Bearing Fault Diagnosis Method Based on Deep Learning and Health State Division

As a key component of motion support, the rolling bearing is currently popular research topic for accurate diagnosis faults and prediction remaining life. However, most existing methods still have difficulties in learning representative features from raw data. In this paper, Xi’an Jiaotong University (XJTU-SY) dataset taken as object, deep technique applied to carry out fault research. The root mean square (RMS), kurtosis, sum frequency energy per unit acquisition period short-time Fourier transform are used health factor indicators divide whole life cycle bearings into two phases: phase phase. This division not only expands but also improves efficiency. Deep Convolutional Neural Networks with Wide First-layer Kernels (WDCNN) network model improved by introducing multi-scale large convolutional kernels Gate Recurrent Unit (GRU) networks. signals classified states trained tested, training testing process visualized, then finally experimental validation performed four failure locations dataset. results show that proposed has excellent noise immunity, can achieve under complex working conditions, greater diagnostic accuracy