Application of Improved HMM in Rolling Bearing Fault Diagnosis

Aiming at the influence of initial parameters of Hidden Markov Model (HMM) on the accuracy of rolling bearing fault diagnosis, this paper applies Particle Swarm Optimization (PSO) algorithm to the training of hidden Markov model library, achieving optimization of initial parameters of hidden Markov model, and completing the construction of PSO-HMM fault diagnosis model library. We conducted fault diagnosis experiments on four types of bearings. First, the vibration signals collected under different bearing states are segmented, and then the time-domain, frequency-domain and wavelet packet energy feature parameters of each segment of signals are extracted to form the feature vector. Then, the compensation distance evaluation technology (CDET) is used to reduce the dimension of the feature vector, and the training datasets and test datasets are constructed. The training datasets are used to train the PSO-HMM model library, and the PSO-HMM model library is used to diagnose the fault of the test data set, and compare the diagnosis result with the real situation. The experimental results show that the accuracy of PSO-HMM model library is higher than that of HMM model library for the normal rolling bearing, the outer ring fault, the inner ring fault and the rolling body fault respectively. The PSO-HMM average fault recognition accuracy is 98.34%, and the HMM average fault recognition accuracy is 77.34%. The PSO-HMM improves the fault diagnosis accuracy by 21.00%, which proves the effectiveness and accuracy of the method.