltage in time on the primary loop and Vi2(t) is the voltage in time on the secondary loop. The index (i) stand for the
first gate (i=1) and second gate (i=2). Due to electromagnetic interferences there are different in time delays respect to
gate antenna's first and second loop voltages and voltages derivatives. The delayed voltages are Vi1(t-τ1) and Vi2(t-τ2)
respectively (τ1≠ τ2) and delayed voltages derivatives are dVi1(t-Δ1)/dt, dVi2(t-Δ2)/dt respectively
1 2 1 2 1 2( ; 0 ; 0 ; , 0) .
Keywords:Double loop antenna, Shifted Gate antennas, Delay Differential Equations (DDEs), Bifurcation, Stability
1. Introduction
In this article, Very Critical and useful subject is discussed: Semi-Passive RFID TAGs system stability.A semi-passive
tags operate similarly to passive RFID tags. However, they contain a battery that enables longer reading distance and also
enables the tag to operate independently of the reader.Semi-Passive TAGs with double loop antennas arranged as a shifted
gate system influence by electromagnetic interferences which effect there stability behavior. The below figure describes
the double loop antennas as a shifted gate in x-direction.
Figure 1. Double loop antennas arranged as a shifted gate in x-direction.
The Semi-Passive RFID TAG with double loop antennas equivalent circuit can be represent as a delayed differential
equations which depending on variable parameters and delays.
2. Semi-passive RFID Tag with Double Loop Gate Antenna Equivalent Circuit and Represent Delay Differential
Equations
Semi-Passive RFID TAG with double loop antenna can be representing as a two inductors in series (L11 and L12 for the
first double loop gate antenna) with parasitic resistance rP1. The double loop antennas in series are connected in parallel to
Semi-Passive RFID TAG. The Equivalent Circuit of Semi-Passive RFID TAG is Capacitor (C1) and Resistor (R1) in
parallel with voltage generator Vs1(t) and parasitic resistance rS1. In case we have Passive RFID TAG switch S1 is OFF
otherwise is ON (Reader/Active RFID system) and long distance is achievable. The second double loop gate antenna is
D
d1
Studies in Engineering and Technology Vol. 1, No. 1; 2014
57
defined as two inductors in series L21 and L22 with series parasitic resistor rP2. Vs2(t) and parasitic resistance rS2 are belong
to the second gate antenna system with another Semi-Passive RFID TAG (Supakit, et al.) .
Figure 2.Double loop antennas in series with parasitic resistance and Semi-Passive RFID TAG.
Figure 3. Equivalent circuit of Double loop antennas in series with Semi-Passive RFID TAG.
L11 and L12 are mostly formed by traces on planar PCB. 2∙Lm element represents the mutual inductance between L11 and
L12. We consider that the double loop antennas parameters values (La1, La2, Lb1, Lb2, a1, a2) are the same in the first and
second gates. Since two inductors (L11, L12) are in series and there is a mutual inductance between L11 and L12, the total
antennainductance LT: LT=L11+L12+2∙Lmand . Lm is the mutual inductance between L11 and L12. K is
the coupling coefficient of two inductors . We start with the case of passive RFID TAG which switch S1 is
OFF. I(t) is the current that flow through double loop antenna. V11 and V12 are the voltages on L11 and L12 respectively. Vm
is the voltage on double loop antenna mutual inductance element.
111 11 12 12 1 1 1 11 12 1 1 ; ; ; 2 ; ; C
CD p m m AB R C CD m C
dVdI dI dIV L V L V I r V L V V V V V V V I C
dt dt dt dt (1)
1 1 1 11 121 1 1 11 12 11 12
1 11 12
1 10 0 ; ; ; C C C CD m
C R
dV V dV dV dVdV dVI I I C I L L I V dt V dt
dt R dt dt dt dt dt L L
(2)
1 11 11 1211 12 1 11 12
11 12 11 12
1 1 ; ;
p pC CD mCD p
r rdV dV dVdV dVI V dt V dt V I r V dt V dt
dt dt dt dt dt L L L L (3)
11 12mL K L L
0 1K
Studies in Engineering and Technology Vol. 1, No. 1; 2014
58
1 1 11 1211 12 11 12 12 11 11 12 11 12
11 12 12 11 11 12 11 12
1 1 1 1 ; ; ;
p pCDr rdV L L dI
V V V V V V I V dt V dt V Vdt L L L L L L dt L L
(4)
12 12 1111 12 11 11
11 11 11
12 2 2 ; 2m
m m
dVL L dVdIV L K L L V K V K
dt L L dt L dt (5)
We get the following differential equation respect to V11(t) variable, are global parameters.
(6)
(7)
' 2' 11 11 11
1 1 1 12 11 2 2 1 1 12 11 1 3 3 11 1 1 11 2( , , , ) ; ( , , , , , ) ; ( , , ) ; ; p p
dV dV d VC L L K C r L L K R L r R V
dt dt dt (8)
(9)
In the same manner we find our V12 differential equation. We get the following differential equation respect to V12(t)
variable, are global parameters.
21 1 112 12 11 11 11 11
1 2 12 3 1 1 2 32
12 12 12 1 12 12 12 1
1 10 ; (1 2 ) ; (1 2 ) ; (1 )
p pC r rd V dV L L L LV C K K
dt dt L L L R L L L R
(10)
(11)
'' '312 2 12
1 1 1 12 11 2 2 1 1 12 11 1 3 3 12 1 1 12 12 12
1 1
( , , , ) ; ( , , , , , ) ; ( , , ) ; ; p p
dV dVC L L K C r L L K R L r R V V V
dt dt
(12)
Summary: We get our RFID double loop antennas system's four differential equations.
' '' ' ' '3 311 2 11 12 2 12
11 11 11 12 12 12
1 1 1 1
; ; ; dV dV dV dV
V V V V V Vdt dt dt dt
(13)
'
11
'
1111
11 14
11 3 32 211 12 33 34 21 43' '
1 1 1 112 1241 44
12
12
; ; ; ; ; 1
dV
dtVdV
Vdt
dV V
dt V
dV
dt
(14)
(15)
The RFID double loop antennas system's primary and secondary loops are composed of a thin wire or a thin plate element
(Figure 2). Units are all in cm, and a1, a2 are radiuses of the primary and secondary wires in cm. There inductances can be
calculated by the following formulas:
1 2 3, ,
21 111 11 12 12 12 12
1 2 11 3 1 1 22
11 11 11 1 11 11
10 ; (1 2 ) ; (1 2 )
pC rd V dV L L L LV C K K
dt dt L L L R L L
1 1 112 3
11 1 1 11 1
1 1 ; (1 )
p pC r r
L R C L R
'' '311 2 11
11 11 11
1 1
; dV dV
V V Vdt dt
1 2 3, ,
' 21 1 '1 12 12 12
2 12 2
12 1 1
1 ; ;
pC r dV dV d VV
L R C dt dt dt
13 14 22 23 24 31 32 41 42 44 0
Studies in Engineering and Technology Vol. 1, No. 1; 2014
59
1 111 1 1 1 1 1 1
1 1 1 1 1 1
2 24 { ln[ ] ln[ ] 2 [ ( )]}
( ) ( )b a c a b
b c b c
A AL L L a l L L
a L l a L l
(16)
2 2 2 22 212 2 2 2 2 2 2 1 1 1 1 1 1 2 2 2 2 2 2
2 2 2 2 2 2
2 24 { ln[ ] ln[ ] 2 [ ( )]} ; ; ; ;
( ) ( )b a c a b c a b a b c a b a b
b c b c
A AL L L a l L L l L L A L L l L L A L L
a L l a L l
(17)
Due to electromagnetic interferences we get a shifted gate RFID system's primary and secondary antennas loops voltages
with delays and respectively. Additionally we get antennas loops voltages derivatives with delays Δ1 and Δ2
respectively.
