KREM: A Generic Knowledge-based Framework for Problem Solving in

Engineering

Proposal and Case Studies

Cecilia Zanni-Merk

INSA de Strasbourg / ICube Laboratory (BFO Team), Illkirch, France

Keywords:

Knowledge Technologies, Ontologies, Reasoning, Experience, Meta-knowledge.

Abstract:

This article presents a generic knowledge-based framework for problem solving in Engineering, in a broad

sense. After a discussion about the drawbacks of the traditional architecture used for deploying knowledge-

based systems (KBS), the KREM (Knowledge, Rules, Experience, Meta-Knowledge) architecture is presented.

The novelty of the proposal comes from the inclusion of experience capitalization and of meta-knowledge use

into the previously discussed traditional architecture. KREM improves the efﬁciency of classic KBSs, as it

permits to deal with incomplete expert knowledge models, by progressively completing them, learning with

experience. Also, the use of meta-knowledge can steer their execution more efﬁciently. This framework has

been successfully used in different projects. Here, the architecture of the KREM model is presented along

with some implementation issues and three case studies are discussed.

1 INTRODUCTION

Nowadays, organizations are becoming increas-

ingly knowledge-intensive and collaborative (Ouer-

tani et al., 2011). However, large parts of useful

knowledge are hidden and not readily available. A

growing number of companies have realized that a

tool that effectively enables the capture, representa-

tion, retrieval, and reuse of knowledge is the key to

supporting various organizational decisions (Daven-

port and Prusak, 2000; Leistner, 2010). And this is

one of the main interests of Knowledge Technologies.

Knowledge Technologies are computer-based

techniques and tools that provide a richer and more

intelligent use of information. Much of their power

comes from the way they combine ideas and applica-

tions from a number of ﬁelds: Psychology, Philos-

ophy, Artiﬁcial Intelligence, Engineering, Business

Studies, Computer Science and Web Technologies

(Milton, 2008).

Knowledge Technologies, as the name suggests,

are about doing things with knowledge. For example:

• Identifying what knowledge is important to an or-

ganisation;

• Deciding what knowledge needs to be captured

to provide an appropriate solution to a real-world

problem;

• Capturing and integrating knowledge from ex-

perts or repositories (such as texts or databases,

for example);

• Representing and storing knowledge in ways that

provide ease of access, navigation, understanding,

maintenance and re-use;

According to (Milton, 2008), Knowledge Tech-

nologies include: Knowledge Based Systems, Nat-

ural Language Processing, Data Mining, Semantic

Technologies, Case Based Reasoning and Intelligent

Agents, among others. We are interested in a subset

of the technologies presented by (Milton, 2008), but

use them in a slightly different way.

One of the main advantages of using Knowledge

Technologies for problem solving in Engineering is

that they facilitate decision-making. Better decisions

can be made when people have the right information

at the right moment. When information is out-of date,

incoherent, irrelevant, hard to ﬁnd and hard to inte-

grate, then the decision-making process is made more

difﬁcult. The information required to make good de-

cisions often exists but tends to be scattered in vari-

ous locations and stored in various formats (such as

databases or plain text). Knowledge Technologies

provide ways of ﬁnding, merging and exposing the

information required to make informed decisions.

Knowledge Technologies are dealt with methods

Zanni-Merk, C..

KREM: A Generic Knowledge-based Framework for Problem Solving in Engineering - Proposal and Case Studies.

In Proceedings of the 7th International Joint Conference on Knowledge Discovery, Knowledge Engineering and Knowledge Management (IC3K 2015) - Volume 2: KEOD, pages 381-388

ISBN: 978-989-758-158-8

381

coming from the ﬁeld of Knowledge Engineering,

whose main goal is the development of Knowledge-

Based Systems (KBSs).

The classical architecture of a KBS (discussed in

Section 2) does not always provide satisfactory results

when solving a problem. In fact, the knowledge mod-

els needed for their implementation are often incom-

plete. Therefore, we propose here a novel architec-

ture to deal with these drawbacks by the inclusion of

the capitalization of the experience learnt with the use

of the KBS. The use of meta-knowledge is also inte-

grated to deal with the whole.

