6/5/2015

1

o n t o l o g y

Model Integration Using Ontology

Input-Output Matching

Linsey Koo, Franjo Cecelja

Process and Information Systems Engineering Research Centre

Chemical and Process System Engineering

University of Surrey, Guildford, United Kingdom

o n t o l o g y

Overview

Challenges in Biorefining Modelling

Existing Framework

Ontological Approach of Model Integration

Demonstration of Input-Output Matching

Conclusion & Future work

6/5/2015

2

o n t o l o g y

Challenges of biorefining modelling

Complexity of biorefining modelling & modelling methods Require a wide range of expertise and time

Independently developed models Different users

Best suited modelling tools

Various ways of developing models Achieve various aims and purposes

Cause inconsistencies of the models

Implicit models Lack of awareness of existing models

Limits the potential of reusability of these models

o n t o l o g y

Existing Framework: CAPE-OPEN

Address the question of standardisation of interfaces Enable interoperability between simulator software components

Framework built around a middleware Computation of physical properties

Simulation of particular unit operation

Numerical solutions for specific mathematical problems

Morales-Rodriguez, P., Gani, R., Déchelotte, S., Vacher, A., and Baudouin, O. (2011) Interoperability between Modelling Tools (MoT) with Thermodynamic Property Prediction Packages (Simulis® Thermodynamics) and Process Simulators (ProSimPlus) Via CAPE-OPEN Standards

Disadvantage:Lack of flexibility

6/5/2015

3

o n t o l o g y

Ontological Approach

Build upon the existing CAPE-OPEN framework Replaces the CAPE-OPEN middleware with a more intuitive and flexible

semantic repository

Semantic repository resides on web

Increase awareness of existing models Increase model reusability and integration

Reduce the developing time, efforts, and the need for expertise.

Service Discovery

Service Invocation

Service Composition & Interoperation

Service

Service Profile

Service Grounding

Service Model

supports(how to access it)

Repository

o n t o l o g y

Domain Ontology

Describe models representing biorefining processes Support explicit description of model and data & enhance consistency

, ∀ 0

, , ∀ 0

, ∶ ∧ ∀ 0

, , , , ,

Raafat, T., F. Cecelja, N. Trokanas and B. Xrisha (2013). An Ontological Approach Towards Enabling Processing Technologies Participation in Industrial Symbiosis, Computers & Chemical Engineering, 59, pp 33-46.

, , ∀ , ∈ ,

, , ∀,

6/5/2015

4

o n t o l o g y

, , , , ,

Domain Ontology

Raafat, T., F. Cecelja, N. Trokanas and B. Xrisha (2013). An Ontological Approach Towards Enabling Processing Technologies Participation in Industrial Symbiosis, Computers & Chemical Engineering, 59, pp 33-46.

dom | → :

rang | → :

Hence binary relationship:• and universal and existential quantifiers over properties of ,

• and , ∈ , cardinality quantifiers over properties of ,• and , ∈ ∨ , equality quantifiers over properties of .

o n t o l o g y

Semantic Model Description

Semantic Web Services (SWS) ontology Describe model/data defining biorefining

ServiceProfile: what it does (outputs, functionality)

ServiceGrounding: environment (software, preconditions)

ServiceModel: requirements of the models (key parameters)

ModelInputs

Precondi tions(environment)

Outputs

(effects)

Model Discovery

Model Invocation

Model Composition & Interoperation

Model

Model Profile

Model Grounding

Model Operation

supports(how to access it)

(1) matchmaking  agent

(1) analysi s of needs; 

(2) composition of descript ions from  mult iple services to perform  a specific task(3) coordination of act ivities of different 

participants(4) monitoring  the execution.