; ' ' ' '
12 12 2 11 11 1 12 12 2( ) ( ) ; ( ) ( ) ; ( ) ( )V t V t V t V t V t V t (18)
. We consider no delay effect on . (19)
The RFID shifted gate system differential equations under electromagnetic interferences (delays terms) influence only
RFID double loop voltages V11(t), V12(t) and voltages derivatives and respect to time, there is no
influence on
'
11
'
11 111
11 14' '11 111 12 11 12
' '
12 12 241 44
12 2
12
( )
( )( ) ( ) ( ) ( ) ; ; ; ;
( )
( )
dV
dtV tdV
V tdV t dV t dV t dV t dt
dt dt dt dt dV V t
dt V t
dV
dt
. (20)
To find equilibrium points (fixed points) of the RFID shifted gate system is by
, , , (21)
. ' '
11 12 11 121 2 1 2
( ) ( ) ( ) ( )=0 ; =0 ; =0 ; =0. t ; t ; t ; t
dV t dV t dV t dV t
dt dt dt dt (22)
(23)
We get four equations and the only fixed point is Since (24)
. Stability analysis: The standard local stability analysis about
any one of the equilibrium points of RFID shifted gate system consists in adding to coordinates
arbitrarily small increments of exponential form , and retaining the first order terms in
. The system of four homogeneous equations leads to a polynomial characteristics equation in the
eigenvalues . The polynomial characteristics equations accept by set the below voltages and voltages derivative respect
to time into two RFID shifted gate system equations.
RFID shifted gate system fixed values with arbitrarily small increments of exponential form are:
i=0 (first fixed point), i=1 (second fixed point), i=2 (third fixed point), etc.,
1 2
11 11 1( ) ( )V t V t
' '
12 12 2( ) ( )V t V t ' '
11 12 11 12 ; ; ; dV dV dV dV
dt dt dt dt
'
11( )V t'
12 ( )V t
11 1 11lim ( ) ( )t
V t V t
12 2 12lim ( ) ( )t
V t V t
' '
11 1 11lim ( ) ( )t
V t V t
' '
12 2 12lim ( ) ( )t
V t V t
1 2 1 2 ( ) ; ( ) ; ( ) ; ( ) , tt t t t t t t t
(0) ' (0) (0) ' (0) (0)
11 11 12 12( , , , ) (0,0,0,0)E V V V V
3 1 12 3 1 340 & 0 0 ; 0 & 0 0
' '
11 11 12 12[ ]V V V V
' '
11 11 12 12[ ] tv v v v e
' '
11 11 12 12 V V V V
' '
11 11 12 12[ ] tv v v v e
Studies in Engineering and Technology Vol. 1, No. 1; 2014
60
' '( ) ' ( ) ' '( ) ' ( )
11 11 11 11 11 11 12 12 12 12 12 12( ) ; ( ) ; ( ) ; ( )i t i t i t i tV t V v e V t V v e V t V v e V t V v e (25)
We choose the above expressions for our and as small displacement from the
system fixed points at time t=0.
' '( ) ' ( ) ' '( ) ' ( )
11 11 11 11 11 11 12 12 12 12 12 12( 0) ; ( 0) ; ( 0) ; ( 0)i i i iV t V v V t V v V t V v V t V v (26)
For the selected fixed point is stable otherwise is Unstable. Our system tends to the
selected fixed point exponentially for otherwise go away from the selected fixed point exponentially.
is the eigenvalue parameter which establish if the fixed point is stable or Unstable, additionally his absolute value (
)establish the speed of flow toward or away from the selected fixed point (Yuri&Jack).
Table 1. Semi-passive RFID TAGs with double loop antennas, variables function of λ eigenvalue and time.
λ<0 λ>0
t=0
;
;
t>0 ;
;
t>0
t→∞
;
;
The speeds of flow toward or away from the selected fixed point for RFID shifted gate system voltages and voltages
derivatives respect to time are
' ' ' '( ) ' ( ) '( ) ' '1 '11 11 11 11 11 11 11 11
110 0 0
( ) ( ) ( ) [ ] [ 1]lim lim lim
ti t t i t t t
e t t
t t t
dV t V t t V t V v e V v e v e ev e
dt t t t
(27)
1 1
' '' '11 12 12 11 1 11 1
11 12 12 11 11
( ) ( ) ( ) ( ) ( ) ; ; ; ; t t t t tdV t dV t dV t dV t dV t
v e v e v e v e e v e edt dt dt dt dt
(28)
(29)
'
11 11( ), ( )V t V t '
12 12( ), ( )V t V t' '
11 11 12 12[ ]v v v v
0, t > 0 0, t > 0
0, t > 0
| |
' '( ) '
11 11 11
( )
11 11 11
( 0)
( 0)
i
i
V t V v
V t V v
' '( ) '
12 12 12
( )
12 12 12
( 0)
( 0)
i
i
V t V v
V t V v
' '( ) '
11 11 11
( )
11 11 11
( 0)
( 0)
i
i
V t V v
V t V v
' '( ) '
12 12 12
( )
12 12 12
( 0)
( 0)
i
i
V t V v
V t V v
' '( ) ' | |
11 11 11
( ) | |
11 11 11
( )
( )
i t
i t
V t V v e
V t V v e
' '( ) ' | |
12 12 12
( ) | |
12 12 12
( )
( )
i t
i t
V t V v e
V t V v e
' '( ) ' | |
11 11 11
( ) | |
11 11 11
( )
( )
i t
i t
V t V v e
V t V v e
' '( ) ' | |
12 12 12
( ) | |
12 12 12
( )
( )
i t
i t
V t V v e
V t V v e
' '( )
11 11
( )
11 11
( )
( )
i
i
V t V
V t V
' '( )
12 12
( )
12 12
( )
( )
i
i
V t V
V t V
' ' | |
11 11
| |
11 11
( , 0)
( , 0)
t
t
V t v e
V t v e
' ' | |
12 12
| |
12 11
( , 0)
( , 0)
t
t
V t v e
V t v e
2 2
''12 2 12 2
12 12
( ) ( ) ; t tdV t dV t
v e e v e edt dt
Studies in Engineering and Technology Vol. 1, No. 1; 2014
61
First we take the RFID shifted gate's voltages V11, V12 differential equations: and adding
coordinates arbitrarily small increments of exponential terms and retaining the
first order terms in .
' ''( ) ' '( 0) '( ) ' '( 0)11 12
11 11 11 11 1 12 12 12 12 2
11 12
; =0 = 1>0 ; ; 0 1 0t i t i t i t iv vv e V v e V v e V v e V
v v
(30)
Second we take the RFID shifted gate's voltages derivatives differential equations:
(31)
and adding coordinates arbitrarily small increments of exponential terms and
retaining the first order terms in .
