Design of adaptable simulation models.

In today's world, with ever increasing competition, modelling and simulation proves to be a very helpful tool. Many methodologies exist to help build a simulation model from scratch. In terms of adaptability, most current attempts focus on either the operational side, ie the automated integration of data into a model, or the creation of new software. However, very few attempts are being made to improve the adaptability of shelved models built in existing simulation software. As a result, there is a certain reluctance, in some areas, to use simulation to its full potential.Based on these facts, it is obvious that anything, which makes reuse of simulation models easier, can help improve the use and spread of simulation as a valuable tool to maintain a company's competitiveness. In order to find such a solution, the following issues are looked at in this thesis: The changes to a simulation model that constitute the biggest problem, ways to minimise those changes, and possibilities to simplify the implementation of those changes. Those factors are evaluated, first by investigating current practices of building adaptable simulation models via a literature review, then the most difficult changes to implement in a simulation model, and the most frequent types of simulation software, are identified by means of interviews and questionnaire surveys. Next, parameters describing the adaptability of a simulation model are defined. In a further step, two of the most widely used simulation packages are benchmarked against a variety of tasks, reflecting the changes most frequent to models. The benchmarking study also serves to define and test certain elements regarding their suitability for adaptable models. Based on all those steps, model building guidelines for the creation of adaptable simulation models are developed and then validated by means of interviews and a framed field experiment. The interviews and questionnaire reveal that deleting is the easiest task and modifying the most complicated, while handling devices are the most difficult element to modify. The results also show that simulators (eg Arena) are the most widespread type of simulation software. The benchmarking showed that Arena is overall more adaptable than Simul8, and confirms the findings from the user survey. Also, it shows that sequencing is very helpful for modifying models, while the use of sub-models decrease the adaptability. Finally, the validation proves that the model building guidelines substantially increase the adaptability of models.