Reducing the Effort for Analysing and Improving Engineering Systems
Andreas Kohn, Maik Maurer, Helena X. Schmidt and Udo Lindemann
Institute of Product Development, Technische Universität München, Boltzmannstraße 15, 85748, Garching, Germany
Keywords: System modelling, System analysis, System improvement, Reuse of knowledge, Structural complexity
management, Multiple domain matrix, Structure computation.
Abstract: This paper presents an approach for combining two actual trends in the engineering domain: ontology-based
knowledge management and structural complexity management. A focussed engineering system can be
analysed and possibilities for improvements can be deduced with low effort by applying structure based
algorithms on already existing ontologies. An overview of the current use of ontologies in the engineering
domain is given for showing the various options for this application of structural complexity management.
Necessary interfaces between ontology-based knowledge management and matrix-based structural
complexity management are deduced by comparing both approaches considering data representation and
analysis capabilities. The proposed approach is applied and discussed by the example of analysing an
ontology originally developed for handling technical solution knowledge in the field of automation industry.
Knowledge management has an increasing influence
on the success of companies. Corresponding to the
overall current trend in knowledge management,
ontologies play an increasing role in knowledge
management in the engineering domain (Kim et al.,
2008). Here, knowledge management systems using
ontologies can be found in the various different
knowledge-intensive applications for modelling,
storing and providing the required knowledge. The
goals of using ontologies in this field correspond to
the overall goals of using ontologies in knowledge
management described by (Tudorache, 2006).
Ontologies in the engineering domain are used for
sharing a common understanding of the respective
domain and enable knowledge sharing between
humans and software applications. As a
consequence, they enable sharing the terminology
defined in the ontology. Furthermore, they make
domain knowledge explicit and make reasoning
about it possible. By having a formal representation,
the knowledge can be used in formal reasoning and
querying algorithms. Finally, ontologies permit the
reuse of existing knowledge and improve the
consistency of the regarded information.
Another actual trend in the engineering domain
is the need for handling the complexity in nowadays
engineering tasks. In almost all relevant sections of
engineering, a steady increase of complexity can be
observed (Lindemann et al., 2009). This complexity
results from the high, possibly time-variant number
of elements and relations that have to be considered.
The Structural Complexity Management (StCM)
methodology offers generic methods for modelling
and analysing the underlying structure of these
complex systems (Lindemann et al., 2009). The
analyses are computed in a so-called Multiple
Domain Matrix (MDM) and are used to identify
potential improvements for the focused system.
The research presented in this paper aims at
combining these two actual trends in the engineering
domain. We claim that StCM methodology can
increase the system understanding for the companies
by applying the structural analysing algorithms on
the knowledge already stored in existing ontologies.
This goes beyond the current use of ontologies in
knowledge management (e.g. for knowledge storing
and querying). Shortcomings and problems of the
respective systems can be revealed by applying
analysis algorithms of StCM, and subsequently lead
to an improvement of the system. As the information
Kohn A., Maurer M., X. Schmidt H. and Lindemann U..
Engineering Systems.
DOI: 10.5220/0003635701950201
In Proceedings of the International Conference on Knowledge Engineering and Ontology Development (KEOD-2011), pages 195-201
ISBN: 978-989-8425-80-5
2011 SCITEPRESS (Science and Technology Publications, Lda.)
needed for the analysis is already stored in existing
ontologies, the additional effort is quite low in
comparison to the possible profit for the companies.
By using existing ontologies, the actual high amount
of time for information acquisition in StCM is
shortened, the required efforts for system modelling
are reduced and the efficient application of the
approach is guaranteed.
According to this main hypothesis, the paper is
structured as follows: first, we will introduce the
steps of StCM methodology and show how systems
can be analysed and improved by applying structural
computation algorithms. Then, we will give an
overview about the use of ontologies in the
engineering domain. This shows the various
possibilities for applying the structural analysis
algorithms on existing ontology-based knowledge
repositories. In the following chapter, we will
compare ontology data representation with the data
representation needed for StCM methodology. This
builds the basis for our solution approach for
transferring information stored in ontologies to the
MDM. The proposed approach will be described and
explained by the example of an ontology used for
storing information about existing technical
solutions in the automation industry. We will
conclude this paper with a discussion of the
presented approach and an outlook on the next steps.
