基于AGEP-DNN的水下聚能装药比冲量预测模型

Prediction Model of Specific Impulse of Underwater Shaped Charge Based on AGEP-DNN

聚能装药比冲量是表征水下爆炸中冲击波对目标破坏作用的重要参数。为实现水下聚能装药比冲量智能预测,提出一种自适应基因表达式编程(adaptive gene expression programming, AGEP)优化深度神经网络(deep neural network, DNN)的聚能装药比冲量预测模型(AGEP-DNN)。考虑装药结构与比冲量数值之间的复杂非线性关系,通过AUTODYN软件建立有限元模型,对水下爆炸过程进行仿真,采用经验公式验证仿真数据的有效性;基于仿真实验数据,设计AGEP算法优化DNN超参数,构建AGEP-DNN模型,对比冲量进行智能预测。实验结果显示,AGEP-DNN聚能装药比冲量预测模型在9种对比智能预测模型中具有最优的预测精度。

The specific impulse of shaped charge is an important parameter in underwater explosions.It is used to represent the destructive effect of a shock wave on a target.In order to predict the specific impulse,a prediction model of shaped charge based on gene expression programming optimized deep neural network(AGEP-DNN)is proposed.Considering the complex nonlinear relationship between the structure and the specific impulse value of the charge,AUTODYN is applied to build finite element models.Empirical formulas are used to validate the data.Based on the simulation experimental data,an adaptive gene expression programming(AGEP)is designed to optimize deep natural network(DNN)hyperparameters.The AGEP-DNN model is constructed to intelligently predict the specific impulse.Experimental results show that among the nine prediction models,AGEP-DNN has the highest accuracy.