Support vector machines model for classification of thermal error in machine tools

Abstract

This paper addresses a change in the concept of machine tool thermal error prediction which has been hitherto carried out by directly mapping them with the temperature of critical elements on the machine. The model developed herein using support vector machines, a powerful data-training algorithm, seeks to account for the impact of specific operating conditions, in addition to temperature variation, on the effective prediction of thermal errors. Several experiments were conducted to study the error pattern, which was found to change significantly with variation in operating conditions. This model attempts to classify the error based on operating conditions. Once classified, the error is then predicted based on the temperature states. This paper also briefly describes the concept of the implementation of such a comprehensive model along with an on-line error assessment and calibration system in a PC-based open-architecture controller environment, so that it could be employed in regular production for the purpose of periodic calibration of machine tools.