Experimental and predictive modelling in dry micro-drilling of titanium alloy using Ti–Al–N coated carbide tools

In a rapidly changing manufacturing environment, accurate and efficient models are necessary to predict cutting force and feature quality in the mechanical micro-drilling process. Micro-drilling is challenging due to high spindle speeds and size effects and, therefore, cannot be considered a scaled-down version of macro-drilling. In this study, micro-holes of Ø 0.4 mm are machined using Ti–Al–N coated carbide micro-drill on Titanium alloy (Cp-Ti grade 2) under dry conditions. The process parameters like cutting speed, feed, and pecking depth are varied in three levels based on the full-factorial design with thrust force, burr height, and radial overcut as responses. Predictive models are developed for responses using two intelligent modelling techniques: generalised regression neural network (GRNN) and adaptive neuro-fuzzy inference system (ANFIS). The experimental data is used to train models, and additional experiments are performed to generate testing and validation data. Later multiple regression analysis (MRA) models are also developed for responses. The results indicated that the predicted responses from GRNN, ANFIS, and MRA errors are within ± 5%, ± 5.5%, and ± 12%, respectively, suggesting that the GRNN and ANFIS models are more reliable than the MRA model. In this research, the GRNN models outperformed the ANFIS models. In continuation of the study, optimal process parameters are ascertained to minimize responses simultaneously. At optimal parameter settings, the performance of uncoated and Ti–Al–N coated carbide micro-drills is also evaluated by experiments. Ti–Al–N coated micro-drill reduces the considered responses with lesser tool wear and a favourable chip formation compared to the uncoated micro-drill.