ONTOLOGY-BASED MODEL ANNOTATION OF

HETEROGENEOUS GEOLOGICAL REPRESENTATIONS

Laura Silveira Mastella,

Yamine Ait-Ameur,

Michel Perrin and

Jean-François Rainaud

Ecole des Mines de Paris, 60, boulevard Saint-Michel, 75272, Paris, France

LISI - ENSMA and University of Poitiers, BP 40109, 86961 Futuroscope Cedex, France

Institut Français du Pétrole, Direction Technologie, Informatique et Mathématique Appliquées

1-4 Avenue de Bois-Préau, 92852 Rueil-Malmaison Cedex, France

Keywords: Model annotation, Ontology, Petroleum reservoir modelling.

Abstract: In this work we study the process of petroleum reservoir modelling as a case of study for applying ontology-

based integration techniques and model annotation. A complete reservoir model gathers information

resulting from several interpretations using heterogeneous representations. We propose that the different

geological interpretations and representations should be annotated, for being integrated. We describe an

approach based on local ontologies for the specific domains, whose concepts are mapped to a shared global

ontology which contains the basic terms used in reservoir modelling.

1 INTRODUCTION

In this work we will study the process of petroleum

reservoir modelling as domain for applying

ontology-based annotation techniques. The models

used for petroleum exploration (earth models)

correspond to final representations of the geological

objects resulted from successive steps of

interpretation operated by professionals from

various earth science domains.

Considering this Reservoir Modelling Workflow,

one difficulty for semantic integration is the

heterogeneity of representation, since one same

geological item is likely to be pictured in many

different ways within the different geomodelling

applications.

To meet this issue, we will propose here an

approach based on semantic annotation of models.

We define local ontologies for annotating the

domain specific models and a global shared

ontology for gathering the vocabulary common to

all these domains. We will focus on a branch of

reservoir modelling activities which will be detailed

in the next section.

2 RESERVOIR MODELLING

WORKFLOW: A CASE STUDY

As a case study, we will consider here the part of

workflow going from the Seismic Interpretation to

the Structural Model building, starting from a real

field data set and choosing real geo-modelling

applications.

Seismic Interpretation: considering the seismic

image of

Figure 1(a), it can be interpreted by

identifying the portions of horizontal traces as

reflectors, or by considering that aligned vertical

traces correspond to interruptions. So, on

Figure

(b), the user identify several reflectors (r1, r2, r3)

and two main interruptions (int1 and int2). The user

then saves these seismic items as “cloud of points”

in data files, that will be the input of the structural

model application.

Structural Modelling: the files corresponding to

interpreted seismic will be imported into a structural

modelling application. Geologists will then apply

definite geological rules to assemble them. For

example, in the Figure 1(c) the faults f1 and f2,

interrupt the horizon h1. The structural model is

saved in specific structural modelling formats.

290

Silveira Mastella L., Ait-Ameur Y., Perrin M. and Rainaud J. (2008).

ONTOLOGY-BASED MODEL ANNOTATION OF HETEROGENEOUS GEOLOGICAL REPRESENTATIONS.

In Proceedings of the Fourth International Conference on Web Information Systems and Technologies, pages 290-293

DOI: 10.5220/0001520302900293

 SciTePress

Figure 1: (a) Raw seismic image, (b) Interpreted seismic

and (c) Structural model.

The practical issue that has to be solved is how

the concepts of the “Seismic domain” can be put in

correspondence with those of the “Structural

modelling domain”. This is not possible at present

since we cannot recover the relation between objects

identified in different phases of the process.

2.1 Geological Objects

Geological objects are identified in the beginning of

the workflow and evolve within the different earth

models. Geological objects appear as the red thread

to which all interpretations and representations in

the workflow should be attached and that can thus

guide most of the modelling process. For this

reason, we believe that the entities considered in the

various representations should all be characterized

as actual geological objects having a unique

identification. The geological objects that we will

use in this case of study are represented in a geology

ontology formerly presented in (Perrin, Zhu,

Rainaud, & Schneider, 2005) and (Mastella, Perrin,

Abel, Rainaud, & Touari, 2007). An extract of the

whole ontology is shown in

Figure 2. It represents

the basic vocabulary shared by all earth science

domains.