2.1 Structural Complexity
This section provides a short introduction to StCM
theory in order to allow the reader to follow the
main topic of this contribution. Details of the
method can be taken from Lindemann et al. (2009).
Figure 1 shows a system description by elements
and relations between them. The elements belong to
three domains (indicated by circles, squares and
triangles). In terms of engineering, these domains
could stand for components, people and functions.
On the left side, the system is visualised by a graph
representation. On the right side, the systems is
modelled in a MDM. This comprehensive matrix
model consists of Design Structure Matrices
(DSMs), which contain relations between elements
belonging to only one single domain and Domain
Mapping Matrices (DMMs), which contain relations
between elements belonging to two different
domains. For example, a DSM describes the links
between system functions, whereas a DMM
indicates which people are responsible for which
components in design.
Figure 1: Structure of a MDM.
The MDM represents the core of the StCM
methodology that provides a five-step procedure to
support users in system definition, information
acquisition, deduction of indirect dependencies,
structure analysis, and the application on the product
design. The initial situation for the application of the
approach is a handling or design problem due to the
system’s complexity. In the first step of system
definition it is clarified which system aspects
(domains and relations) have to be considered in
order to solve the complex problem. Next,
information about specific system elements and their
direct dependencies have to be acquired. This step
represents the highest effort of time within the
approach. At this point, the proposed approach in
this paper can enable benefit in time and quality as
already structure information stored in ontologies
can be easily reused. Whereas only direct
dependencies are acquired in the previous step, now
indirect dependencies are deduced – in case they are
required for solving the initial complex problem. For
example, relations between components of a product
can be deduced on the basis of the performed
functions. Based on acquired and deduced structural
information, now relevant system structures are
analysed. The objective is to identify characteristic
constellations, which allow interpretations and
system optimization in the following. Finally,
findings have to be applied in order to solve the
initial problem. Thus the result of the final step of
the StCM approach is the improved system
management or design due to the application of
gained structural understanding.
2.2 Use of Ontologies in the
Engineering Domain
As possible input for StCM, this section will review
the use of ontologies in the engineering domain and
classify the scope of the ontologies according to
KEOD 2011 - International Conference on Knowledge Engineering and Ontology Development
their general objectives (product, process,
organization) and their area of application.
In a quite exhaustive comparative study, thirteen
approaches for dealing with different types of
engineering problems by applying ontology-based
knowledge management are presented (Kim et al.,
2008). Seven of them focus on product knowledge,
four on process knowledge, and two model both
knowledge about product and process (for details
about the different ontologies and projects, please
see Kim et al. 2008 or the original literature).
Beyond that, several more ontologies in the
engineering domain were identified in the present
research. Most of them focus on the early phases in
the lifecycle of a product, that of product
development. (Gaag et al., 2009) developed a
product-focused ontology for modelling knowledge
about existing technical solutions which supports the
automatic annotating of existing solution documents
and the retrieval of the stored information in the
field of automation industry. Furthermore, an
ontology is developed and used for integrating CAx-
Systems in the step of virtual and physical validation
of parts and prototypes in the automotive sector
(Syldatke et al., 2008). Tudorache proposed a
generic product ontology that is validated in two
scenarios of requirements management and
concurrent engineering (Tudorache, 2006). An
ontology for improving design communication
which contains process-related knowledge was
developed and applied for improving design
collaboration (Uflacker et al., 2009). Furthermore,
Darlington and Culley practically evaluated the use
of ontologies in “requirements development and
capture” as an important phase of engineering design
(Darlington and Culley, 2008). In addition, a
process-oriented ontology was developed to support
the quality-assurance process in the field of
electronics design (Yang, 2005). Although, Anderl
et al. do not present a precise ontology, they also
emphasise the increasing importance of ontologies
for product development (Anderl et al., 2009). They
propose an ontology-based-product development
system that implements the management of access
rights for different user groups and functionalities
for integrating, releasing and storing information.
