Abstract: Objective The aim is to accurately predict the microstructure and properties of evaporator small nozzles after temper bead welding repair, and to assist in formulating the repair process. Methods A general-purpose plug-in for temper bead welding data processing and prediction visualization was developed through Python programming based on Abaqus software. The plug-in not only simplifies the construction and management of neural network prediction models for microstructure and properties of temper bead welding, but also enables the prediction of their microstructure and properties based on the temperature field simulation results of the temper bead welding process. For 18MnD5 low-alloy steel, a dataset of microstructure and hardness for temper bead welding was established by conducting surfacing experiments combined with numerical simulations. Using the aforementioned plug-in, a neural network prediction model for microstructure and hardness of temper bead welding was constructed through sample training. The temperature field, microstructure, and hardness of single-layer multi-pass surfacing were predicted, and accuracy of prediction results was verified by surfacing experiments. Results The comparison results show that the simulated molten pool morphology is in good agreement with the macroscopic morphology of weld cross-section. The prediction results of microstructure and hardness are in good agreement with the results of surfacing experiment. This indicates that temper bead welding data processing and prediction system has high accuracy and can be used for predicting microstructure and properties of temper bead welding. The plug-in was further applied to simulate the excavation and repair process and predict microstructure and hardness of an evaporator small nozzle. The effect of interlayer grinding on microstructure and properties of temper bead welding was analyzed. Conclusion The research results have important guiding significance for formulating repair processes of evaporator small nozzles.