Ontology based Information Management for Industrial Applications

Markus Germann, Constantin Rieder and Klaus Peter Scherer

Institute for Automation and Applied Informatics, Karlsruhe Institute of Technology,

Hermann-von-Helmholtz-Platz 1, Eggenstein-Leopoldshafen, Germany

Keywords:

Ontology Development, Knowledge Graph, Collaboration, Heterogeneous Consistent Knowledge, Semantic

Annotation, Intelligent Reasoning, Semantic Wiki, Ontology Learning.

Abstract:

For industrial applications, intelligent systems are available, helpful and necessary to support the complex

human expert decisions and also the design based construction processes, especially when complex constraints

for the process behaviour are given. Normally, such an intelligent support system consists of a knowledge

based module, which is responsible for the real assistance power, representing the user speciﬁed information,

the reasoning explanation part and the logical reasoning process itself. The interview based acquisition and

generation of the complex expert knowledge itself is very crucial because of differing correlations between

the complex parameters. So, in this project intelligent Wiki based methods are researched and developed for a

quality improvement of an ontology based information system concerning electronic 3D print processes.

1 PROBLEM AND PURPOSE

The establishing guide lines and standards for quality

assurance in machine based manufacturing processes,

especially in the context of print processes (for ex-

ample 3D printing), requires enormous efforts. Many

areas of machine based manufacturing already uti-

lize software systems leveraging knowledge manage-

ment technologies to make complex heterogeneous

relationships accessible towards both the construc-

tion engineer as well as the quality assurance per-

sonal (Kamsu-Foguem and Noyes, 2013). The aim

is to achieve a methodically consistent integrated in-

formation management from the design stage through

the manufacturing process up to the quality control

(Song et al., 2016; Al-Saﬁ and Vyatkin, 2007). In

this proposal a software-based intelligent assistance

and knowledge system is conceptualized, which sup-

ports the human engineer on his acquisition of knowl-

edge, the information inquiry up to the quality con-

trol in the sense of purposeful criteria validation. So

the assistance system is also a support system for the

complex human expert decisions. For this, subject

speciﬁc detailed knowledge has to be represented in a

consistent manner and structured in the partial knowl-

edge domains. A possible approach to reach this goal

can be achieved exploiting the advantages of knowl-

edge graphs (Banerjee et al., 2017; Kamsu-Foguem

and Abanda, 2015). The success of a high performed

support for knowledge acquisition depends also on

special knowledge acquisition structures. Further-

more there’s need for easy-to-use graphical access to

inquire and check the semantics between the individ-

ual knowledge contents. With these goals an enhance-

ment of the semantic ontology network is performed.

Since the knowledge in many existing informa-

tion systems is stored in a decentralized and non well

structured manner and because of that the results are

often not consistent, a user encounters a variety of dif-

ﬁculties by searching and generating information in

the typical classical (Wiki) information systems. The

difﬁculties are:

• Information overﬂow: Information, which is ir-

relevant for ﬁnding a solution in the particular use

case.

• Incomplete information: Missing feature descrip-

tions and necessary details.

• Redundant information: The consequences are a

reduced fault tolerance and problems when incre-

mental changes, i.e. a higher vulnerability to er-

rors.

• Reliable information: Often unveriﬁed knowledge

and fuzzy statements can’t modelled and inte-

grated in the knowledge domain.

• Incorrect information: In the real context contra-

dictions are given to other knowledge contents.

Germann, M., Rieder, C. and Scherer, K.

Ontology based Information Management for Industrial Applications.

DOI: 10.5220/0009095607730779

In Proceedings of the 12th International Conference on Agents and Artiﬁcial Intelligence (ICAART 2020) - Volume 2, pages 773-779

ISBN: 978-989-758-395-7; ISSN: 2184-433X

773

• Disjoint storage of data: This fact leads to an ac-

cess with potential inconsistencies.

• Semantic embedding: Usually not easy accessi-

ble because there exists no ontology network with

semantic relations.

