In this work, the accretion of fine-grained particulate matter into larger objects in high-speed granular flows, due to sudden disturbances in their mean velocity field, is investigated. A multibody collision model is developed whereby the coefficients of restitution and friction, as well as quantities such as the contact area and collision duration time, are implicit functions of the relative collision velocities and surfacial thermochemical reactions during impact. A recursive fixed-point multilayered staggering scheme is developed to simulate the resulting coupled non-linear system. Inverse problems are then constructed whereby transient flow conditions, reaction rates, particulate volume fractions, hardnesses, etc., are sought which deliver prespecified aggregate growth from a base starting particulate size. Classical gradient-based methods perform poorly, to this class of problems due to the fact that the associated objective functions depend in a non-convex and non-differentiable manner on the mentioned starting-state parameters. Furthermore, the results are very sensitive to the size of the control volumes selected. Therefore, due to the lack of robustness of classical gradient-based minimization schemes, a statistical genetic algorithm is developed whereby (I) the starting state-variables are represented by a “genetic string”, and concepts of evolutionary behavior, such as selective reproduction, are applied to a population of such strings in order to determine an optimal set of starting state-parameters and (II) sequences of control volumes, each containing a finite number of particles, are adaptively computed until the sequential change in the ensemble average of a population of control volumes all fall below a given tolerance. Three-dimensional numerical examples are given to illustrate the behavior of the model and the overall solution process.