Construction of Verified Models for Systems

Represented as Networks

Agustin Yagüe

, Pedro P. Alarcón

Juan Garbajosa

, Alexei Lisounkin

, Gerhard Schreck

Universidad Politécnica de Madrid, OEI. E.U. Informática (UPM)

Crtra. De Valencia Km. 7, E-28031 Madrid, Spain

Fraunhofer-Institute for Production Systems and Design Technology (IPK)

Pascalstrasse 8-9 D - 10587 Berlin, Germany

Abstract. This paper presents both a language and process for producing veri-

fied models for systems that can be described as network structures. Analysing

technical systems with interrelated components, a common data model for tech-

nical networks was elaborated and represented formally by UML. The UML

model was transformed to XML and specified by means of XML Schema. The

developed set of XML Schema documents for modelling of network structures

is denoted as the XNetMod language. Use of XML-based language allows to

profit from available XML tools and CAD with XML export. Moreover, XML-

Schema-based specification supports model validation from early phase of

modelling, and tools for the XML-Schema-based validation are available. Ad-

ditionally, model structure verification was provided by algebraic and pattern-

based methods. The design technique was successfully applied to different ap-

plication domains, which were described in terms of network structures.

1 Introduction

This paper presents a language and associated process for producing verified models

for systems which can be described as networks structures (i.e., systems made of

standard components connected to networks). The developed models can further be

used for technical system behaviour simulation in order to highlight issues such as

simulation-based supervision, control and decision-making mechanisms. The ap-

proach taken to meet the intended objectives was as follows: a) an XML-based lan-

guage was defined such that its modelling schemas could be verified by means of

applying algebraic and pattern-based methods specifically developed; b) two model

interpreting processes were implemented: Functional simulation of model that inter-

prets intermediately the language schemas; and derivation of generic plugging (inter-

change) interfaces for communication between the model and an external software.

This paper focuses on how a verified XNetMod modelling document is built. In

the next section, Modelling Language for Network Architectures will give the basics

for the language and verification. Next, the application of the process is presented.

Yagüe A., P. Alarcón P., Garbajosa J., Lisounkin A. and Schreck G. (2004).

Construction of Veriﬁed Models for Systems Represented as Networks.

In Proceedings of the 2nd International Workshop on Veriﬁcation and Validation of Enterprise Information Systems, pages 44-49

DOI: 10.5220/0002685100440049

 SciTePress

2 Modelling Language for Network Architectures

The modelling of technical networks leads us to the field of systems with interrelated

components. Ranging from the very simple and still widely used Chen´s Entity-

Relationship model, to the object-oriented models and functional models, a wide

range of proposals have been issued. However, there is a lack of some basic mecha-

nisms, such as model components relationships definition [1]. Actually, research

attempts are focused on further development of formal and knowledge-based ap-

proaches (see [2], [3]). Though modelling language semantics are obviously essential,

language syntactic capabilities are also important. Here, an XML-based language has

advantages as a language for data interchange and application interaction.

2.1 Abstract Modelling of Network Architectures

A network structured system possesses the following important properties: a) the

processing elements of the system form the nodes of the network architecture; b) the

topological relations among the processing elements are the links in the network and

represent functional relations between the nodes; c) the network structure may pos-

sess a special (pattern based) structure (example – Petri net); d) the connected proc-

essing elements must match some specification with respect to the structure and val-

ues of their attributes.

The structure of the system can be described according to the following data

model, Fig. 1. Here, the class

"model" is the root class for the entire model. Classes

"node<level1>" and "node<level2>" represent node elements of different types

or even hierarchical levels, the number of hierarchical levels is not limited.

1..*

attribute

attributeSingle

attributeVector

attributeMatrix

node <level2>

node <level1>

net

model

...

rule

1..*

Fig. 1. Generic class model of a network structure

We emphasise that the set of node elements may be used for definition of different

kinds of interrelations. One kind of interrelations in the system will be modelled by a

corresponding network (e.g. material flow network, control flow network, etc.). Each

network has its own interpretation procedure (at least one). Obviously, the model

must also be able to represent several networks. Thus, network topologies involved in

the modelling are represented by instances of class

"net". The "pin" elements refer

to the nodes which are directly connected by means of the related network. Also

"rule"-elements, which may define legal network patterns for the relevant network,

are also included in the model.

