An approach to model based fault diagnosis of industrial robots

Abstract

A method for the incipient fault diagnosis of industrial robot mechanics is proposed. It is based on mathematical models expressed in terms of nonlinear differential equations for a robot's different axes. The parameters of these models directly represent characteristic physical quantities (process coefficients), which are calculated by a suitable parameter estimation procedure. Additionally, a simple but efficient approach to the statistical classification of the determined values is devised. The proposed fault diagnosis method needs only measurements from sensors which are necessary for robot control purposes, therefore no additional sensors ae required. Experimental results obtained from an industrial robot show the feasibility of the proposed approach. Practical issues concerning implementation and application are discussed.