SURFACE ROUGHNESS PREDICTION IN CNC TURNING USING ARTIFICIAL NEURAL NETWORK (ANN)

Abstract

Computer Numerical Control (CNC) turning is one of the most widely applied precision machining technologies in modern manufacturing, where surface quality is a key determinant of product performance and reliability. Surface roughness (Ra) is recognized as one of the most critical parameters for evaluating machining results. However, reliance on operator experience in selecting machining parameters often leads to inefficiencies and inconsistent surface quality, indicating the need for more accurate predictive approaches. This study proposes an Artificial Neural Network (ANN)-based model to predict surface roughness in CNC turning using two distinct experimental configurations. The first experiment (Exp1) employs three identical factor variations, whereas the second experiment (Exp2) incorporates different factorial combinations to introduce broader variability. The developed ANN architecture consists of four dense layers with ReLU and LeakyReLU activation functions, complemented by dropout layers to mitigate overfitting arising from the relatively small dataset. The results show that the ANN model effectively learns the nonlinear relationships between machining parameters and Ra values. Furthermore, the model achieves higher predictive accuracy in Exp2, likely due to its more structured parameter variations. Overall, the findings demonstrate that ANN-based prediction provides a promising and efficient approach for enhancing accuracy in surface quality assessment within CNC turning operations