Professional Documents
Culture Documents
net/publication/221269507
CITATIONS READS
9 502
4 authors, including:
All content following this page was uploaded by Darijus Strasunskas on 28 May 2014.
1 Introduction
There is a need for efficient sharing and usage of information. Large traditional indus-
tries, such as oil industry are dependent on internationally agreed standards to main-
tain a semantically equal understanding of shared domains in and across organiza-
tions. However, the international standards currently used are often developed prior to
or independent from the novel technologies. One possible solution is to develop a
language that is powerful enough to express and differentiate the technological ele-
ments as well as the underlying semantics. A further requirement here is that the lan-
guage must be computer and human readable. An initiative to translate the interna-
tional standard ISO 15926 to the semantic Web technology OWL has been taken by
the Norwegian Oil Industry 1 . Here we study and demonstrate an alternative solution,
namely to facilitate controlled semantic transformations for models enabling co-
existence of both technologies. The semantic relationship between the ISO 15926 Part
2 and OWL DL is studied in particular. The objective is to study the semantic rela-
tionship between ISO 15926 and OWL in order to facilitate in- and cross-
organizational co-operation as applied in the semantic Web and Web Service envi-
ronments.
The diversity and multitude of resources and applications in organizations and the
Internet places elaborated requirements on methods and tools for efficient generation,
manipulation and compositional usage of information and services. OWL and ISO
15926 that are in the foci of this study are representing the novel models for reasoning
about and for specifying information systems, respectively.
The purpose of the international standard for Industrial automation and integration –
Integration of life-cycle data for process plants including oil and gas production facili-
ties (ISO 15926) is to facilitate integration of data to support the life-cycle activities
and processes of process plants. It has defined syntax and graphical representation yet
not formal semantics. It is built on EXPRESS (ISO 10303-11) to specify its data
model yet facilitates more accurate distributed specification.
ISO 15926 consists of seven parts as follows. 1 - Overview and fundamental prin-
ciples; 2 - Data model; 3 - Methodology for the development and maintenance of
reference data libraries; 4 - Reference data; 5 - Procedures for registration and main-
tenance of reference data; 6 - Scope and methodology for developing additional refer-
ence data; 7 - Implementation methods for data exchange and integration.
ISO 15926 Part 2 [9] specifies a data model that semi-formally defines the mean-
ing of the life-cycle information in a single context supporting the views that process
engineers, equipment engineers, operators, maintenance engineers and other special-
ists may have on the plant [8]. Table 1 summarizes the main components.
Semantic Web technologies allow the definition of logical relationships against which
reasoning can occur, something EXPRESS does not fully support. EXPRESS does not
formally distinguish between attributes and relationships [12], a feature that is often
required. Hence, a transformation from ISO 15926 to an ontology language is antici-
pated to facilitate reasoning.
OWL was chosen as an ontology language for transformation being a recommen-
dation by W3C showing its relative maturity level and suitability for Web applica-
tions. Furthermore, OWL version of ISO 19526 is recommended by [12] since OWL
offers the capability to better connect with models of individuals and ontologies. It
also offers potential for enhanced distribution of the models through namespace refer-
encing. However, transformation form ISO 15926 to OWL requires a semantic trans-
formation protocol that is controlled and based on semantics.
The long-term objective for the work is to establish a semantic relationship between
ISO 15926 and OWL in order to facilitate in- and cross-organizational co-operation in
traditional domains as applied in the novel semantic Web and Web Service environ-
ments. The semantic gap between ISO 15926 and OWL is seemingly narrow when
considering the object model part of the languages.
The models in the traditional modelling standards such as ISO 15926 have been
used for human interpretation with graphic representation and semi-structured docu-
4 Hakkarainen et al.
mentation. The emerging semantic Web recommendations such as OWL are primarily
intended for machine interpretation, enabled utilizing the high expressiveness of the
syntax when modelling. Thus, the target models should contain more semantics than
can be captured in a pure syntactic transformation process and further, a simple 1:1
mapping between language constructs introduce non-intended semantics to the target
model. This problem has not been recognized in the current body of related work.