' '( ) ' ( ) '( 0) ( 0) 11 1111 11 11 11 12 11 11 11 11 3 11 12 3 11 12' '
11 11
[ ] [ ] ; 0 ; 0 ; 1t i t i t i i v vv e V v e V v e V V
v v
(32)
'( 0) ( 0) ' '( ) ' ( )11 1111 11 3 11 12 3 11 12 12 33 12 12 34 12 12' '
11 11
0 ; 0 ; 1 ; [ ] [ ]i i t i t i tv vV V v e V v e V v e
v v
(33)
(34)
If λ3>0 and λ4>0 then our fixed point is unstable node. If (λ3>0 and λ4<0) or (λ3<0 and λ4>0) or (λ3<0 and λ4<0)
Then our fixed point is saddle point. We define
(35)
(36)
Then we get four delayed differential equations respect to coordinates arbitrarily small increments of
exponential
. (37)
1 2 2( ) ( ) ( )' ' '
11 11 12 33 12 34 12 ; t t tt te v e v e v e v e v
(38)
. In the equilibrium fixed point (39)
. The small increments Jacobian of our RFID shifted gate system is as bellow:
1 1 1
11 11 12 12 13 14 21 22 23 24 31 32 ; ; 0 ; 0 ; ; ; 0 ; 0 ; 0 ; 0e e e
(40)
2 2 2
'
11
11 14
11
33 33 34 34 41 42 43 44'
1241 44
12
; ; 0 ; 0 ; 0 ; ;
v
ve e e
v
v
(41)
' '11 1211 12 ;
dV dVV V
dt dt
' '
11 11 12 12[ ]V V V V' '
11 11 12 12[ ] tv v v v e
' '
11 11 12 12 v v v v
' '
11 12, V V
' '' '11 12
11 11 12 11 33 12 34 12 ; dV dV
V V V Vdt dt
' '
11 11 12 12[ ]V V V V' '
11 11 12 12[ ] tv v v v e
' '
11 11 12 12 v v v v
'( 0) ( 0) 12 1212 12 33 34 4 33 34' '
12 12
0 ; 0 ; 1i i v vV V
v v
1 1( ) ( )' '( ) ' ( )
11 1 11 11 11 1 11 11( ) ; ( )t ti iV t V v e V t V v e
2 2( ) ( )' '( ) ' ( )
12 2 12 12 12 2 12 12( ) ; ( )t ti iV t V v e V t V v e
' '
11 11 12 12[ ]V V V V
' '
11 11 12 12[ ] tv v v v e 1 1( ) ( )' '
11 11 11 12 11
t tte v e v e v
2( ) '
12 12
tte v e v ' ( 0) ( 0)
11 110, 0i iV V
' ( 0) ( 0)
12 120, 0i iV V
Studies in Engineering and Technology Vol. 1, No. 1; 2014
62
(42)
2 2 2 2
2 1
1 1 1 1
2
12 2 1 1 2 1
[ ] [ ] [ ]4
1 2 1 2 12 34 11 34 33 12
( ) ( )2 3
34 12 11 33 33 11
( , , , , ) { }
{ } { }
i j j j
i j j j
j
j
D e e e
e e e e e
(43)
We have three stability cases:
Or or (44)
Otherwise and they are positive parameters. There are other possible simple stability cases:
or ; or (45)
We need to get characteristics equations for all above stability analysis cases. We study the occurrence of any possible
stability switching resulting from the increase of value of the time delays for the general characteristic equation
. If we choose parameter then . The expression for :
2 3
0 1 2 30
( , ) ( ) ( ) ( ) ( ) ( ) ........n
k
n kk
P P P P P P
(46)
The expression for
: (47)
3. RFID Shifted Gate System Fourth Order Characteristic Equation
The second case we analyze is when there is delay in RFID gate's primary and secondary loop antennas voltages (
) and no delay in in gate's primary and secondary loop antennas voltages derivatives (Kuang, 1993; Beretta,
et al., 2002). The general characteristic equation D(λ, τ) is ad follow: 4 3 2 2
33 11 11 33 12 34 11 34 12 33 34 12( , ) ( ) { ( ) ( )}D e e (48)
Under Taylor series approximation: . The Maclaurin series is a Taylor series expansion of a
function about zero (0). We get the following general characteristic equation D(λ, τ) under Taylor series approximation:
. (49)
4 3 2 2 2
33 11 11 33 12 34 11 34 12 33 12 34 12 34 34 12
1( , ) [ ] { [ ] [ ]}
2D e (50)
. The expression for is
2 3 4 4 3 2
33 11 11 330 1 2 3 40
( , ) ( ) ( ) ( ) ( ) ( ) ( ) [ ]n
k
n kk
P P P P P P P
(51)
11 14
41 44
; det | | 0A I A I
1 2 1 2 & 0 1 2 1 20 & 1 2 1 2
1 2 1 2 &
1 2 1 2 ; 0 ; 0 1 2 1 20 ; ; 0 1 2 1 20 ; ; 0 1 2 1 20 ; 0 ;
, ,
( , / / )D ( , ) ( , ) ( , )n m
D eQP ( , )
nP
( , )m
Q 2
0 1 20
( , ) ( ) ( ) ( ) ( ) ........m
k
m kk
Q q q q q
1 2 1 2 & 0
1 2
2 211
2e e
2 211
2e
( , ) ( , ) ( , ) ; 4 ; 2 ; n m
D e n m n mQP ( , )
nP
Studies in Engineering and Technology Vol. 1, No. 1; 2014
63
11 33 33 110 1 2 3 4( ) 0 ; ( ) 0 ; ( ) ; ( ) [ ] ; ( ) 1P P P P P (52)
The expression for
is (53)
2 2
12 34 11 34 12 33 12 34 12 34 34 12 12 340
0
2
11 34 12 33 12 34 12 34 34 121 2
1( , ) ( ) [ ] [ ] ; ( )
2
1( ) ; ( )
2
mk
m kk
Q q q
q q
(54)
The homogeneous system for leads to a characteristic equation for the eigenvalue λ having the form
and the coefficients
Depend on and delay . are any two shifted gate system's parameters, other parameters keep as a constant
.
