A remaining useful life prediction method based on time-frequency images of the mechanical vibration signals

As a key component of the rotating machines, rolling bearings are widely used in mechanical engineering, aerospace and other fields. The health condition is closely related to the safe operation of the equipment. Predicting the degradation trend and remaining useful life of rolling bearings can enable effective preventive maintenance of rotating machinery. Therefore, an attention mechanism based multiscale convolutional neural network prediction model is proposed in this paper. Firstly, the continuous wavelet transform (CWT) is used to transform the one-dimensional vibration signal collected by the sensor into a two-dimensional time-frequency spectral feature map. Secondly, the quadratic degradation function is selected to determine the health indices of the bearings. Thirdly, the multi-scale convolutional neural network (MSCNN) is employed to realize the deep feature extraction. The multi-scale fusion features are constructed by extracting different degradation features of the signal using convolutional kernels of different sizes, and the necessary degradation features extracted are further enhanced and non-essential features are suppressed through a convolutional attention mechanism. Finally, the proposed model is verified on the PRONOSTIA dataset and compared with other prediction methods. The results indicate that the proposed one achieves better performances than other algorithms with the lowest prediction error and the highest prediction score. It is verified that this method can effectively improve the prediction accuracy and generalization performance, which could provide a certain theoretical basis and for RUL prediction of bearings and other equipment.