Simulation Model Validation based on Ontological Engineering
Methods
Elena Zamyatina
1a
, Denis Churin
1
, Viacheslav Lanin
1b
, Lyudmila Lyadova
1c
and Nada Matta
2d
1
Department of Information Technologies in Business, HSE University, Perm, Russian Federation
2
University of Technology of Troyes, Troyes, France
Keywords: Simulation Model, Verification, Validation, Ontology Matching.
Abstract: A task of the simulation models examination (verification and validation, V&V) is considered. At the V&V
process the correspondence degree of the simulation model created by developers to the simulated object, that
description is presented in the form of a conceptual model built by customers, is determined. An ontological
approach is proposed to determine the semantic proximity of the simulation model and the conceptual model,
whose descriptions are presented in the form of ontologies. Matching rules can also be defined with ontology
based on the metrics chosen by the customer. The approach has been tested using the simulation system Triads.
The results of the matching algorithm execution are illustrated by an example. The article provides description
of the simulation model ontology created in TriadNS and conceptual model ontology, developed with MASK
method. The metrics used for proximity assessment are described.
1 INTRODUCTION
It is well-known that there are a great number of
publications devoted to the examination of the
simulation models built by researchers to solve tasks
of designing or forecasting, etc. (Sargent, 2017;
Sargent, 2007; Balchi, 2004). Examination involves
the implementation of such stages as verification and
validation (V&V) of a simulation model. It is
necessary to prove that this model can be trusted that
the quality of information obtained because of a
simulation experiment corresponds to the level that
allows making the right decision.
Before performing simulation experiments,
customers can build a conceptual model, representing
the object (system, process) being simulated, as “a
reference model. To develop a conceptual model, a
suitable environment can be used by customers that
does not require programming skills, the use of
simulation tools. So, verification comes down to
control of the correctness of the transfer of the
a
https://orcid.org/0000-0001-8123-5984
b
https://orcid.org/0000-0002-0650-2314
c
https://orcid.org/0000-0001-5643-747X
d
https://orcid.org/0000-0001-8729-3624
conceptual model, developed by the researcher, into
the simulation environment (Sargent, 2017).
Validation can be defined as examination of the
conceptual model presentation correctness and
checking behavior correctness. Verification is a part
of validation. Verification is carried out during the
construction of the model, while validation is carried
out immediately after the completion of the creation,
the description of the model in a specific specialized
programming language. However, in practice, as a
rule, the processes of verification, validation, as well
as testing and implementation of the model overlap in
time. Along with the concepts of validation and
verification, the literature provides such a concept as
accreditation (VV&A Validation, Verification and
Accreditation). Accreditation may be defined as the
formal certification that a model, simulation, or
combination of models and simulations is acceptable
for use for a specific purpose (Sargent, 2007).
Accreditation is the official certificate of the
customer, which states that the simulation model is
applicable for solving a concrete problem.
Zamyatina, E., Churin, D., Lanin, V., Lyadova, L. and Matta, N.
Simulation Model Validation based on Ontological Engineer ing Methods.
DOI: 10.5220/0011589000003335
In Proceedings of the 14th International Joint Conference on Knowledge Discovery, Knowledge Engineering and Knowledge Management (IC3K 2022) - Volume 2: KEOD, pages 237-244
ISBN: 978-989-758-614-9; ISSN: 2184-3228
Copyright
c
2022 by SCITEPRESS Science and Technology Publications, Lda. All rights reserved
237
Verification and validation should answer the
question of whether the model is sufficiently
accurate. In this case, the purpose of the developed
simulation model should be taken into account. For
example, most of the demonstration models are not
highly accurate, but it is worth remembering that their
purpose is to demonstrate the operation of the
process, object or system under study only in general
terms. If the model copes with the goal set for it, then
we can talk about its compliance with the prototype,
even though the model's accuracy is low. Therefore,
the purpose of the model must be known before
validating it. The customer must develop validation
rules that are appropriate for the purpose.
