Abstract: Aiming at the non-linear characteristics of analog circuit, a fault diagnosis method of analog circuit based on locality preserving projection (LPP) and hidden Markov model (HMM) was presented. Firstly, the high-dimensional space was constructed by extracting the signal characteristics of analog circuit. Secondly, the LPP was used to compress the feature samples and the intrinsic manifold feature in data set was extracted as characteristic vector. Finally, the classification and recognition of all status were implemented with HMM constructed for reflecting the real status of the system. Simulation analysis results show that, compared with other methods, the LPP-HMM can effectively identify incipient faults, and possess higher rate of fault identification.