TOWARDS SOCIAL-SOFTWARE FOR THE EFFICIENT REUSE OF

SOLUTION PATTERNS FOR SELF-OPTIMIZING SYSTEMS

Roman Dumitrescu and Benjamin Kl

opper

Heinz Nixdorf Instiute, University of Paderborn, F

urstenallee 11, Paderborn, Germany

Keywords:

Self-optimization, Mechatronic systems, Solution patterns, Domain-spanning knowledge base, Social soft-

ware.

Abstract:

The conceivable development of information technology will enable mechatronic systems with partial intel-

ligence. We refer to such systems as self-optimizing systems. Self-optimizing systems react autonomously

and ﬂexibly on changing environmental conditions. They have to learn and optimize their behavior during

operation. Hence the design of such self-optimizing systems is an even more interdisciplinary task than the

design of conventional mechatronic systems. Additionally to mechanical, electrical, control and software en-

gineers also experts from mathematical optimization and artiﬁcial intelligence are involved. As a consequence

a domain-spanning methodology is necessary in order to guarantee an effective work ﬂow between the partici-

pating developers from various domains and their domain-speciﬁc methods and solutions. In this contribution

a speciﬁcation for the domain-spanning modeling of solution patterns for self-optimizing systems as well as a

tool-based approach for the domain-spanning use of patterns based on social-software features are presented.

1 INTRODUCTION

The products of mechanical engineering and related

industrial sectors, such as the automobile industry,

are often based on the synergetic interaction of me-

chanics, electronics and software engineering, which

is aptly expressed by the term mechatronics. The

aim of mechatronics is to improve the behavior of

technical systems by using sensors to obtain infor-

mation about the system environment and the sys-

tem itself. The processing of this information enables

the system to react optimally to its current situation.

The conceivable development of communication and

information technology opens up increasingly fasci-

nating perspectives, which move far beyond current

standards of mechatronics: mechatronic systems ex-

hibiting an inherent partial intelligence. We use the

term self-optimization to describe such systems. Self-

optimization enables mechanical engineering systems

that have the ability to react autonomously and ﬂexi-

bly on changing operation conditions.

The design of such systems is a challenge. The

functionality of self-optimizing systems leads to an

increased complexity of their development and re-

quires an effective cooperation and communication

between the developers from different domains dur-

ing the entire development process. In that case,

established design methodologies from conventional

mechanical engineering (Pahl et al., 2007) and also

methodologies from mechatronics (VDI, 2004) do not

fulﬁll those new requirements. Within the Collabo-

rative Research Centre (CRC) 614 ”Self-Optimizing

Concepts and Structures in Mechanical Engineer-

ing” at the University of Paderborn, a new design

methodology is developed. It facilitates engineers

from different domains to model technical systems

in a comprehensive domain-spanning way and en-

ables the reuse of proven solutions in the form of

solution patterns. We distinguish solution patterns,

which relay on physical effects, and patterns, which

serve information processing. However, there are

no solution patterns, which consider the paradigm

of self-optimization. Thus, Active Patterns for Self-

Optimization (AP

) were developed to describe po-

tential solutions for self-optimizing systems (Gause-

meier et al., 2005), (Gausemeier et al., 2007).

In section 2 we describe the paradigm of self-

optimization. Followed by a brief description of

the respective development process we present solu-

tion patterns for self-optimizing systems in section

3. Next, in section 4, we point out the problems oc-

curring, if different domain-speciﬁc languages are in-

volved in the development process of self-optimizing

systems. Section 5 provides a solution approach to

342

Dumitrescu R. and Klöpper B. (2009).

TOWARDS SOCIAL-SOFTWARE FOR THE EFFICIENT REUSE OF SOLUTION PATTERNS FOR SELF-OPTIMIZING SYSTEMS.

In Proceedings of the International Conference on Knowledge Engineering and Ontology Development, pages 342-347

DOI: 10.5220/0002332203420347

 SciTePress

those problems. In section 6 we conclude our work

and describe our future work.

2 SELF-OPTIMIZING SYSTEMS

Figure 1 shows the key aspects and the modes of oper-

ation of a self-optimizing system. The self-optimizing

system determines its currently active objectives on

the basis of the encountered inﬂuences on the techni-

cal system from its environment. For instance, during

operation new objectives are added or existing objec-

tives are discarded. Thus, the system of objectives is

dynamic. Adapting the objectives in this way leads

to an autonomous adjustment of the system behav-

ior in reﬂecting the systems and the environmental

state. This can be achieved by adapting parameters

or the structure of the system. An example for pa-

rameter change is the adaptation of a control param-

eter. Structure adaptations affect the arrangement of

the system elements and thus their relationships, for

instance switching between different controller types.

