Tolga Berkay Şirin

The manufacturing industry could achieve cost effectiveness, energy efficiency, time savings, personal health and safety, as well as economic and environmental sustainability by simple modifications in the design of processes. However, this can be accomplished by identifying the key points to be improved in the process. End milling, a fundamental aspect of the industry, presents an even greater challenge due to its highly dynamic, intricate (complex, nonlinear, and interdependent) and process-dependent nature. The process design comprises numerous parameters (inputs) such as machine tools, cooling and lubrication systems, cutting tools, tool holders, and workpiece characteristics. Following the selection of the workpiece material in accordance with the component to be manufactured, the cutting tool and process parameters (cutting parameters) must be adjusted to match the process. The purpose of this is to ensure the best possible optimization of the desired output. Among many outputs, surface roughness affects many features such as the tribological performance of the part, fatigue and corrosion resistance, stress concentration, and aesthetic appearance. This ostensibly complex process design can be understood through optimization methods that analyze the relationship between the specified inputs and the desired outputs for improvement. The notion that conventional optimization methods, including Taguchi, response surface methodology, analysis of variance, grey relational analysis, and statistical regression, may have limitations in addressing the growing complexity of modern applications has encouraged the use of artificial intelligence approaches as powerful alternatives. Among these sophisticated soft computing models, artificial neural network approaches have emerged prominently due to their exceptional capabilities in delivering high-accuracy processing for intricate pattern recognition and prediction problems. This review evaluates the literature chronologically on the use of artificial neural network approaches in end milling and surface roughness output, which have become popular over time. Furthermore, the objective is to review the effectiveness of ANNs rather than solely evaluating the outcomes.

Optimizing end mill geometry is critical for improving performance and reducing costs in the high-volume manufacturing of tools, dies and molds. This study demonstrates a successful optimization framework for solid end mills machining 1.2379 cold-work tool steel, integrating Finite Element Analysis (FEA), Artificial Neural Networks (ANN), and Genetic Algorithms (GA). The optimized tool geometry, derived from four key design parameters, delivered substantial performance gains over an industrial reference (parent) tool. Our ANN-GA model achieved a remarkable predictive accuracy (R = 0.75–0.98) over the RSM model (R = 0.17–0.63) and identified an optimal design that reduced the resultant cutting force by approximately 11% (to 142.8 N) and improved surface roughness by 21% (to 0.1637 µm) compared to experimental baselines. Crucially, the new geometry halved the tool breakage rate from 50% to ~25%. Parameter analysis revealed the width of the land as the most influential geometric factor. This work provides a validated, high-performance tool design and a powerful modeling framework for advancing machining efficiency in tool, mold and die manufacturing.