超声速喷管性能优化研究与应用

基于重启全局最优化方法和高斯过程（GP）模型，以模型区流场指标为优化目标，对超声速喷管型面进行优化设计。给出0.6m连续式跨声速风洞流场测试结果，提出优化问题并验证了CFD计算的有效性。利用拥挤距离来控制重启局部优化算法的位置，实现更高效的重启全局最优化算法；利用高斯过程模型对喷管设计参数与模型区流场性能指标的关系进行建模，构造替代数学模型来执行优化搜索，以减少实际的CFD评估次数。结果表明:该方法能以较小的代价实现对喷管性能的优化，模型区马赫数方均根偏差由0.012降到0.001，马赫数梯度由0.049降到0。 

Application research on supersonic nozzle performance optimization

With restart global algorithm and Gaussian process (GP) model,supersonic nozzle contour optimization was investigated to improve the flow quality in test section.According to the experimental results in 0.6m continuous transonic wind tunnel,the optimization problem was defined and the CFD results were verified.The crowding distance notation was introduced to form a restart mechanism for a highly efficient global algorithm.The relationship between design parameters and aerodynamic performances was modeled with Gaussian process model,and used in optimization as a surrogate for the real evaluation procedure to reduce the number of CFD evaluation.Result shows that the optimization framework can improve supersonic nozzle performance efficiently,the root mean square deviation of Mach number in model area falls from 0.012 to 0.001,and axial gradient of Mach number falls from 0.049 to 0.