V3I11P14

Development of an AI/ML-Based Predictive Model for Surface Roughness Optimization in CNC Turning Operations

Choure Ashok Apparao1*, Deshmukh P.A.2, Dr. Hulas Raj A. Tonday3

Abstract

In precision manufacturing, surface roughness (Ra) plays a critical role in governing component reliability, tribological performance, and mechanical fatigue life. Conventional empirical models and traditional Response Surface Methodology (RSM) frequently struggle to capture the intricate, non-linear characteristics of lathe turning operations. To overcome this limitation, this study introduces a hybrid predictive framework that combines machine learning regression with Genetic Algorithm (GA) optimization. Utilizing experimental data compiled from diverse machining parameters, four distinct regression algorithms—Multiple Linear Regression (MLR), Support Vector Regression (SVR), Random Forest (RF), and Extreme Gradient Boosting (XGBoost)—were benchmarked. The optimized XGBoost architecture outperformed the others, achieving superior predictive capability with a Coefficient of Determination (R^2) of 0.984 and a Root Mean Square Error (RMSE) of 0.042 \mum. By integrating this surrogate model with a GA-based metaheuristic search, the optimal cutting conditions reduced surface roughness down to 0.62 \mum, representing a 57.2% enhancement relative to standard baseline parameters

Keywords:

CNC Turning, Surface Roughness (Ra), Machine Learning, XGBoost, Genetic Algorithm, Optimization