Surface Roughness Prediction of the Grinding Workpiece Using a Multi-Verse Optimizer

Accurate prediction of surface roughness is important for ensuring machining quality and advancing intelligent manufacturing processes. In this study, we applied data-driven approaches using decision tree, support vector regression (SVR), and extreme gradient boosting (XGBoost) models to predict surface roughness based on key machining parameters. These methods were chosen due to their effectiveness in capturing nonlinear relationships in machining data. The multi-verse optimizer (MVO) was employed to fine-tune hyperparameters. To further enhance model performance and ensure physical interpretability, we employed a physics-guided loss function that constrains the machine learning models based on known physical phenomena. This approach ensures that the model predictions are more consistent with established physical laws. Data were collected from various machining parameters, including rotation speed, feed rate, cutting depth, force, and the abrasive size to ensure a comprehensive analysis of factors affecting surface roughness., and the result shows that SVR achieved the highest accuracy, with an R² of 0.877 and MAPE of 0.119, followed by XGBoost, while the decision tree exhibited the lowest generalization ability. These results confirm that the integration of physics-informed techniques significantly enhances the model’s reliability and physical consistency in predicting surface roughness.