Fragmentation Modeling Using the Expectation Maximization Algorithm

摘要

The modeling of fractures typically uses the Poisson distribution to select the number of particles resulting from a fracture of a solid body. This applies to bullets when striking a plate. This report develops a method to model the number of subparticles using a mixture of Poisson distributions. The goal is to find the mixture with the smallest number of components that fits the data. Rather than attempting to find the distribution of lambda, it is assumed the values of lambda are clustered and the centroids of these clusters constitute a useful model.