摘要
为提高确定炸药JWL参数的效率,提出了一种根据圆筒试验数据快速准确标定炸药JWL参数的三点标定法。三点标定法首先对BP神经网络进行训练,使其可预测能够描述圆筒试验中能量分布和能量转换关系的圆筒能量模型,随后采用遗传算法寻找训练好的BP神经网络的全局最优解,即为炸药JWL参数。结果表明,基于圆筒试验中三个特定点的数据和圆筒能量模型,随机生成2000个样本对双隐含层结构的BP神经网络进行训练,训练好的BP神经网络可较好地预测任意一组R_(1)、R_(2)和ω代表的能量差值;使用三点标定法分别对Φ25.4mm和Φ50mm的圆筒试验数据进行处理,得到了多种炸药的JWL参数并代入有限元软件进行数值检验,结果显示仿真和试验吻合较好,证明了三点标定法适用于两种不同直径的圆筒试验。
To improve the efficiency of determining the JWL parameters of explosives,a three-point calibration method was proposed,which can quickly and accurately determine the JWL parameters of explosives according to the cylinder test data.Firstly,BP neural network was trained by the three-point calibration method to predict the cylinder energy model which can describe the energy distribution and energy conversion in the cylinder test.Then,genetic algorithm was used to find the optimal global solution of the trained BP neural network,which was the set of JWL parameters of explosives.The results indicated that based on the data of three specific points in the cylinder test and the cylinder energy model,2000 samples were randomly generated to train the BP neural network with double hidden layer structure.The trained BP neural network can predict the energy difference of any group of R_(1),R_(2) andωwell.The three-point calibration method was used to process the data fromΦ25.4mm andΦ50mm cylinder test,and the JWL parameters of various explosives were obtained and put into the finite element software for numerical verification.The results showed that the simulation and the test were in good agreement,which proved that the three-point calibration method could be applied to cylinder tests of two different diameters.
作者
崔浩
郭锐
顾晓辉
宋浦
杨永亮
江琳
俞旸晖
CUI Hao;GUO Rui;GU Xiao-hui;SONG Pu;YANG Yong-liang;JIANG Lin;YU Yang-hui(School of Mechanical Engineering,Nanjing University of Science and Technology,Nanjing 210094,China;Science and Technology on Combustion and Explosion Laboratory,Xi′an Modern Chemistry Research Institute,Xi′an 710065,China)
出处
《火炸药学报》
EI
CAS
CSCD
北大核心
2021年第5期665-673,共9页
Chinese Journal of Explosives & Propellants
基金
国家自然科学基金(No.11972197,No.21875109)。
关键词
爆炸力学
BP神经网络
圆筒能量模型
遗传算法
JWL状态方程
圆筒试验
三点标定法
explosion mechanics
BP neural network
cylinder energy model
genetic algorithm
JWL equation of state
cylinder test
three-point calibration method
