A hybrid machine learning-GCO framework for multi-response optimization in AWJ machining of MWCNT infused Mg-FMLs

A hybrid machine learning (ML)-assisted framework was developed to predict and optimize abrasive waterjet (AWJ) machining of multi-walled carbon nanotube (MWCNT)-infused magnesium fibre metal laminates (Mg-FMLs). Thirty experiments were conducted using a four-factor Box–Behnken design, and bootstrap sampling-based augmentation generated 540 additional synthetic observations. Random Forest (RF), Extreme Gradient Boosting (XGB), Decision Tree (DT), and AdaBoost Regressor (ABR) were evaluated for predicting material removal rate (MRR), kerf width (KW), and delamination factor (Fd). RF achieved R2 values of 0.9026, 0.9075, and 0.9140, respectively. Five metaheuristic algorithms were compared, and Germinal Center Optimization (GCO) achieved the highest overall objective value (0.5873). The optimal conditions were 0.6954 wt% MWCNT, 2764.75 MPa waterjet pressure, 784.17 mm/min traverse speed, and 454.83 g/min abrasive flow rate, yielding predicted MRR, KW, and Fd values of 113.22 g/min, 3.8781 mm, and 2.9148, respectively. Confirmation errors were found to be 4.56%, 2.68%, and 4.03%, respectively.