A Semantic Web Engine for Biorefining Model Integration

Edlira Kalemi, Linsey Koo and Franjo Cecelja

Centre for Process & Information Systems Engineering, University of Surrey, Guildford GU2 7XH, U.K.

Keywords: Semantic Engine, Ontologies, OWL-S Services, Biorefining Model Integration.

Abstract: The number of models available to the biorefining community is continuously increasing, there is a need to

provide better ways for their description, categorization, discovery and integration in order to improve

reusability of them. Biorefining models on the other hand are developed using heterogeneous methods, data

format and various environments that makes their reuse challenging. In this paper, we describe a semantic

web engine for the domain of biorefining, which enhances the description of biorefining model by using

semantic web technologies in order to facilitate discovery and integration. In particular, we present how

domain and web service ontologies are used in semantic mapping for the purpose of model integration,

which is achieved by OWL-S (Semantic Markup for Web Services). Whilst the application has been

designed and implemented for a specific domain, this novel design can be applicable to similar problems in

other domains.

1 INTRODUCTION

Mathematical modelling and simulation have been

widely and for long used to represent many aspects

of biorefining i.e. bioenergy production systems,

conversion processes, supply logistics and

environmental impacts (L. Wanga et al., 2015).

Those models can provide powerful tools to design a

biorefining system and evaluate its technical

feasibility, economics and environmental impacts.

As a result, there are a great number of models that

can be shared with the biorefining community and

hence reused. Model integration in biorefining is an

important function, which enables model integration

from unit composition to model orchestration and

invocation. The highest level of integration is the

supply chain integrating lower scale models in

varieties of ways. To this end, biorefining models

are characterized by heterogeneous methods, data

format and different development environments

making their reuse a challenge.

To address this problem, we introduce a semantic

application to enable model sharing, discovery and

integration. Main functionalities of the application

include; i) Model Registration where detailed

information on model functionality, inputs, outputs

and condition of operation are acquired, ii) Model

Publishing which gives the possibility for models to

be shared, iii) Model Discovery which supports the

C6Fermentation">based on different semantic

techniques and iv) Model Composition where the

user has the possibility to integrate his model with

other biorefining models that are discovered through

the application for achieving a more complex goal in

his modelling. Models represent at abstraction level

a biorefining process and are described semantically

by their functionality, inputs, outputs and

preconditions needed for their execution (Figure 1).

The semantic description of biorefining models is

supported by a biorefining domain ontology, as

demonstrated in the following section. CAPE-OPEN

initiative, a widely recognized industry standard that

defines rules and interfaces in chemical and process

systems community, is the only one that has given

significant contribution in model reusability and

interoperability (Belaud and Pons, 2002) of

Computer-Aided Process Engineering (CAPE)

applications or components. CAPE-OPEN

interfaces, provide the opportunity that process

modelling components can be assembled in process

modelling environments at runtime and has a well-

established communication mechanism by which

software components can exchange functionality at

runtime (Braunschweig et al., 2004). Beside the

advantages that CAPE-OPEN offers, the semantics

of interface are implicit and only human providers

and users are able to decide which software

component will best fit for a specific application's

272

Kalemi, E., Koo, L. and Cecelja, F.

A Semantic Web Engine for Bioreﬁning Model Integration.

DOI: 10.5220/0006439902720279

In Proceedings of the 7th International Conference on Simulation and Modeling Methodologies, Technologies and Applications (SIMULTECH 2017), pages 272-279

ISBN: 978-989-758-265-3

need and therefore the task of identifying the most

adequate model from the libraries is a manual

process (Braunschweig et al., 2004).