0 1 2 11 33 3 33 11 40 ; 0 ; ; [ ] ; 1a a a a a (55)
2
0 12 34 1 11 34 12 33 12 34 2 12 34 34 12
1 ; ;
2c c c (56)
Unless strictly necessary, the designation of the variation arguments will subsequently be omitted from P, Q, aj,
cj. The coefficients aj, cj are continuous, and differentiable functions of their arguments, and direct substitution shows that
a0+c0≠0 for , i.e. λ=0 is not a of . We assume that and
can't have common imaginary roots. That is for any real number : 4 3 2
33 11 11 33( , ) ( , ) 0 ; ( , ) ( )n m np i Q i p i i (57)
2 2
12 34 11 34 12 33 12 34 12 34 34 12
1( , ) [ ] [ ]
2mQ i i (58)
4 2 2 3
12 34 34 12 11 33 12 34 33 11
11 34 12 33 12 34
1( , ) ( , ) [ ] ( )
2
[ ] 0
n mp i Q i i
i
(59)
2 8 6 2 4 2 2
33 11 11 33 11 33| ( , ) | {( ) 2 }P i (60)
2 2 2 2 2 2
12 34 11 34 12 33 12 34 12 34 12 34 34 12
4 2 2
12 34 34 12
1| ( , ) | {[ ] 2 [ ]}
2
1[ ]2
Q i
(61)
We need to find the expression for
(62)
( , )m
Q 2
0 1 20
( , ) ( ) ( ) ( ) ( )m
k
m kk
Q q q q q
' '
11 11 12 12 V V V V
4 2
0 0
( , ) ( , ) 0 ; ( ) ; ( )j j
j j
j j
P Q e P a Q c
{ ( , , ), ( , , )}j i k j i ka q q c q q
,i kq q ,i kq q
( , )i kq q
,i kq q ( , ) ( , ) 0P Q e ( , )nP ( , )mQ
2 2( , ) | ( , ) | | ( , ) |F P i Q i
Studies in Engineering and Technology Vol. 1, No. 1; 2014
64
2 2 8 6 2
33 11 11 33
4 2 2 2 2 2 2
11 33 12 34 34 12 11 34 12 33 12 34
2 2 2
12 34 12 34 34 12 12 34
( , ) | ( , ) | | ( , ) | {( ) 2 }
1{ [ ] } {[ ]
2
12 [ ]}
2
F P i Q i
(63)
We define the following parameters for simplicity:
2 2 2 2
0 12 34 2 11 34 12 33 12 34 12 34 12 34 34 12
2 2 2 2 2
4 11 33 12 34 34 12 6 33 11 11 33 8
1 ; [ ] 2 [ ]
2
1[ ] ; ( ) 2 ; 12
(64)
42 2 2 4 6 8 2
0 2 4 6 8 2
0
( , ) | ( , ) | | ( , ) | k
k
k
F P i Q i
(65)
Hence implies and its roots are given by solving the above polynomial. Furthermore
4 2 3
11 33 33 11( , ) ; ( , ) ( )R IP i P i (66)
2 2
12 34 12 34 34 12 11 34 12 33 12 34
1( , ) [ ] ; ( , ) [ ]
2R IQ i Q i (67)
2 2
( , ) ( , ) ( , ) ( , ) ( , ) ( , ) ( , ) ( , )sin ( ) ; cos ( )
| ( , ) | | ( , ) |
R I I R R R I IP i Q i P i Q i P i Q i P i Q i
Q i Q i
(68)
4 2 3 2 2
11 33 11 34 12 33 12 34 33 11 12 34 12 34 34 12
2 2 2 2 2 4 2
12 34 11 34 12 33 12 34 12 34 12 34 34 12 12 34
1{ } [ ] ( ) { [ ]}
2sin ( )1 1
{[ ] 2 [ ]} [2 2
2
34 12 ]
(69)
4 2 2 2 4
11 33 12 34 12 34 34 12 33 11 11 34 12 33 12 34
2 2 2 2 2 4 2
12 34 11 34 12 33 12 34 12 34 12 34 34 12 12 34
1{ } { [ ] } ( ) [ ]
2cos ( )1 1
{[ ] 2 [ ]} [2 2
2
34 12 ]
(70)
That are continuous and differentiable in based on Lema 1.1 (see Appendix A). Hence we use theorem 1.2(see
Appendix B). This prove the theorem 1.3 (see Appendix C).
4. RFID shifted gate system stability analysis under delayed variables in time
Our RFID shifted gate homogeneous system for leads to a characteristic equation for the eigenvalue λ
having the form ; Second case ; .
4 3 2 2
1 2 1 2 33 11 11 33 12 34 11 34 12 33 34 12( , , 0) ( ) { ( ) ( )}D e e (71)
Under Taylor series approximation: . The Maclaurin series is a Taylor series expansion of a
function about zero (0). We get the following general characteristic equation D(λ, τ) under Taylor series approximation:
. (27)
4 3 2 2 2
33 11 11 33 12 34 11 34 12 33 12 34 12 34 34 12
1( , ) [ ] { [ ] [ ]}
2D e (73)
We use different parameters terminology from our last characteristics parameters definition:
( , ) 0F 4
2
2
0
0k
k
k
'
11 11 v v '
12 12 v v
( ) ( ) =0 P Q e 1 2 1 2 0
2 211
2e e
2 211
2e
Studies in Engineering and Technology Vol. 1, No. 1; 2014
65
. Additionally
Then
4 24 3 2
33 11 11 33
0 0
( ) ; ( ) ; [ ]j j
j j
j j
P a Q c P
(74)
(75)
and are continuous and differentiable function of such that . In the
following denotes complex and conjugate. Are analytic functions in and differentiable in . The
coefficients and de pend on RFID
shifted gate system's C1, R1, values and antenna parameters.
(76)
2
0 12 34 1 11 34 12 33 12 34 2 12 34 34 12
1 ; ;
2c c c (77)
Unless strictly necessary, the designation of the variation arguments will
subsequently be omitted from P, Q, aj, cj. The coefficients aj, cj are continuous, and differentiable functions of their
arguments, and direct substitution shows that .
1 2
3 3 11 12 1
21 1 12 12 11 111
11 11 12 12
1(1 )
0
(1 2 ) (1 2 )
pr
L L R
L L L LC K K
L L L L
(78)
i.e is not a root of characteristic equation. Furthermore are
analytic function of for which the following requirements of the analysis (see kuang, 1993, section 3.4) can also be
verified in the present case (Kuang, 1993; Beretta et al., 2002).
(a) If , then , i.e P and Q have no common imaginary roots. This
condition was verified numerically in the entire domain of interest.
(b) is bounded for , . No roots bifurcation from . Indeed, in the limit
2 2
12 34 11 34 12 33 12 34 12 34 34 12
4 3 2
33 11 11 33
1{ [ ] [ ]}
Q( ) 2| | | |( ) [ ]P
(79)
(c) (80)
; ( ) ; ( ) ; 4 ; 2 ; k j k jk j p a q c n m n m ( , ) ( ) ; ( , ) ( )n m
P QQP
2 2
12 34 11 34 12 33 12 34 12 34 34 12
1[ ] [ ]
2Q
0, , n m n m 0, : Rj ja c R
0 0 0a c
" " ( ), ( )P Q
1 1{ ( , ,gate antenna parametrs)ja C R 1 1( , , ,gate antenna parametrs)}jc C R
0 1 2 11 33 3 33 11 40 ; 0 ; ; [ ] ; 1a a a a a
1 1( , , ,gate antenna parametrs)R C
0 0 12 340 ; 0a c
1 , gate antenna parametersC 0 ( ), ( )P Q
i ( ) ( ) 0P i Q i
1 1( , ,antenna parametrs)R C
|Q( ) / ( ) |P | | Re 0
2 2( ) | ( ) | | ( ) |F P i Q i
Studies in Engineering and Technology Vol. 1, No. 1; 2014
66
2 2 8 6 2 4 2 2 2 2
33 11 11 33 11 33 12 34 34 12
2 2 2 2 2
11 34 12 33 12 34 12 34 12 34 34 12 12 34
1( , ) | ( , ) | | ( , ) | {( ) 2 } { [ ] }
2
1{[ ] 2 [ ]}
2
F P i Q i
(81)
Has at most a finite number of zeros. Indeed, this is a polynomial in (degree in ).
(d) Each positive root of is continuous and differentiable with
respect to . This condition can only be assessed numerically.
