Data-driven simulations of flank wear of coated cutting tools in hard turning
Insurance of surface quality and dimensional tolerances in finish turning necessitates the development of accurate predictive models. This study aimed at modeling flank wear of multilayer-coated carbide inserts in finish dry hard turning of AISI 4340 and AISI 52100 hardened steels based on 28 artificial neural networks (ANNs) and the best-fit multiple non-linear regression (MNLR) model. Online-monitored flank wear of multilayer-coated carbide inserts was modeled as a function of the three cutting speeds of 70, 98 and 142 m min-1, and the two workpieces under the constant feed rate and cutting depth of 0.027 mm min-1 and 0.2 mm, respectively. Out of the 28 ANNs, 18 ANNs appeared to be capable of better predictions for tool flank wear than the best-fit MNLR model. Probabilistic neural network (PNN) outperformed all the remaining models based on all the training, cross-validation and testing dataset-related metrics.