Self-optimization within the system takes place as a

series of the three following actions to which we re-

fer as self-optimization process (Gausemeier et al.,

2008b):

1. Analyzing the current situation: The current sit-

uation includes the current state of the system as

well as all observations of the environment that

have been carried out.

2. Determining the system’s behavior: The current

system’s objectives can be extracted from selec-

tion, adaptation and generation.

3. Adapting the system’s behavior: The changed sys-

tem of objectives demands an adaptation of the

behavior of the system. This can be realized by

adapting the parameters and/or by adapting the

structure of the system.

The process of self-optimization leads, reﬂecting

changing inﬂuences, to a new state. Thus, a state tran-

sition takes place. The self-optimization process de-

scribes the system’s adaptive behavior.

3 DEVELOPMENT PROCESS

AND SOLUTION PATTERN

In general we can divide the development of self-

optimizing systems into two sequenced processes: the

domain-spanning conceptual design and the domain-

speciﬁc concretization. Starting point of the concep-

tual design is the development task. An interdisci-

plinary team, consisting of specialists from different

Figure 1: Aspects of Self-Optimizing Systems.

domains, creates the principle solution of the system

to develop. The principal solution does not only de-

scribe just the main physical characteristics of the

system, but also the logical operating characteristics.

This needs to be done in a domain-spanning way in

order to support all participating developers, which

are from different disciplines.

Thus, developers can work nearly independently

during the second phase, the domain-speciﬁc con-

cretization, and specify the system on basis of the

principle solution. Accordingly, they can use the ex-

isting domain speciﬁc methods and tools. An overall

system model, which is based on the principle solu-

tion, guarantees the effective and correct integration

to ﬁnalize the system design.

Within the conceptual design phase, we use a set

of semi-formal speciﬁcation techniques to describe

the principle solution of a self-optimizing system

(Gausemeier et al., 2006). An important intermediate

result during the conceptual design phase is the func-

tion hierarchy, which is primarily derived from the

functional requirements and consists of a logical hier-

archy of all system functions starting from the overall

function at the top. The next step is to ﬁnd solutions

for those functions, which are at the bottom of the

hierarchy in order to describe a system, which real-

izes the overall function. Depending on the type of

functions, different solution patterns are deﬁned. In

general a pattern describes a recurring problem in our

surrounding and the core of a solution to that problem

(Alexander et al., 1977). The solution core is spec-

iﬁed in a solution pattern that deﬁnes the character-

istics of the system’s elements which have to be de-

veloped and the interactions between those elements.

There are solution patterns, which are based on phys-

ical effects, and patterns, which involve data process-

ing. Apart from conventional functions of mechani-

TOWARDS SOCIAL-SOFTWARE FOR THE EFFICIENT REUSE OF SOLUTION PATTERNS FOR

SELF-OPTIMIZING SYSTEMS

343

cal or electrical engineering we investigate functions

of self-optimization such as to plan driving-proﬁle, to

cooperate with other systems, and to optimize behav-

ior. Active patterns for self-optimization (AP

) re-

alize those functions (Figure 2).

Structure

Participating system elements and

their relations

Functions

Function hierarchy, which can be

implemented by the AP

Methods

Methods and algorithms used to

implement the behavior

Methods

Methods and algorithms used to

implement the behavior

Application Scenario

Applications where AP

Application Scenario

Applications where AP

Methods

Methods used to realize the specified

behavior

Behavior

Specification of the Self-Optimizing

Process by activities and state transitions

Active Pattern for

Self-Optimization

Principle Concept

Generally understandable description

of an AP

Application Scenario

Applications where AP

were applied

effectually

starting point

endpoint

Statorausfall

Energy management

To analyze

the situation

To determine

the goals

To adapt the

behavior

To realize

s.o.

NF-System

∆v

Pos.

Neuro-Fuzzy (NF) method

determine

objectives

situation

analyisis

adaption of

behavior

planner

analyzer

monitoring

actuators

sensors

agents

environ-

ment

Hybrid planning tree

a2´

a3´

a4´

a5´

a5´´

Figure 2: Aspects of an Active Pattern for Self-

Optimization.

