Rotation Speed Estimation for Motor Fault Diagnosis Using Bidirectional Long Short Term Memory Networks

Rotation speed estimation is a key step for motor fault diagnosis under variable-speed working conditions. This study investigates a method for speed estimation by fusing the information from vibration and current signals using a bidirectional long short term memory (BiLSTM) deep neural networks. Firstly, the instantaneous frequency (IF) curves are extracted from the signals using an improved time-frequency analysis method. A BiLSTM regression model is trained for IFs fusion and rotation speed estimation. Finally, the rotation angle is computed from the estimated rotation speed curve, and the vibration signal is resampled and processed for motor fault diagnosis. The performance of the proposed method is validated by the experimental data acquired from a switched reluctance motor test rig.