TOWARD A SEMANTIC MANAGEMENT OF GEOLOGICAL

MODELING WORKFLOWS

Nabil Belaid

1,2

, Yamine Ait-Ameur

, St´ephane Jean

and Jean-Franc¸ois Rainaud

IFP Energies nouvelles/R114, 1 & 4 av. de Bois-Pr´eau, 92852 Rueil-Malmaison, France

LISI/ENSMA-UP, 1 av. Cl´ement Ader, 86961 Futuroscope, France

Keywords:

Semantic workﬂow, Semantic service, Ontology, Geological modeling, Meta-model.

Abstract:

Today, the petroleum industry demonstrates a greater concern toward ecology. Actions are undertaken in order

to reduce the CO2 emissions in particular. Storing the CO2 in depleted petroleum reservoirs represents one

of the ways of controlling the CO2 emissions. However, a prior modeling of the reservoirs has to be done

for a secure storage. This geological modeling task follows series of complex workﬂows (business processes)

of data processing services. These workﬂows are built without any methodological rule. It is difﬁcult for

geologists to execute workﬂows or even services that they have not designed. Moreover, it is not possible to

build new workﬂows without having a precise knowledge on the compounding services and sub-workﬂows.

We claim that in order to solve the two stated problems, explicit semantics has to be applied to describe the

workﬂows and the services that compose them.

In this article, we ﬁrst explain how geologists operate today. Then, we show how we enrich geological model-

ing workﬂows, the services that compose them and the data they manipulate with semantic indexations through

ontology-based characterizations (Geological Data and Services Ontologies). Finally, we explain how our ap-

proach supports reuse of existing workﬂows and build new ones.

1 INTRODUCTION

In the petroleum industry, controlling the CO2 emis-

sions has become a serious concern. One of the so-

lutions is to re-inject the produced CO2 in depleted

petroleum reservoirs. However, a prior modeling of

the target reservoirs is necessary for a secure storage.

In order to model the target reservoirs, data which

are obtained thanks to measure campaigns have to be

analyzed in order to build the geological structures of

the target petroleum reservoir. For example, images

on reservoirs, called seismic cubes, can be acquired

by seismic explorations. They can be interpreted by

geophysicists and geologists so that geological struc-

tures can be identiﬁed. Several other types of data

representations are deﬁned to store and exchange the

huge amount of created data, from which we quote

the seismic, the geological and the structural informa-

tion. These data are then processed leading to a ﬁne

reservoir description. This makes it ultimately possi-

ble to realize a precise CO2 ﬂow simulation forecast

in a target reservoir and answer to the question ”can

we store securely CO2 in the given reservoir?”.

This whole reservoir modeling task consists in

complex workﬂows

or processes of services. These

workﬂows, their composing services, their compo-

sitions into more complex workﬂows are not for-

mally modeled in this exploration and production part

of petroleum industry. They are built without any

methodological rule, executed and stored without at-

taching to them any information regarding what they

actually do. Consequently, no resource, except their

names may give information about their meanings

and their roles. Indeed, what would ”workﬂow21

geoReﬂect 15042009.wsdl” mean for a geologist? It

is then problematic for geologists to execute services

or workﬂowsthey havenot created because no seman-

tics is explicitly attached to them. Moreover, it is not

possible to build new workﬂows from scratch without

having a precise knowledge of how to compose their

compounding processing services. We claim that as-

sociating semantic descriptions to services and work-

ﬂows will help the geologists in their daily work.

In order to achieve these objectives, we have es-

tablished an approach based on the deﬁnition of on-

Throughout this article, we designate the atomic data

processing services or software by services and the compo-

sition of data processing services by workﬂows.

282

Belaid N., Ait-Ameur Y., Jean S. and Rainaud J..

TOWARD A SEMANTIC MANAGEMENT OF GEOLOGICAL MODELING WORKFLOWS.

DOI: 10.5220/0003098802820287

In Proceedings of the International Conference on Knowledge Engineering and Ontology Development (KEOD-2010), pages 282-287

ISBN: 978-989-8425-29-4

 2010 SCITEPRESS (Science and Technology Publications, Lda.)

tologies

. These ontologies, that we have contributed

to build, characterize the services, the workﬂows and

the data they consume and produce in the geologi-

cal modeling ﬁeld. Then, we have enriched the ser-

vices, the workﬂows and the data with a semantic

layer by linking them to their ontological character-

izations. This leads to, on the one hand, the easy re-

trieval of existing services and workﬂows, and on the

other hand, the building of new workﬂows.

