Modelling and Simulation of Integrated Systems in Engineering

Issues of Methodology, Quality, Testing and Application

D J Murray-Smith

• Publisher: Woodhead Publishing
• Year: 2012
• Print ISBN: 9780857090782
• eBook ISBN: 9780857096050
• Topic: Negocios y empresa
• Subtopic: Inteligencia empresarial

This book places particular emphasis on issues of model quality and ideas of model testing and validation. Mathematical and computer-based models provide a foundation for explaining complex behaviour, decision-making, engineering design and for real-time simulators for research and training. Many engineering design techniques depend on suitable models, assessment of the adequacy of a given model for an intended application is therefore critically important. Generic model structures and dependable libraries of sub-models that can be applied repeatedly are increasingly important. Applications are drawn from the fields of mechanical, aeronautical and control engineering, and involve non-linear lumped-parameter models described by ordinary differential equations.- Focuses on issues of model quality and the suitability of a given model for a specific application- Multidisciplinary problems within engineering feature strongly in the applications- The development and testing of nonlinear dynamic models is given very strong emphasis

1

The principles of system modelling

Abstract:

This chapter provides an introductory review of processes involved in developing models for engineering applications. Current trends in the design of integrated, multidisciplinary and embedded systems, and the associated consequences for modelling and simulation, are emphasised. Ways of classifying models are presented and the experimental approach to model development through system identification and optimisation is introduced. Issues of model quality are emphasised, together with model reuse, the development of libraries of sub-models, the potential of generic descriptions and use of modelling within procurement. These topics are all considered in more detail in later chapters.

Key words

modelling objectives

classes of model

quality

reuse

1.1 General issues in the development and application of models

For scientific applications, the purpose of a model is usually to explain a complex set of behaviours or to help in the design of experiments as part of the process of hypothesis testing. In such fields, modelling is a central element of the scientific method. Similarly, in some engineering applications a model may be used to describe, analyse or explain the behaviour of a highly complex system, but it is more common to find models being used to support design, to assist in decision-making processes in the context of a specific application, or as a basis for simulators to be used in training or in further engineering developments. A properly tested and well-proven model can reduce development times and costs for many applications.

Integrated systems arise in many application areas including fly-by-wire aircraft, ships, land-based vehicles, energy conversion systems including electrical power generation and distribution systems, chemical plants and even in some household appliances. They typically involve a number of different aspects, disciplines or ‘domains’ (e.g. mechanical, electrical, electronic, control and software) which, ideally, are considered concurrently. For some applications, special forms of integrated system involve digital processors and software in addition to other physical hardware. The fields of aeronautical engineering, automatic control, road and rail vehicle engineering, marine engineering and robotics can provide many examples of such embedded systems.

The importance of integrated systems has been emphasised by the publication by the UK Royal Academy of Engineering (RAE) of a guide entitled ‘Creating Systems that Work – Principles of Engineering Systems for the 21st Century’ [1]. In the press release marking the publication of this guide in 2007, it is stated that the ‘… aim is to demystify the design of large integrated systems, and to give educators, students and practitioners alike six guiding principles that will help them to understand how large projects can be better conceived, designed and delivered’. These six principles can be summarised as:

1. debate, define, revise and pursue the purpose;

2. think holistic;

3. follow a disciplined procedure;

4. be creative;

5. take account of the people; and

6. manage the project and the relationships.

All sections of the RAE report make direct or indirect reference to the importance of appropriate tools for analysis, design and optimisation and the section dealing with the fourth of these principles (‘be creative’) puts special emphasis on modelling and simulation tools and methods. Simulation tools are vital for systems engineers in tackling the trade-offs within the design process, starting from the basic requirements in terms of performance, cost and timescale.

Models can have many benefits in the integrated systems approach to system design, including early assessment of performance, both within and beyond the normal operating envelope. Understanding of parameter inter-dependencies and knowledge of key sensitivities can also be very valuable for design optimisation. The use of simulation models is particularly important and leads to the concept of a virtual prototype which is a software-based implementation of the design, developed prior to any hardware prototype.

The success of virtual prototyping depends on the model quality. A successful model usually results from trade-offs involving several aspects of model performance such as the trade-off between the level of detail included in a model and the speed of solution in the corresponding computer simulation.

The level of detail is linked to model performance and as models are made more detailed, they inevitably become more complex. However, model complexity should never be confused with model quality and a simple description can often be better, in terms of quality measures, than a more complex one. It is also important to separate the processes of modelling from simulation. The development of a computer simulation is one common outcome of a modelling exercise but there are other potential applications for a model, some involving analysis carried out independently of any computer.

