Recognition and Diagnosis of the Incipient Faults in Analog Circuit Using Improved HMM

Abstract

Due to the uncertainties that exist in the running of the analog circuits, the traditional hidden Markov model (HMM) approach is improved through replacing the state transition probability (STP) matrix of the traditional model by time-varying one. An updating control factor is introduced for avoiding the excess updating of the STP in the initial stage of each state. The experimental results indicate that the improved HMM has better fault recognition and diagnosis capability.