Ontological analysis of the lithology data in PPDM well core model

Transcription

1 Ontological analysis of the lithology data in PPDM well core model Mara Abel 1), Alexandre Lorenzatti 2),Sandro Rama Fiorini 1) and Joel Carbonera 1) 1) UFRGS 2) Endeeper INTRODUCTION The quality of petroleum reservoirs (porosity, permeability) is controlled by a large set of parameters captured from different sources in reservoir studies and should be integrated into earth models to produce predictive models. Out of the cost in obtaining the data, the heterogeneity of the information is a major barrier in producing realistic models. A data set is considered heterogeneous when the data are diverse in origin, format, and content and, as a consequence, are not directly or automatically comparable. Heterogeneity can be a consequence of different geologists views or goals when capturing the data, or can be a result of representing the geological models through distinct modeling languages, software or distinct data models. Software and data heterogeneity demand effort and investment in order to translate distinct data formats into a homogeneous representation that can offer a uniform view of information to be manipulated by a single system. However, the biggest difficulty in information integration is not related to representation itself, but to the challenge in preserving the intended semantics of information that was captured by the geoscientist in further data transformation. Even when all models refer to the same geological entities in reality, poor representational artifacts restrict the capability of expressing the original meaning of data and can propagate wrong data and interpretations through the exploration chain. We describe a methodology to analyze legacy models based on the Foundational Ontology view, aiming to identify and formally describe the original semantics of the represented entities. Our aim is to show that many problems related to heterogeneity can be solved or at least simplified by analyzing the models through the view of Ontology (Abel et al., 2015). We based our approach on the same premises of (Bergamaschi et al., 1998): (a) for each source, its conceptual schema, i.e., meta-data, is available; (b) the semantic information is encoded in the schema; (c) a common data model for describing the information to be shared is available; and (d) a partial or total unification of the source schemas is able to be performed. We exemplify our methodology through the ontological analysis of the PPDM lithology data model associated to well data. We base our analysis on the ontological meta-properties of the Ontoclean methodology (Guarino and Welty, 2004), which defines identity, unity and dependence relations about domain entities. In order to orient our conceptualization about the data, we consider that the modeled rock information was intended to support reservoir quality interpretation and feed reservoir models. We expect that our process shows itself as a useful tool for model documentation, interpretation traceability and information integration in the building of the reservoir model. PNEC Conferences 2015, Houston 1

2 PRELIMINARY ASSUMPTIONS Ontology is a logical theory accounting for the intended meaning of a formal vocabulary and the commitment of this vocabulary to a conceptualization taken from someone about the world (Guarino, 1998). Ontology-based approaches deal with what are the entities of the world and what is invariant in them, and can be used to track these entities through time and space (and obviously, through legacy systems). An ontology can be also described by the materialized artifact of the conceptualization: the set of formal models that describes the meaning of a shared vocabulary. Ontology engineering is growing in importance as a conceptual tool for providing interoperability among applications and data in petroleum industry. Any modeling activity is based on a pre-defined set of entity types that are acknowledged by the modeler that applies his/her particular view over this data. He/she generates the mental abstractions of the various entities of the world under analysis to support his/her reasoning process. However, we will assume here a realism philosophical view, in which the various conceptualizations and the various models that are deduced from these conceptualizations all refer to the same portion of the reality that exist in an independent way of the modelers. In addition, the endurantist philosophical view of modeling (Zimmerman, 1996) considers that there exist some entities in reality that exist through time, which means that they are wholly present at more than one time without losing its identity. That is the case of a person that became an adult, a student or a professional without losing the ontological identity. Our modelling strategy starts by giving names and modeling endurant concepts, such as rocks, layers or faults. Our representational framework takes in consideration the referentialist approach, which states that there exists some entity in the real world that materializes the meaning of any named concept of the model (Gärdenfors, 2000). Otherwise, there are some entities that exist in time, like an earthquake, a deposition process, or any geological phenomena. These entities require the existence of endurant entities in order to occur. Our philosophical choices allow us to consider that there is only one reality to be modeled and that geological entities keep their identity through time (even when they are modified by some geological phenomena) until they stop existing. Therefore, a conceptual model is a snapshot of entities that have spatial temporal existence in the world. In a simple way, we mean that for any model, we can anchor the concepts of our model in concrete geological entities of the world that will provide meaning to them. Being endurants, these entities can provide a common reference to integrate the models. This assumption presumes that the modeled concepts refer to entities in reality that have spatio-temporal expression, which is generally correct for Geology, but maybe is not true for all concepts of Geophysics whose models can express also mathematic abstract entities. Even with this intrinsic limitation, the method shows itself a useful framework for integration. In addition, in order to deal with the problem of modelers with different knowledge or goals producing distinct conceptualizations over the same reality, we propose to make them clear by expressing the characteristics of the concepts in a detailed way, using their ontological properties. When we adopt this referential, the differences between the views of a structural geologist, of a stratigrapher and of a petrologist are just related to the differences of ontological choices, i.e. to the choices that each of them makes in considering a particular set of qualities among the large (but finite) number of qualities that are attached to the entities that they intend to model. Therefore, even when interpreted under the view of distinct professionals, the geological entities can have their aspects fully expressed, making their differences being easier detected and providing a common framework through which several models can be integrated. METHODOLOGICAL APPROACH In this paper, we had analyzed the PPDM models conceived to store in databases the rock data obtained from wells: cores descriptions, log profiles, out of operational perforation and location information. Given the space restriction, our focus here shall be on the information regarding to rock and rock properties, reducing our analysis to 49 PPDM lithology tables whose names start with LITH or PNEC Conferences 2015, Houston 2

