Neural control of weld pool in the robotic welding

This paper deals with some problems concerning the controlling of the weld pool shape. The model of the weld pool is represented by using the RC circuit, where the resistance R corresponds to the thermal resistance. The authors try to keep the voltage across the capacitor C constant, regardless of the variation of R, by controlling the applied voltage to the RC circuit. If the knowledge about the variation of the parameter of the plant is known, the performance of the controller may be improved. A neural network controller (NNC) with learning ability is applied to the control of the plant. A performance of NNC depends on the training data, the number of the unit in the hidden layers, and the input variables. A new method based on the expert´s knowledge is proposed to construct the network. That is, the authors determine the input variables from the pole assignment method. The validity of the NNC is verified by using numerical experiments