Machine remaining life prediction based on multi-layer self-attention and temporal convolution network

Abstract Convolution neural network (CNN) has been widely used in the field of remaining useful life (RUL) prediction. However, CNN-based RUL prediction methods have some limitations. The receptive CNN is limited and easy to happen gradient vanishing problem when too deep. contribution differences different channels time steps are not considered, only use deep learning features or handcrafted statistical for These limitations can lead inaccurate results. To solve these problems, this paper proposes an method based on multi-layer self-attention (MLSA) temporal convolution (TCN). TCN extract features. Dilated residual connection adopted structure. efficient way widen field, structure avoid problem. Besides, we propose a feature fusion fuse And MLSA designed adaptively assign weights. Finally, turbofan engine dataset verify proposed method. Experimental results indicate effectiveness