Studying the Quality of Resistance Spot Welding Joints Using Bayesian Networks

Resistance spot welding is used to join two or more metal sheets together, and the technique is used widely in the automotive industry, for example. Furthermore, other metal-to-metal connections, such as wire-to-wire joints in the electronics industry, are accomplished by resistance spot welding. Application-specific measures, such as the diameter of the welding spot, define the quality of the joint. By using non-destructive methods to assure the quality of the welding spot, the need for destructive testing can be reduced, which leads to increasing profits. In this work, the reasons for variation in welding spot diameters were studied using computational methods, mainly Bayesian networks. The features used in the study were extracted from histograms calculated from signal curves measured during the spot welding process. A new approach to modelling the quality of welding joints was developed and evaluated with a large sample test set. The configurations that resulted in failed welding spots proved to be of special importance for the modelling procedure.