In addition, since the coefficients in P and Q are real, we have , and thus ,
may be on eigenvalue of characteristic equation. The analysis consists in identifying the roots of characteristic
equation situated on the imaginary axis of the complex λ – plane, where by increasing the parameters
, Reλ may, at the crossing ,Change its sign from (-) to (+), i.e. from a stable focus
to an unstable one, or vice versa. This feature may be further assessed by
examining the sign of the partial derivatives with respect to and gate antenna parameters.
1 1
1 1 1 1
1 1
Re Re( ) ( ) , R , ,gate antenna parametrs ; (R ) ( ) , , ,gate antenna parametrs
Ri iC const C const
C
(82)
1 1
1 1 11 1 1
1 11
Re Re(R ) ( ) , , ,gate antenna parametrs ; ( ) ( ) , ,R ,
Ri iC const L C const
L
(83)
1 1
11 1 1 12 1 1
11 12
Re Re( ) ( ) , ,R , ; ( ) ( ) , ,R ,i iL C const L C const
L L
(84)
1 1
12 1 1 1 1
12
Re Re( ) ( ) , ,R , ; ( ) ( ) , ,R ,gate antenna parametrs , where .i iL C const C const
L
(85)
1
1 1
Re( ) ( ) , ,R ,gate antenna parametrs , where .i C const
(86)
For the case we get the following results:
(87)
2 2
12 34 12 34 34 12 11 34 12 33 12 34
1( , ) [ ] ; ( , ) [ ]
2R IQ i Q i (88)
2 2 2 2
0 12 34 2 11 34 12 33 12 34 12 34 12 34 34 12
2 2 2 2 2
4 11 33 12 34 34 12 6 33 11 11 33 8
1 ; [ ] 2 [ ]
2
1[ ] ; ( ) 2 ; 12
(89)
42 2 2 4 6 8 2
0 2 4 6 8 2
0
( , ) | ( , ) | | ( , ) | k
k
k
F P i Q i
(90)
Hence
implies (91)
8
1 1( , , ,gate antenna parametrs)R C F( )=0
1 1, , ,gate antenna parametrsR C
( ) ( )P i P i ( ) ( )Q i Q i i
0
1 1, , ,gate antenna parametrsR C
(0) ' (0) (0) ' (0) (0)
11 11 12 12( , , , ) (0,0,0,0)E V V V V
1 1, ,C R
1 2 1 2 & 0
4 2 3
11 33 33 11( , ) ; ( , ) ( )R IP i P i
( , ) 0F 4
2
2
0
0k
k
k
Studies in Engineering and Technology Vol. 1, No. 1; 2014
67
When writing and , and inserting
Into RFID Gate system's characteristic equation , must satisfy the following :
2 2
( ) ( ) ( ) ( ) ( ) ( ) ( ) ( )sin ( ) ; cos ( )
| ( ) | | ( ) |
R I I R R R I IP i Q i P i Q i P i Q i P i Q ig h
Q i Q i
(92)
Where in view of requirement (a) above, and . Furthermore, it follows above and
equations that, by squaring and adding the sides, must be a positive root of
. Note that is dependent of . Now it is important to notice that if (assume
that is the set where is a positive root of and for , is not define. Then for all τ in I
is satisfies that ). Then there are no positive solutions for , and we cannot have
stability switches. For any where is a positive solution of , we can define the angle
as the solution of
2 2
( ) ( ) ( ) ( ) ( ) ( ) ( ) ( )sin ( ) ; cos ( )
| ( ) | | ( ) |
R I I R R R I IP i Q i P i Q i P i Q i P i Q i
Q i Q i
(93)
And the relation between the argument and fo must be
. Hence we can define the maps given by
. Let us introduce the functions ;
That are continuous and differentiable in . In the following, the subscripts and RFID Gate antenna
parameters indicate the corresponding partial derivatives. Let us first concentrate on
,remember in and , and keeping all parameters except one (x)
and . The derivation closely follows that in reference [BK]. Differentiating RFID characteristic equation
with respect to specific parameter (x), and inverting the derivative, for convenience, one
calculates: Remark:
1 ( , ) ( , ) ( , ) ( , ) ( , ) ( , )( )
( , ) ( , ) ( , ) ( , )x x
P x Q x Q x P x P x Q x
x P x Q x Q x P x
(44)
Where etc., Substituting , and bearing i ,
( ) ( ) ( )R IP P i P ( ) ( ) ( )R IQ Q i Q i
2| ( ) | 0Q i ( , )g h R sin
cos
2 2( ) | ( ) | | ( ) | 0F P i Q i ( )F I
0I R ( ) ( )F I ( )
( ) ( , ) 0F ( ) ( , ) 0F
I ( ) ( , ) 0F
( ) [0,2 ]
( ) ( ) I 0( ) ( ) 2 n n
0:n I R
0
( ) 2( ) ; ,
( )n
nn I
I R
0( ) ( ), , n nS I n
1 1, , , R C
1 2 1 2 1 2( , , , , , )a a b bL L L L a a ( )x
1 2 1 2 1 2( , , , , , )a a b bL L L L a a 1 2 1 2 1 2( , , , , , )a a b bL L L L a a
( ) ( ) =0 P Q e
1 1 1 2 1 2 1 2, , , , , , , , .,a a b bx R C L L L L a a etc
,....P
P
i ( ) ( )P i P i ( ) ( )Q i Q i
Studies in Engineering and Technology Vol. 1, No. 1; 2014
68
Then and and that on the surface , one obtains
21 ( , ) ( , ) ( , ) ( , ) | ( , ) |
( ) | ( )( , ) ( , ) ( , ) ( , )
i
x x
i P i x P i x i Q i x Q x P i x
x P i x P i x Q i x Q i x
(95)
Upon separating into real and imaginary parts, with
; (96)
; 2 2 2 ; x Rx Ix R IQ Q i Q P P P (97)
When (x) can be any RFID Gate parameters R1, C1, And time delay etc,.Where for convenience, we have dropped the
arguments , and where
(98)
2 [( ) ( )] ; /x Rx R Ix I Rx R Ix I x xF P P P P Q Q Q Q F F . We define U and V:
( ) ( ) ; ( ) ( )R I I R R I I R R Ix I Rx R Ix I RxU P P P P Q Q Q Q V P P P P Q Q Q Q (99)
We choose our specific parameter as time delay
x = τ. (100)
2 3 4 2 2 2
12 34 12 34 11 33 11 330 ; 0 ; ; ; 2 [2 3 ]R I R I R RP P Q Q P P (101)
5 2 2
33 11 12 34 34 12 11 34 12 33 12 34
13 ( ) ; / ; 2 [ ] ;
2I I R IP P F F Q Q (102)
2
12 34 34 12 11 34 12 33 12 34
12 [ ] ;
2R IQ Q (103)
2 2 2
12 34 34 12 12 34 12 34 34 12
1 12 [ ] [ ( )]
2 2R RQ Q (104)
2
11 34 12 33 12 34[ ] ; 2 [( ) ( )]I I R R I I R R I IQ Q F P P P P Q Q Q Q (105)
(106)
2 2 2
12 34 11 34 12 33 12 34 34 12
12 [ ( )]
2F (107)
4 2 4 2
11 33 33 11 33 11 11 333 ( ) ( ) ; 2 ( ) (2 )R I I RP P P P (108)
2 2
12 34 12 34 34 12 11 34 12 33 12 34
1[ ( )] [ ]
2R IQ Q (194)
2 2
11 34 12 33 12 34 12 34 34 12
12 ( ) ( )
2I RQ Q (110)
( ) ( ) ; 0 ; 0R I I R R I I R R I I RV P P P P Q Q Q Q P P P P (111)
2 2 3
12 34 12 34 12 34 34 12 12 34 11 34 12 33 12 34
1[ ( )] ; [ ]
2R I I RQ Q Q Q (112)
( ) ( )i P i P i ( ) ( )i Q i Q i 2 2| ( ) | | ( ) |P i Q i
; R I R IP P i P Q Q i Q R IP P i P
; R I x Rx IxQ Q i Q P P i P
( , )i x
2 [( ) ( )]R R I I R R I IF P P P P Q Q Q Q
2 2
11 33 33 112 [2 ] ; 3 ( )R IP P
2 [( ) ( )]R R I I R R I IF P P P P Q Q Q Q
Studies in Engineering and Technology Vol. 1, No. 1; 2014
69
. Differentiating with respect to and we get
(113)
21 1
2
2 [ | | ]Re( ) ( ) ; ( ) Re{ } ;
2 [ | | ]i
U P i F F
F i V P F
(114)
21 1
2
2 [ | | ] Re( ) Re{ } ; ; { ( )} {( ) }
2 [ | | ]i
U P i F Fsign sign
F i V P F
(115)
(116)
We shall presently examine the possibility of stability transitions (bifurcations) in a shifted gate double loop RFID
system, about the equilibrium point as a result of a variation of delay parameter τ. The
analysis consists in identifying the roots of our system characteristic equation situated on the imaginary axis of the
complex λ-plane Where by increasing the delay parameter τ, Re λ may at the crossing, change its sign from – to +, i.e.
