Prediction of tensile property of TIG welding joints in GH99 alloy by artificial neural network

Multilayer BP network model based on improved algorithm was established with Matlab 7.0 software to predict the tensile property of TIG welding joints of GH99 superalloy. The input parameters of the model consisted welding current, welding speed, pulse frequency, remelting times, plate thickness, assembly clearance and weld groove. The outputs of the Artificial Neural Network(ANN) model included property parameters, such as tensile strength, yield strength and elongation. The calculated results showed that the multilayer BP network model based on improved algorithm could predict the tensile property of TIG welding joints of GH99 superalloy. The calculated values were in good agreement with measured data, and the average relative errors between calculated values and measured data of tensile strength, yield strength and elongation were -0.76%, 1.71% and 2.30% respectively.