Motor Bearing Remaining Life Prediction Based on Unsupervised Learning

With the development of industry, the machine becomes more and more large and complicate, correspondingly, the requirement of industrial machine reliability is more strict. As an important part of the rotating machine, the safety of the motor bearing is the key for the reliability of the machine. Recently, the development of artificial intelligence industry developed rapidly, it is popular to combine artificial intelligence technology with the reliability of the machine to predict the remaining life or fault of the bearing. Traditionally, the method would be trained by the signals and the corresponding condition first, which means it is the supervised learning method and much labeled data is required. However, it is not easy to obtain too much labeled data. A unsupervised learning method to predict the remaining life of bearing is proposed in this paper. The Fast Fourier Transform (FFT) is used to construct frequency-domain features first, then, moving the clustering centers appropriately to obtain the optimal boundary for remaining life prediction based on Euclidean distance. The superiority of the proposed method is shown by the results of the experiments.