Continuous Hidden Markov Model Based Incipient Fault Monitoring in Filtered Analog Circuits

Abstract

The small variations of components parameters often lead to severe performance degradation in Filtered Analog Circuits (FAC). Most of the researches on soft fault diagnosis in analog circuit are focused on the variations beyond 30% of the nominal parameter in recent years, which is usually unacceptable in FAC. To diagnose the soft fault of small deviation as earlier as possible, the Hidden Markov Model (HMM) was introduced to monitor the FAC. Different from the existing diagnosis approaches based on HMM, in which the variations of components parameters were considered to be random, the continuous variations of the fault component parameter are discretized and modeled by the hidden states of the proposed HMM method. The experiment demonstrates that the proposed HMM approach can model the FAC effectively and recognize the incipient fault earlier.