Consequently there is a need to research and develop

an extendable, easy to handle information manage-

ment system, based on a consistent ontology network

with reasoned correlations which takes into account

the disadvantages of the above mentioned criteria.

One real application scenario is the complex 3D print

process with focus on printed electronics, which is es-

tablished at the Institute of Automation and Computer

Science (IAI) in the Karlsruhe Institute of Technology

(KIT). The following sections 2 and 3 outline the ap-

proach of this application scenario and the methodol-

ogy that is used. Section 4 exempliﬁes further details

regarding the structuring of the knowledge. A con-

crete use case in which the approach is applied and

the models are tested is explicated in section 5.

2 INNOVATIVE APPROACH

In contradiction to classical existing information

management systems, which often use a wiki struc-

ture, in this proposal a semantic wiki with intelli-

gent reasoning methods is required for strong problem

solving capabilities for the development and creation

of the new knowledge. Important research aspects are

the effective linking of heterogeneous knowledge ele-

ments provided by the system, which can range from

semi-structured data (texts, images, videos) to highly

formalized models and rules. Furthermore tools have

to be developed and must be available to simplify

capturing and changing heterogeneous knowledge el-

ements. Ideally, the knowledge should be edited di-

rectly by the domain experts and so the continuous

development of the knowledge base is to be driven

forward. The following Figure 1 shows the architec-

ture of a reason based wiki extension and the interface

between data, semantics and the user.

2.1 Knowledge Structuring

On the ﬁrst level a frame based information system

is performed with all features of the domain ontol-

ogy. The descriptive layer for the interested informa-

tion categories is a concept based network represen-

tation with a reﬁnement from class to subclass along

the relation “subconcept”. For a successful user ac-

cess, useful graphical representations of the hierarchi-

cal knowledge structure exist. For example, Figure 2

Frontend/WebUI

Semantic Wiki

Data

Ontology

Population

Thesaurus

Learning

Term

Extraction

Multimedia

Text

Numerical

Table

Unstructured

annotations

Structured

annotations

Reasoning

Strong

Causalities

Weak

associations

Figure 1: Conceptual overview.

Concept

Subconcept

Subsub-

concept

Figure 2: Circle-Packing visualization of the concept hier-

archy.

shows a hierarchical representation in a bubble man-

ner, called Circle Packing. The whole ontology con-

sists of the declarative part of the semantics, the con-

cept in this knowledge based approach, the sources

(publications, authors and so on) and multimedia rep-

resentations (pictures, tables, and movies) of the con-

cept functionalities. On the next layer the concept

of the semantic dependencies are subdivided into two

types of relations: The knowledge can be correlated

by so called weak links, i. e. semantic associations

between the different concepts. It is a suggestion for

the human user, also in regard to the semantic neigh-

ICAART 2020 - 12th International Conference on Agents and Artiﬁcial Intelligence

774

bored knowledge domain when focusing on a con-

cept. But there exist no strong logical dependencies

between the two knowledge items. The human expert

can be guided from one interest of point to another in

an intelligent manner along the linked correlation. An

additional type of relation is performed by a stronger

relation like a rule in a rule based approach. If the

premises of the ﬁrst domain are given, then the second

knowledge items result as a strong causal conclusion

from the preconditions.

2.2 Knowledge Formalization

Additional to a well structured knowledge domain,

the formalization aspect is important for a correct rea-

soning process. Formalized methods are also respon-

sible for an efﬁcient generation of new knowledge.

One important aim is to generate a system expandable

and consistent in the knowledge domains up from the

starting point. This fact includes a user friendly ar-

chitecture and component development. In the area

of data management, this concerns the access to ﬁle

systems and databases. The subjects of knowledge

acquisition and the use of knowledge are to be con-

sidered as separately architectured components. In

the following diagram (Figure 3), the aspects of the

formalization of knowledge as a continuum with the

intelligent reasoning process are presented.

The architecture of the gained innovative ap-

proach is represented by the following three compo-

nents.

• Wiki knowledge which different aspects

• Intelligent request and user navigation

• Formalized consistent knowledge continuum

Figure 3 summarizes these main components in fur-

ther detail.

