摘要
应用粒子群神经网络模型对黑索今(RDX)基混合炸药冲击波感度的大隔板厚度值进行预测以减少试验量,节约试验成本。选取具有不同密度、空隙率、装药方式、RDX含量等特征的41组RDX基混合炸药,考察炸药实际密度、空隙率、RDX和附加物含量影响因素,通过分析它们与大隔板厚度值的非线性关系,建立大隔板厚度值与上述4个变量之间的粒子群算法优化神经网络模型,采用100进化次数,40种群规模进行计算。计算与试验结果表明:4个变量与大隔板厚度值之间的映射模型良好;模型预测值与试验值吻合良好,相对误差在10%以内。该粒子群神经网络模型预测值对RDX基混合炸药大隔板试验具有一定参考价值。
The large-scale gap thickness value of RDX-based composite explosive shock sensitivity is predicted by particle swarm neural network for reducing the number of tests and saving the test cost. 4l groups of RDX-based composite explosives with different densities, void ratios, charge structures and RDX contents are selected for test. The practical density of explosive, void ratio, RDX and additives content are taken into account as main influence factors. The nonlinear relationship among three influence factors and large scale gap thickness value is analyzed. The neural network model optimized by particle swarm algorithm is established for the above four variables and gap thickness value. The calculation re- sults show that there is a good mapping model between the four variables and large scale gap thickness value; the predicted values are in good agreement with the experimental results, and the relative error is within 10%. The predicted value of the particle sarwm neural network can provide reference for the large-scale gap test of RDX-based composite explosive.
出处
《兵工学报》
EI
CAS
CSCD
北大核心
2014年第2期188-193,共6页
Acta Armamentarii
基金
火炸药国防基础创新项目(20090371)
关键词
兵器科学与技术
混合炸药
冲击波感度
大隔板试验
神经网络
ordnance science and technology
composite explosive
shock sensitivity
large-scale gap test
neural network