3 R_LITH. Our intention is not make a complete analysis, but only to show how the analysis can be done, allowing the reader to reproduce the process. Considering the PPDM model, we begin by identifying the concepts that are potentially able to anchor the modelled concepts to real entities. These concepts have the property of having proper identity and being rigid, in the sense that their instances have their own identity (a specific instance can be distinguish from other instances by its characteristic properties) that will be preserved in space and time. (Guarino and Welty, 2000). In a didactical example, an instance of an entity person cannot stop being a person without ceasing to exist. Otherwise, a student can stop being a student and keeps existing (i.e. as a person). Indeed, the entity person is the one that provides identity to the entity student and there is no instance of student that is not an instance of person. We say that person is a rigid concept with ontological identity, while student is an anti-rigid concept that inherits its identity from person. A deep understanding of the metaproperties of entities can be obtained in (Guizzardi, 2005). There are rigid concepts whose instances can be individualized as a whole, such as a person, a cup or a lake. Other concepts are defined as amounts of matter that cannot be individuated by themselves. This is the case of uncountable entities, like water, wood, petroleum and rock, which require a lake, a cattail, a barrel, or a core to be individuated. In this work, we named rigid countable concepts kind; rigid uncountable concepts as quantity, and anti-rigid concepts as roles. Roles will be countable or uncountable according to the property of the rigid object that provides them with identity. A concept is defined by the set of intrinsic properties (or qualities) and the possible values (quality domain) that a property can assume. An entity in reality instantiates a unique value (or qualia) among the possible values of the concept quality domain. Qualities are important in earth models because they allow us to recognize when two entities with similar names are referring to distinct concepts. Also, interpretation tasks (like defining the quality of a reservoir) are mostly based on their vales. Then, our ontological model will provide accuracy in the definition of properties and quality domains. Considering only the entities under our analysis, most of important concepts in exploration steps are related to the intrinsic properties of rocks and petroleum: texture, porosity, permeability and viscosity. In contrast, many decision tasks in production are concerned with the properties of the entities that individuate rock and petroleum: size, geometry and position of rock unit, boundary between rock and water and its position, etc. Many problems related to information integration result of collapsing the identity of kinds, roles and quantities when representing entities. This might prevent the modeler of recognizing the relevant properties of entities and associating them to the correct instances on the prospect. THE ONTOLOGICAL LITHOLOGY MODEL OF PPDM We start our analysis by defining the concepts that refer to amounts of matter and their intrinsic properties mentioned in the PPDM model. Then we identify the entities that delimit the instances of these concepts and allow them to be individuated and counted. We continue by analyzing some special entities that describe processes or situations. We finish by showing a general view of the relationships among the modelled concepts. The first concept to be described is SEDIMENTARY ROCK, described in Table 1. We prefer using rock instead lithology, because the first term refers to the amount of matter itself, while lithology refers to the description or study of rock properties. Rock is a natural aggregate of mineral or mineraloid grains or crystals. Considering the scale of analysis as macroscopic size, we describe the range of sedimentary rock property values that can be recognized with naked eyes only. The quality domains were completed following the study done by (Abel, 2001). Some attributes were simplified to adjust to the level of detail of PPDM description. PNEC Conferences 2015, Houston 3

