火炸药静电感度试验数理统计方法分析与优化

Analysis and Optimization of Mathematical Statistics Methods for Electrostatic Sensitivity Test of Explosives

为提高火炸药静电感度试验结果的准确性,特别是小概率发火能量计算结果的准确性,通过蒙特卡罗方法数值模拟试验,研究了升降法、兰利法、最优化D法三种典型数理统计方法的原理和特点,探索提出了离散分布试验测点的新思路,对兰利法进行了优化和改进。结果表明,最优化D法在小概率发火能量计算上准确度最高,可在250次试验内获得误差(68.26%置信区间)小于10%的百万分之一概率静电引燃刺激量,且试验结果具有稳定性,但试验程序和数据分析计算过程复杂,可操作性不强,不适用于常规静电感度试验。改进后兰利法试验结果的准确度与最优化D法相当,且其程序和计算过程复杂度远低于最优化D法,可操作性更强,便于常规静电感度试验使用。

In order to improve the accuracy of electrostatic sensitivity test results of explosives,especially the calculation results of the small probability ignition energy,the principles and characteristics of three typical mathematical statistics methods,namely,up-and-down method,Langlie method and D-optimality method,were studied through the Monte Carlo numerical simulation test,where a new idea of discrete distribution test points was proposed,and Langlie method was optimized and improved.The results shows that the D-optimality method has the highest accuracy in the calculation of the small probability ignition energy.The less than 10%error(in 68.26%confidence interval)of one millionth probability of electrostatic ignition stimulus can be obtained within 250 tests,and the test results are stable.However,the test procedure,data analysis and calculation process are complex with poor operability,so it is not suitable for conventional electrostatic sensitivity test.The accuracy of the test results of the improved Langlie method is comparable to that of the D-optimality method,and the complexity of its program and calculation process is far lower than that of the D-optimality method.It has stronger operability and can be conveniently used in the conventional electrostatic sensitivity test.