Modeling for Robot High Precision Grinding Based on SVM

To improve the removal control for robot grinding process, we propose a modeling method based on SVM (support vector machine) regression. By analyzing a group of measurable variables relevant to grinding removal, such as robot's speed, contact force and curvature of the workpiece's surface, a regression model is built using machine learning method to predict the grinding removal. In this way, the analysis on a series of complicated dynamic variables could be avoided. The experimental results show that this method could achieve good performance. The prediction accuracy of the model reaches higher than 90%, which basically meets the demand of practical grinding.