A COMPARISON OF SUPPORT VECTOR MACHINES WITH ARTIFICIAL NEURAL NETWORKS FOR THE PREDICTION OF THERMAL ERRORS IN MACHINE TOOLS

摘要

This paper attempts to analyse a powerful and novel methodology in the effective real-time prediction of thermal errors in machine tools. Most of the conventional error compensation systems have employed standard techniques of error prediction namely multi-variate regression analysis, artificial neural networks etc. However, support vector machines (SVM) has emerged as a powerful data training algorithm extensively used in the field of computer vision. Herein, we make a study of the feasibility of employing this technique to the much-explored field of thermal error prediction. The focus of this paper is to make a detailed comparison of this technique with neural networks, the most widely used method until now. As the accuracy of thermal error compensation systems depends on how well it is able to arrive at a meaningful prediction under conditions not previously encountered during the training or 'learning' phase, the performance of SVM holds special significance.