Artificial neural network to predict toughness parameter CVN of welded joint of high strength pipeline steel

The artificial neural network (ANN)model was developed with VC++6.0 based on multiplayer back propagation (BP) to analyze and predict the Charpy-V notch (CVN)impact toughness parameter of the pipeline steel welded joint.Based on the practical welding parameters of X70 steel, the mean energy input, wall thickness, preheat temperature, welding position and sampling position were used as the input parameters of ANN, which includes one hidden layer with 14 nodes and Sigmoid activation function.The 194 sets of data, obtained from the practical welding, were divided randomly into two parts, in which 150 were used as training data and the other as testing data.The influence of structure of ANN on prediction results was analyzed.The number of the sample whose error is less than 20% is about 77% in the total testing data.