Controlling resistance spot welding using neural network and fuzzy logic

A control scheme able to compensate for variations or errors during automatic resistance spot welding to produce consistently sound welds was developed and demonstrated through simulation. Fuzzy logic control (FLC) was employed to overcome the lack of a precise mathematical model of the process. Electrode displacement, indicative of nugget growth, was used as the feedback signal to create appropriate actions to adjust power delivered in real time. Control action is generated from a rule based system constructed from experimental data for welds made under a wide variety of conditions. A neural network (NN) was constructed to provide process input-output relationships and tune the fuzzy rules off line. The FLC system was evaluated using the NN to describe electrode displacement as a function of the percentage maximum heat input and welding time. Simulation showed the potential of applying this control scheme to deal with the uncertainties of RSW in a typical automated production environment.