Deep Learning for Manufacturing: Surface Roughness Prediction in CNC Turning via 1D CNN

Authors

Amar Awad Mohammad Al-Sumaidaee, Ministry of Electricity, General Company for the Production of Electric Power for the Central Region, Granada Square, 10044 Baghdad, Iraq.Follow

Abstract

Optimisation of machining parameters is crucial to achieve an excellent surface finish, with surface roughness serving as the primary indicator of product quality. This study investigates the application of a one-dimensional Convolutional Neural Network (1D CNN) to predict surface roughness in Computer Numerical Control (CNC) turning processes. In contrast to conventional machine learning methods, the proposed CNN captures local dependencies within the tabulated input features, which include cutting speed, feed rate, depth of cut, and force components (Fx, Fy, Fz, and resultant force). A laboratory data set was used for model training, with the mean square error serving as the loss function. The performance was evaluated using the mean absolute percentage error (MAPE), the mean absolute error (MAE), and the coefficient of determination (R2). The experimental results show that the CNN achieves high prediction accuracy (e.g., R2 = 98.95%, MAE = 0.0515, and MAPE = 3.4087), outperforming traditional baseline models. The results highlight the potential of CNNs as a robust approach for tabular regression tasks in manufacturing and emphasis their value for data-driven optimization of machining parameters for defect-free production.

Recommended Citation

Al-Sumaidaee, Amar Awad Mohammad (2026) "Deep Learning for Manufacturing: Surface Roughness Prediction in CNC Turning via 1D CNN," Engineering and Technology Journal: Vol. 44: Iss. 1, Article 14.