The weld shapes of St12 and St16 deep-drawing steels were influenced by transient temperature distributions. Based on the influence of temperature dependence, latent heat of fusion, convection and radiation, the weld shapes prediction model was built. Additionally, the first and last nodes control method was used in heat loading in order to make sure all nodes could be selected. Experimental results showed that there were no significant differences between the predictive and actual weld shapes. Furthermore, multiple regression prediction models of grain size in weld and heat affected zone (HAZ) were put forward based on PLS. The prediction accuracy of grain size in weld and HAZ was above 95%. Therefore, the proposed models are reasonable and applicable to examine deep-drawing steel for demonstrative purposes. 