2 ONTOLOGY- DRIVEN

APPROACH FOR SEMANTIC

PUBLISHING AND DISCOVERY

Ontologies allow us to define concepts and relations

in the domain of interest, which can be processed by

a machine. OWL (Web Ontology Language) is a

logic based ontology language, which means the

semantics of the language is precisely defined by a

logic. Furthermore, OWL is rich of computation

properties and it is possible to perform computations

automatically using tools, called reasoners (Horridge

et al., 2014). The latest standard ontology languages

is OWL2 (Motik et al., 2009). Reasoners are

software systems that can be used for i) ontology

design time to check certain desirable consequences

follow from what has been stated and ii) application

runtime to provide query answering capabilities,

which ensures queries, such as representing

knowledge is correctly answered. There are many

reasoners, such as FaCT++ (Tsarkov and Horrocks,

2006) and HermiT (Shearer, 2008), that work well

with OWL. A biorefining domain ontology (Koo et

al., 2017) is exclusively developed to describe

biorefining models for the purpose of i) registering

and publishing them based on their semantic

description, ii) performing the process of semantic

discovery, and iii) facilitating the process of

integration. The models in the biorefining domain

ontology are describes in term of functionality,

inputs required for the successful execution of a

model, outputs generated after execution, and the

preconditions of the model, as shown in Figure 1.

Figure 1: Representation of biorefining models.

The ontology is developed in OWL and is often

enriched by adding rules using SWRL (Semantic

Web Rule Language) to provide an additional layer

of expression where OWL lacks expressive power.

The biorefining ontology contains all necessary

concepts and relations for capturing knowledge from

the biorefining domain with a special focus towards

integration of models. Model class is used to collect

information about a biorefining model as an instance

based on functionality, input, output, and biorefining

platform (Figure 2).

Figure 2: Illustration of the hierarchy of classes in the

biorefining domain ontology.

Object and datatype properties are used to define

data inputs and outputs of a model, object properties

are used to further characterize a model and SWRL

rules to define the preconditions which in most cases

in relation to the environment where a model run.

Figure 3 demonstrates an instance of a model in

ontology that belongs to SeparationProcess class.

Figure 3: Separation Process biorefining model instance.

2.1 Registration and Publishing

For publishing and sharing a model, the users will

initially have to register the model by accessing the

semantic web application. In technical terms the

ontology is parsed and checked for consistency by

Hermit reasoner and all classes are given to the user

in the form of a tree (Figure 4a). Figures 4a-c

provides a reduced view of the platform to

demonstrate the operation of the repository to

support the process of registration of the biorefining

models and data.

A Semantic Web Engine for Bioreﬁning Model Integration

273

Figure 4a: (Step 1) Model classification.

In step 1 (Figure 4a), user register a model under a

predefined category by browsing a taxonomy

presented in a form of tree. Every class in the

ontology is correlated with a number of datatype and

object properties, mainly in terms of its functionality

input, output and precondition, which is presented in

Figure 4b and c to complete the model registration

process.

Figure 4b: (Step 2) Name and description the model with

annotations and object properties.

Figure 4c: (Step 3) Complete inputs and outputs of the

model by giving value to the data properties.

Illustration of the registration process under

ModelbyFunctionality category in the biorefining

domain ontology is given in Figure 5. Together with

reasoner, fermentation model is registered in terms

of platform, as well as input and output type.

Figure 5: Registration of a fermentation model in the

biorefining domain ontology.

2.2 Discovery

By registering and publishing the models with the

necessary semantic annotation, biorefining models

are stored in a semantic repository and many experts

in the field can make use of them. Next important

step is to discover the models that the semantic

repository maintains. For discovering the required

model the user can choose among four options:

i) List of all models by different categories:

Browse the models as they are listed

based on different categories (i.e. by

functionality, by output etc.)

ii) Search for the desired model using two

filters: a) functionality and b) output of

the model.

iii) Write a SPARQL query to filter the

required models and run it through the

application.

SIMULTECH 2017 - 7th International Conference on Simulation and Modeling Methodologies, Technologies and Applications

274

iv) Semantic matching between a set of

requirements, which is set by the input

of a requesting model and the list of

models in the semantic repository as

candidate models that meet the

requirements. This method is developed

with an intention to achieve the purpose

of model integration and is further

detailed in Section 2.3.

The first and the second techniques are keyword

based queries. The SPARQL editor is made

available by employing jOWL (Decraene, 2007), a

jQuery plugin for OWL-RDFS documents and is

intended to be used by users who are familiar with

knowledge based systems and SPARQL querying

language.