The Integrated Information Platform (IIP) project [10] extend and formalize ISO
15926 using OWL. OWL Full is chosen for maximum expressiveness in the resulting
representation [3]. The POSC Caesar 2 Intelligent Data Sheet (IDS) project defines
product models for data sheets based on ISO 15926 Part 4. A resulting Reference
Data Library (RDL) 3 is planned to consist of approx. 100,000 EXPRESS standard
classes, be defined in ISO 15926 Part 4, and mapped to Part 7 OWL format.
Further, there are efforts to formalize and ontologize other international standards.
[4] for instance, have taken an approach to ontologize Electronic Data Interchange
(EDI), i.e. to model meaning of the messages and create semantically enabled mes-
sages yet maintaining backward compatibility to traditional systems. [6] specifies an
OWL ontology to cover the Geographic Information – Metadata (ISO 19115) stan-
dard, simply expressing each ISO 19115 as a single OWL construct. These concurrent
approaches seek to define a new language in OWL and assume a 1:1 relationship
between the languages not considering the problem of general modelling purpose and
style.
In order to capture as much of the semantics in the source models as possible, we
use production rules that operate on syntactic markers in the model and produce target
syntax that not always correspond to the source syntax, but try better to transmit its
semantics. In a real world application, we picture the process to be semi-automated,
where the user has the possibility to add semantics that was not in the original model
but is considered its intention. Syntactic markers for such intention may be found e.g.
in natural language (NL) comments and other documentation. NL requires a user to
interpret it, but in the future, an automated NL process can be tailored for interpreta-
tion in individual domains and application areas. In this way, we expect more of the
intended semantics to be preserved, semi-automatically with low labour intensiveness.
A first step in the direction of achieving that long-term objective, namely a study
of possible approaches for transformation of ISO 19526 into OWL with high preser-
vation of human interpreted semantics is presented in the sequel.
3 Semantic Transformations
A method for transforming the basic components of ISO 15926 to their proper repre-
sentation as language constructs in OWL is required. [5] identifies three distinct ways
to perform the transformation, all of which have different interpretations and conse-
quences for the preservation of semantics in the resulting data models. Two of them
are described below, since the third requires access to domain models. Here, only the
2
http://www.posccaesar.com
3
http://www.infowebml.ws
A Semantic Transformation Approach for ISO 15926 5
Transformation method one (TM1) considers the semantic interpretation of the source
language construct when defining a transformation rule to a target construct. Basi-
cally, one ISO 15926 component corresponds to exactly one OWL primitive. The
assumption is that there exists a one-to-one relationship between the models. The
inverse transformation (ITP) using the TM1 method is based on the OWL constructs
that the transformation protocol TM1-TP resulted in, i.e. only the language constructs
occurring as target in TM1-TP are assigned transformation rules in TM1-ITP. The set of
possible target constructs in TM1-ITP is not restricted by the results of using TM1-TP.
The transformation protocols above are defined using two transformation methods
based on semantic and extended semantic interpretation. Accordingly, there is a need
to report the underlying assumptions made for the actual interpretations in this study.
A trial transformation was performed in order to assure the quality of the proposed
transformation protocols. Some basic assumptions underlying the transformation rules
of the above TM1-ITP, TM1-ITP, TM2-TP and TM2-ITP are described in the sequel.
Thing. Every individual in OWL is a member of the class owl:Thing. Thus each user-
defined class is implicitly a subclass of owl:Thing [11]. Similarly, ISO 15926 consists
of 201 hierarchically ordered Entity data types, with thing on top. Thing has six
Attributes. The transformation protocols TM1-TP and TM2-TP map these Attributes to
owl:ObjectProperty or owl:DatatypeProperty. Note that OWL does not allow relating
properties to owl:Thing. Thus, Entity data type thing is mapped to owl:Class thing
which is a subclass of owl:Thing. Further, an Attribute is mapped to
owl:ObjectProperty or owl:DatatypeProperty with the associated Attribute-belongs-to
thing mapped to rdfs:domain thing. These mappings preserve the one-to-one mapping
of the TM1 method.