Ontologies in the engineering domain are also
applied for enhancing manufacturing systems’
intelligence. For example, a cognitive machine shop
is proposed where machines “know” about their
manufacturing capabilities by representing the
relevant knowledge about material, work pieces, etc.
in an machine-interpretable ontology (Shea et al.,
This section provides the basis for the proposed
approach by showing similarities and differences
between knowledge management using ontologies
and StCM. We will discuss the data representation
capabilities of ontologies and MDMs to identify
possible losses during the transformation of
information from an ontology to a MDM. Finally we
will show the analysis and computation capabilities
of StCM and deduce the constraints concerning the
needed information input from ontologies.
3.1 Data Representation
The main elements of domain knowledge in
ontologies are concepts, relations, functions,
procedures, instances, axioms and production rules
(Corcho and Gómez-Pérez, 2000). For each of these
main elements, ontology languages provide various
different features. For example, for describing
concepts in an ontology, features like meta-classes,
definition of attributes and definition of properties of
attributes can be used in most of the languages.
In contrary to this feature-variety in ontology
languages, the data representation used in StCM is –
on the first glance – quite simple-constructed. In a
standard MDM the focus lies on domains, elements
of a domain, relations between the elements and
attributes of elements or relations. This can be
explained by the fact that MDM theory is based on a
matrix and graph representation of the system.
Therefore, features like meta-classes (e.g. for
building a taxonomy of classes) or functional
relations between elements are mostly not focused.
Approaches for modelling this enhanced knowledge
representation that burst the bounds of traditional
matrix-based constraints exist, but are still not
common. For example, logical dependencies (which
can be interpreted as production rules in ontology
modelling) were implemented and evaluated in the
so-called “why-matrix” (Maurer and Braun, 2008).
A hierarchical view of the system and the
introduction of a hierarchy for the domains was
proposed for the reduction of efforts for data
acquisition (Biedermann et al., 2010). Also, the “1.5
Matrix” allows a hierarchical view on the elements
of a system by interpreting attributes of elements as
meta-classes (Eppinger, 2009). A further approach
aims at overcoming the strictly domain-oriented
representation of a system in form of a MDM by
proposing flexible domain modelling (Kohn and
Lindemann, 2010).
Effort for Analysing and Improving Engineering Systems
Summing up this comparison of data
representation, it can be said that not all information
stored in an ontology can be transferred into a
MDM. Nevertheless, all information needed for
analysing complex systems in StCM methodology
(domains, elements and relations) can be found
structured in ontologies. Therefore, ontologies can
generally be used as information input for StCM.
3.2 Analysis Capabilities
In a further step, we will now describe the
computation and analysis capabilities of StCM and
discuss the resulting constraints for the needed
information input from ontologies.
In StCM methodology, there are two main types
of computations. First of all, indirect dependencies
are deduced in the third step of StCM. Here, the
direct dependencies modelled in the DSMs or
DMMs are used for the computation of further
DSMs or DMMs (which then contain indirect
dependencies between their elements). This is done
by applying computation algorithms from the field
of matrix algebra. Several types of indirect
dependencies exist. For example, a DSM can be
computed using only one DMM, or using another
DSM and two DMMs. The degree of indirect
dependency provides information about the strength
and therefore the relevance of the indirect
dependency. Both the direct dependencies and the
indirect dependencies are then used for the second
type of computations in StCM: the structural
analysis algorithms. Using the analysis algorithms,
structural significant subparts of the structure can be
identified. These algorithms are often matrix-based
(e.g. clustering algorithm for static matrices, tearing
algorithm for time-based matrices, etc.) or graph-
based (e.g. leaves of a structure, significant hubs,
etc.). In the next step, the significant subparts of the
structure serve for indicating possible improvement
potential for the system according to the initial goal
of the analysis. Established matrix- and graph-based
visualisation techniques are used for showing and
presenting the results and serve for enhancing the
system understanding.
