Delamination assessment and prediction in ultrasonic-assisted drilling of laminated hybrid composites

This study presents a novel application of the Random Forest (RF) algorithm to predict delamination in ultrasonic-assisted drilling (UAD) of carbon fiber reinforced polymers (CFRPs). It performs a multi-dimensional analysis of factors including graphene nanoplatelet (GNP) addition, ultrasonic vibration, cutting tool type, and feed rate on delamination damage. The RF algorithm was chosen for its ability to handle both regression and categorical tasks. The model demonstrated strong predictive performance, achieving an R² value of 0.9445 on test data, with a root mean squared error (RMSE) of 0.32% and a mean absolute error (MAE) of 0.29% relative to the average values. Analysis of variance (ANOVA), Sobol sensitivity, and Shapley additive explanations (SHAP) analysis were used to assess the impact of input parameters. Sobol identified the cutting tool type and feed rate as the most influential factors, contributing 37.7% and 34.3% to delamination variance, aligning with ANOVA findings. SHAP further confirmed the tooling type and feed rate as key factors, with contributions of 48.75% and 32.61%. The analyses revealed that GNPs increased delamination due to higher thrust forces, while ultrasonic vibration and high-cobalt tools reduced delamination. Optimal conditions were a feed rate of 0.08 mm/rev with an 8% cobalt tool and ultrasonic vibration, excluding GNPs.