The mechanical properties of welding material is correlative with the chemical composition Artificial neural network (ANN) program to predict mechanical properties — yield strength, tensile strength, elongation and average Charpy impact toughness — of welding material is established by Visual C + + 6.0 based on improved BP arithmetic with momentum factor, in which one input layer with 13 nodes, one hidden layer with 23 nodes, one output layer with 4 nodes, and Sigmoid activation function are included. The 20 samples are from the experimental data of semi-automatic welding material of X70. The average maximum relative error of the 4 mechanical properties is less than 0.5%. Based on the program, the influence of the chemical composition, such as C, S, P, Si, Mn, Cr, Ni and Al on the mechanical properties is analyzed. The results show that the different element has different influence on the mechanical properties. For non-metallic elements, the mechanical properties are becoming worse with the increase of composition, in which the influence of C is primary, then P and then S. For metallic elements, the influence is greater and more complex than that of non-metallic ones.

This content is only available via PDF.
You do not currently have access to this content.