关键词: 神经网络, 双相不锈钢, 焊接, 残余应力, 预测

Abstract: With regard to the 2205 duplex stainless steel which is used for special shipbuilding in Chuandong Shipyard, BP neural network forecast models on welding residual stress and shrinkage deformation of 2205 duplex stainless steel ( DSS) have been respectively established, on the basis of the welding experiment of 2205 duplex stainless steel. The proposed forecast model is compared with the multiple linear regression method. The comparison indicates that the proposed forecast model can better simulate the non-linear relationships of welding residual stress and deformation between the process parameters, such as thickness, welding current, arc voltage and welding speed. The forecast results show that BP neural network is more advantageous both in the prediction accuracy and generalization capability than them ultiple linear regression method.

Key words: neural network, duplex stainless steel, welding, residual stress, forecast