The overall structure and speciﬁcation of an AP

is described in various aspects in order to cover a

full self-optimization process and to facilitate the de-

velopers by ﬁnding the AP

, that is appropriate to

the task. The aspect functions gives an overview of

the functions, which can be realized. Therefore the

functions are classiﬁed into the three steps of self-

optimization. The principle concept gives an intu-

itive view on what the AP

can perform. The struc-

ture is a generic speciﬁcation of the mandatory el-

ements and their relations. Within the aspect be-

havior, activities of the elements and resulting state

transitions are described. The aspect methods lists

all possible methods, which can be used for the re-

spective functions. Examples of methods are state-

space search, neuro-fuzzy-systems or multiobjective

optimization. The successful application of an AP

is documented in the aspect application scenario. This

helps the user to understand how this pattern can be

implemented.

4 DOMAIN SPECIFIC

LANGUAGES

The basic idea of solution patterns is the reuse of pre-

viously gathered design experiences. Thus, some kind

of repository is required where the solution patterns or

information about them can be stored and retrieved.

Figure 3 illustrates this basic idea: A knowledge

provider adds some information (knowledge docu-

ment) to the repository. In case of AP

such a knowl-

edge document can be any of the six aspects of the

or some comments or additional information

about them. Especially new application scenarios and

methods are likely to be added frequently to the repos-

itory.

KnowledgeProvider KnowledgeUserPatternRepository

Domain

Knowledge

Document

uer

Analyze

and Utilize

Figure 3: Basic Idea of Information Reuse by Solution Pat-

terns.

On the other side, some knowledge user ex-

plores the repository and analyzes and (if possi-

ble) reuses the experiences provided in the AP

The aspect functions offers the knowledge user a

highly structured way to explore the repository. If

the designer (the knowledge user) identiﬁed self-

optimization functions required in his product, he can

simply ﬁlter out irrelevant patterns. A pattern is ir-

relevant if it supports none of the required functions.

An approach for a methodology to identify required

functions is described in (Gausemeier et al., 2008a).

On the other hand, the designer’s ability to create

an innovation, e.g. a product with new functionality

or performance level, can be supported if he is able

to identify patterns or application scenarios which

exhibit some kind of analogy to his current design

problem. The most efﬁcient and unrestrained method

to explore the repository for analogies is a fulltext

search. It enables the knowledge user to search for

words and phrases which possible indicate possible

analogy. But here arises a problem, which is also

illustrated in Figure 3: If the knowledge provider is

from a domain A while a prospective knowledge user

is from domain B, it is doubtful that the knowledge

user is able to formulate a search query that identiﬁes

the relevant solution pattern or application scenario.

Even if the knowledge user is able to identify the rel-

evant information, he may not be able to understand

it, because it is formulated in the terminology of do-

main A.

Due to the domain-spanning nature of the devel-

opment process of self-optimizing mechatronic sys-

tems such hurdles and gaps in the attempt to reuse de-

sign experiences often occur in this application con-

text. Thus, an information system that implements the

pattern repository must support two features: domain-

spanning search and facilitating the comprehensibility

of information from foreign domains.

KEOD 2009 - International Conference on Knowledge Engineering and Ontology Development

344

5 SOLUTION APPROACH

Building an information system that implements a

repository for active patterns for self-optimization

that supports the two features above is not trivial.

Classical engineering pattern are usually provided in

static catalogues. Static refers to the fact, that the

user of a (maybe electronic) catalogue cannot add or

modify information. These catalogues work well for

longtime established solution patterns with general-

ized applications examples. They will not work in

an entirely new or multidisciplinary domain, which

posses a high degree of innovation. Such a dynamic

linked domain calls for dynamic methods and tools

to manage and link the experiences, knowledge and

success stories. Collaboration between the experts

working on self-optimization is crucial. Especially

due to the domain-spanning nature of self-optimizing

mechatronics it will not be possible to formalize and

generalize all required knowledge into a static cata-

logue.

During the last years a new kind of software was

established: social software (Tepper, 2003). This kind

of software enables user to collaboratively provide

information and information classiﬁcation. Wikis,

blogs and folksonomies are important keywords in

this area. In our opinion, this kind of software is most

promising to build an environment which enables ef-

ﬁcient reuse of design experiences and knowledge.

Many successful examples from knowledge extensive

application domains like health (Boulos and Wheel-

ert, 2007) or academia (Bryant, 2006) are known.

McAffee (MacAffee, 2006) identiﬁed six core

components of social software: search, links, author-

ing, tagging extension and signals. The next section

will introduce our prototype information system or

repository and will subsequently explain how social

software technology can be used to amplify the use of

solutions patterns in the context of self-optimization.