2.3 Semantic Matching

Discovery of candidate models that fully or partially

matches the requirements of a requesting model is a

very important step for a successful model

integration. The requirements of a requesting model

are a set of inputs that the requesting model needs

which has to correspond to the outputs that a model

in the semantic repository has in order to be a

candidate model for integration. This way the

process of matching will result in the process of

comparing the input requirements that the requesting

model has with the outputs that candidate models

offer.

Figure 6: Model input/ output matching.

In our work we have used input/output matching

approach introduced by Cecelja et al. (2015) that

undergoes three-stage matching phases: i)

elimination, where all models that do not satisfy the

critical criteria are excluded from the selection list,

ii) semantic matching by calculating similarity

measures which defines the level of compatibility

between the requested model and the candidate

models and iii) ranking of the candidate models by

similarity measure. The critical criteria is defined by

the user during the formulation of the requesting

model. Semantic matching process calculates the

semantic relevance between the requesting model

and the models published in the repository. This is

calculated as an aggregated value (Algorithm 1) of

distance measure (Algorithm 2) and property

similarity (Koo et al., 2017). Distance measure

signifies the semantic similarity where the ontology

is processed as a graph structure and is calculated by

finding the shortest distance between concepts in the

ontology where paths has different weights. The

quantified distance measure is normalized with the

longest path between concepts in the ontology.

Property similarity is calculated by the average of

cosine and Euclidian similarity using the values of

the properties characterising the models as a vector.

Algorithm 1: Semantic Matching.

Input: O an ontology, M

the requesting

model which is instance of O

Output: S, a set of models ordered

based on the semantic matching

For all models in the repository

Eliminate (O, M

)

For all models in O after

elemination

CalculateSemanticSimilarity(O[M

], M

)

CalculatePropertySimilarity(O[M

], M)

S[i]←DefineMatchingMatching(Ss,

Qm)

Print S in descanted order

Algorithm 2: CalculateDistanceMeasure.

Input: O an ontology with the

reduces number of instances after the

elimination, the requesting model M

Output: K, a number representing the

semantic matching

For all models in the repository

findShortestPath(O, M

)

findLongestPath(O)

CalculateK(findShortestPath(O,

),findLongestPath(O))

Print K

3 SEMANTIC WEB SERVICE

FOR INTEGRATION

One main reason of reusing models in biorefining is

to generate new knowledge from the integration of

heterogeneous data, methods and tools. Semantic

web technologies are considered as the best method

to the support model integration as they use

graphical method to describe data models that

handle unstructured data. The approach we propose

to achieve the model integration in biorefining is

adopted form the semantic web services integration

A Semantic Web Engine for Bioreﬁning Model Integration

275

methodology and the input/output matching. We

consider each biorefining model is an instance of the

domain ontology as a web service and semantically

described with the OWL-S. OWL-S is a well-

established framework with respective service

ontology which, in this case, facilitates a full

automation of model integration. In practical terms,

OWL-S framework (Martin et al. 2004) is

implemented as an ontology with three

interconnected sub-ontologies (Figure 7); Service

Profile (Profile.owl), Process Model (Proces.owl),

and Process Grounding (Service.owl). Typically a

semantic web service has interlink to all the .owl

files in Table 1.

Figure 7: OWL-S upper layer ontology (Martin et al.

2004).

Table 1: C6Fermentation model as OWL-S service.

C6FermentationService.owl Service instance

C6FermentationProfile.owl Profile

C6FermentationProcess.owl Process model

Grounding.owl Grounding instances

Grounding.wsdl WSDL definitions for

grounding

Within the OWL-S framework, Service Profile

delivers a way to describe services, in our case,

models (Martin et al., 2004) and contains two

different types of information i) contact information

of the model provider and ii) functional description

of the model in terms of inputs required by the

requesting model, outputs generated from the

candidate model, preconditions for execution and

expected effects from the execution (IOPE). Below

is part of the content of the service and profile .owl

files of the C6Fermentation model described by

OWL-S. By taking advantage of OWL-S profiles

structure and their references to OWL concepts of

the biorefining domain ontology, the discovery

process find those models that are most likely to

satisfy the needs of a requesting model (Martin et

al., 2007).

