Precipitation Criterion for Inhibiting Austenite Grain Coarsening during Carburization of Al-Containing 20Cr Gear Steels

AlN precipitates are frequently adopted to pin the austenite grain boundaries for the high-temperature carburization of special gear steels. For these steels, the grain coarsening criterion in the carburizing process is required when encountering the composition optimization for the crack-sensitive steels. In this work, the quantitative influence of the Al and N content on the grain size after carburization is studied through pseudocarburizing experiments based on 20Cr steel. According to the grain structure feature and the kinetic theory, the abnormal grain growth is demonstrated as the mode of austenite grain coarsening in carburization. The AlN precipitate, which provides the dominant pinning force, is ripened in this process and the particle size can be estimated by the Lifshitz-Slyosov-Wagner theory. Both the mass fraction and the pinning strength of AlN precipitate show significant influence on the grain growth behavior with the critical values indicating the grain coarsening. These criteria correspond to the conditions of abnormal grain growth when bearing the Zener pinning, which has been analyzed by the multiple phase-field simulation. Accordingly, the models to predict the austenite grain coarsening in carburization were constructed. The prediction is validated by the additional experiments, resulting in accuracies of 92% and 75% for the two models, respectively. Finally, one of the models is applied to optimize the Al and N contents of commercial steel.

Automated summary

AlN precipitates are commonly used to pin austenite grain boundaries in high-temperature carburization of special gear steels. This study quantitatively examined the influence of Al and N content on grain size after carburization through pseudocarburizing experiments, finding that AlN precipitates play a key role in determining grain growth behavior. Models predicting austenite grain coarsening in carburization were constructed and validated with experimental data, showing accuracies of 92% and 75% for the two models, respectively.