关键词: 隐马尔可夫模型, 故障诊断, 线性辨别分析, 模拟电路

Abstract: Due to the uncertainties that exist in the running of the analog circuits, the traditional Hidden Markov Model（HMM） approach is improved. The state transition probability matrix of the traditional model is replaced by time-varying one that more satisfies the actual situation. An updating control factor is introduced for avoiding the excess updating of the state transition probability in the initial stage of each state. The Linear Discriminant Analysis（LDA） is used to reduce dimensionality and remove redundancy of the voltage feature vectors, which are for HMM’s training and testing in order to achieve recognition and diagnosis of the incipient faults in analog circuit. The experimental results indicate that the improved HMM has better fault recognition capability than the traditional HMM.

Key words: Hidden Markov Model（HMM）, fault diagnosis, Linear Discriminant Analysis（LDA）, analog circuit