Automatic Enhancement of Model Parsimony

This paper presents a technique that supports a knowledge engineer in constructing a qualitative model of a physical device . The technique is specifically aimed at achieving a parsimonious model of device behaviour. First we explain why a model that is composed of general model fragments may still contain irrelevant particles . Then we present a technique that removes these irrelevant model particles from the model fragments . The technique assumes that, based on a set of applicable model fragments, the behaviour of the device can be modelled although the description may contain irrelevant model particles . It takes this behaviour description as input and for manipulations of specific parameters identifies the irrelevant particles . Both knowledge concerning the relevance as well as the irrelevance of model particles is used for this purpose . We further employ the distinction between overly detailed and superfluous model particles. Finally, modification of the model fragments yields a behaviour simulation with better computational properties and more clarity . 'The research presented in this paper was partially supported by the Foundation for Computer Science in the Netherlands (SION) with financial support from the Netherlands Organisation for Scientific Research (NWO), project number 612-322-307 . We would like to thank Jaap Kamps for the stimulating discussions on the research presented in this paper.