Identical name for Entity data type and Attribute. Some of the ISO 15926 Attributes
are given lexically identical names to some ISO 15926 Entity data types. OWL does not
allow identical names for owl:Class, owl:ObjectProperty and owl:DatatypeProperty.
In the transformation algorithms in [5] this conflict is simply resolved by starting an
owl:Class with a capital letter, whereas starting owl:ObjectProperty and
owl:DatatypeProperty instantiations with a lower case letter.
Multi-use of properties. Some ISO 15926 Attributes are used in conjunction with
several ISO 15926 Entity data types as Attribute-relates-to with Attribute. A multi-use of
properties arises. OWL does not allow creating owl:ObjectProperty with identical
names. While resolving this conflict the reasoning capabilities of OWL should be as
defined for OWL DL, whereas the semantics in ISO 15926 should be preserved. Thus
the transformation conditions in TM1-TP and TM2-TP above assert that such an Attribute
or Attribute-relates-to is not directly mapped if a respective Entity data type or Attribute
with identical name is already transformed. Note that lexically identical Attributes are
used in conjunction with lexically identical Attribute-relates-to. In [5] the specific
conditions are exemplified for each of the multi-used properties.
An alternative resolution is to rename the lexically identical Attributes in order to
translate each Attribute independently. This strategy is used once where TM1-TP re-
names the Attribute content that is an Entity data type subtype. The exception was neces-
sary since Attribute-relates-to used in conjunction with Attribute content relates to differ-
ent Simple Types for each Entity data type.
Redeclared Attribute-relates-to. An attribute of an Entity data type may be redeclared
for a Subtype of that Entity data type. Here the Attribute-relates-to is a different Entity data
type than its Supertype Entity data type. The term Redeclared Attribute-relates-to is used to
denotate these components.
TM1-TP maps Redeclared Attribute-relates-to to owl:allValuesFrom, while TM2-TP
maps it to either owl:allValuesFrom, rdfs:domain or rdfs:range. The inverse trans-
10 Hakkarainen et al.
INTEGER and BINARY. TM1-TP maps both ISO 15926 INTEGER and BINARY to
OWL xsd:integer. In the inverse TM1-ITP xsd:integer is mapped to INTEGER. As a
consequence, the declarative semantics of BINARY is lost and the number of instances
of INTEGER is increased after an inverse transformation.
An alternative strategy would be to map the ISO 15926 Simple type BOOLEAN to
xsd:boolean, xsd:hexBinary or xsd:integer. However, xsd:boolean restricts the binary
values to 1 or 0 whereas binary may be any combinations of 1 and 0’s, and further,
xsd:hexBinary represents the binary digits as hexadecimal digits. The loss of seman-
tics occurs even here; TM1-TP maps ISO 15926 BINARY to xsd:integer.
BOOLEAN and LOGICAL. Both ISO 15926 BOOLEAN and LOGICAL are translated to
OWL xsd:boolean. The inverse maps xsd:boolean to BOOLEAN. Thus, the declarative
semantics of LOGICAL is lost and the number of instances of BOOLEAN is increased.
TM1 is summarized in Fig.1. 16 ISO 15926 components are mapped to 14 OWL lan-
guage constructs. The overall mapping is straightforward. The detected deviations are
described below.
IMPLICITLY translated components. Both Attribute and Attribute-relates-to may be
implicitly defined (see Fig.1). Some Entity data types have Attributes with identical
names. If the Attribute is already translated we claim that the other Attributes with
identical names are implicitly translated through the first Attribute. Further, Attribute-
relates-to and Attribute-belongs-to are used in conjunction with the Attribute. If the
Attribute is implicitly defined then Attribute-relates-to is also implicitly defined..
