Prediction of Microstructure Behavior During Large-Scale Hot Forging of Ni-Based Superalloy

Abstract

For the production of large heat-resistant components, improving the prediction accuracy of microstructural behavior by the finite element method leads cost and time reduction in process optimization. To perform this task, it is essential to predict the recrystallization behavior accurately, during hot forging and heat treatment of Ni-base superalloys. So far, many researchers have used the Avrami equation to predict the recrystallization behavior. However, this uses experimental coefficients determined empirically for each process, so it is not suitable for predicting the microstructure of large-scale hot forged parts with different complex deformation and thermal histories at different locations. Because the recrystallization behavior is considered to depend on the path, and on the temperature, strain, and strain rate at that time. Therefore, an incremental form of a prediction model based on the Avrami equation has been developed as a solution. Optimized parameters for the incremental model were obtained by using a genetic algorithm which is an adaptive heuristic search algorithm. At this time, it is not enough to use only the results of small tests at the laboratory level as training data for genetic algorithm. This is because the thermal histories and distributions have a large difference between the small specimens and the large parts. In this study, 1,500 and 50,000 t class forging tests were performed, and the parameters in the prediction model were determined using these results as learning data. As a result, the average error of prediction was reduced to about half.