Incorporated with constitutive relationship established by artificial need networks (ANN), a coupled thermoviscoplastic finite element procedure is developed for predicting the microstructure evolution in the hot forging process, considering the factors such as dynamic recrystallization, static recrystallization and grain growth etc. This software system is applied to predict the distributions of the grain size over the crosssection of stigmata, which is found to be in good agreement with the experimental results. The software can provide a fundament for optimizing technological parameters. Incorporated with constitutive relationship established by artificial need networks (ANN), a coupled thermoviscoplastic finite element procedure is developed for predicting the microstructure evolution in the hot forging process, considering the factors such as dynamic recrystallization, static recrystallization and grain growth etc. This software system is applied to predict the distributions of the grain size over the crosssection of stigmata, which is found to be in good agreement with the experimental results. The software can provide a fundament for optimizing technological parameters.