Abstract: Objective Through BP (Back Propagation) neural network, thermal response curve and characteristic parameters of X80 pipeline steel multi-layer welded joints was predicted to solve the problem of difficulty and long cycle in obtaining microstructure distribution by traditional experimental methods, and to lay the foundation for further microstructure prediction. Methods Firstly, a finite element model of gas metal arc welding (GMAW) for X80 pipeline steel was constructed, and the reliability of the model was verified through the collected molten pool and temperature field information from welding experiments. Then, the thermal cycle curve and characteristic parameters of X80 pipeline steel multi-layer welded joints were extracted under different welding parameters with this model, and it was used as training samples for BP neural network to construct a thermal response prediction model based on BP neural network. Results A reliable and effective finite element model for GMAW multi-layer welding of X80 pipeline steel was constructed, and a thermal response characteristic parameter prediction model based on BP neural network was trained with this model. The prediction errors were all within 8%, and the lowest error reached 0.01%. Conclusion In this study, finite element calculations were used instead of experiments to obtain training samples, and a thermal response characteristic parameters prediction model based on BP neural network was constructed. The prediction error was controlled within a reasonable range, which could easily and efficiently predict the thermal response curve and characteristic parameters under different welding parameters. This provided data support for the subsequent prediction of the microstructure of X80 pipeline steel welded joints and had important engineering application value.