论文部分内容阅读
多学科设计优化中应用代理模型可以较好地解决系统分析精度和效率的矛盾,动态代理技术可以根据需要在优化的过程中不断提高代理模型精度.利用三种动态代理方法——最小响应面、最大期望提高和多点加点准则,分别对一维、二维函数和一个高维的协同优化问题中外形/气动学科构造了动态Kriging代理模型,对比了三种加点准则的动态代理效果,并分析了其特点和应用条件.结果表明,最小响应面加点准则是最适合于协同优化中学科优化的动态代理方法.
The application of agent model in multidisciplinary design optimization can better solve the contradiction between accuracy and efficiency of the system analysis.Dynamic agent technology can improve the accuracy of agent model continuously in the process of optimization.Using three dynamic agent methods- Maximum expectation increase and multi-point plus criterion respectively, the dynamic Kriging agent model is constructed for the shape / aerodynamics disciplines in one-dimensional and two-dimensional functions and a high-dimensional collaborative optimization problem, and the dynamic proxy effects of the three additive rules are compared and analyzed Its characteristics and application conditions are discussed.The results show that the minimum response surface plus point criterion is the most suitable dynamic agent method for collaborative optimization of middle school subjects.