4 Meta Type Quantity Table 1 Ontological model of the SEDIMENTARY ROCK concept Name Rock Quality Quality domain Number of values Color The Munsell Rock Color Chart scale. Main grain size Numerical main grain size Second grain size Numerical second grain size Intercalation Sorting Fabric orientation Fabric support [boulder, cobble, pebble, gravel, very coarse sand, coarse sand, medium sand, fine sand, very fine sand, silt, clay] Wentworth scale [boulder, cobble, pebble, gravel, very coarse sand, coarse sand, medium sand, fine sand, very fine sand, silt, clay] Wentworth scale [none, sandstone siltite, sandstone shale, siltite shale, sandstone siltite shale, carbonate sandstone, carbonatesilt, carbonate shale, carbonate sandstone siltite shale] [very well sorted, well sorted, sorted,moderately sorted, poorly sorted] [homogeneous, parallel oriented, imbricated, heterogeneous, chaotic] [grain supported, grain to matrix supported, grain to cement supported, matrix supported, matrix to cementsupported, cement supported] Two values delimiting an interval Two values delimiting an interval Two values delimiting an interval Two values delimiting an interval Fabric main shape [equant, oblate, prolate, spheroidal, rod, blade, disc] Matrix class (For Achie equation calculation) Cement type Porosity type Permeability grade [Diagenetic silica, diagenetic feldspar, diagenetic clays, zeolite, pseudomatrix, iagenetic carbonate, diagenetic sulfate, iagenetic sulphide, iron oxide/ hydroxide, diagenetic titanium, other diagenetic constituents] [Intragranular, intergranular, moldic, grain fracture, rock fracture, shrinkage, oversized,lamellar decompaction] [no porosity, low porosity, moderately porous, porous, very porous] List of values Porosity grade Percentage Consolidation [dense, hard, medium hard, soft, spongy, friable] Sedimentary rock class [conglomerate, sandstone, lutite, carbonate, evaporite, siliceous, phosphatic, iron rich] The instances of rock, as described above, are unlimited portions of matter. However, our integration approach is based on the strategy of anchoring the concepts in their instances on reality. In order to have PNEC Conferences 2015, Houston 4