As stated above, Section 2 presents an introduc-

tion to knowledge-based systems. Section 3 deals

with the capitalization of experience. Sections 4 and 5

presents the KREM model with some notes about its

implementation. Section 6 describes three different

applications of KREM; and ﬁnally, Section 7 gives

our conclusions.

2 KNOWLEDGE-BASED

SYSTEMS

One of the main goals of using an approach based on

Knowledge Engineering methods for problem solving

is the development of KBSs. They have a compu-

tational model of some domain of interest in which

symbols serve as surrogates for real world domain

artefacts, such as physical objects, events, relation-

ships, etc (Sowa, 2000).

A knowledge-based system maintains a knowl-

edge base which stores the symbols of the com-

putational model in form of statements about the

domain, and performs reasoning by manipulating

these symbols. Applications can base their decisions

on domain-relevant questions posed to a knowledge

base. They are programmed to solve problems in a

similar way to that of an expert practitioner, e.g. make

inferences based on the case at hand and adopt the

right strategy to solve the problem; they can deal with

incomplete information by making requests for fur-

ther information or by making intelligent guesses at

the answer (just as an expert has to do when there

is limited information); they have a user interface that

makes intelligent requests for information and can ex-

plain how it has arrived at its answers; and ﬁnally, the

development can be made cost-effective by re-using

generic structures, rules and problem-solving meth-

ods.

The components usually found in a KBS are: a

knowledge base that contains domain-speciﬁc infor-

mation, structures and rules; a working memory (also

known as a blackboard) which holds case-speciﬁc

data (e.g. facts about the initial problem and inter-

mediate results); an inference engine (or reasoning

engine) that controls and directs the solving of prob-

lems by making inferences, i.e. it uses the knowledge

base to alter the contents of the working memory; an

interface to other computer systems and/or to human

users; and an editor that allows a knowledge engineer

or domain expert to inspect and update the knowledge

base.

The knowledge-base of a KBS needs a more ex-

plicit description. One of its components is a formal

conceptual model of the domain of interest. Among

the existing formal conceptual models, ontologies are

formalized representations of vocabularies that are

speciﬁc to a certain area. The dominating deﬁnition of

an ontology is the following: An ontology is a formal

explicit speciﬁcation of a shared conceptualisation of

a domain of interest, based on (Gruber, 1993).

Technically, the principal constituents of an ontol-

ogy are concepts, relations and instances. They are

commonly used with a set of rules that are chained

to simulate the reasoning of a human expert. Rules

come in the form of ”if-then” constructs and allow to

express various kinds of complex statements and al-

low different types of reasoning about the facts in the

working memory based on the domain knowledge in

the ontology.

In summary, a knowledge based system contains a

conceptual yet executable model of an application do-

main. It is made machine-interpretable by means of

knowledge representation techniques and can there-

fore be used by applications to base decisions on rea-

soning about domain knowledge.

The traditional architecture that we have just dis-

cussed has drawbacks, associated, mainly, with the

difﬁculties that appear during the knowledge elici-

tation process from experts; and also with the non-

completeness of the formal conceptual model ob-

tained after the elicitation (Milton, 2008).

In fact, as the knowledge base can be incomplete,

there could be problems that this traditional architec-

ture cannot solve. Reasoning and analysis of this in-

complete knowledge implies that it is needed to take

advantage of the experience acquired from the inter-

ventions of human experts when the traditional sys-

tem does not lead to satisfactory results; that is, some

kind of capitalization of experience is needed.

3 ON THE CAPITALIZATION OF

EXPERIENCE

Set of Experience Knowledge Structure (SOEKS) is

an experience-based knowledge representation that

KEOD 2015 - 7th International Conference on Knowledge Engineering and Ontology Development

382

Figure 1: A SOEKS.

can store uncertain and incomplete data and make

qualitative and quantitative extractions of knowl-

edge from the available information, which can of-

ten be unstructured, semi-structured, fuzzy, and vague

(San

ın and Szczerbicki, 2009; San

ın et al., 2007).

Additionally, SOEKS can be shaped in an exten-

sive understandable and transportable language such

as Extensible Markup Language (XML) or Ontology

Web Language (OWL) (Grau et al., 2008). XML

and OWL representation of SOE allows knowledge

to be exchanged quickly and securely between appli-

cations and systems. The SOEKS has been designed

to collect experiences and knowledge from multiple

applications that are assembled as formal decision

events (San

ın and Szczerbicki, 2009). This collected

knowledge assists organizations in making precise de-

cisions, predictions, and recommendations, and it is a

dynamic structure that is dependent on the informa-

tion and data that it has received.

