Input electrical impedance as quality monitoring signature for characterizing resistance spot welding

Poor quality of resistance spot welding (RSW) often causes quality issues like structural integrity and noise in the car body assembly. Research activities for reliable monitoring methods of RSW quality have therefore been extensive. So far, most of the monitoring methods found in literature are good for off-line utilization only and thus very expensive to apply. This paper introduces into a real-time and in-situ RSW quality monitoring method, which takes the input electrical impedance of the welding system as the monitoring signature. This signature is obtained by probing and processing the input voltage and current throughout the welding process. As input impedance characterizes a dynamic system, its variation with time reveals the conditions of the welding process which result in the final weld quality. By recognizing the pattern of the real part by an artificial neural network, we demonstrate that the weld quality could be classified non-destructively and automatically. Due to the fast signal collecting and processing, the quality monitoring is finished almost in real-time, i.e., classification can be completed before the next welding process is started. Another feature of the method is being in-situ because monitoring action does not jeopardize the welding operation or alter any of the welding parameters in general.