Comparison of Three Control Methods in Penetration Control of Pulsed Gas Tungsten Arc Welding

摘要

An artificial neural network model for backside bead width was established and three control meth-ods PID, fuzzy and neuron were designed, simulated and tested. The test results of bead-on-plate weld ofGTAW indicate that the artificial neural network (ANN) modeling and learning control method have more advan-tages than the conventional method. They show that the ANN modeling and learning control method is an effectiveapproach to real time control of welding dynamics and ideal quality.