战斗部虚拟试验多级模型集成方法研究

Multi-models Integration Approach in Warhead Virtual Experiment

弹药虚拟试验预测模型建模过程需多次调用数值仿真模型,对于运行耗时的大型三维计算模型将面临严重的计算复杂性问题。当计算模型存在多种精度时,提出一种基于自回归模型的预测模型建模方法,可集成多种精度的计算模型,以较少高精度模型和较多低精度模型分别获取训练样本,建立效率与精度平衡的贝叶斯预测模型。设计某型中小口径穿甲弹穿甲威力虚拟试验算例验证该方法的高效性。

Approximation model in ammunition virtual experiment is built on much runs of simulation codes. However, It will be a serious computational problem when the codes are expensive. An autoregressive model combining simulation codes at different levels of sophistication was proposed in virtual experiment, obtaining a more accuracy Bayesian approximation model at lower computational cost. An example demonstrates that the approach is efficient.