Simulation and Model-Based Methodologies: An Integrative View

The aim of this chapter is to explore the possibilities to place simulation in a central position for several scientific disciplines. The following topics are discussed:

In recent years considerable progress has been made in the field of establishing and verifying the theoretical foundations of modelling and simulation. This would most naturally lead the naive observer to expect that equivalent — relevant — progress has been made in the field of unequivocally establishing unique rules for model construction and criteria for evaluating the validity of models with reference to reality.

This chapter can be considered as an attempt both to unify the different concepts which are used in this field and to demistify some exaggerated expectations.

In the first place a comparison of the terminology used in some important chapters that follow in the first section of this book is presented.

Next the issue of “top-down” or “bottom-up” modelling is addressed and some problems with reference to the supposed antagonism between hypothesis verification and feature extraction will be laid to rest.

In another paragraph the question of what can be modelled shall briefly be handled from a more philosophical viewpoint.

In order that the reader would be able to evaluate the implications of what follows, the theoretical material that supports modelling in a positive sense will be reviewed as pieces of evidence, specifically for the validation problem.

Then a framework will be presented which links together the concepts and techniques of model building with concepts and techniques of experimentation and thus might be useful for further clarification of the issue of valid representation of systems.

Following this, three aspect-selection/decomposition paradigms for modelling are confronted with three prototypes of systems and their relation to the valid modelling problem is discussed. The chapter ends by a conclusion which attempts to outline the practical consequences for the model builder.

A conceptual framework for systems problem solving, referred to as general systems problem solver (GSPS), is described. Postulational and discovery approaches to systems modelling are characterized in terms of this framework. The discovery approach, referred to as inductive modelling, is then described in more details.

Considering a model to be a system specification, this chapter reviews the hierarchy of levels at which a system can be specified and the formalisms in which the specification can be done. Such an approach provides a unification of modelling and simulation concepts along both behavior-structure and discrete-continuous lines. Throughout its exposition we point out the utility of the framework for addressing such issues as correctness of simulation programs, valid simplification of models, transformation of models from one formalism to another, and attaining higher levels of model validity.

In this paper is described a comnletely rigorous mathematical theory of system design in which system function analysis, requirements analysis, implementability, modelling and simulation all find their appropriate and completely logical roles.

In this text a — birds’ eye — overview of the relation between modelling and policy construction is presented.

After discussing the rationale for the need of such a relation, the main elements and aspects of policy construction are reviewed in paragraph 1.

Hence, policy evaluation, planning, conflict-resolution/consensus-generation, organizational models and participation are highlighted from a system-oriented, structured-modelling viewpoint in subsequent sections of this chapter.

In paragraph 7 an example, that brings most of these principles together, illustrates them in a practical context.

Finally a note on tools, endeavouring to relate what is needed to what is available, sets out tasks for future research into modelling and model-utilisation utensils.

Model-based simulation support systems aim to provide comprehensive and integrated support of the activities comprising the modelling and simulation enterprise. This chapter discusses several formal structures underlying the design of such support systems including the composition tree, system entity structure, and experimental frame. An example based on Elzas¹ model-based negotiation methodology is employed to illustrate the system-theoretic concepts under discussion.

In this chapter we discuss a basic framework for the design of system-models which are well-structured for their implementation as a simulator. As a tool for it we discuss the concept of a stratified methodology which is built up by the knowledge of different systems specifications and transformations of such specifications into each other on different levels of abstraction. This tool is used to structure the design process to reach system specifications of network type.

Quite often there is a need to extend available qualitative and quantitative knowledge by applying experimentation on a real system, i.e. to influence the system, and to observe and interprete the resulting effect. System experimentation requires that control inputs and observation outputs are present, or can be made available with an appropriate degree of controllability and observability. Experiments can be done by means of a “measuring and control” system, what we will call further on a companion system or briefly cosystem ay. System and cosystem together form the experimentation system esy. An experimental run may yield a trajectory {u(t),y(t)} over some time interval of a control input signal u(t) and the corresponding observation output signal y(t).

GEST is the first model and simulation specification language. Specifications of the model and the experiment are totally separated. The modelling world view is based on the axiomatic system theory of Wymore which provides an excellent basis for simulation modelling and symbolic model processing. This chapter has two aims: 1) To present the GEST language and the robust and rich modelling paradigm it provides even for non-simulation application areas, as well as 2) to foster design and development of other GEST-like modelling and simulation languages which would provide other modelling formalisms within comprehensive modelling and simulation systems.

Continuous models, based on ordinary and partial differential equations, are widely used in the simulation of physical processes. Such processes often involve phenomena which are conveniently modelled by discontinuous change processes (switching, limiting etc.). The basic technique for processing models of this kind on a digital computer is a numerical integration technique which advances the solution of the set of differential equations in a step by step manner. Particular care must be taken when discontinuities are present to ensure that the integration proceeds through the point of discontinuity efficiently and accurately. The use of models of this kind is sufficiently widespread for special simulation languages to be developed to simplify their construction. Simulation languages need to be designed so as to deliver to the modeller accurate numerical solutions of the model equations as simply as possible and also to provide constructs and procedures which facilitate the construction of models and the design of experiments to be performed upon them.

The theory and techniques of discrete event modelling and simulation have advanced substantially over the past two decades. An integrative approach, making use of discrete event formalisms, should now be used when developing computer simulations. An important formalism is the DEVS model — a mathematical representation of the class of discrete event systems. Other formalisms, such as modelling strategies, provide a “world view” in which to conceptualize the simulation model.

