基于CAE和AGRU的滚动轴承退化趋势预测

Aiming at problems of health indictor construction depending on prior knowledge and prediction accuracy being low for rolling bearing performance degradation trend prediction method in rotating machinery, a prediction method for rolling bearing degradation trend based on convolutional auto-encodes(CAE) and attention gated recurrent unit(AGRU) was proposed. Firstly, the method converted the rolling bearing time domain signal into frequency domain signal with fast fourier transform(FFT), and the features were extracted adaptively from frequency domain signal with convolutional auto-encodes. Then, the health indicators werw constructed from encoding features with evaluating and delsting. Finally the health indicators were input into the attention gated recurrent unit mode, and the pruning algorithm optimized the parameters to predict the performance degradation trend of rolling bearings. Results showed that the proposed method can obtain more accurate prediction results for rolling bearing performance degradation trend.