A hybrid method based deep learning approach for Predicting residual Life of Machinery

The remaining useful life (RUL) prediction technology of mechanical equipment plays a critical role in implementing forecast maintenance, which can not only avoid failures, but also effectively save the maintenance costs. However, nowadays, machines become more complicated and enlargement, and it will be difficult to carry out maintenance predictably using the traditional prediction methods. To deal with these complexes maintain tasks, a hybrid model based deep learning approach is proposed in this paper. First, an optimized 1-D convolutional algorithm is used to mine the deep feature implicated in the data. Then, the dilated rate of kernels is set as a large value to get a greater perceptual horizon. In order to solve the gradient vanishing problem, a residual block is introduced between the convolutional layers, which brings a propagation channel directly for the gradient. Finally, the features are entered into long short-term memory (LSTM) network for RUL prediction. In order to demonstrate the effectiveness of the proposed method, a date collected from turbo engines is utilized for the RUL prediction and the results show that the proposed model performs better than the original methods.