NOT MAPPED components. ISO 15926 component Supertype is not mapped in the
transformation. ISO 15926 Subtype is translated to rdfs:subClassOf thus automated
reasoning is still possible and the semantics is preserved even though Supertype is not
mapped.
TM2 is summarized in Fig.2. According to TM2-TP, ISO 15926 Entity data type is trans-
lated to either an owl:Class, an owl:ObjectProperty or an owl:DatatypeProperty.
IMPLICITLY translated components. Both Attribute and Attribute-relates-to may be
implicitly defined (see Fig. 2). If it is Attribute content defined inside Entity data type
EXPRESS_string, EXPRESS_integer, EXPRESS_real, EXPRESS_logical, EXPRESS_boolean
or EXPRESS_binary it is implicitly defined since these Entity data types are translated to
owl:DatatypeProperties that relates to the same Simple Type as Attribute content. Even
though the semantics of Attribute content are lost, it is of no relevance since it is only
used as an intermediary between the Entity data type and the associated Simple Type,
while TM2-TP use the Entity data type itself as the intermediary. This rule does only
apply when the Attribute belongs to an Entity data type relationship or a Subtype of
relationship. Thus, it is a consequence of translating the Entity data type to an
owl:ObjectProperty.
When an Entity data type is translated to owl:Class, and its Attributes are translated to
owl:ObjectProperties, the owl:Properties are defined independent of the owl:Class.
Thus rdfs:domain is used to specify the class that the ObjectProperty belongs to.
However, when an Entity data type is translated to owl:ObjectProperty, for example
Entity data type classification, its Attributes are not translated, but the value they relate to
is specified by rdfs:domain and rdfs:range. Since rdfs:domain and rdfs:range are
used inside owl:ObjectProperty it is not necessary to explicitly define the property
they belong to.
NOT MAPPED components. ISO 15926 Subtype is translated to rdfs:subClassOf thus
automated reasoning is still possible and the semantics is preserved even though
Supertype is not directly mapped. When an Entity data type is translated to
12 Hakkarainen et al.
The major difference between TM1-TP and TM2-TP is that TM1-TP only considers the
semantics interpretation of the ISO 15926 components, while TM2-TP also considers
the semantics of each instance of the components. Consequently a difference is made
in the transformation of the ISO 15926 components.
Using TM1-TP, an Entity data type is translated to an owl:Class. Translation of the
201 Entity data types resulted in 201 owl:Class constructs. The resulting OWL code is
presented in [5]. In TM2-TP an Entity data type is translated to either owl:Class,
owl:ObjectProperty or owl:DatatypeProperty. Translation of the 201 Entity data types
resulted in 144 owl:Class, 51 owl:ObjectProperty and 6 owl:DataProperty [5].
Relationship. Entity data type relationship and Subtypes of relationship are translated to
owl:Class according to TM1-TP and owl:ObjectProperty according to TM2-TP. When
the inverse translations are performed the names of the Attribute is preserved by using
TM1-ITP. By using TM2-ITP the Attributes are created, but a new name has to be de-
fined as the original names are not used.
A Semantic Transformation Approach for ISO 15926 13
The Simple Types. Both TM1-TP and TM2-TP define six Simple Types. While TM1-TP
translates these components into four OWL language constructs, they are translated
into eight language constructs using TM2-TP. The difference is the translation of the
Simple Types used in conjunction with Attribute-relates-to of the Attributes defined within
Entity data type representation_of_Gregorian_date_and_UTC_time. Compared to TM1-
TP, TM2-TP in a larger extent takes advantage of the language constructs in OWL.
However, TM1-TP is preferred if it is necessary to keep hour, minute and second as
separate values.
5 Concluding Remarks
Two views for transforming ISO 15926 into OWL are analysed. The first alternative
considers the semantic interpretation of the ISO 15926 components straightforward.