C6FermentationService.owl

…………

<service:Service

rdf:ID="C6Fermentation">

<!-- Reference to the C6Fermentation

Profile -->

<service:presents rdf:resource="URI

C6FermentationProfile.owl#Profile_C6Fer

mentation "/>

<!-- Reference to the C6Fermentation

Process Model -->

<service:describedBy

rdf:resource="URI\C6FermentationProcess

.owl#C6Fermentation_Process"/>………………

</service:Service>

C6FermntationProfile.owl

…………

<!-- reference to the service

specification -->

<service:presentedBy

rdf:resource="URIC6FermentationService.

owl#C6Fermentation "/>

<!-- reference to the process model

specification -->

<profile:has_process

rdf:resource="URIC6FermentationProcess.

owl#C6Fermentation_Process"/>

<profile:serviceName>C6Fermentationt</p

rofile:serviceName>………………………………….

<profile:contactInformation>

<actor:Actor rdf:ID="C6Fermentation ">

<actor:name>C6Fermentation Process

</actor:name>

<actor:phone>67668788</actor:phone>

<actor:fax>412 268 5569</actor:fax>

<actor:email>email@domain.com

</actor:email>

……………………………….

<profile:hasInput

rdf:resource="URIC6FermentationProcess.

owl#TotalFlow"/>

<profile:hasInput

rdf:resource="URIC6FermentationProcess.

owl#Temperature"/>

<profile:hasInput

rdf:resource="URI/C6FermentationProcess

.owl#Pressure"/>

………………………………………………..

<profile:hasOutput

rdf:resource="URI/C6FermentationProcess

.owl#ComponentFractionEthanol"/>…………………

The Process Model describes a model in the sense of

how it works, the tasks it performs, the sequencing

of this tasks, the intermediate inputs and outputs and

the transformation that happens in each stage. A part

of the C6FermentationProcess.owl file code is

presented below, which expresses the process

model.

SIMULTECH 2017 - 7th International Conference on Simulation and Modeling Methodologies, Technologies and Applications

276

C6FermentationProcess.owl

…………………

<process:AtomicProcess

rdf:ID="C6Fermentation">

<rdfs:comment>

Get inputs for successfully running the

model.

</rdfs:comment>

<process:hasInput>

<process:Input rdf:ID="Get_Input_S1">

<process:parameterType

rdf:datatype="http://www.w3.org/2001/XM

LSchema#anyURI">

anyURI/BiorefiningConcepts.owl#ModelFun

ctionality </process:parameterType>

</process:Input>

</process:hasInput>

………………………

<process:hasOutput>

</process:parameterType>

</process:Output>

</process:hasOutput>……….

<process:inCondition>

<expr:SWRL-Condition>

<rdfs:label>hasPlatform</rdfs:label>

<expr:expressionObject>

<swrl:AtomList>

<rdf:first>

<swrl:IndividualPropertyAtom>

<swrl:propertyPredicate

rdf:resource="#hasPlatform"/>

<swrl:argument1 rdf:resource="#GAMS"/>

</swrl:AtomList>

</expr:expressionObject>

</expr:SWRL-Condition>

</process:inCondition>

OWL-S Grounding contain all necessary details for

invoking a model (Pedrinaci et al., 2011) and the

main function is to show how the inputs and outputs

of process are defined concretely as messages,

which carry those inputs and outputs in some

specific transmittable format. To achieve this

purpose OWL-S makes use of the Web Services

Description Language (WSDL) which is a well-

defined language for the purpose.

The model integration phase in the application

starts at the stage of model discovery. The user owns

a model (the requesting model) and tries to discover

other models that could be integrated with his

model, such as models that fully or partially match

the requirement of the model. This phase is

supported by input/output matching approach

explained in section 2.3. The user is given a rank of

available models that are matching fully or partially

to the requesting model. Based on the criteria, user

then selects one of the candidate models.

