Abstract: The performance prediction in spot welding of the galvanized steel sheets are very important in the automobile body manufacturing.So it was studied with galvanized steel sheet GMW2 and DP600.Artificial neural networks(ANN) are used to describe the mapping relationship between welding parameters and welding quality. After analyzing the limitation of standard BP networks, the original model was optimized based on lots of experiments.Then a lot of experimental data about welding parameters and corresponding spot welding quality were supplied to the ANN for training. The results indicate that the improved BP network model can accurately predict the influence of welding currents on welding nugget diameters, depth of indentation and the tension-shear strength of welding spots.That is to say, the model can effectively predict the spot welding performance of the galvanized steel sheets.The forecasting precision is high enough to meet the practical need of engineering and has some application value.