Manufacturing process diagnosis using functional regression

The quality of many manufacturing processes depends on preset process conditions and is strongly affected by process faults. Fault diagnosis is therefore crucial in modern manufacturing practice. This paper presents a new fault diagnosis approach based on the functional regression method. Fault patterns in manufacturing processes are treated as functional responses of both the preset process conditions and process faults. With these time domain functions, a mathematical model is developed, based on which a nonlinear optimization approach is applied to fault diagnosis. The proposed diagnosis approach considers the interaction effects among the faults and process settings. Therefore, it is capable of classifying multiple simultaneous faults with various process conditions. The new diagnosis method is demonstrated using an example from the resistance spot welding process. An overall success rate of 88% has been achieved.