Digital twin-enabled surface quality prediction and optimization in dry turning of Ti6Al4V using ANFIS and genetic algorithm

Abstract Ensuring high surface quality while minimizing tool wear remains a significant challenge in machining titanium alloys, particularly for applications in the aerospace, biomedical, and precision manufacturing industries. This study presents a novel integrated approach that combines experimental design, statistical analysis, intelligent modeling, and optimization to predict and minimize surface roughness (Ra) during the dry turning of Ti-6Al-4 V. Spindle speed (SS), feed rate (FR), and depth of cut (DoC) were selected as machining parameters, while cutting force (Fz), surface roughness (Ra), chip thickness (CW), and flank wear served as response variables. A Design of Experiments (DoE) with 20 trials was generated using Response Surface Methodology (RSM), and each trial was repeated thrice for accuracy using a fresh tool tip to eliminate tool wear effects. Analysis of Variance (ANOVA) identified significant parameters, and regression models were developed with R2 values of 97.4% (Fz), 86.64% (Ra), 88.35% (CW), and 83.96% (flank wear). A Fuzzy Inference System (FIS) dataset was used to train an Adaptive Neuro-Fuzzy Inference System (ANFIS) model, which was validated through 12 additional experiments. The ANFIS model demonstrated high prediction accuracy, with RMSE values ranging from 0.015 to 0.038 µm for Ra. Finally, a Genetic Algorithm (GA) was applied to identify optimal machining conditions: SS = 660.10 rpm, FR = 30 mm/min, and DoC = 0.3 mm, resulting in a minimum predicted surface roughness of 0.3392 µm. This integrated framework provides a robust solution for predictive quality control and process optimization in intelligent manufacturing environments.