Self organizing fuzzy linguistic control with application to arc welding

Presents a self organizing fuzzy linguistic control strategy that is based on on-line modification of the control rules according to the extent of deviation of the process output from the output of a given reference model. Accordingly, the learning/adaptation algorithm, which is based on the hill climbing approach, modifies the parametrized characteristic functions of the fuzzy subsets describing the control rules, such that the meaning of each rule is iteratively changed to reflect new information regarding the behavior of the process. Simulation results show that this technique improves the overall response of the system in presence of asymmetric dynamic characteristics of the process, proving that this strategy can be a suitable alternative to ordinary fuzzy control. In order to study this self organizing scheme, the authors used a simulation model with similar characteristics as the gas metal arc welding process. This model simulates the variation of the peak surface temperature of the workpiece directly underneath the weld bead in response to changes in the arc current, or more accurately, changes in the electrode wire feedrate