A Hybrid Modeling and Optimization Framework for Finish Milling Using SVR, NSGA-II, and Entropy-Based TOPSIS

Abstract

This study presents a hybrid methodology for optimizing finish milling processes by integrating predictive modeling, evolutionary algorithms, and multi-criteria decision-making techniques. The target output responses include surface roughness (Ra), Material Removal Rate (MRR), and cutting force (Fc), modeled as functions of cutting speed (Vc), feed per tooth (fz), axial depth of cut (ap), and radial depth of cut (ae). Support Vector Regression (SVR) models yielded high accuracy for Ra (R² = 0.926) and MRR (R² = 0.999), while a second-order polynomial regression model excelled for Fc (R = 0.938, Radj 2 = 0.869). These models were integrated into Non-dominated Sorting Genetic Algorithm II (NSGA-II), generating a Pareto front of 100 optimal solutions. Entropy-weighted Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) ranked these solutions, identifying the best trade-off at Ra = 0.698 μm, MRR = 5228.24 mm³/min, and Fc = 522.17 N, with a TOPSIS score of Ci = 0.878. This solution significantly enhances productivity while maintaining acceptable surface quality and cutting force. The workflow was implemented in MATLAB, demonstrating the efficacy of this hybrid approach for advanced manufacturing. This hybrid framework provides a practical tool for real-time process optimization and decision support in smart manufacturing environments.