5.1 The Active Pattern Knowledge Base

Once a self-optimization process has been speciﬁed

and implemented successfully, the information needs

to be documented in order to enable the reuse in dif-

ferent applications. This is necessary not only if the

developer could not refer to an existing AP

(in this

case a new AP

has to be described), but also if an

was implemented (in that case at least a new ap-

plication scenario has to be described). Apparently it

is necessary to use a database, which stores the infor-

mation in such a manner, that developers are able to

recognize an appropriate AP

. Therefore we devel-

oped an ergonomic knowledge base for the systematic

management of AP

, called Active Pattern Knowl-

edge Base (APKB). The graphical user interface of

the APKB is illustrated in Figure 4.

Figure 4: GUI of the Active Pattern Knowledge Base.

The Active Pattern Knowledge Base links all the

aspects of an active pattern to the respective partial

models, which are speciﬁed in Microsoft Visio. How-

ever, the user can see a preview of the partial model.

Furthermore, information regarding the various meth-

ods is stored directly in the knowledge base. Both,

the active pattern and the methods are stored as a list

to enable a quick access. To ﬁnd an appropriate ac-

tive pattern, a fulltext search is available, which looks

for matches between the search word and the notions

used in the partial models.

5.2 Social Software Features for

Solution Patterns

In this subsection we explain how the six core com-

penents of social software can be used to support the

knowledge reuse in the design of self-optimizing sys-

tems.

5.2.1 Search

In (Kl

opper et al., 2008) we introduced a domain

spanning search process. The domain spanning

search process relies on the deﬁnition of domain spe-

ciﬁc ontologies. The domain spanning-search process

relies on the deﬁnition of domain-speciﬁc ontology.

In their application to information systems, ontolo-

gies are regarded as designed artefacts consistent of

a speciﬁc shared vocabulary used to describe entities

in the domain, as well as a set of assumptions about

the intended meaning of the term in the vocabulary

(Chandrasekaran et al., 1999).

The domain spanning search is split up into two

phases: the maintenance phase and the search pro-

TOWARDS SOCIAL-SOFTWARE FOR THE EFFICIENT REUSE OF SOLUTION PATTERNS FOR

SELF-OPTIMIZING SYSTEMS

345

cess. During the maintenance phase a number of do-

main ontologies is deﬁned. By translating the search

phrase into other ontologies, active pattern speciﬁed

from other domains can be found. Search results are

presented with a percentaged quality. A manual or

tool supported process (information regarding ontol-

ogy matching tools cf. (Noy and Musen, 2002) de-

ﬁnes translation dictionaries between the domain on-

tologies. These dictionaries are used to translate the

search query entered by the user into different do-

mains.

5.2.2 Authoring

Wikis are an example of social writing software (Leuf

and Cunningham, 2001). A wiki allows a user to

add new content and to modify existing content.

They are used for sharing knowledge (encyclopaedia-

style wiki) or to run community projects. The wiki-

functionality is very interesting for the active pattern

knowledge base. It would enable the knowledge user

to add comments and thus facilitate the comprehensi-

bility for users from the same domain.

5.2.3 Links

Links created by the user would be very useful in

the active pattern context. For instance, user could

add links to application scenarios which facilitate the

understanding of certain aspects of an active pattern.

Or the implementation of certain behavior by a new

method could be documented just by linking them.

5.2.4 Tagging

Ontological engineering, which is the basis of the do-

main spanning search process is a complex task. It is

hardly possible to grasp an entire domain in the ﬁrst

try. Especially the extraction of relevant terms from

domain expert is unlikely to be complete in a single

attempt.

Tagging offers an interesting alternative to enable

domain spanning enquiry. Tagging enables the user

to attach simple one-word descriptions. By attaching

these tags, the users categorize the content. The result

is a so called folksonomy which offers an alternative

way to explore the solution pattern in a domain span-

ning fashion.

5.2.5 Extensions

Extensions are algorithms which automate the catego-

rization and pattern matching to certain extend. Ama-

zons recommendations are an excellent example.

Obviously, such automated recommendations

could be very useful in the active pattern knowledge

base. Especially in order to explore application exam-

ples and methods recommendation could be useful. A

recommendation algorithm for the pattern knowledge

base could work on similarity of application scenar-

ios, nouns used in the descriptions or the tags added

by users.

5.2.6 Signals

Signals refer to the user option to register for certain

changes or updates. For instance, a user could regis-

ter for all new application scenarios related to certain

application domain. Thus, signals enable the user to

conveniently track all update and changes relevant for

his every day work.

6 CONCLUSIONS

This contribution introduced self-optimization as a

new powerful paradigm for mechatronic systems. In

addition, it was pointed out that the design of self-

optimizing systems is a challenge, since various do-

mains are participating in.