For the import of information from an ontology,
one critical point has to be mentioned concerning the
analysis capabilities: all computations done in StCM
analysis are based on the elements in a MDM and
their relations. Without the elements, the algorithms
and therefore the analysis would be obsolete. In
contrast, ontologies do not need the instances for
reasoning or querying an ontology. They can
perform computation only on the taxonomic level of
the ontology (also referred to as T-Box – among
others (Struckenschmidt, 2009)). Thus, structural
analysis algorithms can only be applied on
ontologies that contain instances and have a well-
populated T-Box. But, as current trends in ontology
design goes from pure taxonomic ontologies towards
descriptive and instantiated ontologies (Kim et al.,
2008), this restriction becomes less significant.
The initial question when using StCM on the basis
of existing ontologies is whether the information
modelled in the ontology can serve for solving a
handling or design problem. Therefore, it has to be
clarified if the concepts and relations of the ontology
and the contained elements suffice for the intended
analysis of the system. In some cases, only part of
the information stored in the ontology can be used
and, consequently, further information sources have
to be taken into account. The relevant and useful
concepts and relations for the initial problem are
then taken for building the meta-model.
Subsequently, the elements and their relations in the
T-Box of the ontology can be transferred into the
matrices of the MDM in the step of information
acquisition. The next three steps of deducing indirect
dependencies, analysing the structure and
application on the system are equal to the original
StCM methodology.
By the example of the application in a use case in
the field of the automation industry, we will
illustrate the presented approach and discuss the
benefit of the further analysis capabilities described
above. The developed ontology is an OWL-DL
ontology and contains information about existing
technical solutions (Gaag et al., 2009). This
ontology is used for supporting the annotation of
existing solution documents and enables their
subsequent retrieval by providing different
abstraction principles for the similarity of technical
solutions (Kohn and Lindemann, 2011).
The following Figure 2 shows an excerpt of the
main concepts and properties of the ontology with
an exemplary instantiation. A company provides a
technical solution that is used in an industrial sector.
The technical solution uses function owners that
perform certain functions. For example, a robot
performs the function “transfer bottle” and the
cylinder performs the function “fill bottle”.
KEOD 2011 - International Conference on Knowledge Engineering and Ontology Development
Figure 2: Ontology for describing solution documents with
exemplary instantiation.
In our example, we use parts of this ontology to
identify possibilities to modularise the technical
system by reorganising the function owners into
modules. In the engineering domain, this can be
done by reorganising and grouping functions owners
according to the performed functions. Therefore, we
will transform the needed information for this design
problem into the meta-model in form of a MDM.
The concepts of the ontology are interpreted as
domains in the MDM and the properties are the
relations between the domains (see Figure 3).
Figure 3: Meta-model of the system in form of a MDM.
In our case, we only need information about
function owners and functions (in Figure 3 labelled
as number 1). In a next step, the instances and the
relations between instances are transferred from the
ontology to this matrix. Properties of instances
become relations of elements in the system. For
transferring the information in the ontology, the
property “perform” of the function owner “robot” to
the function “transfer bottle” becomes a cross in the
appropriate field in the DMM “function owner
performs function” (Figure 4 a). In a further step,
indirect dependencies can be deduced for showing
the dependencies between function owners. In this
example, indirect dependencies between function
owners were computed from the initial DMM and
stored in a new DSM with to the relation “two
function owners are related when they perform the
same function” (in Figure 3 labelled as number 2).
In a next step, this newly calculated DSM with
indirect dependencies can be used for analysis
algorithms. Here, the result of a clustering algorithm
is shown in a matrix-based visualisation (Figure 4 c).
In this case, two clusters are clearly visible. In terms
of engineering the clusters indicate possibilities for a
functional modularisation of these function owners,
as they perform the same functions. Beyond the
matrix-based analysis, graph-based analysis can be
used (Figure 4 d). The elements of the system are
visualised as the nodes of the graph and the relations
as edges. As structural significant system elements, a
bridge element connects the two clusters identified
above. If this element fails, both clusters would be
affected. For our modularisation approach, this
element can build the hub between the two modules.
Figure 4: Applying computation, analysis and
visualisation algorithms on the MDM.