With the use of such semantic Wiki concepts the fol-

lowing advantages have been established:

• Knowledge management: Easy and consistent

handling of the knowledge concepts and the re-

lationships between them.

• Ontology engineering: Integrated comprehensive

state description with semantic network models.

• Flexibility: Case based driven updates in the

knowledge domain by comfortable and useful

knowledge acquisition methods.

• Reasoning: Strong causality in the sense of predi-

cate logic second order and additional weak asso-

ciations as weak links.

• Query languages: Easy user accessibility to the

knowledge concepts distributed in the network

structure.

Figure 3: Composition of the target features for acquiring

and modelling of knowledge.

Based on these networking and standardized ontol-

ogy concepts, enormous advantages are expected for

the information management performance concerning

users and the process developers as domain experts.

This results also in an intelligent human decision sup-

port system for the given industrial application (here

the 3D print domain as use case).

3 ADVANCED KNOWLEDGE

REPRESENTATIONS

A Wiki is a web-based system used for collaborative

development and maintenance of content. A seman-

tic wiki extends the ﬂexibility of a conventional wiki

when editing texts by the creation of ontologies (con-

cepts and their semantic linking) (Baumeister et al.,

2011). In this work a high performed concept for a se-

mantic wiki for the representation and structured use

of 3D print processes (expert knowledge) with ontol-

ogy based methods shall be conceptualized. Besides

the semantic linking of multimedia contents (texts and

images), the system should also be able to process

strong problem solving knowledge in the sense of a

reasoning process for a required functionality, e.g. as

rules or references.

The knowledge will be available directly from the

Wiki, so that decisions can be automatically sup-

ported immediately after the creation of the knowl-

edge base. Consequently images, explanatory texts or

videos can be made available directly next to formal-

ized rules for human construction requests within the

print process. The semantic wiki system KnowWE

Website of KnowWE, visited 2019-11-07: https://

www.d3web.de/Wiki.jsp?page=KnowWE

Ontology based Information Management for Industrial Applications

775

(”Knowledge Wiki Environment”) is in representa-

tion mode similar to the well known free encyclopedia

Wikipedia, but there are additional features: there are

semantic links between the contents, i.e. specialized

articles can be placed in certain relationships to other

and they can be evaluated according to the resulting

ontology (the graph system). The ontology network

to be developed is to be modelled with the help of

this software in order to provide an intelligent infor-

mation management, based on the use case of the 3D

print process.

Each fact (knowledge concept), i.e. each technical

term, is assigned to an article that can be integrated

into any content such as textual explanations, images

or other media contents (see Figure 4). The infor-

mation concepts can be associated with each other

via different relations (partially in the sense of pred-

icates). Reversely, these n-digit relations (linking of

n concepts via predicates) can read and electronically

processed. The corresponding visualizations are also

realized in this manner. A visualization of the knowl-

edge base can be accessed and regarded from any con-

cept. The combination of the given data structures and

the visualization with functions such as Overview,

Zoom, Details-on-Demand enables a fast user con-

trolled data exploration (Shneiderman, 1996). As an

effect, the user can get a very fast overview of the con-

text of a certain term, fact or circumstance very efﬁ-

ciently. The user can navigate through the knowledge

base textually from article to article like in a classical

Wiki system, or he can use the visualization if nec-

essary to proﬁt from the advantage of the graphical

illustration with direct access to the relationships via

the network structure (Figure 5). The visualizations

can also extended to the view design proﬁles of com-

plex products or processes like presented in (Keivan-

pour and Kadi, 2018). An example of the textual and

graphical representation can be seen in Figure 4.

4 ONTOLOGY & INFORMATION

BASED MANAGEMENT

An intelligent use of the existing information can be

supported by deﬁning an ontology. The ontology can

then provide the following features:

• Terminology for the description of content.

• Axioms that deﬁne terms used by other terms.

• For certain statements automatic conclusions can

be performed (e.g. transitivity).