Automation of the candidate model selection and

Figure 8: Layered model composition architecture adopted

from Xiang et al 2015.

composition is avoided to give to the user flexibility

of choosing different pathways. Selection of the

most suitable model will call the composition and

invocation process to apply the model and data

integration. Figure 8 gives the model composition

layered architecture where: i) Base layer handles

model registration and invocation ii) Atomic model

layer contains the so called atomic models (simple

model) and serves as the base for planning and

invocation of models iii) Composite service layer

contains composite models (complex model that can

be further decomposed in atomic models) iv) Service

abstraction layer serves as the basis of model

composition v) Model Composition layer composes

models based on their semantic description vi)

Model planning layer provides the sequence of

models to be invoked based on a planning algorithm

vii) Model Invocation layer invoke models based on

the sequence generated from the previous layer.

4 APPLICATION

IMPLEMENTATION

The application is mainly developed using semantic

technologies but is also supported by other web

technologies. Figure 9 displays the semantic

A Semantic Web Engine for Bioreﬁning Model Integration

277

application architecture which is conceptualized in

three main pillars; i) Back end or Service Layer ii)

Middleware and iii) Front End. At the back end, the

semantic web services and the domain ontology are

stored. The core part of the system is developed in

Java and represents the middleware layer, which is

responsible for i) reasoning and interacting with the

knowledge domain ii) match making engine and iii)

interaction of front end interface and services. The

Front End, being a web application, will be accessed

through any browser and this part of the system has

been developed using Bootstrap and Java Server

Pages technologies. Through this interface the users

can i) search for specific models based on specific

criteria (i.e. the functionality of the model or output

of the model or for a set of criteria) ii) registration

of a model and iii) composition (Figure 10) of the

requesting model with the models found in the

repository (candidate models). As the knowledge

base expands the system is updated in real time with

the new knowledge and services. At the middleware

layer a key role plays OWLAPI (Horridge and

Bechhofer, 2011), a Java library for manipulating

OWL ontologies. In collaboration with the reasoner

(i.e. HermiT) and other supporting Java classes,

OWLAPI, are checking the domain knowledge for

consistency, parsing the inferred ontology for

passing it to further elaboration from the front end,

expanding the knowledge base with new instances of

models or integrated models. As the highest aim of

the repository is model integration the input/output

matchmaking module is a crucial component and a

processor step to our end use, the model integration.

For the development of the match maker OWL-S

Java API for parsing, serializing, validation,

reasoning, and execution services for OWL-S (Sirin

and Parsia, 2004).

Figure 9: Application architecture.

The Semantic matching algorithm makes use of the

JENA library (Dickinson et al., 2004), especially for

measuring distances between concepts

(findShortestPath()). Figure 9 shows the stage of the

application where a list of biorefining models is

offered to the user. The user can decide to test the

integration of his model by choosing any of the

candidate models even if the matching between his

model and the candidate ones are not full matching

(100%). Another instrument that can support the

user in his decision in the “Details” section for

every candidate model, where a justification of the

semantic similarity is given. Justification is given by

emphasising matching and non-matching properties

between the requesting and candidate model and the

semantic relevance that the requesting model has

with each of the candidate models.

Figure 10: Part of the application where a list of candidate

models is presented to the user for proceeding with the

selection and integration steps.

5 CONCLUSIONS

A semantic web engine for supporting the

biorefining community in model sharing and reuse

has been introduced. Semantic web technologies are

used to describe biorefining models semantically in

order to assure model discovery and integration

based on the content, instead of just syntax. All the

models are described by the biorefining domain

ontology and further treated and instances of OWL-

S ontology. Users of the application can be experts

of the biorefining community with different levels of

expertise. The application is a complex one and

build upon advanced technologies but this

complexity is hidden to the user and they can benefit

only from the advantage of it. The datatype

properties, inputs and outputs of a biorefining

model, are given by the owner of the model during

registration stage. In some cases the user might need

to test the integration for different data inputs.

Future work will be focused on designing an

intelligent algorithm that will facilitate the

SIMULTECH 2017 - 7th International Conference on Simulation and Modeling Methodologies, Technologies and Applications

278

integration of biorefining models with dynamic data

properties.