from a stable focus E(*)
to an unstable one, or vice versa. This feature may be further assessed by examining the sign of the
partial derivatives with respect to τ,
(117)
1
1 1
Re( ) ( ) , ,R ,gate antenna parametrs ; where .i C const
(118)
For our stability switching analysis we choose typical RFID shifted gate parameters values: L11=4.5mH, L12=2.5mH,
C1=23pF, R1=100kOhm=105, , K=0.6, 2·Lm=0.004 ( ).
12 5 12 5
1 2 3 1 256.22 10 ; 2.49 10 ; 222.42 ; 101.2 10 ; 4.492 10 (119)
5 12 5 5 12322 3 1 2 3 11 12
1 1
2.49 10 ; 222.42 ; 101.2 10 ; 4.492 10 ; 400.4 ; = 4.42 10 ; 3.95 10
(120)
5 123233 34 21 43 13 14 22 23 24 31 32 41 42 44
1 1
= 4.43 10 ; = 3.95 10 ; 1 ; 0 ; 0
(121)
Then we get the expression for for typical RFID shifted gate parameters values.
2 2 8 6 10 4 20 24 2 12 2
2 17 24 2 24 24 2 12 48
( , ) | ( , ) | | ( , ) | 39.16 10 {383.17 10 [7.8 10 7.9 10 ] }
{[34.94 10 15.6 10 ] 31.2 10 [7.8 10 7.9 10 ]} 243.39 10
F P i Q i
(122)
We find those values which fulfill . We ignore negative, complex, and imaginary values of for
specific values. and we can be express by 3D function . Since it is a very complex
function, we recommend to solve it numerically rather than analytic. We plot the stability switch diagram based on
different delay values of our RFID double gate system. Since it is a very complex function we recommend to solve it
numerically rather than analytic.
2 2 21 1
2 2 2 2
2 [ | | ] 2 { ( ) ( )}Re Re( ) ( ) Re{ } ; ( ) ( )
2 [ | | ] 4 ( )i i
U P i F F V P F U P
F i V P F V P
(123)
( , ) 0F
0 ; F
F F IF
1
2{ ( )} { } { }
| |
U V
sign sign F signP
(0) ' (0) (0) ' (0) (0)
11 11 12 12( , , , )E V V V V
1 Re( ) ( ) i
1 100pr Ohm11 122 2mL K L L
( , )F
, ( , ) 0F
[0.001..10] ( , ) 0F
Studies in Engineering and Technology Vol. 1, No. 1; 2014
70
The stability switch occurs only on those delay values ( ) which fit the equation: and is the solution of
when if only is feasible. Additionally when all RFID double gate
system's parameters are known and the stability switch due to various time delay values is describe in the following
expression:
1
2
( ( )) ( ( )) ( ( )){ ( )} { ( ( ), )} { ( ( )) ( ) }
| ( ( )) |
U Vsign sign F sign
P
(124)
Remark: we know implies it roots and finding those delays values which is feasible. There are
values which is complex or imaginary number, then unable to analyse stability (Kuang, 1993; Beretta, et al. 2002).
5. Results of RFID Shifted Gate System Stability Switching under Delayed Variables in Time
We find those ω, τ values which fulfil F(ω, τ)=0. We ignore negative, complex, and imaginary values of ω for specific τ
values. and we can express by 3D function F(ω, τ)=0. We define new MATLAB script parameters:
τ→Tau, Gi(i=1,..,10): G1=39.16e10; G2=383.17e20; G3=7.8e24; G4=7.9e12; G5=34.94e17; G6=15.6e24;
G7=31.2e24;G8=7.8e24;G9=7.9e12;G10=243.39e48, Ξj→Phij ; j=8, 6, 4, 2, 0, Running MATLAB script for τ values
gives the following results:
MATLAB script:
Tau=10;G1=39.16e10;G2=383.17e20;G3=7.8e24;G4=7.9e12;G5=34.94e17;G6=15.6e24;G7=31.2e24;G8=7.8e24;G9=
7.9e12;G10=243.39e48;Phi8=1;Phi6=G1;Phi4=G2-(G3*Tau*Tau+G4).^2;
Phi2=-((G5-G6*Tau).^2-G7*(G8*Tau*Tau+G9)); Phi0=-G10;p=[Phi8 0 Phi6 0 Phi4 0 Phi2 0 Phi0];r=roots(p)
Results: We plot 3D function F(ω, τ)=0. τ:0→10; ω:0→1e13. We define additional MATLAB script parameters: ω→w,
τ→t. We get some possible real values for ω which fulfil F(ω, τ)=0, F(ω=0 or ω=2.1437, ω=2.7928, ω=1.0e+006,
ω=1.0e+009, ω=1.0e+010, ω=1.0e+011, ω=1.0e+012, ω=1.0e+013, τ)=0; . Next is to find those ω, τ
values which fulfil
sinθ(τ)=…. and cosθ(τ)=.. (125)
(126)
Table 2a. Semi-passive RFID TAGs with double loop antennas, ωi(τ).
ωi τ=0 τ=0.001 τ=0.01
ω1 1.0e+006 1.0e+009 1.0e+010
ω2 -0.0000+3.105i -2.7928 -2.7928
ω3 -0.0000-3.105i -0.0000+2.792i -0.0000+2.7928i
ω4 -2.1437 -0.0000-2.792i -0.0000-2.7928i
ω5 -1.528+0.0823i 2.7928 2.7928
ω6 -1.528-0.0823i 0.0000+0.0000i -0.0000+0.0000i
ω7 2.1437 0.0000-0.0000i -0.0000-0.0000i
ω8 1.528+0.0823i -0.0000+0.0000i 0.0000+0.0000i
ω9 1.528-0.0823i -0.0000-0.0000i 0.0000-0.0000i
( )
( )
( )
sin ( ) ... ; cos ( ) ... ( )
( , ) 0F ( )i i
i
[0.001..10]
[0.001..10]
[0.001..10]
2sin( )
| |
R I I RP Q P Q
Q
2 2 2
2
(cos( ) ; | |
| |
R R I IR I
P Q P QQ Q Q
Q
Studies in Engineering and Technology Vol. 1, No. 1; 2014
71
Table 2b. Semi-passive RFID TAGs with double loop antennas, ωi(τ).