• Unconventional conclusions are possible (making

implicit information explicit).

Figure 4: Wiki representation with hierarchical class struc-

ture, a feature list and images.

Water

Resistant

Flexible

Polypropylene

Material

Properties

has

type

Printer-

Filament

Type

has

name

has

name

has

name

Figure 5: Polypropylene is both ﬂexible and water-

repellent.

Furthermore, the architecture of a 3D printing ontol-

ogy can be used to create a consistent information

model with the following representative properties:

• Deﬁnition and Classiﬁcation of the individual el-

ements and components (of substrates and ﬁla-

ments).

• Relations between the deﬁned elements repre-

sented by semantic associations.

• Links to describe elements using their elementary

characteristics.

With regard to the development of an ontology for 3D

printing materials various aspects must be considered.

Figure 5 shows the possibility of classifying materials

according to different types. Another possibility is the

categorisation according to the properties of materials

(Figure 6). Annotated elements in the ontology can

be considered in relationship with each other. For in-

stance, the material ”Polypropylene” is both ﬂexible

and water repellent. In Figure 5 (see above) the log-

ICAART 2020 - 12th International Conference on Agents and Artiﬁcial Intelligence

776

ical extension of a classical wiki is shown. Free de-

scribed relations between the different concepts (from

different partial hierarchical trees) can used for logi-

cal reasoning. The propylene has properties from dif-

ferent class structures an the user has the possibility

to ask for feature combinations, which is not possible

in classical wikis.

4.1 Ontology Concepts and Relations

In order to build a semantic net and provide the users

more expressive types of knowledge, more additional

relation types can be deﬁned for example:

performsProcess:

A relation that represents the possibility of any

class or individual to perform any speciﬁc or type

of processes.

fulﬁllsPurpose:

This relationship represents the fulﬁlling of a cer-

tain purpose by a class or individual by another.

compatibleWith:

For instance one substrate can be compatible with

a speciﬁc ink type or not.

The resulting ontology is represented in RDF(S). In

cooperation with the 3D print experts the require-

ments for further relations and concepts have to be

identiﬁed. In conclusion the following useful aspects

were regarded:

• Obtain an overview of the 3D print knowledge

base by reducing the complexity.

• Obtain an overview of the processes and depen-

dencies between the procedure steps.

• Browsing through the entire knowledge base to

identify interesting spots.

• Retrieve detailed information on special relations

between concepts and procedure steps on demand.

• Help the user to ﬁnd quickly the category of a con-

cept inclusive the sub-concepts.

By deﬁning an ontology, the existing knowledge can

be classiﬁed. This can be restrictive because often the

several users use different terminologies. The users

must formulate their own ontology of terms, so these

can linked with each other. On this way, the contained

concepts can directly mapped if all possible designa-

tions or labels are entered. If for example, any user

searches for ”Polypropylene”, the entries with ”PP”

are also taken into account (Figure 7).

The look-up of information can be extended. In case

a substrate with several properties is required, it will

be simpliﬁed by using an ontology. The following

UV Resistant

Impact

Resistant

Water

Resistant

Flexible

Polypropylene

Material

Properties

Printer-

Filament

subconcept

owl: hasProperties

subconcept

Figure 6: Property assoziations.

example illustrates this fact. The search for a mate-

rial that is both ﬂexible and waterproof would deliver

polypropylene as a result (see Figure 6). In an on-

tology, these properties were deﬁned and implicitly

interlinked for PP.

ABS

Nylon

Polycarbonat

TPE

Polypropylene

Printer-Filament

Acrylnitril-

Butadien-Styrol

Thermoplastic

Elastomer

rdfs: label

Figure 7: Typical labeling concept for the printing domain.

4.2 Self Learning Methods for

Ontologies

In industrial applications and more speciﬁcally in

print technologies, semantic information systems are

in progress but there is a lack of solution approaches

covering the necessary and permanent maintenance of

the knowledge. This is shown in Figure 8.

