CAResNet Based Aero-engine Remaining Useful Life Prediction

The aero-engine is one of the core components of an aircraft. Failures in aero-engine are possible to lead to devastating consequences. Therefore, accurately predicting the remaining useful life (RUL) of the engine is crucial for ensuring the stable operation of the aircraft. In order to accurately predict the RUL value of aero-engines, we propose a model called CAResNet that combines improved channel attention and residual blocks to achieve accurate RUL prediction. Firstly, a channel attention module based on convolution operations in two dimensions is constructed. Attention weights for different information are obtained by summing the results of the two pooling methods. Secondly, a feature extraction module is built based on stacked residual blocks, which composed entirely of linear layers, achieves effective feature extraction while maintaining a simple network structure. Finally, an RUL prediction module consisting of dilated convolution blocks is utilized to predict the RUL value of the engine. To validate the effectiveness and superiority of this method, we conduct simulation experiments using the C-MAPSS turbofan engine dataset and compare the results with other relevant studies. The experimental results demonstrate that CAResNet performs well in RUL prediction tasks.