Aero-engine’s remaining useful life estimation model based on multi-attention mechanism and selective kernel network

Abstract Remaining useful life prediction of aero-engines is one of the research hotspots, and making full use of complex sensor data is the key challenge. To this end, a prediction model utilizing selective kernel convolution and a multi-head attention mechanism is introduced. The model combines the multi-head attention mechanism and selective kernel convolution, aiming to analyze complex sensor data in multiple scenarios. The multi-head attention mechanism improves the model’s ability to extract multi-dimensional features and capture time series dependencies, while the selective kernel convolution dynamically selects the convolution kernel size to effectively integrate local and global features. Experiments are conducted on the publicly available CMAPSS dataset, and the results show that the RMSE and Score metrics are reduced by 0.63 and 63, respectively, which validates the effectiveness of the model.