Model-free damage prediction of brittle materials based on particle swarm optimization coupled with a probabilistic fission method

This study aims to develop a model-free damage prediction algorithm to predict directly local mesoscopic elasticity modulus and damage distribution of brittle materials based on only the monitored global structural response such as force-displacement curve, which can overcome complex theoretical modeling material behavior. The algorithm is established based on the modification of particle swarm optimization algorithm by establishing a probabilistic fission method, which can consider the physical mechanism that damage is correlated positively with the stress level in the optimization process. A representative numerical example of a Brazilian disk test is implemented to verify the algorithm. The damage prediction result is compared with the result obtained from the traditional particle swarm optimization algorithm and the corresponding experimental result. As can be seen from the comparison results, the ability and effectiveness of the algorithm are supported. Furthermore, the effect of different divisions of representative volume elements in the algorithm is analyzed to support the conclusion that fine division of representative volume elements can obtain refined prediction results, but it can significantly increase the prediction error and computation time. The recommended size of representative volume element is about 2 mm for the numerical case in this study.