A NEW APPROACH TO QUANTITATIVE AND CREDIBLE DIAGNOSIS FOR MULTIPLE FAULTS OF COMPONENTS AND SENSORS (REPRINTED FROM J JAPAN SOC ARTIF INTELL, VOL 9)

Many practical applications of system diagnosis require the credible i dentification of multiple faults of nonlinear components and sensors i n quantitative measures. However, the state of the art of diagnosis te chnique is considered to be still insufficient to meet these severe re quirements. The approach of diagnosis using the traditional linear sys tem identification theory can diagnose the disturbed parameters of a s ystem in detail and evaluate the quantitative amplitude of the disturb ance. However, it hardly provides the diagnosis of the multiple faults and the diagnosis of the components having high nonlinearity. On the other hand, some recent model-based diagnosis approaches can diagnose the multiple faults even for highly nonlinear components, though they do not provide the detailed diagnosis of elements indivisibly involved in components and the quantitative amplitudes of the faults. The meth od proposed in this paper provides an efficient remedy to achieve all of the practical requirements, i.e., the credible, detailed and quanti tative diagnosis of multiple faults of nonlinear components and sensor s. Our study newly proposes the frameworks of optimal constraints and causal ordering of physical systems. Also, a systematic and strict the ory to synthesize these frameworks together with the model-based diagn osis is provided to characterize an optimal consistency checking metho d in diagnosis and to evaluate quantitative amplitudes of faulty distu rbances. First, the detection of faulty behaviors of an objective comp onent is performed based on the quantitative consistency checking betw een observations and the optimal constraints, called as ''minimal over -constraints'', consisting of first principles in the components. Seco nd, once if some inconsistencies are detected, a mathematical operatio n of model-based diagnosis derives the candidates of faulty elements a nd functions even under multiple fault conditions. Third, the anomalou s quantities directly disturbed by the faulty elements are identified systematically based on causal ordering. Furthermore, the quantitative deviations of these quantities are evaluated by using the minimal ove r-constraints. The performance of the proposed method is demonstrated through an example to diagnose an electric water heater. The ability o f this diagnosis has been confirmed for the multiple faults in nonline ar and dynamic systems. (C) 1997 Elsevier Science B.V.