The remainder of this article is structured as fol-

lows. Section 2 describes a case study in geological

modeling: the seismic interpretation workﬂows. We

present, in section 3, our ontology-basedapproach de-

scribing services and workﬂows semantics. Then, in

section 4, we outline the description of the ontology-

based database OntoDB which we use as a seman-

tic repository to store the services, the workﬂows and

their ontological descriptions. We show, in section

5, the implementation of the case study in OntoDB.

In section 6, we describe the Semantic Workﬂow con-

cept that allows building new workﬂows. Finally, we

conclude and discuss some perspectives.

2 SEISMIC INTERPRETATION

The goal of the seismic interpretation is to build a

structural model from basic geological objects. It is

a complex process and can be modeled by multiple

complex workﬂows. Figure 1 shows a candidate seis-

mic interpretation workﬂow

as it is modeled by ex-

perts in geology.

Figure 1: Simpliﬁed fragment of a potential seismic inter-

pretation workﬂow.

In this ﬁgure, we start from a SeismicCube which

represents the seismic images of a given reservoir.

Then, we apply a Reﬂector Extraction in order to ex-

tract the Reﬂectors which are small visible parts of

geological elements. Two services Horizon Reﬂec-

tor Association and Fault Mirror Detection are per-

An ontology is a formal and shared representation of

the knowledge of a given domain by a set of concepts and

the relationships between those concepts.

This workﬂow has been simpliﬁed for illustration.

formed in parallel to obtain respectively Horizons and

Faults etc.

The choice for this case study is motivated by the

fact that it deﬁnes a complex chain involving different

compositions of services. Moreover, precise informa-

tion on this part through the work of (Verney et al.,

2008) can be accessed.

3 FOR A NEW MANAGEMENT

OF WORKFLOWS

The main idea behind our approach (Ait-Ameur,

2009) is to create ontologies in order to index se-

mantically existing workﬂows and services descrip-

tions and the engineering models they reference. Here

lays the difference with the other approaches that

start from speciﬁc existing workﬂows and services

description languages such as ontology-based ap-

proaches (OWL-S (Martin et al., 2004), WSMO (Ro-

man et al., 2005) etc.) that intend to deﬁne the se-

mantics of the terms related to the services and work-

ﬂows or the annotation-based approaches (SAWSDL

(Farrell and Lausen, 2007), WSDL-S (Miller et al.,

2004) etc.) that intend to enrich services and work-

ﬂows descriptions. Our approach allows to: (i) store

heterogeneous descriptions of workﬂows and services

in a repository and (ii) add a semantic layer through

ontologies in order to bring semantics to these stored

workﬂows and services and enable their complex se-

mantic management.

First, consider the two horizontal layers (a) and (b)

of Figure 2. The principle consists in building models

that represent the data and the services and workﬂows

(Figure 2.(a)). Then, create the ontologies that char-

acterize them (Figure 2.(b)). Finally, link the models

to the ontologies level through indexations.

Figure 2: An approach for semantic handling of workﬂows.

TOWARD A SEMANTIC MANAGEMENT OF GEOLOGICAL MODELING WORKFLOWS

283

3.1 Data Stack

Engineering Models. Data are instances of engi-

neering models whose types or formats are usually

deﬁned in XML-Schemas. For example, integer and

string are simple data models and formated ﬁles such

as SEG-Y (representing seismic images) are complex

engineering models.

Domain Ontology. A domain ontology gives ex-

plicit semantics to the concepts of a given domain.

For example, there are ontologies that characterize the

medical ﬁeld or the mechanical parts. The domain on-

tology concepts index the engineering models enrich-

ing them semantically through an independent layer

without overloading them.

3.2 Services Stack

Services Model. The service models are the de-

scriptions of workﬂows and services. In our case

study, we have the example of a Web service that ex-

tracts Reﬂectors (produces a Hollow Matrix File from

a SEG-Y ﬁle).

Services Ontology. The ontologies of services are

semantic descriptions of services and workﬂows in a

given domain. They deﬁne hierarchies and composi-

tions of services. A concept of an ontology of services

is a semantic service and is independent of speciﬁc

implementations. Figure 3 shows a fragment of the

ontology that characterize the seismic interpretation

services.

Figure 3: Fragment of the seismic interpretation services

ontology and the referenced domain ontologies.

The ontology of services shown in Figure 3 has

GeoService as a root concept and more speciﬁcally

the concept SeismicInterpretation. From this concept,

three concepts are derived: Extraction, Association

and Merging which respectively characterize the ser-

vices for elements extractions, services that associate

elements and services for elements merging etc.

