When quantitative knowledge is incomplete or unavailable (e.g. during design), qualitative models can be used to describe the behavior of chemical plants. Qualitative models were developed for several different process units with controllers and recycle, including a nitric acid plant reactor unit, and simulated using QSIM. In general, such systems produce an infinite number of qualitative states. Two new modeling assumptions were introduced, perfect controllers which respond ideally to a disturbance and ignore dynamics in controller variables, and pseudo steady state which ignores transients in all variables. Redundant constraints, reformulated equations, and quantitative information were also used to reduce ambiguity. A library of general physical and chemical phenomena such as reaction and heat flow was developed in the Qualitative Process Compiler (QPC) representation and used to automatically build qualitative models of chemical plants. The phenomenon definitions in the library specify the conditions required for the phenomena to occur and the equations they contribute to the model. Given a physical description of the equipment and components present, their connectivity and operating conditions, the automatic model builder identifies the phenomena whose preconditions are satisfied and builds a mathematical model consisting of the equations contributed by these active phenomena. Focusing techniques were used to ignore irrelevant aspects of behavior. A dynamic condenser model was automatically generated illustrating QPC's ability to create a new model when a new phase exists. Based on the ability to automatically build and simulate qualitative process models, a prototype hazard identification system, Qualitative Hazard Identifier (QHI), was developed which works by exhaustively positing possible faults, simulating them, and checking for hazards. A library of general faults such as leaks, broken filters, blocked pipes, and controller failures is matched against the physical description of the plant to determine all specific instances of faults that can occur in the plant. Faults may perturb variables in the original design model, or may require building a new model. Hazards including over-pressure, over-temperature, controller saturation, and explosion were identified in the reactor section of a nitric acid plant using QHI.
Supervisor: Lyle H. Ungar. Thesis (Ph.D. in Chemical Engineering) -- Graduate School of Arts and Sciences, University of Pennsylvania, 1993. Includes bibliographical references.