基于POD代理模型的叶栅气动优化设计理论及方法研究

项目名称： 基于POD代理模型的叶栅气动优化设计理论及方法研究

项目编号： No.51206003

项目类型： 青年科学基金项目

立项/批准年度： 2013

项目学科： 工程热物理与能源利用学科

项目作者： 罗佳奇

作者单位： 北京大学

项目金额： 25万元

中文摘要： 随着CFD技术的日臻成熟和优化算法的发展，基于CFD的叶轮机优化设计成为了可能。然而，目前基于CFD的优化设计仍需解决两个问题：如何准确地得到全局最优解；如何缩短设计周期。本项目通过引入POD代理模型，快速求解设计外形的流场，避免了传统优化设计中的流场重复计算，显著提高优化效率。首先采取两种措施来完善POD代理模型，包括：提高POD基的表达能力；设计过程中POD基的更新。之后通过基于遗传算法的叶片反设计来验证本项目所发展的气动优化设计方法的可靠性。最后采用POD方法对叶轮机叶片气动优化设计中的复杂问题展开研究：发展一种基于梯度信息的Pareto解求解方法，实现基于梯度方法的多目标气动优化设计；考虑不同几何参数对气动参数的影响，采用POD方法代替反方法，实现真正意义上的气动/几何参数匹配优化设计；采用POD方法对多排叶片掺混面上展向气动分布进行优化设计，提高多排叶片的工作性能。

中文关键词： 本征正交分解；优化设计；代理模型；自适应抽样；多排叶片

英文摘要： The design optimization of turbomachinery blade rows based on CFD has become to be feasible and active, because of the mature CFD technologies and optimization methods. However, at present for the CFD-based design optimization it still require much efforts on: 1)how to obtain the accurate global optimals; and 2)find advanced means to reduce the design time without loss in accuracy. In the present project a flow surrogate, POD model will be introduced to support the flow solutions required in the design after only several algebraic steps, subsequently the design time can be significantly reduced and the design should be faster due to the elimination of the iterative flow calculations required in the traditional design optimization. In order to complete the POD model, two strategies are investigated in the project firstly, which includes: enhance the expression ability of POD bases and update the POD bases in the design process. Then the inverse designs and design optimizations of turbine blades are performed to validate the feasibility of the design method proposed in the project. Finally several complex design cases in the design optimizations of turbomachinery are studied by using POD model: 1)Perform multi-objective designs by using gradient-based optimization methods after completing an efficient gradient-bas

英文关键词： proper orthogonal decomposition；design optimization；surrogate model；adaptive sampling；multi stage