The numerical simulation of forging process allows to know the temperature and strain evolution in every point of the part. The metallurgical models are able to account of these evolutions in order to predict the grain size everywhere in the work piece. The aim of the present study is to predict the recrystallized fraction in various processes (hot die forging, ring rolling, or hammering). The recrystallized fraction is strongly dependent on the stored energy (work-hardening or dislocation density) contained in the material. As a matter of fact a predictive recrystallized fraction model based on dislocation density is developed. It predicts successfully the recrystallized fraction on small samples and parts. The dislocation density is connected with the broadening of (400) X-ray line profile. The line profile broadening measurement is related to the conventional 0.2% yield stress. The correlation between predicted and measured work-hardening will be done in a following work, which will enable to link the microstructure to mechanical properties and then to predict the yield stress within a numerical simulation of the forging process.