Consequently, model instances are strongly associated with domains they repre-

sent. An application domain does prescribe the model architecture and validation

principles and rules.

2.2 Basic Concept of the XNetMod Language

Using the natural structure of the XML grammar, the model structure as it is shown in

Fig. 1 could easily be transformed into a linguistic object with four main semantic

parts: a) a set of functional elements in a network (

"node"-elements) – nodes in a

graphical representation of the model; b) a set of networks (

"net"-elements) with

their connections (

"pin"-elements) – edges in graphical representation of the model;

c) a set of verification rules (

"rule"-elements) which may define legal and illegal

network patterns for the relevant network; and d) a set of attributes (

"attr"-

elements) related to node elements, network definitions, and connection descriptions.

The network modelling language defines the structure – topology – of the technical

facility or process and provides the highest abstraction level for the description of the

process functional relations.

As mentioned above, an application domain provides an impact on syntactic and

semantic aspects of a modelling language, and, of course, defines substantially func-

tionality of associated interpreting tools. Thus, the specification of the

"node"-

element is given as

abstract and was separately extended for specific application

fields. From this point of view we can speak about a language family. Such semantic

aspects as validity intervals for model attributes can be treated only in connection

with a chosen application domain.

Taking into consideration this dependency of the language on a application do-

main, the decision was taken to concentrate the semantic/syntactic structure of the

language on the representation and validation of network properties in general and

connection features in local.

Consequently, the structure of the XNetMod language was developed in order to

support the development of network interpreters. For this reason, the definition of the

model has three parts – the definition of data model, the definition of relational

model, and the definition of verification model. The data model component provides

the interpreter with context information. The relational model – definition of model

topology – provides the interpreter with information for construction of algebraic

terms. The verification model possesses information for the term interpretation. Con-

sidering the activities related to modelling of ontologies (http://www.w3c.org/), we

would like to emphasise, that the developers of the XML language followed a similar

strategy in defining logical relations between the data entities. There, the separation

of data model and logical relations allowed modelling of complex semantics. The

interpretation of the semantic was also done by an interpreter – inference machine.

2.3 Model Verification Approach

Obviously, the system topology correctness is crucial for adequate processing of

simulation-based tasks. Therefore, special efforts were made in order to elaborate

proper methods for the network structure validation. The approach developed in-

cludes two mechanisms: a) the use of an XML specific tool – XML Schema – for the

verification of model elements; and b) the use of rules, which define allowed model

patterns, for verification of model structure.

For the XML-Schema-based mechanism, the definition of sophisticated XML

schemas for model components and relation between them must be provided. The

verification procedure can be implemented by means of available XML parsers, for

example by Xerces of the Apache Software Foundation (http://www.apache.org/).

It is not possible to verify the semantic of XML documents based only on the

XML Schema functionality. The core idea of this technique is the application of alge-

braic and pattern based methods for the model verification. The use of network alge-

bra for mapping of a network topology into a proper algebraic term can be found in

[4]. Two model check tasks were considered: a) model configuration check – check

of an attribute appearance and values, and b) network topology check – identification

of non-valid connections.

The formal description of network configuration check can be provided using the

set algebra. Let us consider a network G defined by sets of nodes N and links L :

G :=

〈N, L 〉, N := {N

, …, N

}, L := {L

, …, L

(1)

The nodes N

, i=1(1)|N|, possess type attribute t

and simulation attributes a

:= 〈t

, {a

, …, a

i|N|

}〉, t

∈T, T := {T

, …, T

}, a

:= 〈b

, v

〉.

(2)

Here b

∈B and v

∈IR are attribute name and value. Sets T :={T

, …, T

} and B :=

, …, b

|B|

} collect application field dependent node types and attribute names. To-

pological structure of the network is given by links L ⊂ N×N . Nodes N

, N

∈N are

connected if ∃ L∈L, L=〈N

, N

〉. Additionally, a set of verification rules is also given

P ⊂ T × T × O × {0,1}, O :={>

⊗, ⊗<, ⊕}.

