Machine Learning-Based Prediction of Surface Roughness in Electrical Discharge Machining

Electrical discharge machining (EDM) and wire EDM (WEDM) involve complex nonlinear interactions, making surface roughness prediction challenging with traditional models. This paper introduces a machine-learning framework for pre-dicting arithmetic surface roughness (Ra), comparing regression, ANN, ANFIS, SVR, ELM, GPR, RF, and boosting methods. Literature shows no universally optimal model due to variations in materials, regimes, and datasets. The proposed ensemble pipeline emphasizes Random Forest regression, supported by SVR, GPR, ANN, and boosting benchmarks. Key elements include design-of-experiments data, physics-informed features, nested cross-validation, hyperparameter tuning, permutation im-portance, uncertainty analysis, and independent validation to ensure generalization beyond specific datasets.