The concepts of this ontology of services have

properties that reference domain ontologies. For ex-

ample, the concept Reﬂector Extraction has as input a

Seismic Cube. Notice that if a semantic service sub-

sumes a second semantic service, the input and output

of the ﬁrst one subsume respectively the input and the

output of the second one (co-variance relationship).

Moreover, these ontologies are different from other

ontologies since they characterize particular elements

which have input, preconditions etc. Thus, they are

designed following a speciﬁc model of ontologies.

4 ONTODB: A SEMANTIC

REPOSITORY FOR SERVICES

Our approach cuts through three modeling levels: (i)

meta-models (model of ontologies), (ii) models and

ontologies and (iii) instances; and two stacks: (1) data

and (2) services. For a full exploitation of the pro-

posed approach descriptivepower, we need a database

architecture that enables a semantic based storage and

search. Classical database architectures are limited in

the way that they neither enable meta-modeling rep-

resentations nor offer a query language for querying

simultaneously the different database levels.

Recently, several approaches and systems were

proposed to store in the same database data and

the ontologies describing their meanings. We

call these databases, ontology-based databases

(OBDB).OntoDB is one of these approaches (Dehain-

sala et al., 2007). It stores explicitly in the database

not only the data, but also the conceptual model deﬁn-

ing the structure of data and the ontology representing

the semantics of data (see Figure 4).

Figure 4: The OntoDB architecture.

Our approach requires the deﬁnition of the con-

cept of semantic service not available in the classi-

cal ontology models. Since OntoDB allows accessing

KEOD 2010 - International Conference on Knowledge Engineering and Ontology Development

284

and enriching the meta-model level, we have chosen

this OBDB as a persistence model for our workﬂow

descriptions. Moreover, OntoDB offers the possibil-

ity of querying the databases at the ontology level

thanks to the query language OntoQL (Jean et al.,

2008). This language is also used to create and/or ma-

nipulate concepts of meta-models (models of models

or ontologies) using OntoQL statements (Figure 4.A)

of the form:

CREATE ENTITY #[name of the concept of ontology model] ([list of attributes

deﬁnitions]).

Example: CREATE ENTITY #Class (#deﬁnition STRING).

In the example, we enriched the ontology model

by adding the attribute ”#deﬁnition” to the entity

”#class”. Once the model of ontologies is created and

enriched, it is possible to create ontologies concepts

(valuating the new attribute) using OntoQL state-

ments (Figure 4.B) of the form:

CREATE #[name of the concept of ontology model] (DESCRIPTOR [name

of the ontology concept] PROPERTIES [list of attributes valuations] [list of

properties deﬁnitions]).

Example: CREATE #Class Person (DESCRIPTOR (#deﬁnition=’Humain

being’) PROPERTIES (uri STRING, name STRING));.

In the example, we have created the ontology con-

cept ”Person” with the properties ”uri” and ”name”.

It is ﬁnally possible to create instances of ontologies

concepts using OntoQL statements (Figure 4.C) of the

form:

INSERT INTO [name of the ontology concept] [list of properties to be

valuated] VALUES [list of properties valuations].

Example: INSERT INTO Person (uri, name) VALUES (1728, Belaid).

In this example, the ”Person” with the name ”Be-

laid” is instantiated. An example of each type of these

queries is given in next section and applied to our case

study.

5 IMPLEMENTATION

We present the application of our approach to the seis-

mic interpretation. The implementation of the data

stack is outlined in the work of (Mastella et al., 2008).

Here, we show how we have implemented the ser-

vices stack. We recall that the services ontologies are

particular characterizations. Thus, theses ontologies

are designed following a speciﬁc model of ontolo-

gies which we will describe with its implementation.

Then, we present the implementation of the ontology

of services shown in Figure 3 according to the speci-

ﬁed model. Next, we show examples of services and

workﬂows corresponding to a particular semantic ser-

vice and how the indexation is done. Finally, we ex-

plain how queries can be made on OntoDB in order to

perform semantic searches.

5.1 Ontology Model of GeoServices

Semantic services are characterized by different at-

tributes such as their inputs or their preconditions.

These attributes are not available in usual ontology

models as they are speciﬁc to services. As a conse-

quence, we have specialized the class constructor of

the usual ontology model to represent services. This

extension is depicted in Figure 5 and is stored in the

A part OntoDB (Figure 4.A).

Figure 5: Fragment of the model of services ontologies.

The following statement implements this exten-

sion on OntoDB:

CREATE ENTITY #SemanticService UNDER #Class (

#input REF (#SemanticData) Array,

...