An important aspect of the development of tech-

nical systems is the use/reuse of existing solution pat-

tern. Thus, we introduced active pattern for self-

optimization as a new type of solution pattern to en-

able a reuse of once successfully implemented self-

optimizing concepts. The different domain-speciﬁc

languages of the developers result in two challenges

for an active pattern repository: the implementation

of a domain-spanning search and the support of un-

derstanding information from another domain.

To answer these challenges the Active Pattern

Knowledge Base was developed. It supports devel-

opers in searching for active pattern adequate to the

design task. In order to match all the requirements for

the efﬁcient development of self-optimizing mecha-

tronic systems, which are mostly due to their inter-

disciplinary design ﬂow, we propose the use of social

software features. We strongly believe that those fea-

tures, in the way we presented them, will improve the

collaborative work between developers from different

domains effectively.

ACKNOWLEDGEMENTS

This contribution was developed and published in

the course of the Collaborative Research Centre 614

KEOD 2009 - International Conference on Knowledge Engineering and Ontology Development

346

”Self-Optimizing Concepts and Structures in Me-

chanical Engineering” funded by the German Re-

search Foundation (DFG) under grant number SFB

614. One author is supported by the International

Graduate School of Dynamic Intelligent Systems,

which is state-aided by the Ministry of Innovation,

Science, Research and Technology of the federal state

North Rhine-Westphalia, Germany.

REFERENCES

Alexander, C., Ishikawa., S., Silverstein., M., Jacobson.,

M., Fiksdahl-King., I., and Angel, A. (1977). A Pat-

tern Language. Oxford University Press, New York.

Boulos, M. and Wheelert, S. (2007). The emerging web

2.0 social software: an enabling suite of sociable tech-

nologies in health and health care education. Health

Information and Libraries Journal, 24:2–23.

Bryant, T. (2006). Social software in academia. Educause

Quaterly, 2006(2):61–63.

Chandrasekaran, B., Josephson, J., and Benjamins, V.

(1999). What are ontologies, and why do we need

them? IEEE Intelligent Systems, 14(1):20–26.

Gausemeier, J., Dumitrescu, R., and Podlogar, H. (2008a).

Implementing cognitive functions with active patterns

in self-optimizing systems. In Proceedings of 3rd Asia

International Symposium on Mechatronics 2008.

Gausemeier, J., Frank., U., and Schmidt, A. (2005). Active

patterns of self-optimization as a means for the design

of intelligent systems. In Proceedings of the 15th In-

ternational CIRP Design Seminar 2005.

Gausemeier, J., Frank, U., and Steffen, D. (2006). Speci-

ﬁying the principle solution of tomorrows mechanical

engeneering products. In Proceedings of the Design

2006.

Gausemeier, J., Kahl, S., and Pook, S. (2008b). From

mechatronics to self-optimizing systems. In Proceed-

ings of he 7th International Heinz Nixdorf Symposium,

volume 223, pages 3–35. HNI-Verlagsschriftenreihe.

Gausemeier, J., Zimmer, D., U. Frank, B. K., and Schmidt,

A. (2007). Using active patterns for the conceptual de-

sign of self-optimizing systems exampliﬁed by an air

gap adjustment system. In Proceedings of 27th ASME

International Computer and Information in Engineer-

ing Conference.

opper, B., Podlogar., H., Witting, K., and Gausemeier,

J. (2008). Domain spanning search for solution pat-

terns for conceptual design of self-optimizing sys-

tems. In Proceedings of the 1st International Work-

shop on Data Modeling in Virtual Engineering.

Leuf, B. and Cunningham, W. (2001). The Wiki Way. Quick

Collaboration on the Web. Addison-Wesley.

Noy, N. and Musen, M. (2002). Evaluating ontology map-

ping tools: Requirements and experience. In Proceed-

ings of the Workshop on Evaluation of Ontology Tools

at EKAW02.

Pahl, G., Beitz., W., Feldhusen, J., and Grote, K.-H. (2007).

Engineering Design A Systematic Approach. Springer

Verlag.

Tepper, M. (2003). The rise of social software. netWorker,

7(3):18–23.

VDI (2004). Vdi- guideline 2206 design methodology

for mechatronic systems. Technical report, Verein

Deutscher Ingenieure (VDI).

TOWARDS SOCIAL-SOFTWARE FOR THE EFFICIENT REUSE OF SOLUTION PATTERNS FOR

SELF-OPTIMIZING SYSTEMS

347