5 a concrete instance of rock, it is necessary to model the concepts that delimit the uncountable instances, especially for rock. Instances of concepts like outcrop, cores, samples and thin sections will individuate a portion of rock for analysis or handling, i.e. these instances will be constituted by instances of rocks. One sensible problem for integration is that the geologist has fully freedom to choose the portion of domain he/she wants to describe and model. This choice would consider both the scale of analysis and the properties to be emphasized. The result may include different portions of the same rock or even two rocks in a same portion, producing records that are hardly integrable. In order to deal with this problem, Lorenzatti proposed in (Lorenzatti, 2010) a method for core and outcrop description based on the separation of the portion of rock to be described in distinct rock facies. A rock facies is a portion of rock that has homogeneous lithological class, color, texture, structure and fossil content (Prothero, 1990). This means that any modification in a single parameter will define a new facies and, as a consequence, a new portion to be described. The method has two important advantages. The first is that it will help the definition of automatic interpretation of the genetic conditions of rock creation, since facies reflect the depositional or chemical process of rock formation. The second is that it will provide better conditions for automatic correlation, since different geologist would tend to describe the same portions of a rock occurrence. We follow this proposal in the definition of rock portions in our ontological model. We start our definition with the LITHOLOGICAL UNIT described in Table 2, which delimits in 3D space some amount of rock. The relationship between LITHOLOGICAL UNIT and ROCK is defined as CONSTITUTED_BY, since any unit is made of rock and two lithological units can share the same rock properties. Having spatial expression, the lithological unit is defined by its geometric parameters and location, while the instance of rock that constitute a lithological unit is defined by its internal properties (texture, composition). The separation allow us to interpret automatically that two different lithological units are constituted by the same rock, and deal with geometric properties of rock bodies (size, volume, positions) in a separated way from internal properties of rock (porosity and permeability) that are derived after the study of several different portions of rocks. Meta Type Kind Table 2 Ontological model of the LITHOLOGICAL UNIT concept Name Lithological unit Quality Quality domain Number of values Top depth Bottom depth Stratigraphic unit identifier Stratigraphic name set identifier Biotrubation index Fossil presence Number that measure that distance between the top of the unit and the head of the well Number that measure that distance between the bottom of the unit and the head of the well unique identifier for the stratigraphic unit Unique identifier for the stratigraphic name set. [no bioturbation, sparse bioturbation, low bioturbation, moderate bioturbation, high bioturbation, intense bioturbation, complete bioturbation] [porifera, cnidaria, brachiopoda, bryozoan, mollusca, arthropoda, echinodermata, vertebrata, trace fossil] One name One name List of values Modification [altered, deformed, dissolved, fractured, recrystallized] Oil stain [abundant, common, rare, trace] PNEC Conferences 2015, Houston 5

6 The model is completed with the properties that are related to this particular portion of rock, such as, bioturbation, alteration, presence of oil, fossils, and others. The quality domains described in Table 2 were completed following the study done by (Abel et al., 2012). One important characteristic of sedimentary rocks are expressed by their sedimentary structures. Sedimentary structures are the external visual aspects resulting of the internal spatial arrangement of the grains in the rock, which is a consequence of the genetic process of the creation of this rock (Collinson et al., 2006). Carbonera in (Carbonera, 2012) classified the concept sedimentary structure in the ontological metatype category, meaning that it groups several distinct kinds of entity by the common property of being individuated by their internal organization of grains. The sub-kinds of the category of sedimentary structure are top-surface features, base-surface features, deformation, chemical / diagenetic, biogenic and depositional. For simplicity and by its importance, we will detail only the properties of DEPOSITIONAL SEDIMENTARY STRUCTURE on Table 3. Table 3 Ontological model of the DEPOSITIONAL SEDIMENTARY STRUCTURE concept. Meta Type Kind and Inseparable part Name Depositional structure Quality Quality domain Number of values Shape of laminae [parallel, planar, sigmoid, tangential, channeled, truncated wave, wave] Angularity [horizontal, low angle, high angle] Thickness of layer [small size, medium size, large size} Thickness of laminae [thin, medium, thick] Mud presence [present, absent] Thrust angle [subcritical, critical, Supercritical] Type of structure [Climbing ripple (critical), Climbing ripple (subcritical), Climbing ripple (supercritical), Climbing ripple crosslamination, Cross stratification, Current ripple crosslamination, Flaser bedding, Herringbone (reverse cross bedding), Heterolithic bedding, Hummocky crossstratification, Intraclasts, Irregular bedding, Irregular lamination, Linsen bedding, Low angle crossstratification, Massive], Normal grading, Parting lineation, Planar cross bedding, Plane parallel lamination, Plane parallel stratification, Reverse grading, Scattered granules, Sigmoidal cross strata, Swaley crossstratification, Tangential cross stratification / Tangential cross bedding, Trough cross bedding, Trough crosslamination, Wave ripples lamination] The next concept to be modelled refers to the LITHOLOGICAL BOUNDARY between rock units. Regarding its ontological properties, objects like pores and boundaries are inseparable parts of rigid concepts (Guizzardi, 2005). This means that, even having their own essential properties and objective existence, the instances of these objects cannot exist without some other entity. It is the case of a hole in PNEC Conferences 2015, Houston 6