ωi τ=0.1 τ=1 τ=10
ω1 1.0e+011 1.0e+012 1.0e+013
ω2 -2.7928 -2.7928 -2.7928
ω3 0.0000+2.7928i 0.0000+2.7928i 0.0000+2.7928i
ω4 0.0000-2.7928i 0.0000-2.7928i 0.0000-2.7928i
ω5 2.7928 2.7928 2.7928
ω6 -0.0000+0.0000i -0.0000+0.0000i -0.0000+0.0000i
ω7 -0.0000-0.0000i -0.0000-0.0000i -0.0000-0.0000i
ω8 0.0000+0.0000i 0.0000+0.0000i 0.0000+0.0000i
ω9 0.0000-0.0000i 0.0000-0.0000i 0.0000-0.0000i
Case I: ω=0 then typical RFID shifted gate parameters values: L11=4.5mH,
L12=2.5mH, C1=23pF, R1=100kOhm=105, , K=0.6, 2·Lm=0.004 ( ).
12 12 243 312 34 12 34
1 1
0 ; 0 ; 3.95 10 ; = 3.95 10 ; 15.6 10R R RQ Q Q
(127)
sin(ω∙τ)=… fulfil and cos(ω∙τ)=… can't fulfil since .
Case II: ω≠0 ; ω>0 ; ω=2.1437, 2.7928, 1.0e+006, 1.0e+009, 1.0e+010, 1.0e+011, 1.0e+012, 1.0e+013 which can fulfil
expressions sinθ(τ)=… and cos(ω∙τ)=… Finally we plot the stability switch diagram based on different delay values of our
RFID shifted gate system.
4 2 10 3 5 24 2 24 2 1219.58 10 ; 8.85 10 ; 15.6 10 [7.8 10 7.9 10 ]R I RP P Q . (128)
3 5 24 2 24 2 12 17 24 248.85 10 ; 15.6 10 [7.8 10 7.9 10 ] ; [34.95 10 15.6 10 ] ; 15.6 10I R I IP Q Q Q (129)
2 24 24 24 2 24 2 1215.6 10 ; ( ) ; 15.6 10 {15.6 10 [7.8 10 7.9 10 ]}R R I I R R IQ V Q Q Q Q Q Q (130)
3 24 17 24 2 5 2 1015.6 10 [34.95 10 15.6 10 ] ; 26.55 10 ; 2 (2 19.58 10 )I R I RQ Q P P (131)
12 2 12 17 242 (7.8 10 7.9 10 ) ; Q 34.94 10 15.6 10 ; ( ) ( )R I R I I R R I I RQ U P P P P Q Q Q Q (132)
We plot the function:
2 21 1
2 2 2
2 { ( ) ( )}Re Re( ) ( ) ( ) ; ( ) ( ) ( )
4 ( )i i
F V P F U Pg g
F V P
(133)
2 21
2 2 2
2 { ( ) ( )}Re[ ( )] [ ( )] [( ) ] [ ]
4 ( )i
F V P F U Psign g sign sign sign
F V P
(134)
Since 2 2 2 1 2 24 ( ) 0 [ ( )] { ( ) ( )}F V P sign sign F V P F U P (135)
1 2 2 1 /[ ( )] {[ ] [( ) ( )]} ; ; ( )
/
F F Fsign sign F V P U P
F F F
(136)
1 2 2 1
2 2
1[ ( )] {[ ] [ ]} ; [ ( )] {[ ] [ ] [ ]}
V Usign sign F V U P P sign sign F
P P
(137)
; (138)
; (139)
(Table 3)
(140)
12 340; 0; ; 0R I R IP P Q Q
1 100pr Ohm11 122 2mL K L L
0cos( ) | 0
1
2 2
1[ ( )] {[ ] [ ] [ ]}
V Usign sign F
P P
1
2 2
1[ ] 0 [ ( )] {[ ] [ ]}
V Usign sign sign F
P P
1
2 2
1[ ] 0 [ ( )] {[ ] [ ]}
V Usign sign sign F
P P
1
2[ ( )] [ ] [ ]
V Usign sign F sign
P
2 [( ) ( )]R R I I R R I IF P P P P Q Q Q Q
Studies in Engineering and Technology Vol. 1, No. 1; 2014
72
We check the sign of according the following rule: If sign[Λ-1
(τ)] > 0 then the crossing proceeds from (-) to (+)
respectively (stable to unstable). If sign[Λ-1
(τ)] < 0 then the crossing proceeds from (+) to (-) respectively (unstable to
stable). Anyway the stability switching can occur only for ω≠0; ω>0 ; ω=2.1437, 2.7928, 1.0e+006, 1.0e+009,
1.0e+010, 1.0e+011, 1.0e+012, 1.0e+013 and . Since it is a very complex function we recommend solving
it numerically rather than analytic. We plot the stability switch diagram based on different delay values of our
Semi-passive RFID TAGs with double loop antennas system.
Table 3. Semi-passive RFID TAGs with double loop antennas stability switching criteria.
+/- +/- +
+/- -/+ -
6. Discussion
In this paper we consider Semi-passive RFID TAGs with double loop antennas. Due to electromagnetic interferences
there are delays in time for voltages and voltages derivatives in the first and second loop. These delays cause to stability
switching for our Semi-passive RFID TAGs with double loop antennas. We draw our Semi-passive RFID TAGs with
double loop antennas circuit and get system differential equations. Our variables are first and second loop antennas
voltages and voltages derivative. Our system dynamic behaviour is dependent on circuit overall parameters and parasitic
delays in time. We keep all circuit parameters fix and change parasitic delays over various values . Our
analysis results extend that of in the way that it deals with stability switching for different delay values. This implies that
our system behaviour of the circuit cannot be inspected by short analysis and we must study the full system. Several very
important issues and possibilities were left out of the present paper. One possibility is the stability switching by circuit
parameters. Every circuit's parameter variation can change our system dynamic and stability behaviour. This case can be
solved by the same methods combined with alternative and more technical hypotheses. Moreover, numerical simulations
for the Semi-passive RFID TAGs with double loop antennas model studied in references suggest that this result can be
extended to enhance models with more general functions.
7. Conclusion
Semi-passive RFID TAGs with double loop antennas environment is characterize by electromagnetic interferences which
can influence shifted gate system stability in time. There are four main RFID double loop antenna variables which are
affected by electromagnetic interferences, . Each loop antennas voltages variables under
electromagnetic interferences are characterize by time delay respectively1 2 1 2 1 2( ; 0 ; 0 ; , 0) . The two time
delays are not the same but can be categorized to some subcases due to interferences behavior. The first case we analyze
is when there is delay in RFID first gate's primary loop antenna voltage and no delay in secondary loop antenna
voltage.The second case we analyze is when there is delay in RFID gate's primary and secondary loop antennas voltages (
) and no delay in in gate's primary and secondary loop antennas voltages derivatives (Kuang, 1993; Beretta, et
al., 2002).The third case we analyze is when there is delay in RFID gate's primary and secondary loop antennas voltages (
) and delay in in gate's primary and secondary loop antennas voltages derivatives (Steven; Kuang, 1993).
