Performance Assessment And Life Prediction Of Servo Motor Bearings Based On Reduced SOM And LSTM

With the continuous development of science and technology,servo motors have been widely used in machinery,metallurgy,transportation,petrochemical industry,shipbuilding,aerospace and scientific research and other fields.During the long-term operation of the motor,the bearing performance will gradually decline.Once a serious failure occurs,the system's operating conditions will deteriorate sharply and cause production accidents.If the motor bearings are repaired and replaced blindly,it will cause detection or over-repair.Therefore,performance evaluation and life prediction of rolling bearings have become an increasingly important research topic.Traditional performance assessment and life prediction methods are often insensitive to initial degradation,failure thresholds are difficult to determine,low prediction model accuracy,insufficient long-term prediction capabilities,and lack of uncertainty expression,etc.,gradually fail to satisfy accurate and efficient equipment health management requirements.Therefore,in order to solve the problems of insensitivity to initial degradation,difficulty in determining the failure threshold,and low accuracy of the prediction model,this paper proposes a performance assessment method based on reduced SOM to obtain an assessment curve,and uses the LSTM network on the basis of the assessment curve predict the remaining life.This paper takes the rolling bearing as an example,starting from two cases of complete data and incomplete data,the performance assessment and life prediction method research based on the data driving method is carried out.First,the typical fault types and their effects of servo motors are analyzed,and it is determined to use vibration acceleration data for feature extraction to obtain a feature set that includes time domain features,frequency domain features,and time-frequency domain features.Secondly,taking the feature set extracted from the rolling bearing vibration data as the research object,comparing the existing performance degradation assessment methods,a performance assessment algorithm based on reduced SOM is proposed,and the algorithm is verified on the IMS bearing data set,and the results are given.The assessment indicators of health factors are compared with PCA and SOM methods.Finally,the obtained rolling bearing performance degradation curve is taken as the research object,compared with the existing remaining life prediction model,a life prediction model based on LSTM is proposed,and the model verification is carried out in two cases of complete and incomplete data,The evaluation index of the prediction result is given,and at the same time,it is compared and analyzed with AR and BP methods.As the research results show,compared with the traditional performance assessment method,the proposed rolling bearing performance assessment method based on reduced SOM can construct a performance degradation curve with only a few normal samples as training data,and the time correlation,monotonicity and robustness is good.The obtained health factor is also more sensitive to initial damage.The assessment model is established in an unsupervised manner and has good versatility.Based on the obtained performance degradation curve,according to the two conditions of whether the data is complete,the LSTM network model is used to predict the remaining life of the rolling bearings.Due to the consideration of the correlation of different time steps,the model's long-term prediction ability is enhanced.At the same time,the 95% confidence interval of the health factor is calculated to quantify the expression of uncertainty in the prediction.