A design approach of model-based optimal fault diagnosis

The main purpose of fault diagnosis is to detect faults rapidly and accurately, decide the types, sizes and trends of faults, furthermore, separate the fault and make a proper decision to avoid fault. A model-based optimal fault diagnosis method is proposed in this paper. Modeling for a class of typical faults whose dynamic characteristics were known and the initial conditions were unknown with methods in linear system theory, and a reduced-order state observer was designed for fault system. In order to realize fault state of on-line optimal estimation, an optimal fault diagnosis method was proposed by optimal control theory and duality principle, meanwhile, observation error and control energy was optimum. In addition, threshold value is used to decide whether the fault occurred. Simulation results demonstrate the optimal fault diagnosis system can detect the typical faults on-line.