A SOEKS has four components (Figure 1) (San

ın

and Szczerbicki, 2009): variables, functions, con-

straints, and rules. Each formal decision event can

be stored in a combined structure of those four com-

ponents of the SOEKS.

Variables usually involve representing knowledge

using an attribute-value language. This is a tradi-

tional approach from the origin of knowledge repre-

sentation. Variables are related among them in the

shape of functions. Functions, the second component,

describe associations between variables. Therefore,

the set of experience uses functions and establishes

links among the variables constructing multi objec-

tive goals. Constraints are another form expressing

relationships among the variables. A constraint is a

restriction of the feasible solutions in a decision prob-

lem, and limits the performance of a system with re-

spect to its goals. Finally, rules are suitable for repre-

senting inferences or for associating actions with con-

ditions under which the actions should be performed.

They are conditional relationships of the universe of

variables.

4 THE KREM MODEL

Taking into account the points discussed in the pre-

vious sections, we propose a modular architecture,

called KREM (Knowledge, Rules, Experience and

Meta-Knowledge), to manage the complexity of de-

veloping KBSs, to try to solve the aforementioned

problems and to incorporate the capitalization of ex-

perience with the goal of improving decision-making.

Because to be effective, decision-making must re-

sult from reasoning and analysis of this knowledge,

also taking into account the experience and expertise

of decision-makers. As a consequence, it is needed

to capitalize them to take advantage of the experi-

ence acquired from the interventions of human ex-

perts when the traditional system does not lead to sat-

isfactory results.

The use of meta-knowledge to steer the ex-

ecution of the whole system is also necessary.

Meta-knowledge is knowledge about domain knowl-

edge, about rules or about experience. This meta-

knowledge can take the form of context, culture or

protocols to use this knowledge. Context is infor-

mation that characterizes a situation in relation to in-

teraction among human-beings, applications and their

environment, and can be of four types: identity, place,

status or time (Dey et al., 2001). Culture meta-

knowledge tries to take into account the fact that de-

cisions are made differently depending on the coun-

try or culture (Meyer, 2014). And ﬁnally protocols

may include strategies or problem-solving heuristics

for the task to be done (for example, in the case of

medical diagnosis, the protocols used by physicians

change according to the type of symptoms or the sus-

pected illness).

Therefore, the proposed components of the archi-

tecture are (Figure 2):

• The Knowledge component that contains the do-

main knowledge to operate, by means of different

domain ontologies to be developed.

• The Rules component that allows different types

of reasoning (monotone, spatial, temporal, fuzzy,

or other) depending on the application.

• The Experience component that allows the capi-

talization and reuse of prior knowledge.

• The Meta-knowledge component, including

knowledge about the other three bricks that

depends on the problem. For example, in a

medical diagnosis problem, this component may

include clinical protocols used by physicians.

The way the domain knowledge is formalized

will shape the way the rules are expressed. Experi-

ence will come complete the available knowledge and

KREM: A Generic Knowledge-based Framework for Problem Solving in Engineering - Proposal and Case Studies

383

Figure 2: The KREM architecture with its four interre-

lated components: knowledge, rules, experience and meta-

knowledge.

rules. Finally, meta-knowledge will directly interact

with the rules and the experience to indicate which

rules (coming from experience or from the initial rule

set) need to be launched according to the context of

the problem to solve.

A modular architecture, such as KREM, is one

of the main architectural design pattern for large and

complex systems. In this pattern, each module or

component has a speciﬁc functionality providing sep-

aration of concerns that, in turn, support reuse or re-

placement (i.e. changes in a single module would

not affect the others, permitting the continuous op-

eration of the system). Moreover the communication

between modules needs to be based on well deﬁned

interfaces to provide low coupling.

The KREM model has already been successfully

used for very speciﬁc applications; the semantic anal-

ysis of urban satellite images, the diagnosis of small

and medium enterprises or the formalization of the

Lean enterprise. We will discuss these applications in

Section 6.