#subtypeof REF (#SemanticService))

We enrich the ontology model by creating a new

entity that enables the deﬁnition of semantic services

and their attributes.

5.2 Ontologies of GeoServices

The characterization of the services and workﬂows in

the seismic interpretation ﬁeld leads us to the ontol-

ogy of services shown in Figure 3. The semantic ser-

vices that compose the ontology of services is stored

in the B part OntoDB (Figure 4.B). The following On-

toQL statement represents an example of the creation

of a semantic service Reﬂector Extraction as a con-

cept of the ontology of services:

CREATE #SemanticService ReﬂectorExtraction (

DESCRIPTOR ( #input = ’SeismicCube’,#output = ’Reﬂector’)

PROPERTIES ( uri STRING,description STRING,url STRING,

inputM DataType,outputM DataType))

We create a newsemantic service ReﬂectorExtrac-

tion and valuate its attributes. We also attach to it

TOWARD A SEMANTIC MANAGEMENT OF GEOLOGICAL MODELING WORKFLOWS

285

some properties such as its description language, the

input data type etc.

5.3 Geological Modeling Services

The services and workﬂows are indexed by deﬁning

them as instances of semantic services. They are

stored in the C part OntoDB (Figure 4.C). Table 1

shows an example of the deﬁnition of some services

and workﬂows corresponding to the semantic service

Reﬂector Extraction.

Table 1: Some services and workﬂows descriptions.

uri descr. url inputM outputM

uri1 BPEL ../RMark.bpel SEG-Y WITSML

uri2 WSDL ../getRef.wsdl SEG-Y XYZFile

... ... ... ... ...

Each service is described by values of the proper-

ties deﬁned on the semantic service it belongs to. As

an example, the service getRef in the second row of

Table 1 is a Web service described in WSDL and lo-

cated in ../getRef.wsdl. It takes as input a SEG-Y File

and produces an XYZ File as output. The following

OntoQL statement represents the indexation of this

service by instantiating the semantic service Reﬂec-

tor Extraction.

INSERT INTO ReﬂectorExtraction (uri, description, url, inputM, outputM)

VALUES (’uri1’, ’WSDL’, ’../getRef.wsdl ’, SEG-Y Format Type, XYZFile

Format Type)

We index a processing service by instantiating the

semantic service ReﬂectorExtaction and valuating its

properties. The service getRef is described in WSDL,

takes as an input data ﬁle a SEG-Y File etc.

5.4 A Semantic Repository for the

Handling of Geological Services

Our implementation consists in extending the ontol-

ogy model deﬁnition and then storing services, work-

ﬂow and data in a single repository: OntoDB. Stor-

ing all these data in a single repository enables com-

plex semantic queries. For example, the following

OntoQL query retrieves services that perform an ”Ex-

traction”, and produces as output a ﬁle in an ”XYZ

File” format:

SELECT SE.uri, SE.url FROM Extraction SE

WHERE SE.outputM = Typeof(XYZFile Format Type)

This query uses the semantic service subsumption

relationship. Indeed, services which are instances of

”Extraction” such as ”ReﬂectorExtraction” and ”Re-

ﬂectorGapExtraction” will be retrieved.

An interesting query consists in retrieving data

processing services that may replace another one.

For example, the following query retrieves the ser-

vices that could replace the service whose uri is uri25

(GS2). This query consists in searching all services

GS1 instances of the same semantic service as GS2

and that are not GS2:

SELECT GS1.uri, GS1.url FROM GeoService GS1, GeoService GS2

WHERE (TYPEOF(GS1).#oid = TYPEOF(GS2).#oid)

AND (GS2.uri LIKE ’uri25’) AND NOT (GS1.oid = GS2.oid)

The complexity of the previous queries lies in the

fact that they query the database at different levels

(meta-model or model of ontologies), different stacks

(data and services) and different conceptual repre-

sentations (models and ontologies).Theses queries are

hidden to the geologists thanks to a well adapted user

interface.

6 SEMANTIC WORKFLOWS

Building semantic workﬂows can be done by com-

posing semantic services. Such a task should be nat-

ural for ”experts in the given domain” since seman-

tic services represent explicitly the semantics of ser-

vices and composing semantic services into semantic

workﬂows is only a matter of ”knowing” what is the

order of execution of simple tasks in order to perform

a more complex task.

In what follows, we show how we describe seman-

tic workﬂows from semantic services. We then ex-

plain how a processing workﬂow can be derived from

the constructed semantic workﬂow.

6.1 Description of Semantic Workﬂows

The deﬁnitions of semantic workﬂows require differ-

ent operators to compose semantic services. Figure 6

presents the semantic workﬂow model we have estab-

lished.

