Searching Optimal Welding Parameters with GA and BP

Traditional method used to determine best welding parameters is depended mainly on experience and is lack of precision. So a new hybrid method combined GA with BP to search proper arc-welding parameters automatically is proposed here, the main contents are described as follows. Testing schema is designed according to orthogonal form, data used for later work are obtained at first; then a BP neural network which is treated by GA is used to simulate welding procedure, special combination method of GA and BP is introduced in detail according to practical problem; in the third, GA is used for globally searching with the help of BP network gotten above. Because the standard GA algorithm aims at global optima, which may lead to negative welding parameters sometimes, an adaptive penalty coefficient is put into use during heredity operation and satisfied parameters are found out at last. Results manifest the correctness and effectiveness of this method. Method used here can also be put into use in similar optimizing problem.