5 SOME NOTES ABOUT THE

IMPLEMENTATION

The implement the KREM architecture, it is neces-

sary, at least, to combine rules formalisms with de-

scription logics (this is the classic architecture of a

KBS). There are several possibilities to accomplish

this. But our choice has been to have hybrid systems

interfacing logic programming systems and descrip-

tion logic systems.

As explained before, the knowledge module con-

tains the domain ontologies that have been developed

for each project. These developments are made in

OWL (Grau et al., 2008). The Web Ontology Lan-

guage (OWL) is an integral component of the Seman-

tic Web, as it can be used to write ontologies or formal

vocabularies which form the basis for semantic web

data mark-up and exchange.

The Prot

editor in several versions has been

our preferred choice for the design of the ontologies.

Prot

e is usually used to construct domain mod-

els and knowledge-based applications with ontologies

because of possessing a Java-based Application Pro-

gramming Interface (API) and being easily extended

by a plug-in architecture.

Concerning the rules, we have also tested several

approaches. The ﬁrst natural choice was the use of

SWRL (Semantic Web Rule Language).

SWRL is a proposal by the W3C to extend OWL

with rules. It is based on a combination of some

sub-languages of OWL with the Unary/Binary Data-

log RuleML sub-languages of the Rule Markup Lan-

guage

. The proposal extends the OWL abstract syn-

tax to include the syntax of these rules and the OWL

model-theoretic semantics to provide a formal mean-

ing for ontologies that include rules written in this

syntax (Golbreich, 2004).

However, some of the constraints of SWRL

(mainly the lack of input/output features and the im-

possibility to create new individuals in the ontologies

“on the ﬂy”) made us begin investigating other possi-

bilities. And our choice went to Jess.

Jess (Java Expert System Shell) is a rule engine for

the Java platform, developed by Ernest Friedman-Hill

of Sandia National Labs since 1995. Jess supports the

development of rule-based expert systems which can

be tightly coupled to code written in Java language.

Compared with the traditional matching mechanism -

only one loop for all - the Jess rule engine continu-

ously matches facts against rules to infer conclusions,

which result in actions. Thus rules can modify the

collection of facts (Hill, 2003).

Unfortunately, Jess also suffers from some issues,

mainly regarding its licensing, so we are currently

investigating some other open-source inference en-

gines, such as Drools

Concerning the implementation of experience, we

are using the SOEKS in its ontology form. Figure

3 shows part of the SOEKS class hierarchy with the

four components involved in decision making events:

variables, functions, constraints and rules.

What is important to remark concerning the inter-

action between the experience module and the others

is that the reasoning process in the rules component

enables the evolution of the initial set of rules with

experience, as it was proposed by (Toro et al., 2012).

More precisely, every time the native reasoner is ex-

ecuted, the system will store both the output of the

reasoner and the ﬁnal decision, if any, made by the

http://protege.stanford.edu

http://www.ruleml.org.

http://www.drools.org

KEOD 2015 - 7th International Conference on Knowledge Engineering and Ontology Development

384

Figure 3: The SOEKS ontology.

expert. Decisions that do not follow the proposition

of the reasoner produce an evolution of the set of ex-

perience rules.

Finally, concerning Meta-Knowledge, we are cur-

rently investigating the implementation of context, in

a broad sense. Context has been studied carefully for

a long time in areas like knowledge-based and ubiqui-

tous systems, either for handling the complex knowl-

edge in a dynamic manner (McCarthy, 1993; Sowa,

1995; Giunchiglia et al., 2012) or to provide smarter

human interfaces (Dey et al., 2001). However, there

is no agreement on a concise deﬁnition of context in

these areas. (Bazire and Br

ezillon, 2005), for exam-

ple, proposed the components of context as the user

and the observer involved, as well as the items, the en-

vironment, and other related contexts. (Porzel, 2010)

states that context helps when detecting semantic re-

lations to provide extra information and correct inter-

pretations for applications.

Diverse representations of context exist in differ-

ent research areas. (McCarthy, 1993) uses a term c

representing context, and the formal representation

of a proposition p is true in the context c is repre-

sented as ist(c, p). As described above, (Dey et al.,

2001) deﬁne context as any information that char-

acterizes a situation for context-aware applications,

while (Porzel, 2010) refers to a context work, where a

model of context contains components and the differ-

ent relations of the components. The components are

the user, an item, and the observer in the environment.

