Identification, classification and management of errors in automated component assembly tasks

Manufacturing industries are turning to advanced technologies in order to improve performance. While this is feasible in many industrial applications, it creates new problems in assembly. In particular, handling errors in automated assembly is crucial to the development of viable automatic assembly systems. Although many writers have recognized this as a problem, few have attempted to deal with it in a systematic fashion. In this preliminary paper the subject of errors in automated assembly is discussed, and a scheme for the classification of errors applicable to automated assembly systems is presented. It is suggested that errors are context dependent, and it is plausible to classify errors in close relation to a generic classification of assembly operations. Task analysis methods are used to define specific operational contexts, and which allow us to postulate likely errors. A worked example of this methodology and classification scheme is presented, using the traditional ‘peg-in-hole’ problem. The methodology facilitates a structured analysis of error conditions associated with particular assembly tasks, and by accounting for system characteristics it will then be possible to develop appropriate error management strategies. This work is currently being applied to characterize the Alvey Robotic Assembly Cell, as developed by Lucas Engineering and Systems Limited, and currently sited here at Birmingham. The overall aim of the work is towards the development of the EMA (error management) expert system that will assist the designer and go towards providing intelligent control systems capable of detecting imminent errors, taking corrective actions and/or recovering from them.