Prediction of mechanical properties of welded joints based on RBF neural network

A RBF neural network model on the welding parameters and the mechanical properties of TC4 titanium alloy joints welded by TIG welding was established.The 27 sets of experimental data are used to train this model, and other 9 sets are used to simulation.The results show that the welding parameters including welding current, welding speed and argon gas flow rate as network input parameters can predict mechanical properties including tensile strength, bend strength and ductility.The efficiency and accuracy of the RBF network predictions have improved comparing with common standard BP neural network, which overcome the BP network's disadvantage of long time to train and plunge in part smallest easily.