Multi-Head Attention Network with Adaptive Feature Selection for RUL Predictions of Gradually Degrading Equipment

A multi-head-attention-network-based method is proposed for effective information extraction from multidimensional data to accurately predict the remaining useful life (RUL) of gradually degrading equipment. The features desired equipment were evaluated using a comprehensive evaluation index, constructed discrete coefficients, based on correlation, monotonicity, and robustness. For extraction, optimal feature subset, determined by adaptive selection method, was input into multi-head temporal convolution network–bidirectional long short-term memory (TCN-BILSTM) network. Each individually mined avoid loss information. effectiveness our RUL prediction verified NASA IMS bearings dataset C-MAPSS aeroengines dataset. results indicate superiority compared other mainstream machine learning methods.