Prediction Of Performance Degradation Of Rolling Bearings Based On Attention-LSTM

Rolling bearing is one of the most widely used and vulnerable parts in rotating machinery.Its operating state is closely related to the overall working performance of the equipment.By ways of predicting the rolling bearing change trend and accurately judges its performance degradation stage,which can not only effectively avoid dangerous accidents,but also strengthen equipment health management Focusing on the theme of rolling bearing performance degradation prediction,the research on signal denoising,performance degradation state evaluation and performance prediction is carried out step by step.The main research contents are as follows:Firstly,aiming at the problems in traditional noise reduction methods,such as mode aliasing and signal edge information loss,proposing a signal denoising method based on CEEMD-wavelet adaptive threshold.The CEEMD method is used to decompose the vibration signal into modal components of different frequency bands.And then,by the means of analyzing the cross-correlation between basic modal components and original signals,determining the content of noise in each frequency band component.Furthermore,using the wavelet adaptive threshold method to denoise the high-frequency components with high noise content,and the noise is further suppressed.The high-frequency components and other components are reconstructed after noise reduction,thus completing the whole noise reduction process.The simulation and measured fault signal analysis shows that the proposed method can effectively suppress the noise of high-frequency signals and ensure the integrity of the effective signal to the greatest extent,and the noise reduction effect is better.Secondly,aiming at the problems in the lack of performance degradation indicators in the equipment performance degradation assessment process and the difficulty of visually characterizing the bearing performance degradation characteristics by the monitoring signal,proposing a performance degradation assessment method based on step steady state.Three indicators of correlation,monotonicity and robustness were used to screen performance degradation indicators to reduce the uncertainty of human observation selection.The original characteristic index curve is divided into hi curve and residual curve by means of fixed window mean processing separation method,which has better time correlation and monotonicity;proposing the concept of step steady state to reflect the degree of deterioration of bearing performance quantitatively and intuitively.The effectiveness of the degradation assessment method is verified by the measured full life cycle data set of the bearing.Then,aiming the traditional method rely too much on expert experience and signal processing technology to deal with the low precision of complex time series prediction,proposing a performance degradation prediction method based on Attention-LSTM.Establishing the rolling bearing performance degradation prediction model,using the attention mechanism to improve the learning ability of the data around the performance step,and the sensitivity of the prediction model to the decay characteristics is enhanced.Analyzing the effects of different input and output,activation function and optimization function on prediction accuracy and convergence speed,and determining the appropriate network structure and parameter configuration.By the ways of predicting to the bearing full life cycle data set,compared with other prediction methods,verifying this prediction method has higher prediction accuracy,robustness and generalization ability.Finally,establishing the fault simulation test rig,and taking the single-row angular contact ball bearing as the research object to verify the noise reduction method proposed in this paper.Establishing the full life cycle performance degradation test bench,taking the cylindrical roller bearing as the research object to predict the performance degradation trend by this method and evaluate the performance state of each stage of the bearing.