The traditional fault diagnosis methods can not locate the fault location accurately. Therefore, an improved particle swarm optimization hidden Markov mode (HMM) fault diagnosis method is proposed. The HMM l is used to identify the fault mode of the integrated transmission device, the fuzzy set is used to define the mode to study the voltage signal feature extraction method, and the optimal selection is made according to the sensitivity of the eigenvalue; 3 HMM are applied to analyze the fault signals of the integrated transmission device under different operating conditions. And the diagnosis results are compared. The results show that the improved particle swarm optimization HMM model can quickly and effectively identify the wear, damage and other fault modes in the integrated transmission, and has good applicability, which can be applied to the fault diagnosis of the actual integrated transmission system.