Relations here include not only the relations between

the components, but also the relations to other con-

texts.

We choose to represent context in the meta-

knowledge module in a way similar to the one used

by (McCarthy, 1993). An ontology enables the iden-

tiﬁcation of the context for a certain agent in the ap-

plication, and this agent will reason with a subset of

all the rules in its Rules or Experience modules, that

will be chosen with the help of the above mentioned

predicate ist, once a judicious interpretation of it has

been chosen.

6 SOME APPLICATIONS OF

KREM

In this section, we will present some applications of

KREM in diverse ﬁelds, showing the degree of devel-

opment of the components and giving pointers to our

publications for the interested reader.

6.1 Diagnosis of SMEs

The MAEOS project (modeling of business advice for

organizational and strategic development of SMEs in

Alsace) is the result of a collaboration between AEM

Conseil (Alsatian SME working on business advice

to small and very small companies) and our labora-

tory. Its main goal was to develop a software pro-

totype for acquisition and processing of information

collected by AEM Conseil from its clients to improve

its business advice.

To do this, a set of knowledge bases relating to

business management has been built and used to char-

acterize the client companies. This characterization

was then augmented by expert knowledge provided

by AEM Conseil obtained during its professional

practice. As this is a multidisciplinary project about

Management Science and Knowledge Engineering,

one of the main obstacles is the combined use of di-

verse and sometimes contradictory knowledge, this is

inherent to models that come from Management Sci-

ence; different authors (or the same author in different

publications) may express different views on the same

situation.

Several domain ontologies were developed in the

Knowledge module of KREM and each ontology rep-

resents a conceptualization of the knowledge of a spe-

ciﬁc ﬁeld in the context of SME (e.g. organization,

KREM: A Generic Knowledge-based Framework for Problem Solving in Engineering - Proposal and Case Studies

385

production, strategy, ﬁnance, etc.). Although an on-

tological study was conducted to provide the neces-

sary theoretical foundations (Renaud et al., 2009; Re-

naud et al., 2010), the majority of the existing ontolo-

gies provide formal exhaustive models that could be

applied to our purposes, but not many among them

include reasoning rules to permit analysis or diagno-

sis. Consequently, our own ontologies and rules bases

needed to be developed.

This Rules brick of KREM manages different

types of rules to allow to reason about the instances

of the classes in the domain ontologies of the knowl-

edge module. To exploit these rules, a prototype of

a multi-agent system (Zanni-Merk et al., 2011), capa-

ble of manipulating modular ontologies and reasoning

with the Jess inference engine, has been developed.

Three types of rules appear in this component:

• Rules for diagnosis to obtain the analysis of the

current situation of the SME.

• Recommendation rules providing hints to im-

prove the current situation of the company.

• ”Bridge” rules to ensure the semantic equivalence

among concepts belonging to different ontologies.

Our latest works permitted the formalization of

the ﬁrst three bricks of the KREM architecture. We

have also begun to explore meta-knowledge that

could be included into the fourth module (Gartiser

et al., 2014).

In fact, making a diagnosis of the situation of

SMEs involves identifying a set of characteristics that

are ”present” in its current state and also identifying

another set of ”missing” features (the desired state).

In this project, we need to effectively assess the cur-

rent state and decide on actions necessary to achieve

the desired state. The current and desired states are

formalized as a SOEKS, and speciﬁc similarity mea-

sures to compare the current and ideal situations of

the SME have been developed. The Meta-Knowledge

module includes diagnosis models (e.g., about correct

or incorrect behavior of companies) and criteria rec-

ommendation according to the experience of the con-

sultant. In addition, in their daily tasks, consultants

follow certain unwritten protocols (theoretical or not)

to diagnose businesses that are also formalized here.

6.2 Semantic Analysis and

Interpretation of Remote Sensed

Images

The availability of very high resolution satellite im-

ages allows to envisage the identiﬁcation of complex

urban objects such as composite objects (e.g. neigh-

borhoods, residential areas).

To extract and classify these urban objects, it is

relevant to design image analysis methods based on a

”thematic” modelling of these objects to extract and

label them. Using formal ontologies seems a judi-

cious choice. Our previous works (Cravero et al.,

2012; di Sciascio M. et al., 2013; de Bertrand de Beu-

vron et al., 2013) show that this can limit the involve-

ment of experts in time-consuming and unrewarding

tasks such as labelling a large set of objects examples.