For simplicity of our analysis we consider in the third case all delays are the same (there is a difference but it is neglected
in our analysis). In each case we derive the related characteristic equation. The characteristic equation is dependent on
double loop antennas overall parameters and interferences time delay. Upon mathematics manipulation and [BK]
theorems and definitions we derive the expression which gives us clear picture on double loop antennas stability map. The
stability map gives all possible options for stability segments, each segment belong to different time delay values
segment. Double loop antennas arranged as a shifted gate's stability analysis can be influence either by system overall
parameters values. We left this analysis and do not discuss it in the current article.
Appendix A: Lemma 1.1
Assume that is a positive and real root of
1( )
[0.001..10]
[ ]sign F 2[ ]V U
signP
1[ ( )]sign
[0.001..10]
'
11 11( ), ( )V t V t'
12 12( ), ( )V t V t
1 2
1 2 1 2
( ) ( , ) 0F
Studies in Engineering and Technology Vol. 1, No. 1; 2014
73
Defined for , which is continuous and differentiable. Assume further that if , , then
hold true. Then the functions , are continuous and differentiable on I.
Appendix B: Theorem 1.2
Assume that is a positive real root of defined for , and at some , . For
some then a pair of simple conjugate pure imaginary roots
of (141)
exist at which crosses the imaginary axis from left to right if and cross the imaginary axis from
right to left if where
* *
* * *
( )
( )Re( ) { | } { ( ( ), )} { | }n
i
dSdsign sign F sign
d d
(142)
Appendix C: Theorem 1.3
The characteristic equation has a pair of simple and conjugate pure imaginary roots real at if
for some . If , this pair of simple conjugate pure imaginary roots crosses the
imaginary axis from left to right if and crosses the imaginary axis from right to left if where
(143)
If , this pair of simple conjugate pure imaginary roots cross the imaginary axis from left to right if
and crosses the imaginary axis from right to left If where
(144)
If then and , the same is true when . The
following result can be useful in identifying values of where stability switches happened.
References
Aluf, O. (2008). RFID TAGs COIL’s Dimensional Parameters Optimization As Excitable Linear Bifurcation Systems,
(IEEE COMCAS2008 Conference, May 2008).
Aluf, O. (2011). RFID TAGs Coil's System Stability Optimization Under Delayed Electromagnetic Interferences, (IEEE
COMCAS2011 Conference, November 2011).
Beretta, E., & Kuang, Y. (2002). Geometric stability switch criteria in delay differential systems with delay dependent
parameters. SIAM J. Math. Anal., 33, 1144-1165.
Jack, K. H. (1991)Dynamics and Bifurcations. Texts in Applied Mathematics, 3, 170-484.
Kuang, Y. (1993). Delay Differential Equations with applications in Population Dynamics. Academic Press, Boston.
Steven, H. S. (1994). Nonlinear Dynamics and Chaos(Pp. 123-284). Westview press.
Supakit, K., Chuwong, P., & Danai, T. (2009). Novel design of double loop antennas by Using a shifted Gate for the LF –
RFID system. Faculty of Engineering and Technology, Asian University, Thailand.
Yuri, A. K. Elelments of Applied Bifurcation Theory. Applied Mathematical Sciences. 112.
This work is licensed under a Creative Commons Attribution 3.0 License.
I i R
( , ) ( , ) 0, Rn nP i Q i 0( ), nS n N
( ) ( , ) 0F 0, I I R * I *( ) 0nS
0n N
* * * *( ) ( ), ( ) ( )i i
( , ) 0D * *( ) 0
*( ) 0
* *( ), ( ) * I
* * *( ) ( ) 0n nS 0n N * *( ) ( )
*( ) 0 *( ) 0
* *
*
( )
( )Re( ) { | } { | }n
i
dSdsign sign
d d
* *( ) ( )
*( ) 0 *( ) 0
* *
*
( )
( )Re( ) { | } { | }n
i
dSdsign sign
d d
* * *( ) ( ) ( ) *( ) 0 *( )
Re{ | } 0
i
dsign
d
' *( ) 0nS
Studies in Engineering and Technology
Vol. 1, No. 1; February 2014
ISSN 2330-2038 E-ISSN 2330-2046
Published by Redfame Publishing
URL: http://set.redfame.com
74
Reviewer Acknowledgements
Studies in Engineering and Technology (SET) would like to thank the following reviewers for reviewing manuscripts
from October 15, 2013, to December 31, 2013. Their comments and suggestions were of great help to the authors in
improving the quality of their papers. Many authors, regardless of whether SET publishes their work, appreciate the
helpful feedback provided by the reviewers. Redfame Publishing appreciates the following reviewers’ rigorous and
conscientious efforts for this journal. Each of the reviewers listed below returned at least one review during this period.
Alexander Russell
Halil Karahan
Hossam Zaqoot
Isaac Atuahene
Roohollah Kalatehjari
Sonia María Gómez Puente
Martin Jaeger
Quamrul H. Mazumder
Shahab Bahrami
Andrew Walter
Editorial Assistant
Studies in Engineering and Technology
-------------------------------------------
Redfame Publishing
9450 SW Gemini Dr. #99416
Beaverton, OR 97008, USA
Tel: 1-503-828-0536 ext. 504
Fax: 1-503-828-0537
E-mail: [email protected]
URL: http://set.redfame.com
This work is licensed under a Creative Commons Attribution 3.0 License.
Paper Selection and Publishing Process
a) Submission Acknowledgement
When you submit a manuscript online, you will receive a submission acknowledgement letter sent by the online
system automatically. For email submission, the editor or editorial assistant sends an e-mail confirmation to the
submission’s author within one to three working days. If you fail to receive this confirmation, please check your
bulk email box or contact the editorial assistant.
b) Basic Review
The editor or editorial assistant determines whether the manuscript fits the journal’s focus and scope. Next a
check for the similarity rate is done using CrossCheck, powered by iThenticate. Any manuscripts out of the
journal’s scope or containing plagiarism, including self-plagiarism, are rejected.
c) Peer Review
We use a double-blind system for peer review; both reviewers’ and authors’ identities remain anonymous. The
submitted manuscript will be reviewed by at least two experts: one editorial staff member as well as one to three
external reviewers. The review process may take two to four weeks.
d) Decision Is Made
The decision to accept or reject an article is based on the suggestions of reviewers. If differences of opinion
occur between reviewers, the editor-in-chief will weigh all comments and arrive at a balanced decision based on
all comments, or a second round of peer review may be initiated.
e) Notification of the Result of Review
The result of review will be sent to the corresponding author and forwarded to other authors.
f) Publication Fee Is Paid
If the submission is accepted, the authors revise paper and pay the publication fee.
g) Print Journals Are Mailed
After publication, two hard copies of the journal will be sent to the corresponding author. If you want more
copies, please order at http://home.redfame.com/subscription-information/
h) Publication Notice
The authors and readers will be notified and invited to visit our website for the newly published articles.