Ensemble Prediction Framework for Abrasive Aqua-Jet Cutting of TiAFL with RSM Integration

Abstract

This work comprehensively assesses the machinability of basalt/flax fibre laminates that were compressed-molded with titanium and then cut via abrasive aqua-jet cutting (AAJC). The focus is on four key AAJC parameters including; abrasive pressure, cutting speed, distance from nozzle and rate of abrasion to assess how each parameter affects the cut quality and also seeks to develop improved machine learning prediction capabilities for increased precision in machining. An ensemble machine learning model (EML), integrating Ridge-Regression (RR), Random-Forest (RF), XG-Boost (XGB), and Neural-Networks (NN), was constructed and evaluated against Response Surface Methodology (RSM). ANOVA demonstrated that AP was the dominant parameter, contributing 70.0% and 71.4% of the observed variation in SR and MRR, respectively (p < 0.001). The developed quadratic RSM models achieved R 2 values of 0.9568 for MRR and 0.9763 for SR, with predicted R 2 values of 0.9164 and 0.9433, respectively. The EML models achieved R 2 values of 0.9769 for SR and 0.9810 for MRR. RSM optimization reduced SR from 2.65 to 2.15 μm and increased MRR from 144.18 to 149.92 mm 3 /s, with prediction errors of only 0.93% and 0.94%, respectively. The experimentally validated EML Pareto solution achieved 3.69 μm SR and 151.25 mm 3 /s MRR, closely matching predictions of 3.65 μm and 150.13 mm 3 /s, with errors of 1.08% and 0.74%. Scanning Electron Microscopy, Atomic Force Microscope Images and 3D Surface Topography were used to evaluate the microstructure and surface of the Hybrid TiAFL in order to show that different morphologies of the hybrid TiAFL are formed as a result of erosion processes. The results of this investigation provide an advanced machinability understanding of Hybrid TiAFL and deliver an ensemble-based predictive model for optimizing machining parameters on an industrial scale with improved performance.