Almost every discipline in aerospace, from Guidance, Navigation, and control to Propulsion and Structures, has yielded itself to the power of computational intelligence. In this study, computational intelligence is applied for optimization of star grain geometry of a solid rocket motor missile to achieve maximum range under the constraint of axial overload. A simple Genetic Algorithm is shown capable enough to evolve to the optimal solution. Some techniques for optimization efficiency are introduced. Improved average convergence is achieved by utilizing Design of Experiments technique to create the first generation of population of candidate solutions, instead of randomly generated population. Computational time is then drastically reduced by incorporating pre-trained Neural Network as a Meta Model to replace the star grain regression and trajectory simulation modules. However, Neural Network was trained by exact solutions of some space filling candidate designs selected by Latin Hypercube Sampling technique.
展开▼
