Martin Sachenbacher and Peter Struss
Technische Universität München, Department of Computer Science Oreansstr. 34,1667 München
E-Mail: {sachenba, struss} @in.tum.de
Abstract: In this paper, we deal wih the problem of abstracting behavior models such that their level of granularity is as coarse as possible, but still fine enough to carry out a given behavior prediction or diagnosis task. The focus is on determining task-dependent distinctions with in the domains of varibles - i.e. qualitative values - that are both necessary and sufficient, given a model composed from a library, a granularity of possible observations, and a granularity of desired results.
We present a formalization of the problem, present fundamental results regarding the existence and characterization of solutions to task-dependent qualitative abstraction, and devise a method for automatically determining qualitative values based on a hierarchical representation of the device model that allows to exploit its specific structure.
A principled application is to turn real-valued models, as commonly used in industry, into qualitative models to make the accessible to model-based reasoning methods.The resulting tool set thus enhances the ability to use a behavior model of an engineered device as a common basis to support different tasks along its life cycle.