The data of welding seam width and height were obtained by TIG welding experiments. Two models were established using BP neural networks: One can predict the weld seam dimensions by inputting the welding parameters, and the other model can perform oppositely. Originally by inputting given welding seam dimensions to the model one, the welding parameters can be predicted. Then change the parameters a little properly and input them to model two, the welding seam dimensions which would reflect the

more »

... eld parameters whether were needed can be acquired. So appropriate weld parameters can be chosen to control the weld seam dimensions by the two models, and good experiment results were acquired.