Whatever the use made of a model, it is important that its development should build upon previous experience. Attention must be given both to tools available for the development of computer-based modelling and simulation programs, and also to support systems for model management. Reuse of model components is important and some commercial modelling and simulation systems provide libraries of reusable models. Model management is also very important for applications involving large teams of developers, especially when these include multidisciplinary groups and geographically distributed teams. New developments in cloud computing are likely to have a significant influence on how simulation models are used within many organisations in future, but this is an area in which rapid changes are taking place and it is not possible, at the time of writing, to make more detailed predictions.

Developing a model requires careful examination of information about the real system. From this, inconsistencies or gaps in the available knowledge can be found which may result in further testing of the real system or a prototype, or some reconsideration of requirements. Donald Rumsfeld’s much-quoted statement, made during a US Department of Defense news briefing in February 2002 [2], has direct relevance to issues of model quality and uncertainties:

… as we know, there are known knowns; there are things we know we know. We also know there are known unknowns; that is to say we know there are some things we do not know. But there are also unknown unknowns – the ones we don’t know we don’t know.

His statement was much ridiculed at the time but those words certainly apply to the processes of developing models. The ‘unknown unknowns’ in modelling are vitally important and have to be exposed by whatever means possible, including experimentation and testing.

Since a model is only an abstraction of the system it represents, perfect accuracy is impossible. The key issue is one of determining the model quality levels needed for the application in question. This implies reducing errors to defined levels for specified regions of the operating envelope of the system and balancing appropriate measures of accuracy against other measures of performance, such as solution speed.

In applications involving design, it is usual to base the structure of models on prior physical, chemical or biological knowledge. However, some sub-models may be based purely on input-output descriptions derived from tests on the corresponding elements of the real system (i.e. ‘black box’ models). Models thus range from completely ‘transparent’ descriptions, based on the application of recognised and accepted scientific or engineering principles, to purely empirical ‘black box’ forms, which are opaque. Between these extremes there is an important group, sometimes referred to as ‘grey box’ models, involving some empirical information found experimentally but with the structure of the model based on well-established physical laws and principles.

In summary, therefore, it can be said that mathematical modelling is an important tool for decision making and for engineering design. If models are developed correctly, they can then be applied over the range of conditions for which the descriptions are held to be accurate representations of the real system. An extensive programme of system testing and the creation of a solid base of experimental data are important steps in establishing how a given model may be used. When used within the concurrent approach to design, well-proven and tested models can lead to useful virtual prototype systems based on simulation software or to hardware-in-the-loop simulation involving a combination of simulation models and real system hardware.

1.2 Classes of model for engineering applications

Models used in science and engineering often involve variables that are continuous functions of time, such as position, velocity, acceleration, temperature or pressure. These are continuous-variable models and involve ordinary or partial differential equations or differential-algebraic equations. This is the main class of model considered in this book.

A second type of model that can be important in engineering involves what are known as discrete-event descriptions. In discrete-event models, all the variables remain constant between events that mark changes in the model. These changes take place at discrete time instants, either periodically or in a random fashion. A digital processor or computer used for real-time control is a good engineering example of a discrete system involving periodic changes. In this case, a continuous variable is sampled periodically through an analogue to digital converter and calculations are carried out using the discrete values obtained from the converter. Output from the processor may be converted back to continuous variable form using a digital to analogue converter. In modelling this type of component within some larger engineering system, we cannot use differential equations because of the discrete nature of the events within the processor and the associated converters and a difference-equation based approach is necessary.

Problems in which events occur in a more random fashion, as in road traffic flow or communications network modelling, lead to another approach known as discrete-event simulation. This is an important area, especially in engineering manufacturing and production, but this book is more concerned with modelling and simulation of continuous systems. Hybrid systems involving representations that are mainly continuous but do involve some discrete-event elements are discussed. Further details of discrete-event modelling and simulation techniques and their applications may be found in texts that deal with this area (see e.g. [3], [4]).

1.2.1 Conventional continuous-variable models

Within the class of continuous variable models we can distinguish between models of data and physically based models of systems.

A model of data involves a description fitted to measured responses, usually from a real physical system, leading to a model that expresses an observed relationship between two or more variables. It c...

Table of contents

• Cover image
• Title page
• Table of Contents
• Copyright
• List of figures
• List of tables
• List of abbreviations
• Acknowledgements
• Preface
• About the author
• Chapter 1: The principles of system modelling
• Chapter 2: Integrated systems and their significance for system modelling
• Chapter 3: Problem organisation
• Chapter 4: Inverse simulation for system modelling and design
• Chapter 5: Methods and applications of parameter sensitivity analysis
• Chapter 6: Experimental modelling: system identification, parameter estimation and model optimisation techniques
• Chapter 7: Issues of model quality and the validation of dynamic models
• Chapter 8: Real-time simulation, virtual prototyping and partial-system testing
• Chapter 9: Model management
• Chapter 10: Further discussion
• Appendix A1: models of an unmanned underwater vehicle (UUV)
• Appendix A2: numerical methods for the solution of ordinary differential equations
• Index