The Hybrid Phenomena Theory

The Hybrid Phenomena Theory (HPT) is a framework for formalizing how dynamic state space models of physical systems are built from first principles. The HPT descends from the Qualitative Process Theory (QPT), [Forb us, 1984], from which it inherits basic concepts like views, phenomena and influences. However, the HPT redefines some of these concepts in a more strict manner in order to represent knowledge of physics with the accuracy needed to develop full parametric models. Specifically, influences may specify quantified non-linear functions of several variables. A mechanism denoted subsumption is introduced to ensure consistency in the emerging models when different simplifying assumptions are made. The HPT has been implemented in CLOS. 1 Background and Motivation Qualitative reasoning, [Bobrow, 1984; Weld and de Kleer, 1990], has provided valuable insight and methods for analyzing physical systems. Yet, several fundamental problems have been identified. The inherent ambiguity described by Kuipers [1986] poses a problem in some potential areas of application. The problems with the qualitative mathematics pointed out by Struss [1990] imposes limitations on what can be accomplished using any form of qualitative simulation. It is something of a paradox that while Hayes [1985] argued that it was time for AI to move away from the toy problems commonly addressed by AI research at that time and tackle a real world problem, he simultaneously argued for a purely qualitative approach. But a large class of problems involving the physical world depends on quantitative knowledge. By choosing a purely qualitative approach, many researchers have disqualified their methods from a vast number of potential application areas in science and engineering. For all the merits of qualitative simulation, it is the authors view that the greatest accomplishment of the * qualitative reasoning community is the development of methodologies enabling reasoning about how structural relationships between objects comprising a physical system impose constraints between the parameters and variables describing that system. This reasoning captures essential parts of the reasoning performed by physicists and engineers when building mathematical models of physical systems. However, a predominant fraction of those models incorporate parameters and variables intended to have a quantitative interpretation. The Hybrid Phenomena Theory (HPT) was developed with monitoring and diagnosis applications in the process industry in mind. The HPT builds on the framework of the QPT [Forbus, 1984; Forbus, 1990]. But contrary to the QPT, the HPT derives parametric state space models. The parameters may have a quantitative …