超临界翼型设计中的多响应代理模型

摘要

To consider the correlations between the multi-responses in surrogate model for supercritical airfoil design , this paper studies the surrogate model based on the multi -output Gaussian process (MOGP) regression and verifies the significant correlations between lift coefficient ,drag coefficient and moment coefficient through experimental data .We establish the MOGP-based surrogate model for super-critical airfoil design and compare our model with kriging and other two neural network model , BP and RBF.The results show that the MOGP can significantly improve the prediction accuracy and stability com-pare with kriging ,BP and RBF when the outputs are significantly correlated in supercritical airfoil design .%为考虑超临界翼型设计中代理模型多响应之间的相关性，研究了基于多输出高斯过程回归的代理模型。通过实验数据验证了升力系数、阻力系数及力矩系数之间显著的相关性，建立了超临界翼型设计中的基于多输出高斯过程的多响应代理模型，并与kriging模型、BP和RBF两种人工神经网络模型进行了对比。结果表明，当超临界翼型设计中多输出端口之间存在显著的相关性时，多输出高斯过程相比kriging、BP和RBF能有效显著提高预测精度，并获得更加稳定的预测性能。