Efficient 3D reconstruction of random heterogeneous media via random process theory and stochastic reconstruction procedure

This paper introduces an algorithm for reconstructing three-dimensional (3D) heterogeneous media, in which the randomly distributed individual entities, such as aggregates, pores or other material phases with different sizes and shapes can be explicitly represented. The reconstruction procedure achieves high efficiency and reliability by integrating random process theory and stochastic reconstruction procedure. Specifically, the random process theory is applied to generate an initial model based on the statistics obtained from the digital image of heterogeneous material sample and then this coarse model is evolved by the simulated annealing algorithm to match the same statistical descriptors of sample structure. Two 3D models of different sizes are reconstructed. The autocorrelation function is applied to validate the accuracy of the reconstructed models. The comparison between the reconstructed models and the real model shows good agreement. The proposed method is suitable for reconstructing large size 3D random media for its high computational efficiency. (C) 2019 Elsevier B.V. All rights reserved.