OPTIMIZING STOCHASTIC PROCESS FOR EFFICIENT MICROSTRUCTURE RECONSTRUCTION

摘要

To improve the efficiency in microstructure reconstruction is critical to build a high resolution statistical stable representative volume element practically. Using correlation function to reconstruct image using simulated annealing is revisited in this study. Different cooling schedules in simulated annealing were utilized and compared. Dramatic increase of computation efficiency has been achieved by optimizing cooling schedule, making it feasible for future computation intensive study on upscaling prediction.