Genetic Algorithm Optimisation of a TNT Solidification Model

Abstract

The control of the solidification process of energetic materials is important to prevent manufacturing defects in high explosive ammunitions. The present work aims to propose an optimisation procedure to determine the value of the model parameter, avoiding the traditional trial and error approach. In this work, the solidification of TNT has been numerically modelled employing apparent heat capacity method and the model parameter was optimised using genetic algorithm. One dimensional numerical model has been solved in Comsol Multiphysics Modeling Software and the genetic algorithm code was written in Matlab. The Neumann's analytical solution of the solidification front was used as a reference to build the fitness function, following the inverse problems concepts. The optimum model parameter has been predicted after 20 generations and among 30 candidate solutions for each generation. The numerical solution performed with the optimised model parameter has agreed with the analytical solution, indicating the feasibility of the proposed procedure. The discrepancy was 3.8 per cent when maximum difference between analytical and numerical solutions was observed.