Thus, the task is to develop a flexible validation
tool that allows not only to create models, but also to
adjust requirements for specific purposes of model
examination.
This paper presents approach for developing
software tools implementing an ontological approach
to verification and validation of simulation model.
This software tools includes portal for the customers
and modelers interaction. Customers must submit a
conceptual model ontology (O1) and simulation
model developers also submit their version of the
conceptual model ontology (O2). Using an algorithm
for determining the proximity of ontologies, a
comparison of two ontologies is performed and
inconsistencies in them are revealed.
System TriadNS is chosen as the simulation tools
for performing experiments. This system uses
ontology to customize simulation models to specifics
of domains. It is proposed to extend the purpose of
the ontology for solving model validation tasks.
2 RELATED WORKS
It should be noted that there are different forms of
validation:
validation of the conceptual model (checking
the degree of detail of the model);
data validation (checking the accuracy of the
data);
white-box validation (detailed (micro)
validation of the model, determining the
accuracy of the model components);
black box validation (general (macro)
verification of model performance, in which it is
determined whether the general model provides
a sufficiently accurate representation of the real
world).
It is well-known that nowadays many simulation
systems use ontologies for simulation model design
(Jain, 2016). It allows to integrate simulation models
(Benjamin, 2006), to adjust the system to a specific
subject area quickly and flexibly, to determine one or
another mathematical abstraction of simulation
model (Queue theory, Petri Nets and so on).
Using the capabilities of the TriadNS system to
use ontologies, it is proposed to implement validation
tools based on the scheme (Balchi, 2004) shown in
Figure 1.
Figure 1: The scheme of verification and validation of
simulation model.
Any process can be viewed from different angles
and points of view. For example, the process of
customer service from the manager's point of view
looks like a sequence: call the customer, find out the
need, meet the need, complete the service. While the
same process for the customer will look like a
sequence: choose the right ticket, take a ticket, stand
in line, come to the service, voice the need, get
service, complete the service. Accordingly, different
viewpoints influence what will be considered model
validity and how goals are achieved. Different
perspectives do not make a model invalid; they only
reflect different interpretations of the process.
Situations of developing a model from different
viewpoints, as in the situation of setting multiple
goals, will require a lot of resources because of the
high level of detail. Therefore, one model may be
valid for one viewpoint and completely invalid for
another viewpoint.
The model validation process may involve many
tests using data. For example, it may examine what
results the model will produce when real data arrive.
A model run on the same data as the object of its
representation is expected to behave similarly to the
real-world object (process or system). In a fairly
simple experiment, a number of problems may arise:
1. Lack of real data. The experimenter may not
have real data to start the simulation process. For
example, when simulating queues to ATMs,
KEOD 2022 - 14th International Conference on Knowledge Engineering and Ontology Development
238
information on the average customer flow to bank
branches may not be collected. Expert knowledge can
be used as a solution to this problem.
2. Inaccuracy of real-world data. Due to
inaccurate data, model results are not reliable and not
suitable for research. That is why data validation is
considered as one of the key elements of simulation
model building life cycle.
3. Unrepresentative data sampling. If the
experimenter obtains real world data, such as data
from an electronic queuing system, it does not mean
that it is representative. For example, the analyst has
data on customer flows to ATMs during the holiday
season. However, this sample is not representative,
and the conclusions are statistically insignificant
because the weekday clients flow is very different
from the holiday data. Therefore, for current data the
model will show results comparable only to the days
for which these data are collected. To prevent the
problem in question, you should use static tests when
validating the data, but the result of the model will
have to be seen as the probability of the event
occurring. Critical values can be passed to the model
to test it and assess the adequacy of the conclusions.
As mentioned before, the process of validation
and verification should take place at each stage of
model building and should be repeated if the model
undergoes changes. Given the complexity of models
and the increasing cycle of its construction, validation
and verification may require a long period of time,
during which the process in the real world may also
change. For example, during the validation of the
final computer model of the customer service process,
in the real-world service standards have changed and
the model under study is no longer relevant. To
minimize the impact of this problem, the analyst
should calculate the time and ensure that the
validation and verification as often as possible while
evaluating the model as a whole and its elements.
