Neural-network-based self-organized fuzzy logic control for arc welding

Abstract

Fuzzy logic control (FLC) is becoming an attractive technique to control processes in welding mainly due to its ability to solve problems in the absence of an accurate mathematical model. In this paper, a novel technique, that combines both FLC and neural network (NN) techniques is presented to control the gas tungsten arc welding (GTAW) process. This technique overcomes limitations such as the dependency on the experts for fuzzy rule generation and non-adaptive fuzzy set. The adaptation of membership function as well as the self-organizing of fuzzy rule are realized by the self-learning and competitiveness of the NN. This approach facilitates the automatic determination of the fuzzy rule and in-process adaptation of membership function for an advanced welding process control. This overcomes the limitations of a fixed membership function, which cannot guarantee the required performance in a highly time-variable environment such as an arc-welding process. The proposed technique has been verified to be highly effective in an arc-welding process in which the welds bead width is regulated. Computer simulations confirm that the characteristics of the system have improved notably when compared with the currently available methods.