An Inverse RBF Neural Network for Plasma Arc Welding Control

The primary issue of this research is to create an inverse neural network model for automatic welding parameter control. In this research, there have a new model of inverse radius basis function neural network (IRBFN) proposed, which is an indirect approach. The development processes first build a feed-forward radius basis function neural network (FRBFN), which learns the influence of the input to the output by the training cases and then generates the IRBFN algorithm based on the FRBFN. According to the delta learning rule, the variables (or connection weights) are adjusted based on the error obtained from the comparison of the outputs from the IRBFN with the desired outputs in order to find the parameter design for the satisfaction of the quality requirements. In this research, an inverse backpropagation neural network (IBPN) was developed for the comparison to the IRBFN. From the experiments, it shows that the performance of the IRBFN is much better than that of the IBPN in accuracy and time-taken. Besides, the IRBFN has better stability of accuracy than the IBPN, if their standard deviations of MSE are compared.