Verification and validation methods are described
in more detail in the work of the American researcher
Osman Balchi (Balchi, 2004). One is to test the
construction of a conceptual simulation model.
The developed conceptual model describes those
components of the real-world system that should be
included in the model (and those that should be
excluded from the final model), and is expressed
either formally (for example, using activity cycle
diagrams) or informally (for example, in the form list
of items).
To create a conceptual model, developers must
analyze all the information received and come to an
optimal solution. A project specification or terms of
reference can be used to validate the conceptual
model. In addition, it is necessary to get estimates
from experts who are versed in the subject area and
similar systems on the compliance of the conceptual
model with the requirements described in the
documentation.
So, then the stage of verification and validation
should be carried out jointly, both by the developers
of the model and by the customers who need to solve
a specific problem.
As described earlier, the model verification and
validation processes affect all stages of the
development of a simulation model. The process of
checking the adequacy and accuracy of the simulation
model includes:
structural testing (structural validity
determining the correspondence of the structure
of the simulation model, the list of objects and
their interrelationships to the researcher's
intentions);
testing the functions of the model;
testing the behavior of the model (operational
validity checking the correspondence of the
functionality of the simulation model to the
concepts of the researcher).
In addition, simulation model validation should be
performed at every stage of simulation model design.
If an error is found in the simulation model, it is
necessary to return to the previous stage of model
verification to identify the discrepancy between the
constructed model and the customer's intentions.
A comparison of two ontologies is performed and
inconsistencies in them are revealed.
This article focuses on the structural validity only.
Verification is performed by comparing the
ontologies representing the conceptual models
received from customers and from modelers
designing simulation models.
3 SIMULATION MODELING IN
TriadNS SYSTEM
TriadNS is a simulation system which was developed
based on the simulation system Triad. Triad is
intended for the design and simulation of computer
systems. Simulation models are described using
special language named “Triad”.
It has a three-layer representation of the
simulation model: M = (STR, ROUT, MES), where
STR is the structure layer, ROUT is the routine layer,
MES is the message layer.
A layer of structures is a collection of objects
interacting with each other by sending messages.
Simulation Model Validation based on Ontological Engineering Methods
239
Each object has poles (input and output poles), which
serve, respectively, for receiving and transmitting
messages. The basis of the representation of the layer
of structures is a graph. Separate objects should be
considered as the nodes of a graph. The arcs of the
graph define the relationships between objects. The
structure layer is described as a parameterized
procedure and allows to flexibly change the number
of nodes in the graph, etc. One may change the input
parameters both before the start and during the
simulation experiment. In the second case, the
override is performed in the special unit of simulation
model description: “the conditions of simulation.
Model objects act according to a specific
scenario, which is described by routine”. The state
of the object is determined by the values of the
variables of the routine. The simulation system
TriadNS is event-driven. Events are described in
routines. A routine, like an object, has input and
output poles. The input poles are used for receiving
messages, and the output poles serve for transferring
messages. Message receiving is an event.
The message layer is intended to describe
messages of complex structure (it may be a program,
for example).
The collecting information on the simulation
experiment is carried out with information
procedures. Information procedures work like
sensors and monitor the changes in the values of
variables, the arrival and sending of messages, and
the execution of events. The list of information
procedures is specified in a special program unit – the
conditions of simulation. The conditions of
simulation determine the initial conditions during the
simulation experiment, the moment of completion of
the simulation and determine the algorithm for the
study of the simulation model. These tools can be
used to validate model checking in simulation
experiments, in dynamics. But in this paper, the focus
is only on static structural characteristics.
4 TriadNS SIMULATION SYSTEM
ARCHITECTURE
Modelers can use different tools to develop models in
the TriadNS system. They can create models with text
editor using Triad language to describe structure,
routines, and other elements of the model. Modelers
can use a graphical editor to create visual models of
the simulated systems (the developed models are
transformed into a textual representation in the Triad
language and translated into executable code to
perform simulation experiments). Created models
and components of models (structures, routines,
information procedures, modeling conditions) can be
stored in a model repository and reused for new
experiments. The generalized structure of the
TriadNS system repository is shown in Figure 2.
