Research On Bearing Fault Diagnosis Method Based On Hidden Markov Model

Bearing is the most widely used component in the rotating machinery equipment and also the most easily damaged part. Whether the bearing can be in normal working condition directly affects the efficiency of mechanized production. When the bearing failure occurs, the machine will produce abnormal vibration and noise. The key content of this paper is how to extract the characteristic value of the vibration fault signal and diagnose the fault accurately. This paper proposed a series of bearing fault diagnosis method based on the hidden Markov model.The hidden Markov model is very suitable for fault vibration signal analysis of non stationary rolling bearing since the characteristics of strong classification ability, less training samples and fast calculation speed expressed in bearing fault diagnosis. In this paper, it trains the hidden Markov model with appropriate state number based on the collected vibration signal features, calculates the similar probability, and finally determines the type of fault though the appropriate state number. The specific research contents as follows:1. Before the fault signals of rolling bearing were analyzed, establish the hidden Markov model based on the extracted feature data of bearing fault signal, and verify the feasibility of the hidden Markov model in fault diagnosis.2. Proposed a fault recognition method based on singular value decomposition and hidden Markov model. Firstly use singular value decomposition to extract fault features of bearing fault vibration signal, then and then send the fault features into the trained hidden Markov model to diagnose the fault. The experimental test the method based on singular value decomposition and hidden Markov model is effectiveness.3. Proposed a fault recognition method based on S-transform and hidden Markov model. Firstly use the S-transform for the fault signal data of rolling bearing, and do singular value decomposition for the time-frequency spectrum matrix decomposed by the S-transform. Finally, send the fault features into the trained hidden Markov model to diagnose the fault and effectively judge the fault type. In this part, the effectiveness of S-transform and HMM method in bearing fault diagnosis is verified by experiments.4. Design the vibration signal acquisition and analysis system, the system is contained by the vibration sensors PXR02, the digital conversion chip AD9225, the master controller STM32 and the corresponding wireless transmission module. Rely on the VS to receive data because of its powerful serial port function. Finally, use Matlab to analysis the data. In this way, the real-time monitoring of the running state for the bearing is accomplished.