Year: 2026
Type: article
Abstract: Hard turning is a precision machining process employed as a finish operation on hardened alloy steels (42–68 HRC), widely used as an alternative to cylindrical grinding. The complexity and non-linearity of the hard turning process—governing surface roughness, cutting forces, tool wear, and cutting temperature—make traditional analytical modeling inadequate. Artificial neural networks (ANNs) have emerged as powerful data-driven tools capable of modeling and optimizing these intricate relationships with high predictive accuracy. This review systematically surveys peer-reviewed research published between 2005 and 2025, with emphasis on studies from 2018–2025, examining the application of ANN-based models to optimize hard turning parameters. The review covers ANN architectures employed (feedforward, radial basis function, deep neural networks, and hybrid approaches such as ANN–GA, ANN–PSO, and CNN-LSTM), training algorithms (back-propagation, Levenberg–Marquardt, Bayesian regularization), workpiece materials (AISI 52100, AISI H13, AISI D2, AISI 4340), cutting tool types (CBN, PCBN, coated carbides), and performance indicators (surface roughness Ra, cutting forces Fx/Fy/Fz, tool flank wear VB, cutting temperature). Key findings confirm that ANN models consistently outperform regression-based models in capturing non-linear input–output relationships, with prediction errors frequently below 5%. Recent trends toward hybrid ANN–genetic algorithm models, explainable machine learning (XAI), and real-time CNN-LSTM architectures signal a move toward intelligent and autonomous hard turning optimization systems.
Authors: Mangesh Patil, Munna Verma, Pundlik Patil
Institution: Bhagwant University
Language: English
Subfield: Mechanical Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7202200641