Figure 2: The generalized structure of the TriadNS system.
KEOD 2022 - 14th International Conference on Knowledge Engineering and Ontology Development
240
The ontology is the core of the system repository
(Zamyatina, 2018). The ontological representation of
the simulation model has a multilevel structure. The
upper level (the level of general concepts) is an
ontology that describes general concepts for various
mechanisms of the simulation system. The second
level of the ontology is the level of concepts
represented in domain ontologies. Ontologies of this
level include concepts specific for systems and
processes related to different domains. This level
allows to set the semantics of the models, facilitate
their understanding by experts in the relevant research
areas, and the interpretation of the results of
simulation experiments. The third level is the level of
simulation models. Models and their components are
described at this level. Model representations can be
associated with specific areas described by domain
ontologies. Model components are associated with
domain concepts. The lower level of the system
ontology (rule level) contains descriptions of specific
rules that supplement the descriptions available at the
previous level. This layer specifies the design rules,
the basic requirements that the simulation model must
meet.
Ontologies of conceptual models created by
customers are used to solve simulation models
examination tasks. The rules that are used for model
comparison are also be represented in the ontology as
descriptions that can be defined for specific modeling
projects. The ontology of a conceptual model can be
built automatically based on the model created by
developers with using a visual model editor or with a
text model editor (Figure 3). The customer's
conceptual model ontology (Figure 4) can be created
with an ontology editor.
5 ONTOLOGY CREATION
METHOD
The main problem of applying the proposed model
examination method is the development of an
ontology by customers who do not have the skills of
knowledge engineers.
To translate the requirements and logic of the
conceptual model into a machine-readable form of
ontology, it is necessary to resort to knowledge
extraction tools (Matta, 2002). At the moment, two
approaches to knowledge extraction are widely
known:
REX is a method of direct knowledge extraction
using Data Mining and Text Mining tools from
regulations and documents.
MASK is a knowledge gathering method based
on interviewing experts.
One of the significant advantages of the MASK
method over a similar one is a significant saving of
time when interviewing an expert, analyzing, and
modeling, due to joint work with an expert and using
his experience in a specific subject area.
Within the framework of this work, to build the
ontology of the conceptual model, the MASK method
is mainly used to accurately determine the
requirements and vision of the network model by an
expert. MASK knowledge gathering helps the expert
to focus on his subject area to describe it by
emphasizing its main characteristics.
The list of questions is determined by the
specifics of the modeling domain. In this case when
interviewing an expert to create a conceptual model
ontology, he was asked the following questions:
1. What network elements are used to build the
model?
2. How many network elements of each type are
used in the model?
3. What are the characteristics of the components
of a model?
4. What are the relationships between the
components of the model?
5. What is the routing algorithm for the model
nodes?
6. What are the parameters of the routing
algorithm in this model?
7. What are the conditions of simulation
experiment?
8. What is the limit time of simulation?
Depending on the answers of experts, the list of
questions is detailed, the answers are specified. For
example, for the of the routing algorithm the
following parameters were obtained:
two parameters (ST11 and ST12) affect the
message processing time (T1+Random(T2));
two parameters (ST21 and ST22) affect the
frequency of sending messages to nodes;
parameter SQueueLen defines the buffer size;
parameter SBBCon defines whether there is a
broadband connection;
parameter STFlops specifies the performance or
computing power of the node, measured in flops
(the number of floating-point operations per
second);
and so on.
One may see these parameters at Figure 4 where
the conceptual model ontology of a customer is
shown (the basic ontology enriched by subclasses for
SBARC algorithm).
Simulation Model Validation based on Ontological Engineering Methods
241
Figure 3: The basic ontology enriched by subclasses for SBARC algorithm (the conceptual model ontology of a developer).
