基于伴随方法、梯度增强Kriging方法的涡扇发动机进气道减噪高效优化方法研究

发动机进气道减噪优化设计问题中涉及大量设计变量,使用常规的Kriging模型来求解时计算量巨大且计算效率低下。为了高效设计低噪声进气道,本文首次发展了用于管道声学问题的声传播方法、伴随方法、梯度增强Kriging方法的混合方法,提出了一套基于声学伴随方法和梯度增强代理模型的高效优化设计框架。本文利用伴随方法高效求解设计变量的梯度信息,并加入Kriging代理模型中,显著提高了代理模型的精度,阐述了梯度增强Kriging方法和伴随混合方法在低噪声涡扇发动机进气道设计中的应用潜力。首先通过两个标准函数,测试了梯度增强的Kriging方法的寻优性能。最后,应用发展的混合方法进行了高达24个设计变量的典型进气道低噪声设计。结果表明,在额外的梯度信息下,混合方法可设计出满足气动性能约束的低噪声进气道,验证了设计框架的有效性和高效率。

Research on the Efficient Noise Reduction Optimization of Turbofan Engine Intake Based on the Adjoint Method and Gradient Enhanced Kriging Method

There are many design variables concerned in the optimization design of engine inlet noise reduction. When the noise reduction optimization is carried out by using the conventional Kriging model, the computation cost is huge and the computation efficiency is low. In order to efficiently design low noise inlet, this paper first developed a hybrid method of an adjoint method and gradient enhanced Kriging for duct acoustic problems, and proposed an efficient optimization design framework based on acoustic adjoint method and gradient enhancement Kriging model. In this paper, the gradient information of all the design variables were efficiently solved by using adjoint method derived from sound propagation equations, and the accuracy of surrogate models were improved significantly with the adding all the gradient information. The potential of the hybrid method for the design of low noise turbofan engine inlet was illustrated. Firstly, the optimization performance of the hybrid method was validated by two standard functions. Then, a low noise design for typical inlet was carried out with up to 24 design variables. The results show that with the additional gradient information, the low noise inlet can be designed successfully by the hybrid method within aerodynamic performance constraint, which validates the effectiveness and efficiency of the design framework.