Figure 6: Description model of semantic workﬂows.

A semantic workﬂow can be derived in a compos-

ite semantic service (semantic sequence, a semantic

parallel or a semantic loop) or a basic semantic ser-

vice. A semantic loop is composed by one semantic

KEOD 2010 - International Conference on Knowledge Engineering and Ontology Development

286

workﬂow while a semantic sequence or parallel can

be composed of multiple semantic workﬂows etc.

6.2 Building a Processing Workﬂow

When a semantic workﬂow has been built, it is possi-

ble to ”deduce” a processing workﬂow by selecting

a processing service for each semantic service that

composes the semantic workﬂow. For example, start-

ing from the semantic workﬂow described in Figure

1, Figure 7 represents a processing workﬂow obtained

when selecting a processing service for each semantic

service that composes the whole semantic workﬂow.

Figure 7: A potential seismic interpretation processing

workﬂow.

In this ﬁgure, we start from a SEG-Y ﬁle which

represents a 3D-image. Then, the service getRef.wsdl

is executed to obtain a Hollow Matrix ﬁle. Next, the

workﬂow Hormerg.bpel and the service Serv27.wsdl

are performed in parallel to obtain two XYZ Files

which represent respectively Horizons and Faults etc.

Although the semantic services of a semantic

workﬂow compose, it may not be the same for the

obtained processing workﬂow. Indeed, if in the pre-

vious example, the workﬂow ChenauxWF.bpel had as

output an IJK File, it would not be composable with

the following service which take as an input an XYZ

File. In this case, an adapting processing service that

converts the IJK File to an XYZ File is searched in the

database and is automatically added to the workﬂow.

7 CONCLUSIONS

Throughout this article, we have described an ap-

proach based on ontologies of services for the se-

mantic handling of services and workﬂows. We have

indexed the services, the workﬂows and the manip-

ulated data through ontologies concepts. On the

one hand, this semantic enrichment enables seman-

tic queries over existing geological modeling services

and workﬂows. On the other hand, it enables the de-

sign of new workﬂows based on semantic workﬂows

ﬁrst. This methodology is more natural for experts

in geology. This work represents a signiﬁcant step

toward the tool

we are currently designing and that

intends to assist geologists in searching, building and

executing their geological modeling workﬂows.

In future work, we plan to identify templates of

queries that are needed to perform geologists tasks.

These queries will be parameterized, integrated in the

graphical-user interface and thus hidden to geologists.

We will also extend the deﬁnition of semantic services

to take into account more detailed and crucial infor-

mation when referencing services, particularly non

functional requirements such as preferences, avail-

ability or quality of service.

REFERENCES

Ait-Ameur, Y. (2009). A semantic repository for adaptative

services. In Int. conf. SHWS’09.

Dehainsala, H., Pierra, G., and Bellatreche, L. (2007). On-

todb: An ontology-based database for data intensive

applications. In DASFAA’07.

Farrell, J. and Lausen, H. (2007). Sawsdl. Technical report,

W3C Recommendation.

Jean, S., Pierra, G., and Ait-Ameur, Y. (2008). Querying on-

tology based database using ontoql. In ODBASE’06.

Martin, D., Burstein, M., Hobbs, J., Lassila, O., McDer-

mott, D., McIlraith, S., Narayanan, S., Paolucci, M.,

Parsia, B., Payne, T., Sirin, E., Srinivasan, N., and

Sycara, K. (2004). Owl-s: Semantic markup for web

services.

Mastella, L. S., Ait-Ameur, Y., Perrin, M., and Rainaud,

J.-F. (2008). Ontology-based model annotation of

heterogeneous geological representations. In 4th Int.

Conf. WEBIST. Madeira, Portugal.

Miller, J., Verma, K., Rajasekaran, P., Sheth, A., Aggar-

wal, R., and Sivashanmugam, K. (2004). Wsdl-s.

http://www.w3.org/Submission/WSDL-S/.

Roman, D., Keller, U., Lausen, H., de Bruijn, J., Lara, R.,

Stollberg, M., Polleres, A., Feier, C., Bussler, C., and

Fensel, D. (2005). Wsmo. IOS Press.

Verney, P., Perrin, M., Thonnat, M., and Rainaud, J.-F.

(2008). An approach to seismic interpretation based

on cognitive vision. In 70th EAGE, Rome.

GWE (Geological Knowledge Editor)

TOWARD A SEMANTIC MANAGEMENT OF GEOLOGICAL MODELING WORKFLOWS

287