This consistent maintenance has to be carried out for

structured data (measurements, results, digital pat-

terns) as well as for textual information like reports,

design steps and guidelines. There are two essential

problems concerning the learning models:

• The semantic approaches use highly specialized

knowledge and the established knowledge is given

only by a few experts.

• The currently used methods are only available for

small data sets and it is a big challenge to pro-

cess different corpora concerning other document

structures, different terms or word variations.

Ontology based Information Management for Industrial Applications

777

initial ontology

enhanced learned

ontology

thesaurus word

synonymous

numerical data

association rules

structuredunstructured

Figure 8: Enhancing an ontology by learning.

Therefore an innovative objective for intelligent in-

formation processing is the research and development

of self learning semantic information systems (Furth

and Baumeister, 2014; Karthikeyan and Karthikeyani,

2015). These learning methods must be integrated in

the ontology network. In this context, the construc-

tion and the support of the knowledge base is the cen-

tral point. And the methods must be supported by

knowledge engineering tools. Additional to the es-

tablished knowledge representations like RDF(s) and

OWL there exist also tools like KnowWE, Onto stu-

dio, Prot

e and semantic Media Wiki, but none of

these tools serves for the maintenance of the knowl-

edge base. The focused self learning system should

be powerful to reprocess the concepts of the knowl-

edge base by modern knowledge engineering meth-

ods. This means especially delete, add, reprocess and

conﬁrm of concepts of the ontology. The functions

will lean on basic data base operations Create, Read,

Update and Delete.

5 USE CASE AND A FIRST

APPLICATION SCENARIO

Since 3D printing technologies has gained a high pop-

ularity by becoming accessible for a broad range of

users, a lot of experiments have been conducted on the

materials that can be used. This includes, but is not

limited to, ink types, substrates and printing systems

of varying character (e.g. thermal ink-jet or piezo

printing systems). On top of that scientiﬁc projects

aiming for speciﬁc use cases of printing technologies

have been under development. Among them some

target the purpose of printing electric circuits with

electrically conductive ink. This kind of scientiﬁc re-

search leads in the past time to a vast amount of in-

formation, knowledge and experiences gained in the

process. The knowledge has been collected, sorted

and ordered by many scientiﬁc researches and orga-

nizations and therefore been stored decentralized and

heterogeneously. We propose a system that enables

the collaborative sharing of the knowledge to make it

accessible to all participating parties and enrich the

knowledge collected leveraging automated ontology-

based reasoning methods utilizing semantic annota-

tions of the retrieved data. This is an effective as-

sistance power for communicating with the domain

experts.

6 CONCLUSION AND OUTLOOK

The ontology based information model with gained

application to the complex 3D print process as a use

case is regarded and developed at the Karlsruhe In-

stitute of technology (KIT). So, the heterogeneous

knowledge of an industrial process (here the complex

3D print process) can be modelled in a consistent log-

ical network with the power of intelligent reasoning

processes. This is supported by a formalized, com-

puter based communication process. The knowledge,

hidden in structured tables or lists can be used for log-

ical reasoning processes and to extend the ontology

network and the knowledge domain. The computer

based representation and visualization for the print

knowledge is suitable and necessary for a high per-

formed communication of the experts because they

can learn concepts and relations in a semantic net-

work and they can enhance the printing development

steps. That means an effective assistance power for

communicating with the domain experts. So, for the

use case of 3D print processes in terms of intelligent

human experiments can be enhanced by a formal-

ized, computer based (intelligent) information man-

agement system.

In order to verify the approach for this applica-

tion scenario both performance and user acceptance

have to be tested and the results need to be analyzed

for possible improvements. Another consequence of

these tests can be a follow-up ﬁne tuning of the mod-

els as well as an enrichment or advancement to facili-

tate the usage in other industrial applications. With

the model growing and ﬁne adjustment it could be

possible to detect the stable invariants of such an on-

tology development.

ACKNOWLEDGEMENT

The work presented in this article is supported and

ﬁnanced by Zentrales Innovationsprogramm Mittel-

stand (ZIM) of the German Federal Ministry of Eco-

nomics and Energy. The authors would like to thank

the project management organization AiF in Berlin

for their cooperation, organization and budgeting.

