Prediction of Aerodynamic Forces at the Tip of the Compressor Blades Based on Multi-scale 1DCNN Combined with CBAM

•The aerodynamic pressure at the blade tip is predicted by a data-driven model.•Convolutional Block Attention Module is added to the multi-scale one-dimensional CNN model.•Train the model using datasets with different stagger angles and twist angles to improve its robustness. The compressor is a crucial component of aircraft engines, and the blades are the critical factor affecting the performance of the compressor. Based on multi-scale one-dimensional convolution neural network (1DCNN) with Convolutional Block Attention Module (CBAM), a data-driven model is proposed for predicting the aerodynamic characteristics of the blade tips. The model is trained using the Adam with decoupled weight decay (AdamW) optimizer and a staged learning rate scheduling strategy. Due to the distinct aerodynamic pressure distributions on the suction and pressure sides, separate models are constructed in order to reveal the aerodynamic performance of the blade tips accurately. During the model validation, Root Mean Square Error (RMSE) and the coefficient of determination (R2) are taking as evaluation criterions, where high reliability is demonstrated compared to Computational Fluid Dynamics (CFD) results.