基于代理模型的制导子弹气动外形优化设计

目的 针对传统气动外形优化设计依赖于高精度仿真耗时过长的问题，为有效提高气动优化设计的效率，提出一种基于代理模型的气动外形优化方法。方法 以12.7 mm制导子弹为研究对象，通过CFD仿真，分析尾翼收缩段长度、弹底半径和尾翼收缩扩张段交界处半径对制导子弹飞行过程中阻力系数的影响。基于代理模型技术，综合运用试验设计、参数化建模、CFD技术和四阶响应面模型，构建制导子弹气动外形优化设计代理模型，以制导子弹飞行过程中的最小阻力系数为优化目标，结合遗传算法对制导子弹外形参数进行优化。结果 对比四阶多项式响应面和Kriging模型建立的代理模型预测精度，与测试样本点CFD仿真计算结果相比，四阶响应面代理模型阻力系数预测值的平均误差为0.386%。表明四阶响应面代理模型能够有效替代CFD仿真，可用于预测不同外形参数下制导子弹的阻力系数。结论 相对于最初的制导子弹外形，制导子弹的阻力系数下降了14.59%，有效减少子弹飞行过程中的能量损失和缩短了优化时长。该方法在保证精度的前提下，减少了制导子弹的设计周期，为相关工程应用与研究提供了一定的参考。

Aerodynamic Shape Optimization Design of Guided Bullet Based on Surrogate Model

Aiming at the problem that the traditional aerodynamic shape optimization design relies on high-precision simulation and takes a long time, the work aims to propose an aerodynamic shape optimization method based on a surrogate modelin order to effectively improve the efficiency of aerodynamic optimization design. The 12.7 mm guided bullet was taken as the research object. Through the CFD simulation of the guided bullet, the effect of the length of the tail contraction section, the radius of the bullet base, and the radius of the junction of the tail contraction expansion section on the drag coefficient of the guided bullet during flight were analyzed. Based on the surrogate model technology, the surrogate model of aerodynamic shape optimization design of guided bulletwas constructed by experimental design, parametric modeling, CFD technology, and a fourth-order response surface model. The minimum drag coefficient during the flight of the guided bullet was taken as the optimization objective, and the genetic algorithm was used to optimize the shape parameters of the guided bullet. The prediction accuracy of the surrogate model established by the fourth-order polynomial response surface and the Kriging model was compared. Compared with the calculation results of the CFD simulation of the test sample points, the average error of the predicted value of the drag coefficient of the fourth-order response surface surrogate model was 0.386%, which indicated that the fourth-order response surface surrogate model could effectively replace the CFD simulation to predict the drag coefficient of guided projectile under different shape parameters. Compared with the initial shape of the guided bullet, the drag coefficient of the guided bullet is reduced by 14.59%, which effectively reduces the energy loss of the bullet during the flight and shortens the optimization time. This method reduces the design cycle of the guided bullet under the premise of ensuring accuracy and provides a certain reference for related engineering applications and research.