ICAART 2020 - 12th International Conference on Agents and Artiﬁcial Intelligence

778

REFERENCES

Adrian, B. (2012). Information Extraction on the Semantic

Web. doctoralthesis, Technische Universit

at Kaiser-

slautern.

Al-Saﬁ, Y. and Vyatkin, V. (2007). An ontology-based

reconﬁguration agent for intelligent mechatronic sys-

tems. In Ma

ık, V., Vyatkin, V., and Colombo,

A. W., editors, Holonic and Multi-Agent Systems for

Manufacturing, pages 114–126, Berlin, Heidelberg.

Springer Berlin Heidelberg.

Banerjee, A., Dalal, R., Mittal, S., and Joshi, K. P.

(2017). Generating digital twin models using knowl-

edge graphs for industrial production lines. In Work-

shop on Industrial Knowledge Graphs, co-located

with the 9th International ACM Web Science Confer-

ence 2017, Troy, NY, USA.

Baumeister, J., Reutelshoefer, J., and Puppe, F. (2011). En-

gineering intelligent systems on the knowledge for-

malization continuum. International Journal of Ap-

plied Mathematics and Computer Science, 21(1):27–

39.

Benjamins, V. R., Fensel, D., and G

omez-P

erez, A. (1998).

Knowledge management through ontologies. In

Reimer, U., editor, PAKM 98, Practical Aspects of

Knowledge Management, Proceedings of the Second

International Conference, Basel, Switzerland, Octo-

ber 29-30, 1998, volume 13 of CEUR Workshop Pro-

ceedings. CEUR-WS.org.

Buitelaar, P., Cimiano, P., and Magnini, B. (2005). On-

tology Learning from Text: Methods, Evaluation and

Applications. IOS Press.

Furth, S. and Baumeister, J. (2014). TELESUP-Textual

Self-Learning Support Systems. In Proceedings of the

LWA 2014, FGWM Workshop, RWTH Aachen Univer-

sity, Aachen, Germany, volume 1226, pages 277–286.

Kamsu-Foguem, B. and Abanda, F. H. (2015). Experience

modeling with graphs encoded knowledge for con-

struction industry. Computers in Industry, 70:79 – 88.

Kamsu-Foguem, B. and Noyes, D. (2013). Graph-based

reasoning in collaborative knowledge management

for industrial maintenance. Computers in Industry,

64(8):998 – 1013.

Karthikeyan, K. and Karthikeyani, V. (2015). Ontology

based concept hierarchy extraction of web data. In-

dian Journal of Science and Technology, 8(6):536.

Keivanpour, S. and Kadi, D. A. (2018). Strategic eco-design

map of the complex products: toward visualisation of

the design for environment. International Journal of

Production Research, 56(24):7296–7312.

Schaffert, S., Bry, F., Baumeister, J., and Kiesel, M. (2007).

Semantic wiki. Informatik-Spektrum, 30(6):434–439.

Shneiderman, B. (1996). The eyes have it: a task by data

type taxonomy for information visualizations. In Pro-

ceedings of the 1996 IEEE Symposium on Visual Lan-

guages, VL ’96, pages 336–343, Boulder, CO, USA.

IEEE Comput. Soc. Press.

Song, H., Wang, H., Liu, T., Zhang, Q., and Gao, B.

(2016). The design and development of manufactur-

ing process knowledge base system based on ontol-

ogy. In Luo, Y., editor, Cooperative Design, Visual-

ization, and Engineering - 13th International Confer-

ence, CDVE 2016, Sydney, NSW, Australia, October

24-27, 2016, Proceedings, volume 9929 of Lecture

Notes in Computer Science, pages 9–16.

Studer, R., Benjamins, V., and Fensel, D. (1998). Knowl-

edge engineering: Principles and methods. Data &

Knowledge Engineering, 25(1-2):161–197.

Ontology based Information Management for Industrial Applications

779