The example provides a short glance on the
possible application of the presented approach. It
showed that information stored in an ontology that
was originally developed for a different purpose can
be successfully used for analysing the underlying
system. Possibilities for system improvements
concerning a certain design problem can be
identified. As the needed information is already
stored in the ontology, the effort for the otherwise
time-consuming information acquisition is very low.
function owner
transf er bottle
bottling line
industrial sector
fill bottle
comp any
comp any provides
function owner
function ownerfunction
function owner
function owner
function owner
Indirect dependencies
Initial DMM
Clustering algorithm Graph visualisation
a b
Effort for Analysing and Improving Engineering Systems
A study of related work revealed that combining
matrix methodologies and ontologies was presented
by different authors before. However, they
combined matrix methodologies and ontologies each
in a very specific use-case. They did not consider
and evaluate the general possibilities for interaction
of the two approaches as presented in this research.
For example, Yang transforms information stored in
an ontology to a Domain Structure Matrix in a
process framework for consumer electronic design.
He uses the computation capabilities in the DSM for
analysing activity dependencies in an ARIS
(Architecture for Information Systems) process for
removing cyclic dependencies in form of interaction
loops (Yang, 2005). Syldatke et al. propose the
combination of ontologies and DSM-modelling
techniques. They show the possibilities of enhancing
DSM-modelling techniques by using semantic
technologies (e.g. ontology languages and reasoning
capabilities), but did not go into detail considering
the conditions for the proposed combination
(Syldatke et al., 2008). Finally, existing research
reveals a study of using ontologies for the
integration of DSM analysis techniques into the
planning of construction projects. Here, once again
the analysis capabilities of matrix-based algorithms
are emphasised and evaluated in the specific use
case (Masera, 2007).
Summing up, two different ways of combining
the two approaches can be found. First, – as shown
in this paper – structural analysis algorithms on the
basis of DSM, DMM or MDM applications can be
used for enhancing current ontology-based
knowledge modelling approaches. The other
possibility aims at the opposite direction. Here,
achievements of established semantic technologies
are proposed to be used for enhancing representation
and modelling capabilities of matrix-based
approaches. Both approaches can be useful
depending on the intended purpose and reveal
interesting options for the combination and
integration of both approaches.
Regarding the comparison between ontology-
based knowledge management and StCM done in
this research, the general different scope of these
two fields has to be discussed. As shown above, the
StCM methodology is directly focused on the
structural analysis of complex systems. In contrast,
ontologies and ontology development in general
have no real focus on a certain application per se. It
depends strongly on the individual specification and
the desired objectives. Therefore, the capabilities of
ontologies are more generic and have to be
individually specified corresponding to the
application in the respective use-case. This
generality can be both benefit and disadvantage. For
experts in knowledge engineering this is certainly an
advantage. However, for people in the engineering
domain, who are often not experienced in building
and using ontologies (Kim et al., 2008), predefined
analysis criteria as proposed by the StCM
methodology could be helpful. Also, the
visualisation part and the comprehensibility have to
be taken into account. Matrix-based and graph-based
visualisations have proved to increase system
understanding and help people handling complex
As shown in the previous chapter, an existing
ontology can only be used for design or handling
problems that require information which is – either
direct or indirect – available in the ontology. The
quality of the ontology influences the results of the
analysis. If the concepts or relations do not provide
the needed information or not in satisfying quality,
the ontology cannot be used for the specific
problem. Then, the definition of the meta-model and
the information acquisition has to be done according
to the standard StCM methodology.
The research presented in this paper gives an insight
into possibilities for the combination of knowledge
management with ontologies and StCM in the
engineering domain. By applying analysis
algorithms on already existing ontologies, the
system understanding can be enlarged and
possibilities for system improvement according to
the analysis results can be deduced. The literature
review on existing approaches using ontologies for
handling engineering problems reveals the
increasing use of ontologies in the engineering
domain. This provides the basis needed for the
presented approach and confirms its further
application possibilities.
The transformation of information stored in
ontologies to MDM needed for the structural
analysis is shown by comparing ontologies and
StCM according to their data structure and analysing
capabilities. An example from the automation
industry illustrated the presented approach. Here, an
ontology originally designed for capturing
KEOD 2011 - International Conference on Knowledge Engineering and Ontology Development
knowledge about already existing technical solution
was exemplarily analysed by selected structural
analysis algorithms for the identification of
modularisation potential in the system. This
application showed that structural analysis
algorithms can enhance the analysis of systems
already modelled in an ontology with very low
information acquisition effort.
