In this paper we address the problem of fault diagnostics in industrial robots. The goal was to develop a method that automatically, accurately and in a generic way could identify and classify faults once they occur for any type of industrial robot used. Although a large number of diagnosis methods and relevant applications for industrial equipment already exist, the current research in the area of fault diagnosis of industrial robotic manipulators is rather poor, due to the large variability of faults, the unsteady and non-uniform operating conditions, the small amount of sensors used in industrial manipulators and the rather limited time records of the equipment. These restrictions present key challenges of the current research to be undertaken. In this paper we present a novel approach to perform fault diagnostics of industrial robotic systems using Support Vector Machines (SVM) and Discrete Wavelet Transform based feature extraction. Experimental results are obtained from an industrial manipulator used in the semi-conductor industry.
Comparison of Feature Selection Methods in Support Vector Machines
