Integrating qualitative and quantitative modeling

Second generation expert systems are characterized by the use of models of the application domain at different levels of detail and focused on different aspects of the same process. In many engineering applications, sophisticated numerical models are available, but they are inadequate as a basis for reasoning. On the other side, qualitative models have proved effective as deep representations for knowledge-based systems, but too underdefined in their behavioral prescriptions. For both the qualitative and the quantitative approaches, problems grow with the dimension of the simulated system. We present a method to integrate these two modeling paradigms. We assume that the system to be modeled can be partitioned in two parts interacting through shared variables. One part is described qualitatively, the other one, quantitatively. The parallel evolutions of the two simulation/monitoring processes are based on a qualitative interpretation of the numerical course of the shared variables. The resulting qualitative simulation is enriched with quantitative information about both time and variable values.