In this chapter, the formalisms are first described. Next, detailed case studies of simulations within three problem domains are considered: (1) insect population dynamics; (2) nuclear waste management; and (3) computer communication networks. For each case study, the formalisms are shown as intimately intertwined in the model formulation and simulation development.

In the text the structure characterization problem is discussed. Trends in mathematical modelling and simulation are revealed. The utilization of mathematical representations in various scientific disciplines has created a need for new methodology addressing modelling problems that formerly were solved in standard deductive manners. The structure characterization problem is presented here as a hierarchical decision problem. Inductive methods are required in order to solve the problem mentioned. A framework for those inductive techniques is given. Pattern recognition is chosen as a viewpoint to analyze existing procedures and to suggest new approaches. The minimum variance technique, the AIC-procedure and an algebraic method are presented while an example is given belonging to the discipline of microbial fermentation.

The purpose of reconstructability analysis is to provide systems investigators with useful methodological tools for dealing with the various questions regarding the relationship between overall systems and their various subsystems. The term “system” is viewed in reconstructability analysis as a characterization of certain type of fuzzy measures by which the constraint among variables of interest is described.

Two complementary problems are involved in reconstructability analysis: (i) given an overall system, determine which sets of subsystems can be used to reconstruct it adequately (reconstruction problem); (ii) given a set of systems characterized by the same kind of measure, derive from it as much knowledge as possible regarding the unknown overall system (identification problem).

This chapter gives an overview of the main issues involved in reconstructability analysis and its current state of development.

The main intent of this paper is to demonstrate the use of systems methodology and systems theory in the light of its applicability. One way to accomplish this is by translating the ideas, concepts, theoretical algorithms and propositions into a useful tool. Such a tool can be implemented on a computer using software. The Systems Approach Problem Solver (SAPS) as a package is such an accomplished tool. We apply it here in the context of inductive modelling. We concern ourselves with the identification of data systems, behavioral systems of different forms, structural systems and metasysterns. For each, ample material from applications in recent consulting and research is provided.

The determination of optimal values for parameters is often an important aspect in both the formulation of mathematical models for systems and in their subsequent use in simulation studies. Optimization subproblems are therefore intimately associated with model-based studies. The objective of the paper is to explore some features of the interface between these two problem classes and to provide an overview of some of the numerical procedures that are available for solving such parameter optimization problems.

A basis for the taxonomy of concepts related with quality assurance in modelling and simulation is provided. It has three parts:

The comprehensive categorized list of related terms given can be the starting point

This chapter describes different means to improve the robustness of simulation software (languages, compilers, and run-time systems) with respect to products which are currently available on the software “market”.

It is shown how these improvements can help to ameliorate the robustness of models and of their coded counterparts: the simulation programs. Model robustness forms a part of the total validity picture, while simulation program robustness partly covers the correctness verification assurance.

This chapter addresses itself primarily to the simulation software designer. It is hoped that these considerations may help future software developers in producing more reliable simulation software.

This chapter presents the state of the art in model validation and some research topics in validation.

This paper considers the major issues involved in judging a model’s worth. The emphasis is on developing ways to judge the validity of model structure in terms of model purpose. Topics addressed include the dimensions of simplification, types of purpose, diagnosis and explanation, types of validity, types of structure, alternative modelling methodologies, a framework for evaluation, choice of system boundaries and simplification, and modelling objectives. The paper concludes with a few observations on the role and purposes of models in the study of socio-economic problems.

The application of computers to business administration has ranged from startling successes to appalling failures. The latter class includes the application of Management Information Systems (MIS) to support high level management decisions in an organization (Dearden 1972). The inadequacy of the MIS approach has been largely attributed to its philosophy of “management by computer,” which tends to exclude the manager from the decision process. The Decision Support Systems (DSS) approach differs from that of MIS in the change of philosophy from “management by computer” to “computer-aided management,” i.e., to provide computer aid for a manager to achieve appropriate decisions more effectively (Keen and Morton 1978, Sprague and Carlson 1982).

One particular simulation language may be better suited than another, or than any other, for the implementation of a simulation model because its “world view” more closely matches that of the applications are or modelling technique chosen; in fact, many simulation languages were designed to address a particular problem type, and their world view is, therefore, often restricted only to problems of this type. Simulation languages which include higher-level, general purpose programming facilities (possibly inherited from a host language) allow enrichment of the environment by addition of new objects and operations, and can be used to explore proposed modelling techniques and languages. SIMULA was intentionally designed to be extensible from within the language itself, and is ideally suited for prototyping of this sort.

An inductive scheme for systems modelling has been developed in this context of a comprehensive hierarchy of the epistemological levels of systems such as so called source systems, data systems, generative systems, structure systems, and so on (Cavallo and Klir 1978). According to the inductive direction in this systems hierarchy, we should start with a source system, a collection of the variables to characterize an object of our specific interest. A source system, together with its corresponding data, define a data system. For a data system, its corresponding generative system, local state trans- ition mechanism (data generator), is identified by mask analysis, which is an evaluation and selection procedure of local state transition rules based on the criterion of minimizing the conditional uncertainty in the determination of next states given previous states. In mask analysis, therefore, measures of uncertainty (information) are essential.

My topic is qualilty assurance. However, I believe the importance of that subject can best be appreciated if we first consider the modeling and simulation methodology in which it is embedded. So let’s back off from the trees, take a look at the forest, and again consider that oft-asked question—Why simulate?