Research On Robust Optimal Design Theory, Methods And Their Applications To Solid Rocket Motor Design

This dissertation made a systematic research on robust optimal design theory, the correlative research methods and their applications to solid rocket motor. The concepts and fundamentals of the robust optimal design were expounded, and the theory framework along with the implementation flow was proposed. The theoretical problems, such as response surface modeling, global sensitivity analysis, robustness analysis, multiobject optimization and sequential optimization based on response surface model, were researched in detail.The precision and efficiency of diverse response surface fit functions were analyzed by four test functions, and the applicability of different fit functions was deduced. On that basis, the sequential response surface modeling method and its sequential sampling strategies were studied. With the proposed strategy, the model fit precision and the computational cost of response surface modeling could be balanced.The definition of tensor product function was expounded, and its characteristic of convenient for probability integral was analyzed. Based on that characteristic, tensor product function was applied to robustness analysis and global sensitivity analysis, to overcome the computational complexity of robustness analysis in robust optimization processes, and to improve the efficiency of filtering variables.Based on the convenience of physical programming method in expressing design preferrence, the construction method of multiobject optimization model was studied. To get a well distributed Pareto set, the standard genetic algorithm was improved in two aspects. Simultaneously, the search method of Pareto frontier and the decision-making method of Pareto set were researched.The response surface model based sequential optimization method of single object optimization problems was studied. On that basis, the sequential optimization method applicable to multiobject optimization problems was researched. With the proposed method, the convergency of the response surface model based optimization result could be guaranteed.Finally, based on the theory and techniques researched above, the robust optimal design model of solid rocket motor was built, and a set of robust solid rocket motor design were derived from that model. Comparison between robust design result and traditional design result illustrated the scientificity and effectivity of robust optimal design.