Further work will focus on the above mentioned
bidirectional combination and the integration of
ontology-based knowledge management and StCM
in combination with further empirical studies. Using
structural analysis algorithms and therewith enabling
the corresponding system improvement possibilities
directly with an ontology-based system seems to be
very promising. Therefore, the already existing
analysis algorithms StCM have to be translated into
ontology-interpretable defined classes for enabling
reasoning or the appropriate ontology queries.
Part of this work has been funded by the German
Federal Ministry of Economy and Technology
(BMWi) through THESEUS.
Anderl, R., Mecke, K., Sprenger, A., and Weitzmann, O.,
2009. Ontology Support for Product Development -
Successful Application of Ontologies in Product
Development. In KEOD 2009, Madeira, Portugal.
Biedermann, W., Strelkow, B., Karl, F., Lindemann, U.,
and Zaeh, M. F., 2010. Reducing Data Acquisition
Effort by Hierarchical System Modelling. In 12th
International DSM Conference, Cambridge, UK, 309-
Corcho, O. and Gómez-Pérez, A., 2000. Evaluating
knowledge representation and reasoning capabilities of
ontology specification languages. In ECAI 2000
Workshop on Applications of Ontologies and
Problem-Solving Methods, Berlin.
Darlington, M. J., and Culley, S. J., 2008. Investigating
ontology development for engineering design support.
Advanced Engineering Informatics, 22.
Eppinger, S. D., 2009. Integrating the Product, Process,
and Organization Views of Complex System
Development. In 11th International DSM Conference,
Greenville, SC.
Gaag, A., Kohn, A., and Lindemann, U., 2009. Function-
based Solution Retrieval and Semantic Search in
Mechanical Engineering. In 17th International
Conference on Engineering Design, Stanford,
California, USA.
Kim, S., Bracewell, R., and Wallace, K., 2008. Some
Reflections on Ontologies in Engineering Domain.
TMCE 2008, Izmir, Türkei.
Kohn, A., and Lindemann, U., 2010. Approach towards a
more flexible handling of domains in complex
systems. In 12th International DSM Conference,
Cambridge, UK, 249-261.
Kohn, A., and Lindemann, U., 2011. Search for similar
technical solutions by object abstraction using an
ontology. In ICED2011, Kopenhagen, Denmark.
Lindemann, U., Maurer, M., and Braun, T., 2009.
Structural Complexity Management, Springer, Berlin.
Masera, M., 2007. An ontology applied to the
management of the construction process representing
an IDEFo metamodel integrated in a planning/design
structure matrix. In CIB W102 3rd International
Conference 2007, Stuttgart, Germany, 395-403.
Maurer, M., and Braun, T., 2008. The Why-Matrix. In
10th International DSM Conference, Stockholm,
Shea, K., Ertelt, C., Gmeiner, T., and Ameri, F., 2010.
Design-to-fabrication automation for the cognitive
machine shop. Advanced Engineering Informatics,
24(3), 251-268.
Struckenschmidt, H., 2009. Ontologien - Konzepte,
Technologien und Anwendungen Springer, Berlin.
Syldatke, T., Lutz, M., Chen, W., and Hess, C., 2008.
Managing Dependencies in the Product Development
Process with Semantic Technologies. 10th
International DSM Conference, Stockholm, 105-114.
Tudorache, T., 2006. Employing Ontologies for an
Improved Development Process in Collaborative
Engineering, Berlin, TU Berlin.
Uflacker, M., Skogstad, P., Zeier, A., and Leifer, L., 2009.
Analysis of virtual design collaboration with team
communication networks. In International Conference
on engineering design, ICED'09, Stanford, California,
Yang, Q. Z., 2005. Methods and tools for analysing and
controlling design quality assurance processes.
SIMTech technical reports, 6(3).
Effort for Analysing and Improving Engineering Systems