Fault Detection Method of Motor Bearing Based on Improved Random Forest Algorithm

Once the rotating machinery equipment fails, the result often causes serious economic losses and safety accidents. However, the occurrence of bearing faults is not sudden, and the early stage of the fault has little impact on the system performance, which is difficult to judge manually, but the vibration signals of different faults are different from the normal situation. Therefore, the fault diagnosis method of motor bearing with high recognition accuracy and low false negative rate has become a research hotspot at home and abroad. The high-speed operation of the motor makes it very demanding on the reliability of the motor system. Rolling bearing is one of the most important parts of electrical equipment, and it is also one of the most vulnerable parts in electrical equipment. Singular value decomposition is used to diagnose the early faults of rolling bearings, and the fault characteristic frequencies are extracted from the envelope spectrum of the component signal, and the algorithm mentioned is verified by engineering experiments. Therefore, it is necessary to carry out working condition monitoring and fault diagnosis, and use predictive maintenance or discretionary maintenance to replace the traditional scheduled maintenance, which can not only maximize the working potential of the bearing. And it can reduce or eliminate the occurrence of accidents, so that the production can be carried out in a reasonable and orderly manner. Among them, the time-domain features used more include root mean square value and kurtosis value, while the frequency-domain and time-frequency features usually need signal transformation, such as wavelet transform, fast Fourier transform and empirical mode decomposition. Random forest algorithm is widely used in cryptography, bioinformatics, ecology and other fields because of its strong anti noise ability, few adjustable parameters and strong adaptability. The operating condition of the bearing can be analyzed by the change trend of the bearing pad temperature, the oil temperature at the inlet and outlet of the bearing and the distribution of the temperature field at the base. The improved random forest algorithm is used to detect the fault of rolling bearing, and the proposed algorithm is verified by experiments. In order to improve the recognition accuracy of motor bearing fault, a bearing fault diagnosis method based on adaptive multi-scale dispersion entropy and improved random forest algorithm is proposed in this paper.