Figure 4: The basic ontology enriched by subclasses for SBARC algorithm (the conceptual model ontology of a customer).
6 STRUCTURAL VALIDATION IN
TRIADNS SYSTEM
The ontological approach is actively used in the
TriadNS simulation system (Zamyatina, 2018).
Ontologies are used to customize model to the
domain specifics. TriadNS system has a basic
ontology consisting of basic classes TriadEntity (any
named logic entities), Model (simulation model),
ModelElement (a part of simulation model and all the
specific characteristics of a node of structure layer),
Routine (node behavior), Message and so on.
Let us consider, as an example, illustrating the
possibilities of the ontological approach to model
checking, the SDN network (Software-Defined
Network). A data transferring control level is
separated from the data transferring devices, and it is
implemented in software (Zamyatina, 2017).
The basic ontology of the TriadNS simulation
system was extended to model SDN. The modelers
KEOD 2022 - 14th International Conference on Knowledge Engineering and Ontology Development
242
got the opportunity to operate with such objects as a
router, workstation, super-node, node, router etc.
For the experiment, the SBARC algorithm is
simulated (Zhiyong, 2008). Using the graphical editor
of the TriadNS system, a network structure was
created, including six nodes and connections between
them. To implement the SBARC algorithm in model,
it is necessary to add new elements into the base
ontology. A subclass SDNNode is to be added to class
“Nodes”. All elements of the network simulation
model (nodes and connections) are described in the
ontology O
1
(Figure 3). New class is created for each
node of the network simulation model (Node_1,
Node_2, Node_3, Node_4, Node_5, Node_6). This
was done due to the fact that each node in the network
has its own unique set of attributes and links to other
nodes. These relationships are described as properties
of objects. The onto-graph for the conceptual model
created by customer (O
2
) is shown in Figure 4.
The final stage of the research is the comparison
(determination of the proximity measure) of the
ontologies built by the developer and the customer
(checking the conformity of the logic of the conceptual
model). The model built by the developer is valid if it
meets the expectations and requirements of the expert
set out in the conceptual model. The differences in the
structure of ontologies would point specifically to the
part of the model in which the error was made.
Ontology comparison rules are based on the
approach described in the paper (Ngom, 2018).
As an example, a comparison method is
implemented that includes the following steps:
1. Defining sets of classes: O
1
\ O
2
a set of
classes represented in the O
1
ontology and not
represented in the O
2
ontology; O
2
\ O
1
– a set
of classes represented in the O
2
ontology and
not represented in the O
1
ontology; O
1
O
2
a set of classes, presented in both ontologies.
2. Evaluating the semantic proximity measure
between the concepts of each set.
3. Increasing the O
1
and O
2
ontologies using the
set O
1
O
2
. At this step, for each class X from
O
1
O
2
(X O
1
and X O
2
) a search is made
for their descendants in O
1
(or in O
2
,
respectively), which are added as descendants
of the class X into O'
2
(or into O'
1
, respectively),
if they don’t exist in this ontology.
4. Defining set of classes O'
1
O'
2
, which are a
set of common concepts of the ontologies O'
1
and O'
2
.
5. Assessment of similarity between ontologies.
The formula
is used to evaluate the ontologies proximity measure.
The following designations are used in the formula:
is the ontology O
1
integrity factor (n
2
is
the number of O
2
concepts added to extend O
1
);
is the ontology O
2
integrity factor (n
1
is
the number of O
1
concepts added to extend O
2
);
𝑆𝐼𝑀
is a function of the average value of the
ontology concepts similarity degrees determined with
the formula:
,
where c
3
is the closest common parent of c
1
and c
2
;
ϑ, ω, α and β are parameters that allow to consider
the value of proximity in relation to the number of
concepts of sets O'
1
and O'
2
and the number of
concepts of ontologies O
1
and O
2
:
ϑ =
__(

)
__(
) 

,
ω =
__(

)
__(
) 

,
α =
__(
\ 
)
__(
)
,
β =
__(
\ 
)
__(
)
.