The second alternative considers extended semantic interpretation, i.e., it also consid-
ers the semantics of each instance of the components. A third alternative that is based
on contextual semantic interpretation on a domain model was identified in addition
yet not further elaborated. The alternative views were tested through definition of
alternative transformation protocols with corresponding inverse transformation proto-
cols and analysed according to their ability to preserve semantics.
The analysis shows that Transformation Method one (TM1) results in a seemingly
direct representation of ISO 15926 in OWL, and enables full specifications, whereas
14 Hakkarainen et al.
Transformation Method two (TM2) takes more advantage of the language constructs
in OWL. The requirements one places at the resulting OWL representation are deci-
sive for selection of a transformation method and for assessing the appropriateness of
the proposed transformation protocols. TM1 is most appropriate if the goal is to
achieve a representation of ISO 15926 in OWL where the quantity of Entity data
types is equal to the number of the corresponding OWL language construct, i.e. where
a one-to-one mapping is desired. TM2 is most appropriate if the transformation is
performed in order to take advantage of the reasoning facilitates provided by OWL,
thus adding functionality not natively present in ISO 15926.
As not having access to complete domain models in ISO 15926, the transformation
method three (TM3) fell out of scope for this study. However, the method should be
tried out with both semantic and extended semantic transformation method in the
future. TM3 is based on the contextual semantic interpretation on existing domain
models. A transformation to target language is based on mapping the components of
the domain model in source language to semantically equal language constructs in
target language. In this case not solely analytic, but empirical observations could be
made in conjunction with appropriate statistical analysis.
References
1. Antoniou, G., van Harmelen, F. Web Ontology Language: OWL. Staab, S. and Studer, R.
(Eds.), Handbook on Ontologies in Information Systems, Springer Verlag (2003)
2. Bechofer, S., van Harmelen, F., Hendler, J., Horrocks, I., McGuinness, D.L., Patel-
Schneider, P.F., Stein, L.A. OWL Web Ontology Language Reference. W3C Recommenda-
tion
3. Christiansen, T., Jensen, M., Valen-Sendstad, M. Defining ISO 15926-4 Reference Data
Library classes in OWL. The Norwegian Oil Industry Association
4. Foxvog, D., Bussler, C. Ontologizing EDI: First Steps and Initial Experience. In proceedings
of International Workshop on Data Engineering Issues in E-Commerce, 2005, pp. 49-58
5. Hella, L., Tuxen, S.M. Semantic Web Technology and ISO 15926 – A Semantic Transforma-
tion Approach. MSc thesis, IDI, NTNU, Trondheim, Norway, 2004
6. Islam, A.S. et al. Ontology for Geographic Information - Metadata (ISO 19115:2003). URL:
7. McGuinness, D.L., van Harmelen, F. OWL Web Ontology Language Overview. W3C Rec-
ommendation
8. Price, D., Sandsmark, N. Industrial automation systems and integration - Integration of life-
cycle data for process plants including oil and gas production facilities - Part 1: Overview
and fundamental principles. ISO/Final Draft International Standard FDIS 15926-1:2003(E)
9. Price, D., Sullivan, J. Industrial automation systems and integration - Integration of life-cycle
data for process plants including oil and gas production facilities - Part 2: Data model.
ISO/Final Draft International Standard FDIS 15926-2:2003(E)
10. Sandsmark, N., Mehta, S. Integrated Information Platform for Reservoir and Subsea Pro-
duction Systems. In Proceedings of the 13th Product Data Technology Europe Symposium
(PDT 2004), October, Stockholm 2004
11. Smith, M.K., Welty, C., McGuinness, D.L. OWL Web Ontology Language Guide. W3C
Recommendation
12. Wix, J., Liebich, T. Final Draft CWA3 Proposal European eConstruction Meta-Schema
(EeM), Task T2 of the eEurope Pilot Project "SPICE" towards the CEN/ISSS Workshop on
eConstruction