火箭弹气动学科代理模型构建方法研究

Research on constructing surrogate model of rocket aerodynamic discipline

针对火箭弹多学科优化体系中气动学科采用计算流体力学方法计算时间过长的问题,提出了一种综合运用CFD技术、试验设计技术、径向基函数神经网络技术构建火箭弹气动学科代理模型的方法,并对其流程进行了详细分析说明。通过算例分析,证明了该方法的可行性和有效性。该方法在保证一定精度前提下大大降低了火箭弹气动学科的计算周期。

Aiming at long calculation time when computational fluid dynamics method was used for optimization of rocket aerodynamic multidiseipline, a new method for constructing surrogate model for rocket aerodynamic discipline was put forward by means of computational fluid dynamics（ CFD ）, experiment design and radial basis function （RBF） neural network techniques ,and its flow chart was analyzed in detail. Through example analysis, the method was proven to be feasible and effective, Under the high-precision precondition, the RBF neural network surrogate modeling method can greatly reduce computational time.