Observer-based supervision and fault detection in robots using nonlinear and fuzzy logic residual evaluation

A high degree of automation in flexible production units require powerful tools for supervision and fault detection to maintain quality and productivity. In this paper, an observer-based fault detection method is proposed which makes use of non-measurable process information instead of installing as many sensors as possible. The detection method is reviewed and applied to the fault detection problem in an industrial robot, using a dynamic robot model. The robot model is enhanced by the inclusion of nonlinear friction terms. A new residual evaluation approach of model-based fault detection methods is investigated for processes which exhibit unstructured disturbances, arising from model simplification. The present analytical approaches are applicable only to structured approaches. In this paper a fuzzy-logic approach is presented which is capable to address unstructured disturbances as well. Finally, some practical results for an industrial robot example are presented