Abstract: To study the effective ways to improve the firing dispersion and solve its optimal design problem considering random factors, a calculation method of parameters interval oriented by the predefined indicators is proposed by using robust design theory based on random model and random optimization. A large caliber howitzer data is used to calculate firing dispersion based on Monte Carlo simulation and six degrees of freedom exterior ballistics model, and the statistical properties of firing dispersion are analyzed. A random model of robust design is built by calculating the objective criteria of robust design including unqualified probability and sensitivity index. Based on multi-objective genetic algorithm (NSGA-II) and BP neural network, the random model is optimized to determine the parameters interval of random factors which meet the requirements. The analysis results show that the proposed method can ensure the optimality and robustness of firing dispersion, and is feasible for the prediction of firing dispersion.

Key words: ordnance science and technology, firing dispersion, robust design, random factor, multi-objective genetic algorithm, neural network