In order to improve the accuracy of the remaining life prediction task,a remaining life prediction model was proposed based on attention long and short term memory network and multi-headed self-attention mechanism (ALSTM-MHA),which could extract the importance of feature dimensions and correlation information of time dimensions under the condition of using data temporality.The model was experimentally validated using the C-MAPSS dataset and analyzed in comparison with other methods.The results show that the ALSTM-MHA model can effectively extract the attention information in feature and time dimensions,and the root mean square error and asymmetric evaluation indexes are reduced by at least 0.3% and 20.48%,respectively,compared with other methods,which verifies the feasibility and effectiveness of the model.