In the car body assembly process, welding robots must weld various parts of the body, involving complex motion of the robot arm. Some programs place such a heavy load on the robot cables that they break, bringing the production line to a halt. In this study, Mahalanobis-Taguchi techniques were used to create a system that can diagnose a robot motion program in the programming stage and accurately predict whether the motion load placed on the welding cables will cause them to break. The techniques used included alternative characterization of time-series data and optimization of the unit space by parameter design. For a motion program diagnosed as a cable breaker, impact characteristics are extracted by using signal-to-noise ratios and are graphically represented to provide guidelines for improvement. The development of this diagnostic system has improved the efficiency of program development and reduced production line outage due to broken cables.
