Abrasive water jet machining (AWJM) is a versatile and non-conventional machining process that is particularly conducive for machining brittle materials, e.g., borosilicate glass, which presents a great obstacle when it is machined with traditional machining processes because of its high brittleness and its high susceptibility to damage. This scientific research ensures to address an identified research gap by applying an integrated statistical-machine learning (ML) framework to the AWJM of borosilicate glass, with particular focus on elucidation of the influence of key process parameters (traverse speed (TS), abrasive flow rate (AFR), and stand-off distance (SOD)) with respect to the kind of kerf taper angle (θ). A hybrid methodology as a combination of Response Surface Methodology and Box-Behnken Design (RSM-BBD) and k-Nearest Neighbors (kNN) classification model has been developed. RSM analysis showed that TS and AFR are statistically significant, where TS has a direct relationship and AFR has an inverse relationship with the taper angle, while SOD showed negligible influence in the tested range. The optimal combination for minimizing taper was found to be TS= 80 mm/min, AFR= 300 g/min, and SOD= 1 mm; the composite desirability was 1.000, and the validation error was below 14%. The implemented kNN model has obtained a prediction accuracy of up to 93.75% (for k = 1) in the category of taper angle (high/low). As such, this research presents a new research area at the intersection of good traditional optimization and intelligent predictive modeling, providing a robust and replicable methodology for improving the precision and repeatability of AWJM on brittle materials.
Gawade et al. (Fri,) studied this question.