7 a piece of wood, which is only possible to exist if the wood is present. A BOUNDARY is defined by the relationship with the two lithological units that limit it, which is described in the Figure 1, and by its intrinsic property, detailed in the Table 4. Meta Type Kind and Inseparable part Table 4 Ontological model of the LITHOLOGICAL BOUNDARY concept. Name Lithological boundary Quality Quality domain Number of values Type of contact [erosive, sharp, gradational, faulted, undefined, lag deposit, cryptic] The concept LITHOLOGIC COMPONENT that appears in the PPDM model was not modelled here, because it seems more of an ontological category that groups several different ontological concepts (mineral, fossil, structure) than an instance of some particular meta-type. To describe the fossil content in the rock unit, we modelled FOSSIL PRESENCE as an attribute of LITHOLOGICAL UNIT, since the fossil specie and its position is relevant for stratigraphic interpretation. This solution is useful in macroscopic scale. Mineral components, if described, should deserve a proper concept that would allow individuating their intrinsic properties. In a similar way than rocks and lithological units were modelled, the modelling of the quantity MINERAL should be done in a separate way of its physical expression in space, the kinds GRAIN or CRYSTAL. Since it is quite difficult to recognize minerals in macroscopic scale and this is not considered in PPDM model, we will not include these concepts in this article. The lithology content of PPDM model includes also the description of DEPOSITIONAL ENVIRONMENT. A depositional environment is the physical and chemical conditions in which some sediment was transported, deposited and consolidated to become a rock. Being a perdurant, the instances of a depositional environment have no concrete expression in reality. They are recognized through the interpretation of its effects over other instances, such as the instances of grains or rocks that were transported or consolidated under the particular physical and chemical conditions. In ontological terms, a depositional environment is a perdurant concept that can be described as a situation that affects a set of endurant instances during some slice of time. There is no concrete instance of a situation in reality, out of the modifications caused to the set of affected instances. Considering the level of subjectivity in the interpretation of a depositional environment, the concept should not be modelled or instantiated in the same way that endurant concepts are. We propose that any geological/geophysical concept whose instances have no concrete expression in reality and are the result of a subjective interpretation process would be modelled by the description of affected instances. This is not a trivial way of modelling concepts and we suggest the reading of (Carbonera, 2012; Carbonera et al., 2013) to obtain a better understanding of ontological models built for supporting automatic depositional environment interpretation. Also, the model should preserve the relationship with the interpreter in order to allow traceability. In any case, perdurant concepts should not be part of the ontological model intended for information integration. The Figure 1 describes the hierarchical and partonomic relations among the concepts described in the tables 1 to 4. The partonomic relations are specialized in their types, exemplified by the CONSTITUTED-BY relation and the INSEPARABLE PART relation. PNEC Conferences 2015, Houston 7

8 Figure 1 - Our ontological model of lithology information in PPDM Lithological unit <constituted by > Rock Sedimentary rock <inseparable part > Siliciclastic rock Chemical rock Conglomerate Sandstone Lutite Carbonate Evaporite Silicieous Phosphatic Iron Rich Lithological Boundary <inseparable part > Erosive Sharp gradational faulted undefined lag deposit cryptic <category> Sedimentary strtucture Top surface features Base surface features deformation Diagenetic Biogenic Depositional Mud cracks... Birdseye... Cross stratification... <hierarchical relationship> <part of relationship> CONCLUSION This work describes the ontological analysis of the PPDM lithology data model associated to well data, based on the ontological meta-properties of the Ontoclean methodology (Guarino and Welty, 2004) that define the identity, unity and dependence relations of these entities. As a result, we propose the hierarchical and partonomical scheme of concepts, ontological definition of qualities and main definitional relationships of every geological object described in the data model. Our model includes the set of properties of the main entities that affect in reservoir quality and are present the data model, such as, rock type, texture (grain size), fabric, sedimentary structures and diagenesis. The description of cores provides essential information for the understanding of depositional environments and systems, as well as of their stratigraphic evolution, which constitute the basis for the construction of realistic reservoir models. Nevertheless, in the whole reservoir modeling workflow, the most subjective and non-automatic data acquisition stage involves the analysis of cores and the derived interpretation of sedimentary facies and facies associations. As a consequence, the rock information has delivered little contribution in increasing the quality of reservoir models. Accurate data models should be able to correctly express the meaning of represented data (or the conceptual model behind the data schema), in order to orient the system development and further information integration. That is the main aim of using ontological analysis when building petroleum data models. PNEC Conferences 2015, Houston 8

