Simulation models are very popular in many areas of science. The ability to predict the behavior of a complex system is important in decision making in industry, government and in many fields of scientific research. For example, simulation models are used to help set the allowable emissions from nuclear power stations or to predict the changes in efficiency produced by design modifications to prototype engines.
Traditionally, the modeling of complex systems has been the domain of the highly skilled mathematical programmer requiring a large investment of time and resources and most often resulting in Fortran programs. The lack of an easier way to model complex systems has left scientists with smaller budgets and more modest needs with few options. This has effectively prevented the development of mathematical models in many areas of scientific research. For example, in the field human pathology, where there is a strong tradition of animal experimentation, relatively few mathematical models of physiological processes have been produced. Research has been held back through a lack of customizable models and cost-effective tools.
In recent years several software products have been developed which strip away the veils of mathematical complexity and provide the modeller with the tools necessary to design and construct numerical simulation models without having to write programs. The increased availability of powerful personal computers also means that scientists comfortable with spreadsheets and word-processors are now taking the opportunity to develop models for themselves. No longer burdened with mathematics or complex code, they can use their knowledge of the system they're modeling directly without needing to express the model in mathematical and then programming terms. Better yet, results can be obtained almost immediately.
Using Microsoft Windows or the Macintosh, these modeling packages make full use of the graphical user interface, providing drawing tools for placing and linking different model components to create a model diagram. These diagrams illustrate the structure of the model showing its various elements and the relationships between them. Each package has a different vocabulary of model components but they can all be used to represent systems of differential and non-differential equations as well as stochastic and event-driven systems. Once the model has been constructed and its values calculated, the results can be presented as tables, graphs or data and exported to other packages for further analysis or presentation, if necessary.
In addition to the simple calculation of model values, many of the modeling packages offer tools which allow the in-depth analysis of model characteristics. These tools range from model optimization where the model is compared with and adjusted to fit experimental data, through confidence interval calculation and sensitivity analysis to spectral analysis using fast Fourrier transforms. With these many powerful tools, one package can be used for the whole of an experimental modeling procedure industry.
By defining a user-friendly and open environment for the model development, these modelling packages also provide a powerful vehicle for scientific collaboration. For example, a typical biosphere model will contain sub-models to represent climate, plants, animals and soil water and nutrients. Each of these elements is a field of study in its own right. By using a modeling application, each sub-model can be developed by a different specialist, allowing the whole model to be constructed with ease from its constituent parts. Alternative sub-models for each element could be substituted easily as the need arises.
The use of numerical simulation modeling packages has four main benefits when compared to programming:
Without the need for programming skills and in-depth knowledge of mathematical techniques, even an inexperienced user can be productive immediately.
The graphic presentation of the model structure and elements makes the whole model accessible both to the original developer and to colleagues and collaborators who wish to adapt and expand it. It is much easier to perform 'what-if' analyses.
Many of the packages provide a suite of powerful tools to analyze the characteristics of the model and to aid the experimental modeling process.
Programming mathematical methods is beset with dangers. By using a modeling application, a wide range of tried and tested mathematical techniques become available.
Modeling tasks which used to take days or weeks can now be performed in hours or minutes.
--This article was contributed by Andrew Walker, developer of Modelmaker, published by Cherwell Scientific Publishing.
© 1996 Scitech International, Inc. All rights reserved
Back to articles menu
Go to next article