论文部分内容阅读
采用Monte Carlo模拟仿真方法,研究了分布选择对D-优化法参数估值精度的影响,进而分析了估值结果对火工品99.9%的可靠性外推结果的影响。仿真结果表明:在D-优化法感度试验中,无论样本的真实分布如何,选择错误的其它分布对均值的影响不大,尤其在样本量较小的情况下;同时,无论样本的真实分布如何,选择错误的其它分布对标准差的影响都较大,这就导致了估值结果对99.9%的可靠性外推结果的影响较大。综合分析仿真结果,得出在不了解样品感度分布时,应尽量选择正态分布或对数正态分布进行感度试验,从而对99.9%可靠性外推结果的的影响最小。
The Monte Carlo simulation is used to study the effect of distribution on the accuracy of D-optimization method. The effect of the estimation on the extrapolation of 99.9% of the reliability of the product is analyzed. The simulation results show that in the sensitivity experiment of D-optimization method, no matter what the true distribution of the samples, the other distributions of the selection errors have little effect on the mean value, especially in the case of small sample size; meanwhile, regardless of the real distribution of the samples , The other distributions with the wrong choice have greater impact on the standard deviation, which leads to the greater impact of the evaluation results on the extrapolation of the reliability of 99.9%. A comprehensive analysis of the simulation results shows that when we do not understand the sensitivity distribution of samples, we should try our best to select the normal distribution or lognormal distribution for sensitivity test, which will have the least impact on the extrapolation results of 99.9% reliability.