Neural Networks in GTA Weld Modeling and Control

Solutions to modeling the Gas Tungsten Arc(GTA) Welding process using a non-conventional technique is presented here. This approach is a non-linear modeling technique employing neural networks which has exhibited the potential to learn to model the time response of a non-linear, multivariable system. This paper examines the feasibility of this approach an alternative to existing techniques Potential problems with this approach are also discussed. A control architecture using a second neural network is also suggested.