Abstract
Abstract
Robot failures are costly and difficult to diagnose. Breakdown data for robot-automated production lines, collected from automotive applications, showed that nearly half of robot failures are caused by positional error. A further quarter were attributed to drive failures. Positional error may be caused by a number of mechanical failure modes or by poor tuning of the control system. Testing of repeatability or absolute position in the work place is hard because the robot moves quickly, allowing little time for measurement. Measurement may be required in up to six axes. The end-effectors of a robot, e.g. welding guns, grinding wheels and sealant applicators, are powerful and unguarded, and are not compatible with conventional contact, or close proximity non-contact, transducers.
This paper presents significant new industrial evidence of the need for advanced diagnostic techniques in industrial robotics. Historical records for about 200 robots in automotive applications were analysed, identifying the major failure modes. Existing maintenance and condition monitoring methods were audited, and the problems of selecting condition monitoring methods for robots were analysed, identifying characteristics that differ from application in calibration. Measurement methods are briefly reviewed in the paper. It is concluded that non-invasive, non-contact methods are a prerequisite for operational effectiveness.