Research On Remaining Useful Life Prediction Method Of A Rolling Bearing Combining CNN And LSTM

A rolling bearing is one of the basic components of many rotating machinery equipment.Its running state plays an important role in the safe and reliable operation of the equipment.Once a fault occurs,it will not only cause huge economic losses,but also lead to catastrophic casualties.Therefore,the accurate prediction of the remaining useful life(RUL)of rolling bearings can provide a basis for preventive maintenance decision-making,facilitate the formulation of maintenance strategies,improve the reliability and safety of rotating machinery equipment,and avoid accidents.In order to excavate fully the information of bearing running state contained in the original vibration signal of rolling bearings,the trend quantification health indicators are independently constructed to predict the RUL of rolling bearings.Combined convolution neural network(CNN)and long short-term memory(LSTM)neural network,a remaining useful life prediction method for rolling bearings is proposed in this paper.Firstly,in order to improve the recognition ability of the vibration signal of the rolling bearing,obtain more useful information,extract fault characteristics and improve the overall prediction ability of the model.The vibration signal of the rolling bearing is preprocessed in frequency domain,and the original vibration signal is transformed into frequency domain amplitude signal by fast Fourier transform.Then,the frequency-domain amplitude signal obtained by pre-processing is normalized as the input of CNN.CNN has the characteristics of convolution and weight sharing.The local abstract information of frequency-domain amplitude signal is extracted independently,and the deep features can be obtained,which can avoid the problem of traditional feature extraction methods excessively relying on expert experience.Finally,the deep features are input into LSTM network,and the trend quantification health indicators are constructed independently according to the percentage of lifetime,and the failure threshold is determined at the same time.The construction method of health indicators abandons the traditional idea of feature extraction and feature fusion,avoids the difficulty of determining failure threshold,and reduces the consumption of human resources and time cost.Meanwhile,the moving average method is used to smoothen the local oscillation,and then the polynomial curve fitting is used to predict the future failure time and realize the RUL prediction of rolling bearings.The experimental results show that,under the two modes,the trend quantification health indicators constructed by the proposed method all have good monotonous trend and prediction results can better approach the real life values.