(1) communication protocol(2) message formats

(3) Port No. and/or software(4) unambiguous way of 

exchanging  data elements for each I/O 

6/5/2015

5

o n t o l o g y

Semantic Model Description

Semantic Web Services (SWS) ontology Describe model/data defining biorefining

ServiceProfile: what it does (outputs, functionality)

ServiceGrounding: environment (software, preconditions)

ServiceModel: requirements of the models (key parameters)

Data

Precondi tions(environment)

Outputs

(effects)

Model Discovery

Model Invocation

Model Composition & Interoperation

Model

Model Profile

Model Grounding

Model Operation

supports(how to access  it)

o n t o l o g y

Ontological Approach

Process of model integration Register as an instance of the domain ontology (Semantic Web Services)

Publish in the purposely built public repository for I-O matching

Use Ontology Web Service Description (OWL-S) framework Discovery stage

Selection stage

Composition stage

Execution stage

6/5/2015

6

o n t o l o g y

Partial Matching

Facilitate the flexibility of model integration through I/O partial matching Strong synthesis capabilities and functions

Invite degrees of freedom

Methods Semantic Matching

Property Matching

o n t o l o g y

Semantic Matching

Measure semantic similarities: Input/Output types & properties Facilitate the flexibility of model integration

Partially satisfying the input criteria

6/5/2015

7

o n t o l o g y

Property Matching

Distance measurement method: Numerical Vector Comparison

Cosine Similarity

Euclidean Similarity

Aggregated Similarity

, ∙ ∑ , ,

∑ , ∑ ,

, 1∑ , ,

∑ , ,

, ,

2N. Trokanas, F. Cecelja, T. Raafat, 2014, Semantic input/output matching for water processing in industrial symbiosis, Computer & Chemical Engineering, vol. 66, pp. 259-268.

o n t o l o g y

Backward Matching Process

Develop knowledge based input-output (I-O) matching technique An efficient and accurate model and data discovery

Avoid the process of expansion, compared to forward matching

Forward Matching Process

(Resource as a requestor)

Requestor

Model 1.1

Model 1.2

Model 1.3

Model 1.k

…

Model n.1

Model n.2

Model n.3

Model n.k

…

…

…

…

…

Requestor

Model n.1

Model n.2

Model n.3

Model n.k

…

Model 1.1

Model 1.2

Model 1.3

Model 1.k

…

…

…

…

…

Backward Matching Process

(Solution as a requestor)

6/5/2015

8

o n t o l o g y

Demonstration

Co-Fermentation (Requestor) using the bacterium Zymomonas Mobilis

Produce 315M litre of ethanol per year from corn stover

Operating temperature 41 degree C

Resident time of 1.5 days

Pre

trea

tmen

t

o n t o l o g y

Demonstration: Semantic Measure

Semantic Similarity Material Type

60%

20%

6/5/2015

9

o n t o l o g y

Demonstration: Property Measure

Property Similarity Measure Physical Property

Quantity, Temperature, Pressure, Composition

Registration Collect explicit information by parsing the ontology

Technology Type Pre-treatment

Model Function Dilute Acid and Enzymatic Hydrolysis

Model Output Sugar

Model Input BarleyStraw

Output Quantity 51 t/h

Operating Temperature 50ºC

Input Composition 38.1% Glucan, 26.9% Xylan2.6% Arabinan

Conversion 96% Glucose yield57% Xylose yield

Technology Type Pre-treatment

Model Function Ammonia Fibre Expansion

Model Output Sugar

Model Input Switchgrass

Output Quantity 46 t/h

Operating Temperature 37ºC

Input Composition 34.2% Glucan, 22.1% Xylan3.1% Arabinan

Conversion 93% Glucose yield70% Xylose yield

Technology Type Conversion Reaction

Model Function Co-Fermentation

Model Output Ethanol

Model Input Sugar from CornStover

Output Quantity 56 t/h

Operating Temperature

41ºC

Input Composition(CornStover)

37.4% Glucan, 22.1% Xylan

Conversion 99.7% Glucose yield94.3% Xylose yield

63%50%

o n t o l o g y

Demonstration: I-O Matching

OutputCo-Fermentation(Requestor)

Dilute Acid and Enzymatic Hydrolysis

(DAEH)

Ammonia Fibre Expansion

(AFEX)

Aggregated Similarity

Input

Output

Output

Input

Input

Semantic Similarity: 60%Property Similarity: 63%

Semantic Similarity: 20%Property Similarity: 50%

Result

6/5/2015

10

o n t o l o g y

Conclusions

Semantic Web Service descriptions for conversion technologies and resources are established to enable Input-Output matching

Model Integration through Input-Output matching was implemented in OWL-S Framework

Backward matching process for model discovery

Facilitate the flexibility of model integration using partial matching

Research continues towards identifying key parameters and understanding of the impact of partial matching

o n t o l o g y

Acknowledgement