Figure 2: Extract of Basic Geology ontology

Another problem is that current software systems

are not able to take into account the fact that the

successive categories of data have been interpreted

as corresponding to the same geological object.

To address this issue, we believe that each of

these geological objects should also be linked to its

specific representation along the modelling chain.

2.2 Geological Applications’

Metamodels

Each task of the workflow uses a different earth

modelling application, which represent the

geological objects in a different way. So as to

identify the objects within applications, we need to

semantic annotate their metamodel, which

represents the primitives used by the application to

represent a geological object. For example, the

metamodel of a seismic interpretation analysis tool

is as shown in

Figure 3.

Figure 3: Metamodel for the seismic interpretation

This metamodel stipulates that an object has

different seismic associated properties (frequency,

etc), which will be useful to identify the concept in

the ontology later on. We will describe in the next

section the approach that we propose.

3 ONTOLOGIES FOR MODEL

ANNOTATION

Ontology-based annotation of resources allows to

assign explicit meanings to objects and features

interpreted by an observer. In this work, we intend

to use ontologies to annotate domain specific

models. We describe here an annotation architecture

that helps users to make explicit their interpretation

about the geological models. Unlike the common

methods, the annotation architecture in this work are

not automated; it is expected that human users will

provide the detailed annotations of the models,

subject to the contents and constraints of the

ontology. The goal of the completed annotations is

to offer a knowledge base (knowledge = geological

models' data + annotations) which stores the

geological interpretation.

ONTOLOGY-BASED MODEL ANNOTATION OF HETEROGENEOUS GEOLOGICAL REPRESENTATIONS

291

3.1 Model Annotation Architecture

We are choosing ontologies because it is a

consolidated approach to solve the problem of

integration of heterogeneous information (Noy,

2004; Uschold & Gruninger, 2005). The work of

(Lin, Strasunskas, Hakkarainen, Krogstie, &

Solvberg, 2006) describes an approach of semantic

annotation of process templates, for better reuse of

this process in the business workflow using a

general ontology.

We intend to set up the hybrid approach of

ontologies (general + local) to the problem of the

reservoir modelling workflow. Indeed, we have

different specific knowledge domains and one pivot

field that is shared by the others. For this reason, we

propose an annotation methodology resting on (1)

local ontologies (LO) which represent the concepts

of the local domains of expertise or activities, such

as Seismic (an extract represented in UML-like class

diagram is shown in Figure 4), Structural Geology,

that are required for annotating each specific

representation; (2) a global ontology (GO), the

Basic Geology Ontology (section 2.1), which links

concepts used in the local ontologies; (3) application

metamodels, for specifying how computing

applications represent geological objects.

Figure 4: Extract of Seismic Interpretation LO

The objective is not to integrate the concepts of

the local ontologies inside the global ontology, but

to establish subsumption links (isA and isCaseOf)

between the local concepts and the shared concepts.

Accordingly, we require to each LO concept

instance to be an instance of at least one subsuming

concept in the GO. We will use LO concepts to

annotate the specific applications metamodels by (i)

creating an interpretation link between an

ontological concept and a metamodel entity and (ii)

assigning a unique identifier to the metamodel entity

that is interpreted. The objective is to allow the

ontological manipulation of the application aspects.

This implies that each instance of the LO can be

referred and accessed from the GO level without any

specific knowledge nor expertise of the LO level.

In the moment when the geologist performs an

interpretation, he assigns a unique identifier to the

instance of the local ontology, which is the same of

the entity in the modeling tool. Figure 5 shows how

to annotate an entity of the seismic metamodel

(Cloud of Points) with a concept of the seismic local

ontology (Reflector).

Figure 5: Seismic interpretation: Cloud of Points

annotated as Reflector.

We have defined an approach on metamodels

and ontological concepts. Next section shows a

complete case study involving instances of the

concepts described above.

4 APPLICATION TO THE CASE

STUDY

Let us consider the two tasks shown in section 2: the

Seismic Interpretation and the Structural Model. We

will see in Figure 6 the architecture that represents

how to annotate the files that represent geological

objects.

The objects recognized in the seismic

interpretation are the reflectors and interruptions.

When this interpretation is saved in the specific

computing application, they are saved as clouds of

points. In order to annotate the seismic files, we

create a link from the metamodel entities to the

Seismic Local Ontology (Figure 6 (a)).

