A SIMPLE DESIGN STRATEGY FOR FAULT MONITORING SYSTEMS

A simple strategy is developed for the design of model-based fault-monitoring systems. Several new features allow this approach suitable for practical applications. The core of the suggested monitoring system is a parallel parameter estimation method designed to reduce the chance of bias. Two important criteria, diagnostic observability and diagnostic resolution, are adopted to evaluate the performances of alternative designs. By following asystematic procedure, it is possible to determine the best sensor locations in a complex chemical process based only on the structural informations of the system model. After selecting the most appropriate measurement variables, diagnostic resolution can be enhanced further by exploiting additional insights gained from analysis of the physical meanings of the model parameters and by comparing the fault patterns in diagnosis. Several effective on-line implementation techniques are also proposed to minimize computation for the estimation algorithm. The correctness and efficiency of the suggested design strategy are demonstrated by the simulation results in examples.