Inconsistencies identified when determining the
semantic proximity of two ontologies are shown in
Figure 5. The proximity measure of the network
model ontology and the conceptual model ontology
according to the described method is 81%.
In an ideal situation, they should be identical, but,
in practice, this is practically unattainable due to the
possible features of special modelling environments.
The conformity equal to 100% is the ideal and desired
result when validating the simulation model.
To increase the flexibility of the system, the rules
for matching ontologies and determining the integral
proximity estimate can be specified by users in the
ontology of the system.
7 CONCLUSIONS
The developed method of the simulation model
structural validation based on ontologies comparison
was tested. The customer view and created simulation
model are described with ontologies. The fixed rules
for calculating the integral proximity score are used.
Experiments showed the practical significance of the
proposed approach.
Suggested approach and developed software tools
can be useful when developing complex simulation
models.
Simulation Model Validation based on Ontological Engineering Methods
243
Figure 5: Inconsistencies identified when determining the semantic proximity of two ontologies.
The next step is to increase the flexibility of the
approach via providing users with the ability to
develop their own metrics, to create rules for
comparing models, customizable to specific
modeling tasks and domains, with using ontology.
ACKNOWLEDGEMENTS
The reported study was funded by RFBR and the
Krasnodar region Administration, project number
19-47-230003.
REFERENCES
Sargent, R.G. (2015). An Introductory Tutorial On
Verification And Validation Of Simulation Models. In
Proceedings of the 2015 Winter Simulation Conference.
Pp.1729–1740
Sargent, R.G. (2007). Verification and Validation of
Simulation Models. In Proceedings of the 2007 Winter
Simulation Conference. Pp. 124–137.
Balchi, O. (2004). Quality Assessment, Verification, And
Validation Of Modeling And Simulation Applications.
In Proceedings of the 2004 Winter Simulation
Conference. Pp. 246–249.
Jain, A., Robinson, D., Dilkina, B., Fujimoto, R. (2016). An
Approach to Integrate Inter-Dependent Simulations
Using HLA With Applications To Sustainable Urban
Development. In Proceedings of the 2016 Winter
Simulation Conference. Pp. 1218–1229.
Benjamin, P., Patki, M., Mayer, R. (2006). Using
Ontologies for Simulation Modeling. In Proceedings of
the 2006 Winter Simulation Conference. Pp. 1161–
1167.
Robinson, S. (1997). Simulation Model Verification and
Validation: Increasing the Users’ Confidence. In
Proceedings of the 1997 Winter Simulation Conference.
Pp. 53–59.
Zamyatina, E., Mikov, A., Lanin, V. (2018) Automation of
Simulation Steps Using Ontological Approach. In
Proceedings of the 10th International Joint Conference
on Knowledge Discovery, Knowledge Engineering and
Knowledge Management (IC3K 2018). Pp. 223–230.
Zamyatina, E., Maltcev, G., Mikov, A. (2017) Simulation
of Routing Algorithm for SDN and SON Networks. In:
2017 IEEE 11th International Conference on
Application of Information and Communication
Technologies. Vol. 2. IEEE. Pp. 79–84.
Zhiyong, X., Yiming, H. (2008). SBARC: A Supernode
Based Peer-to-Peer File Sharing System. Department of
Electrical & Computer Engineering and Computer
Science University of Cincinnati. Cincinnati. Pp. 1–12.
Ngom, A., Kamara-Sangare, F. (2018). Assessing
Similarity Value between Two Ontologies». In IC3K
2018. Proceedings of the 10th International Joint
Conference on Knowledge Discovery. Pp. 343–350.
Matta, N., Ermine, J.L., Aubertin, G., Trivin J.Y. (2002).
Knowledge Capitalization with a Knowledge
Engineering Approach: The Mask Method. In
Knowledge Management and Organizational
Memories. Springer, Boston, MA. Pp. 17–28.
KEOD 2022 - 14th International Conference on Knowledge Engineering and Ontology Development
244