In a second phase, the reflectors are interpreted

by the geologist as portions of horizons; and the

aligned interruptions are interpreted as faults. In this

case, the user is using a Basic Geology vocabulary.

So, the link is made between the concepts of the

Seismic Local Ontology to the concepts of the

Geology Global Ontology (

Figure 6 (b)). This

represents the basic subsumption link between the

concepts of the two ontologies: Reflector is_a

Horizon, Interruption is_a Fault.

When the user passes to the structural modelling

phase, his/her interpretation consists in identifying

structural objects from the image. The geologist may

identify several portions of horizons that are likely

WEBIST 2008 - International Conference on Web Information Systems and Technologies

292

to be parts of one same structural horizon. And

he/she can specify the structural category of the fault

(normal, reverse), by observing the way it affects the

structure. This illustrates the subsumption link

between the concepts of the GO and the concepts of

Structural Geology LO: NormalFault is_a Fault

(

Figure 6 (c)).

Finally, when the user stores the structural

model, the objects are saved in binary files that

represent a specific structural interpretation

metamodel (such as ICarre metamodel

), whose

entities are annotated with the concepts of the

Structural modelling LO. (Figure 6 (d).

Figure 6: Links between GO, LO and Metamodels.

Setting up this approach, we are able to answer

to queries that refer to different domains. Such a

query would be, for example, by which seismic

reflectors is formed the structural horizon H1? An

structural horizon is represented as an

ICarreHorizon (d), which is annotated as

StructuralHorizon (c), which is subsumed by the

concept Horizon in the GO. All those instances have

the same ID. It is then easy to retrieve the instances

of Reflector in the Seismic LO that are subsumed by

the GO Horizon and that are used to annotate the

looked-for “cloud of points” files.

5 CONCLUSIONS

We have presented here an approach based on

ontologies to annotate specific domain conceptual

models. The application domain is the workflow of

oil reservoir modelling, which is a multi-

representation multi-interpretation domain. In this

process, an interpretation can be considered as

putting in correspondence concepts belonging to

different specialized domains. So, our architecture

proposes to create semantic annotations from the

specific metamodels to the local ontologies. Then,

the local ontologies concepts are subsumed by the

global ontology concepts, which is the pivot of the

modelling process.

Creating correspondences between the models is

likely to enable us to answer queries that cannot be

addressed at present, because we cannot recover the

relation between objects identified in different

phases of the process.

The next steps in this work will be to automate

most complex mapping rules, which will represent

inferences that can be made within specific domains.

Moreover, we expect to scale up the proposed

approach to deal with large file size interpretations.

We plan to use persistent models with ontology

based databases.

REFERENCES

Lin, Y., Strasunskas, D., Hakkarainen, S., Krogstie, J., &

Solvberg, A. (2006). Semantic Annotation Framework

to Manage Semantic Heterogeneity of Process

Models. In S. B. Heidelberg (Ed.), Advanced

Information Systems Engineering (Vol. 4001/2006,

pp. 433-446): Springer.

Mastella, L., Perrin, M., Abel, M., Rainaud, J.-F., &

Touari, W. (2007). Knowledge Management for

Shared Earth Modelling. Paper presented at the EAGE

Conference & Exhibition, London.

Noy, N. F. (2004, December 2004). Semantic Integration:

A Survey Of Ontology-Based Approaches. SIGMOD

Record, Special Issue on Semantic Integration, 33.

Perrin, M., Zhu, B., Rainaud, J.-F., & Schneider, S.

(2005). Knowledge-driven applications for geological

modeling. Journal of Petroleum Science and

Engineering, 47(1-2), 89-104.

Uschold, M., & Gruninger, M. (2005). Architectures for

Semantic Integration. Paper presented at the Dagstuhl

Seminar Proceedings: Semantic Interoperability and

Integration, Dagstuhl, Germany.

. UML-like class diagram from TopBraid Composer tool

(http://topbraidcomposer.com).

. UML class diagram.

. OpenFlow ICarre proprietary modelling application of French

Institute of Petroleum (http://www.ifp.fr).

ONTOLOGY-BASED MODEL ANNOTATION OF HETEROGENEOUS GEOLOGICAL REPRESENTATIONS

293