9 Our methodology is based on a set of modelling principles and rules that can orient the modeler when producing conceptual models and the geologists and engineers when using them. We emphasize the more important principles here: - Any data model has a conceptual schema (or meta-data) that describes the meaning of modelled data; - a common ontological model can be used as the conceptual schema for integrating and sharing data, if the legacy models refer to the same reality (or model the same entities of reality); - the ontological model should be expressed around a basic framework composed by concepts that preserve three properties: provide proper ontological identity; conserve their existence through the passing of time; and have space-temporal expression through their instances in reality; - the ontological model will be completed by the definitional relationships among these concepts. The resulting model provides a support for information integration along the exploration chain. We aim that the model can be used as an initial step for automatic integration of geological data repositories. ACKNOWLEDGEMENT The authors are grateful for the support of the Brazilian Research Council, CNPq; PRH PB-17 program (supported by PETROBRAS); and ENDEEPER Co. for the financial support to this work. They would like to also thank PPDM organization for the access to the PPDM data model. REFERENCES Abel, M., 2001, The study of expertise in Sedimentary Petrography and its significance for knowledge engineering (in Portuguese).Doctoral Thesis ]: Federal University of Rio Grande do Sul, 239 p. Abel, M., Lorenzatti, A., Ros, L. F. D., Silva, O. P. d., Bernardes, A., Goldberg, K., and Scherer, C., 2012, Lithologic Logs in the Tablet through Ontology-Based Facies Description, AAPG Annual Convention & Exhibition Long Beach, CA, AAPG. Abel, M., Perrin, M., and Carbonera, J., 2015, Ontological analysis for information integration in geomodeling: Earth Science Informatics, v. 8, no. 1, p Bergamaschi, S., Castano, S., De Capitani di Vimercati, S., Montanari, S., and Vincini, M., 1998, An Intelligent Approach to Information Integration., in Guarino, N., ed., Formal Ontology in Information Systems, FOIS'98: Trento, Italy, IO Press. Carbonera, J., 2012, Reasoning over visual knowledge (in Portuguese)Master Dissertation]: Federal University of Rio Grande do Sul. Carbonera, J. L., Abel, M., Scherer, C. M. S., and Bernardes, A., 2013, Visual Interpretation of Events in Petroleum Geology, IEEE International Conference on Tools with Artificial Intelligence (ICTAI): Washington DC, IEEE Collinson, J., Mountney, N., and Thompson, D., 2006, Sedimentary Structures, Dunedin Academic Press Ltd, 302 p.: Gärdenfors, P., 2000, Conceptual Spaces: the geometry of thought, Cambridge MIT Press. Guarino, N., 1998, Formal Ontology in Information Systems in Guarino, N., ed., Formal Ontology in Information Systems, FOIS'98: Trento, Italy, IO Press, p. 6-8 June Guarino, N., and Welty, C., 2000, A Formal Ontology of Properties, The ECAI-2000 Workshop on Applications of Ontologies and Problem-Solving Methods.: Berlin, Germany, IOS Press. Guarino, N., and Welty, C. A., 2004, An overview of OntoClean, in Staab, S., and Studer, R., eds., Handbook of Ontologies: Berlin, Springer p Guizzardi, G., 2005, Ontological Foundations for Structural Conceptual Models, Enschede, The Netherlands, Universal Press, CTIT PhD Thesis Series, 410 p.: PNEC Conferences 2015, Houston 9

10 Lorenzatti, A., 2010, Ontology for Imagistic Domains: combining textual and pictorial primitives (in Portuguese)Master Dissertation ]: Federal University of Rio Grande do Sul. Prothero, D. R., 1990, Interpreting the Stratigraphic Record., W.H. Freeman & Co. Zimmerman, D., 1996, Persistence and Presentism: Philosophical Papers, v. 25 no. 2, p PNEC Conferences 2015, Houston 10