A method for rapidly evaluating reliability and predicting remaining useful life using two-dimensional convolutional neural network with signal conversion

Real-time monitoring and rapid evaluation of bearing operating conditions, especially for the reliability evaluation and remaining useful life (RUL) prediction, are major challenges in the rotating machinery field. A two-dimensional (2-D) deep convolution neural network (CNN) is proposed for rapidly evaluating reliability and predicting RUL, in which a signal conversion method is proposed for converting the one-dimensional signal into the 2-D image to satisfy the input requirements of the 2-D CNN. Different activation functions are employed to implement the conversion of the input data to the output data for each layer of the network, and dropout is only adopted in the hidden layer to change the network structure to prevent overfitting. The maximum correlation entropy with regular terms is employed as the loss function of the model to obtain better training performance compared with the mean square error (MSE). Then, the rolling bearing degradation vibration data is applied to the proposed model to verify the accuracy and rapidity. The results show that the proposed method has good accuracy and fast calculation ability in bearing reliability evaluation and RUL prediction, especially, its time consumption is shorter than that by other deep learning networks.