(3)

Configuration check: The validator verifies if attributes of nodes (2), which are in

direct connection, mach with patterns defined by rules (3). Here, attribute names,

types, values and other defined characteristics (e.g. physical units) can be addressed.

Topology check: Topology verification was realized using an algebra-based ap-

proach. Abstract operations "apply left" (>⊗), "apply right" (⊗<), "join" (⊕), etc.,

were defined for mapping of a network graph (1) in a proper algebraic term. The

verification is provided by a specialised interpreter which is able to interpret the alge-

braic term symbolically or numerically with respect to rules and operations (3).

3. Modelling Process Example

The process of a model building can be divided into two activities: definition of a

model on semantic level and its formalization using an interpretable modelling lan-

guage. The language chosen for the semantic was UML. The process shown is sup-

ported by a real application domain: modelling of forest fire extinguishing tasks.

The importance of forest fire prevention, extinguishing, and management is well

known, especially in the Mediterranean region. In case of forest fire, there are re-

sources that must be supervised: human, land and air resources. Every kind of re-

source has its own properties, and, in case of fire, is implemented in a different way.

A simulation based training system for forest fire officers was developed in frame-

work of

XnetMod research project (see the Acknowledgements).

node

resources

human equipment

vehicle aircraft

location

observer

office airbase

water population

crossroad

Fig. 2. Resource and location node hierarchy diagram

Cr1

Cr3 Cr4

Cr2

Ab1P1

Cr5

A1V1

Cr6

Fig. 3. Road and resources network example

The modelling of the system requires three main types of nodes – resources, loca-

tions, and crossroads. Fig. 2 models the relationship among the classes. Crossroads

connected by roads and provides the main system network. Resources and location

nodes related to the road network. In Fig. 3, a road network example with resources

and location nodes is shown. Here, Cr1 … Cr6 are crossroads, P1 and P2 are popula-

tions, Ab1 is an airbase, A1 is an aircraft, V1 is a vehicle. The figure depicts also

allocation links: P1 to Cr2, Ab1 to Cr4, V1 to P1, and A1 to Ab1. The following veri-

fication rules for this system can be formulated: a) a population must be linked to a

crossroad, b) vehicle must be linked to a road or crossroad, and c) aircraft must be

linked to an airbase.

Once the application domain model was defined in terms of UML, it was specified

using XnetMod language. Some code fragment of an XnetMod document is given:

</Node>

<From>Node:Population</From><To>Node:Crossroad</To>

</Rule></Net></Model>

Conclusion

A language and process for producing verified model schemas was presented for

systems that can be described as networks structures, that is, systems made of interre-

lated components. The developed language could serve as the simulation basis for

such application related issues as supervision, control, and decision-making. XML

Schema technique was used to specify the language. One of the reasons for this was

the number of available facilities to deal with XML structures. The language defined

on an abstract level with adaptation possibility with respect to a relevant application

domain. The approach has been tested with such various domains as forest fire extin-

guishing tasks (as described within this paper) and gas/water distribution automation.

Acknowledgements

This work is part of the XnetMod (XML Based Modelling Language for Simulation of Techni-

cal Networks ) project (CRAFT IST-2001-52057) partially funded by the European Commis-

sion. Contractors of the project are: ELPRO Prozessindustrie und Energieanlagen GmbH (co-

ordinator) (Germany), SIMPOWER Simulator Systeme GmbH (Germany), Investigacion y

Programas S.A. (Spain), ACEBO, C.B. (Spain), and RTD performers: Fraunhofer Institute for

Production Systems and Design Technology (Germany) and Universidad Politécnica de Ma-

drid (Spain).

References

1. Broy, M.: Object-oriented programming and software development – critical assessment. In

Programming Methodology. Springer-Verlag. Berlin Heidelberg New York (2003) 211-222

2. Fritzson P. Principles of Object-Oriented Modelling and Simulation with Modelica 2.1

Wiley-IEEE Press (2004)

3. Abdullah M.S, Benest I., Evans A., And Kimble C.: Knowledge Modelling Techniques for

Developing Knowledge Management Systems. In Proceedings of the 3rd European Confer-

ence on Knowledge Management. Dublin, Ireland, September 2002 (2002) 15-25

4. Stefanescu G.: Network Algebra. Springer-Verlag. Berlin Heidelberg New York (2000)