Taking advantage of those previous works, we

have proposed the formalization of a new semi-

automatic semantic approach for classiﬁcation of ur-

ban objects from satellite images, using the KREM

architecture (Zanni-Merk et al., 2015).

The Experience and Meta-Knowledge compo-

nents of KREM are particularly useful for this appli-

cation. The Experience component capitalizes the ex-

perience acquired from experts in the knowledge do-

main. After running the classiﬁcation step and fail-

ing to identify some segmented regions, the state-

ments contained in this brick should be queried in

order to derive non-usual membership. For instance,

if an unidentiﬁed region corresponds to a swimming

pool in a backyard, then some additional knowledge

based on the expert’s experience could help to rec-

ognize it in this last instance. The Meta-Knowledge

module includes knowledge about the use of the other

components. In fact, according to the context of

each image (latitude, longitude, season), different sets

of rules could be launched to take into account that

context. For example, the radiometric characteristics

that allow the identiﬁcation of vegetation in an image

change according to the season.

6.3 Towards a Formal Model of the

LEAN Enterprise

The goal of this project is to describe the character-

istics of the Lean Enterprise and make the case for

modelling it in order to reproduce its successful prac-

tices more easily.

The literature contains many good descriptions of

the Toyota Production System and Lean in general

(Womack et al., 1991; Byrne and Womack, 2012),

but no formal model that we can build upon. We

propose, then, to follow the KREM model: the K

(Knowledge) component will include domain knowl-

edge about Lean in the form of several ontologies, the

R (Rules) component will be expressed by probabilis-

tic rules, the E (Experience) component will describe

the practices (Kata) and the M (Meta-knowledge)

component will describe the context of the applica-

tion of Lean (different types of companies or cultural

environments, for example).

KEOD 2015 - 7th International Conference on Knowledge Engineering and Ontology Development

386

Our ﬁrst works (Masai et al., 2015) permitted

to verify the feasibility of the approach, by devel-

oping two complementary ontologies of Lean: the

HOT (House Of Toyota production system or House

Of TPS) ontology, which structures the core issues

in Lean Management, and the ontology for Hoshin

Kanri, which represents the entities necessary to

model this speciﬁc process. The Hoshin Kanri pro-

cess is of particular interest because it displays the

behaviour of the agents (the employees in the house

of Toyota production system) at various levels in the

organisation going back and forth.

Concerning the Rules component, because the be-

haviour of enterprises experiences a high level of vari-

ability according to the current context while comply-

ing with speciﬁc management rules, they should be

modelled as stochastic processes. In the previously

cited work, we have chosen to model these stochastic

processes with graphical representations of Markov

Chains (Norris, 1998). We are working on the transla-

tion of these probabilistic graphical models into prob-

abilistic rules (Malhotra, 2001), in order to be exe-

cuted by a probabilistic inference engine.

We are also currently developing the Meta-

Knowledge brick of KREM, because in the present

world, it is not rare that not only two, but multiple cul-

tures coexist in large, multinational companies and in

global projects spanning multiple geographies. The

misunderstandings arising from these cultural differ-

ences are responsible for many failures, which are

then blamed by each nationality on the others, with-

out trying to understand those differences ﬁrst and ad-

dressing them later.

We have developed a strategic ontology of cul-

ture, based on the works of (Meyer, 2014), who iden-

tiﬁes eight dimensions: Communicating, Evaluating,

Persuading, Leading, Deciding, Trusting, Disagree-

ing and Scheduling.

From the speciﬁc point of view of Knowledge

Engineering, this ontology is being formally aligned

with the Upper Ontology of Culture or UOC (Blan-

chard et al., 2010) as a framework. The cultural con-

cepts formalized in this way in the Meta-Knowledge

component of the KREM model can be used in prac-

tice using the HOT ontology. Cultural aspects in the

Meta-Knowledge component will also steer the be-

haviour of the agents that is modelled in the Rules

component of KREM.