ACKNOWLEDGEMENTS

The authors wish to acknowledge the financial

support by the Marie Curie Initial Training

Networks Program, under the RENESENG project

(FP7-607415).

REFERENCES

Belaud, J.P., Pons, M., 2002. Open software architecture

for process simulation: the current status of CAPE-

OPEN standard, Comput. Aided Chem. Eng., 10 (C),

pp. 847–852

Braunschweig, B., Fraga, E., Guessoum, Z., Marquardt,

W., Nadjemi, O., Paen, D., Pinol1, D., Roux, P. Sama,

S., Serra, M., Stalker, I.,Yang, A., 2004. CAPE web

services: the COGents way, Comput. Aided Chem.

Eng., 18 (C) , pp. 1021–1026

Carroll,J., Dickinson,I., Dollin,C., Reynolds, D., Seaborne,

A., Wilkinson, K. , 2004. Jena: implementing the

semantic web recommendations, 04 Proceedings of the

13th international World Wide Web conference on

Alternate, Pages 74-83, doi>10.1145/1013367.

1013381

Decraene, D. 2007. jOWL. [online] Available at:

http://ontologyonline.org/ [Accessed 20 March 2017].

F. Cecelja, N. Trokanas, T. Raafat,M. Yu, 2015. Semantic

algorithm for Industrial Symbiosis network synthesis.

Computers &Chemical Engineering, Online.

Horridge, M., Bechhofer S., 2011. The OWL API: A Java

API for OWL ontologies. Semant. web 2

Horridge, M., Brandt, S., Parsia, B. and Rector, A. L.,

2014, A Domain Specific Ontology Authoring

Environment for a Clinical Documentation System,

IEEE 27th International Symposium on Computer-

Based Medical Systems, New York, NY, pp. 329-334.

doi: 10.1109/CBMS.2014.84

Koo, L., Trokanas, N., and Cecelja, F. , 2017. A semantic

framework for enabling model integration for

biorefining. Computers & Chemical Engineering.

http://doi.org/10.1016/j.compchemeng.2017.02.004

Martin, D.,. Burstein, M., Lassila, O., Paolucci, M., Payne,

T. and McIlraith, S., 2004. Describing Web Services

using OWL-S and WSDL. http://www.daml.org/

services/owl-s/1.1/owl-s-wsdl.html

Motik, B., Patel-Schneider, P. F., Parsia, B., 2009. OWL 2

Web Ontology Language structural specification and

functional style syntax. W3C - World Wide Web

Consortium, W3C Recommendation, [Online].

Available: http://www.w3.org/TR/owl2-syntax/

Pedrinaci, C., Domingue,J., Sheth, P. A.:Semantic Web

Services. Handbook of Semantic Web Technologies

2011: 977-1035

Shearer, R., Motik, B., Horrocks, I., 2008. HermiT: A

Highly-Efficient OWL Reasoner in Proc. of the 5th

Int. Workshop on OWL: Experiences and Directions

(OWLED 2008 EU)

Sirin, E., Parsia, B., 2004. The OWL-S Java API,

Proceedings of the Third International Semantic Web

Conference

Tsarkov, D., Horrocks, I., 2006. FaCT++ description logic

reasoner: System description in Proc. of the Int. Joint

Conf. on Automated Reasoning (IJCAR 2006),

Lecture Notes in Artificial Intelligence, vol. 4130.

Springer, 2006, pp. 292-297. [Online]. Available:

download/2006/TsHo06a. pdf

Wanga, L., Agyemangb, S. A., Aminib, H., Shahbazia, A.,

Mathematical modeling of production and biorefinery

of energy crops, Renewable and Sustainable Energy

Reviews, Volume 43, March 2015, pp 530-544, ISSN

1364-0321, http://dx.doi.org/10.1016/j.rser.2014.11.

008.

Xiang, D., Xie, N., Ma, B. and Xu, K., 2015.The

Executable Invocation Policy of Web Services

Composition With Petri Net. Data Science Journal, 14:

5, pp. 1-15, DOI: http://dx.doi.org/10.5334/dsj-2015-

005

A Semantic Web Engine for Bioreﬁning Model Integration

279