Research On Integrated Design Methods Of Range Test Based On Bayesian Theory

Range test plays an important role in national defense research and development of weapon system, which constitutes the foundation for verifying equipment design philosophy, evaluating tactical and technical specifications, and aiding decision-making of equipment acquisition. Traditional test and evaluation(T&E) mode is on the basis of classical statistics, which needs large sample size of field test data, and consumes a huge number of test resources. So it no longer satisfies the requirements of equipment development. Therefore, integrated test and evaluation(IT&E) mode with combined use of simulation and field test is getting extensive recognition, which has become the future developmental trend of equipment T&E by now. Under IT&E mode, there are three major issues need to be addressed, namely rational experimental design with comprehensive consideration on different roles of simulation test and field test, sample size determination(SSD) as well as fusion and assessment of range test.Based on Bayesian theory, this dissertation makes systematic investigation on the above three major issues of equipment range test under IT&E mode. The primary contributions are listed as follows:Firstly, considering the features of range tests that mass of simulation data provides prior information for field test, power prior is put forward for weighting of simulation data with probable distribution heterogeneity to field test data, whose rationality in the design and assessment of range test is illustrated by analyzing optimality properties of both standard and normalized power priors in posteriror inference. Thus, it forms the foundation for subsequent researches.Secondly, Bayesian regularization based experimental design method is proposed for design of experiment(DOE) of range test, which solves the ill-posed problem in analysis of effects of factors. The DOE problem in range test is described through regression model. Regularization method is sugessted to use in factor effects analysis considering its stability in solving ill-posed problem, which lead to the proposed Bayesian regularization based factor effects analysis method. As to experimental design with certain information about underlying effects of factors on response, an optimal design with additional test points for possible factor effect items is put forward. While for experimental design with no information about underlying effects of factors on response, a space-filling design is suggested. The effectiveness of Bayesian regularization based DOE method in range test is verified through illustrative use in gun firing table test.Thirdly, a modified power prior based SSD method is put forward for sample sizes calculation of simulation test and field test in range test, which overcomes the contradiction between credibility risk and assessment precision while using standard power prior. The impact of prior sample size on prior variance of parameter of interest is analyzed, which lead to the index of test effect. Hence, a modified power prior elicitation method is given out based on test effect equivalence. With modified power prior framework, optimization equations for SSD of range test scheme are given out according to Bayesian SSD criteria and pre-posterior distribution features for parameter of interest. The validity of the proposed Bayesian SSD method for range test is verified with application example.Fourthly, normalized power prior based Bayesian fusion and assessment methods are given out for range test with multi-source information, which improves fusion architecture of multi-source histotical data in range test assessment. Generalized linear mixed model is used to describe fusion and assessment problems in range test. Considering mutual effects and weight constraints of multi-source information, different structural styles of normalized power prior elicitation methods are suggested. As to sample size assessment of range test, a normalized power prior based posterior variance analysis method is given out, whose effectiveness is illustrated with sample size assessment on design point. While for DOE assessment of range test, a normalized power prior based method by analysis of effects of factors is put forward with verification by gun firing precision test.Finally, the research achievements on three fundamental problems of range test under IT&E mode are summarized and used in the integrated design and assessment of some air-to-ground guided missile range test. The effectiveness of our proposed methods are demonstrated by analysis and assessment on integrated design of test.