7 CONCLUSIONS

These article proposes a modular architecture called

KREM (Knowledge, Rules, Experience, Meta-

Knowledge) to integrate the capitalization of previ-

ous experience in the deploying of a knowledge-based

system. In fact, as in general, knowledge elicitation

from experts produces incomplete expertise models,

sometimes, knowledge-based systems are not able to

provide pertinent results. In this case, it is necessary

to call in an expert to provide an answer. It is inter-

esting, therefore, to capitalize that answer along with

the cognitive process (rules or others) that the expert

carries out to develop it.

Well known technologies exist to implement the

knowledge and the rules components of KREM

(OWL + SWRL). But the notion of “experience” has

not been fully addressed in the literature. The SOEKS

(set of experience knowledge structure), initially pro-

posed by the KERT team from the University of New-

castle (Autralia)

is the natural choice for formalizing

the experience capitalization.

Modern knowledge engineering practices support

the early criteria held in the eighties, based on a mod-

ular architecture for the development of intelligent

systems, which separates domain knowledge from the

reasoning mechanism. This division is thus held in

the proposal of the KREM architecture, where the

K module corresponds to knowledge stored in the

form of logic statements, and the R module con-

tains the necessary rules to manage and exploit such

knowledge. To achieve a better performance of the

knowledge-based system, we have incorporated an

additional brick, which represents the experience of

experts in the knowledge domain (this experience

capitalization permits, to a certain extent, to have ac-

cess to some tacit knowledge from the experts). Fi-

nally, the use of Meta-Knowledge allows the rational

use of the other three components.

The main interest of the KREM architecture is that

it is ﬂexible and modular. Hence, each project pre-

sented in the last section was described within the ar-

chitecture according to the degree of its development

and maturation, showing its adaptability to cope with

different Engineering application domains in a broad

sense.

REFERENCES

Bazire, M. and Br

ezillon, P. (2005). Understanding context

before using it. In Modeling and using context, pages

29–40. Springer.

Blanchard, E. G., Mizoguchi, R., and Lajoie, S. P. (2010).

Structuring the cultural domain with an upper on-

http://www.newcastle.edu.au/research-and-innovation/

centre/engineering-built- environment/

knowledge-engineering-research-team.

KREM: A Generic Knowledge-based Framework for Problem Solving in Engineering - Proposal and Case Studies

387

tology of culture. The Handbook of Research on

Culturally-Aware Information Technology: Perspec-

tives and Models, pages 179–212.

Byrne, A. and Womack, J. (2012). The Lean Turnaround:

How Business Leaders Use Lean Principles to Create

Value and Transform Their Company. BusinessPro

collection. McGraw-Hill Education.

Cravero, M., de Bertrand de Beuvron, F., Zanni-Merk, C.,

and Marc-Zwecker, S. (2012). A description logics

geographical ontology for effective semantic analysis

of satellite images. In KES, pages 1573–1582.

Davenport, T. H. and Prusak, L. (2000). Working Knowl-

edge: How Organizations Manage What They Know.

Harvard Business School Press, Boston, Mass.

de Bertrand de Beuvron, F., Marc-Zwecker, S., Puissant, A.,

and Zanni-Merk, C. (2013). From Expert Knowledge

to Formal Ontologies for Semantic Interpretation of

the Urban Environment from Satellite Images. KES

Journal - Innovation in Knowledge-Based and Intelli-

gent Engineering Systems, 17:55 – 65.

Dey, A. K., Abowd, G. D., and Salber, D. (2001). A

conceptual framework and a toolkit for supporting

the rapid prototyping of context-aware applications.

Hum.-Comput. Interact., 16(2):97–166.

di Sciascio M., Zanni-Merk, C., C., W., S., M.-Z., and

de Beuvron F. (2013). Towards a semi-automatic se-

mantic approach to satellite images analysis. In KES,

pages 1388–1397.

Gartiser, N., Zanni-Merk, C., Boullosa, L., and Ana (2014).

A semantic layered architecture for analysis and diag-

nosis of sme. In Elsevier, editor, Procedia Computer

Science, pages 1165–1174. KES 2014.

Giunchiglia, F., Maltese, V., and Dutta, B. (2012). Domains

and context: ﬁrst steps towards managing diversity in

knowledge. Web Semantics: Science, Services and

Agents on the World Wide Web, 12:53–63.

