Machine learning-enhanced prediction of energy absorption in AL/CFRP hybrid tubes under oblique compression

This study investigates the performance of adhesively bonded square aluminum sections under structural loading, with the aim of optimizing energy absorption (EA) and preventing connection failures. To enhance structural integrity and reduce uneven deformation, the design incorporates carbon fiber-reinforced polymer (CFRP) reinforcement applied in three configurations: fully wrapped tubes, outside local reinforcement (OLR), and inside local reinforcement. A machine learning (ML) model based on an artificial neural network (ANN) was developed to predict specific EA (SEA) values based on geometrical and material parameters. The ANN was trained on 341 samples derived from both finite-element simulations and experimental results. The dataset covered a wide range of reinforcement schemes, CFRP ply orientations and thicknesses, and oblique loading angles from 0° to 30°, ensuring sufficient diversity for robust model training. The trained model achieved high predictive performance with a coefficient of determination R2 = 0.79, validating its ability to generalize to unseen data. The results show that the inside-corner reinforcement configuration retained 97% of the EA of the fully wrapped tube while using only 20% of the CFRP material, and the outer-corner configuration maintained 83%, demonstrating superior efficiency. The highest EA value, 1374.57 J, was recorded in the inner-corner configuration with a [90°,0°] fiber stacking sequence.