Golbreich, C. (2004). Combining rule and ontology reason-

ers for the semantic web. In Lecture Notes in Com-

puter Science (including subseries Lecture Notes in

Artiﬁcial Intelligence and Lecture Notes in Bioinfor-

matics), volume 3323, pages 6–22.

Grau, B. C., Horrocks, I., Motik, B., Parsia, B., Patel-

Schneider, P., and Sattler, U. (2008). OWL 2: The

next step for OWL. Web Semantics: Science, Services

and Agents on the World Wide Web, 6(4):309 – 322.

Semantic Web Challenge 2006/2007.

Gruber, T. R. (1993). Toward Principles for the design of

Ontologies used for Knowledge Sharing. Int. J. Hum.-

Comput. Stud., 43:907 – 928.

Hill, E. F. (2003). Jess in Action: Java Rule-Based Systems.

Manning Publications Co., Greenwich, CT, USA.

Leistner, F. (2010). Mastering Organizational Knowledge

Flow: How to Make Knowledge Sharing Work. Wiley.

Malhotra, Y. (2001). Expert systems for knowledge man-

agement: crossing the chasm between information

processing and sense making. Expert Systems with

Applications, 20(1):7–16.

Masai, P., Parrend, P., and Zanni-Merk, C. (2015). Towards

a formal model of the lean enterprise. In Elsevier, ed-

itor, Procedia Computer Science (Proceedings of KES

2015), volume 60C, pages 226–235.

McCarthy, J. (1993). Notes on formalizing context.

Meyer, E. (2014). The Culture Map: Breaking Through the

Invisible Boundaries of Global Business. PublicAf-

fairs.

Milton, N. (2008). Knowledge Technologies. Polimetrica,

Milano, Italy.

Norris, J. R. (1998). Markov Chains. Cambridge University

Press.

Ouertani, M.-Z., Ba

ına, S., Yesilbas, L. G., and Morel, G.

(2011). Traceability and management of dispersed

product knowledge during design and manufacturing.

Computer-Aided Design, 43(5):546–562.

Porzel, R. (2010). Contextual computing: models and ap-

plications. Springer Science & Business Media.

Renaud, D., Zanni-Merk, C., Bouche, P., Gartiser, N., and

Rousselot, F. (2010). A strategy for reasoning with

close knowledge bases. Business Informatics - Data

Mining and Business Intelligence, 16(104):25–33.

Renaud, D., Zanni-Merk, C., and Rousselot, F. (2009).

A compound strategy for ontologies combining. In

KEOD, pages 200–205.

San

ın, C. and Szczerbicki, E. (2009). Experience-based

knowledge representation: Soeks. Cybernetics and

Systems, 40(2):99–122.

San

ın, C., Szczerbicki, E., and Toro, C. (2007). An owl

ontology of set of experience knowledge structure. J.

UCS, 13(2):209–223.

Sowa, J. F. (1995). Syntax, semantics, and pragmatics of

contexts. In Conceptual Structures: Applications, Im-

plementation and Theory, pages 1–15. Springer.

Sowa, J. F. (2000). Knowledge Representation: Logi-

cal, Philosophical and Computational Foundations.

Brooks/Cole Publishing Co., Paciﬁc Grove, CA, USA.

Toro, C., Sanchez, E., Carrasco, E., Mancilla-Amaya, L.,

San

ın, C., Szczerbicki, E., Gra

na, M., Bonachela, P.,

Parra, C., Bueno, G., et al. (2012). Using set of expe-

rience knowledge structure to extend a rule set of clin-

ical decision support system for alzheimer’s disease

diagnosis. Cybernetics and Systems, 43(2):81–95.

Womack, J. P., Jones, D. T., and Roos, D. (1991). The

Machine That Changed the World: The Story of Lean

Production. The MIT international motor vehicle pro-

gram. HarperCollins.

Zanni-Merk, C., Almiron, S., and Renaud, D. (2011).

A multi-agents system for analysis and diagnosis of

smes. In KES-AMSTA, pages 103–112.

Zanni-Merk, C., Marc-Zwecker, S., Wemmert, C., and

de Beuvron, F. d. B. (2015). A layered architecture

for a fuzzy semantic approach for satellite image anal-

ysis. International Journal of Knowledge and Systems

Science (IJKSS), 6(2):31–56.

KEOD 2015 - 7th International Conference on Knowledge Engineering and Ontology Development

388