【作者】 汪光文；

【导师】 胡骏；

【作者基本信息】 南京航空航天大学 ， 航空宇航推进理论与工程， 2009， 博士

【摘要】 数值气动优化设计方法是计算流体力学和数值最优化理论相结合的一种跨学科自动设计方法,它较少地依赖设计人员的经验,能实现多目标优化设计和多设计参数组合优化设计,但是该方法对流场计算精度和计算机速度要求较高,直到近些年,随着计算机技术的发展得以快速发展。本文风扇/压气机的数值气动优化设计重点研究四个关键技术:研究耗时少、计算较准确的叶片通道流场数值计算方法;研究具有良好寻优能力的数值最优化方法;利用计算机辅助几何设计技术,研究三维叶片造型的参数化方法;研究以气动性能为目标的多目标函数构建方法。本文编制大量程序源代码实现了上述气动优化设计关键技术,并在此基础上展开下列具体工作:研究了叶片通道流场计算方法。针对雷诺平均三维NS方程流场计算耗时较长,本文利用Denton的粘性体积力流场计算方法编制程序,对叶片通道流场进行了仿真计算,大大缩短了流场计算耗时。通过美国NASA 67和37号转子计算算例,探索出该程序经验参数设置规律;并将粘性体积力方法的计算结果与实验数据、NUMECA软件计算结果进行比较,验证了粘性体积力方法的准确性。研究了高效率的并行遗传算法。采用实数编码、自适应算子、小生境技术等对基本遗传算法进行改进,提高了搜索效率和全局寻优能力;采用WinSock通讯机制、多线程技术、主从结构体系,实现了基于局域网多台计算机或服务器多CPU并行优化,大幅度缩短了优化时间。研究了风扇/压气机三维叶片基于修改量的参数化方法。这里提出了以叶片积叠线、叶片型面、子午面流道和叶片弦长都作为设计变量的多变量组合优化设计思想。在叶片型面参数化中,引入借鉴多重网格法和Beizer递推算法的多层参数化方法,并与遗传算法结合,达到提高寻优效率的目的。为了统一规范积叠线参数化描述,将叶片的弯曲设计、前后掠设计和叶片弦长设计与计算机辅助几何设计方法的三维旋转、平移和缩放建立对应关系,构造叶片参数化方法的矩阵公式。本文采用了基于修改量的参数化方法,在优化设计前,需预先给定初始叶片。通过对任意回转面与平面叶栅流动分析,明确了两者流动差别,最终选择基于回转面流动方法进行叶型设计。初始叶片由任意回转面叶型沿径向叠加形成。将风扇/压气机优化设计各个关键技术研究形成的模块整合成优化设计软件。软件各模块通过全局变量执行数据传输,既保证了各模块的相对独立,又便于软件编写、调试和功能扩展。软件具有使用简便的用户界面;具有二维、三维叶片并行优化设计功能;并有较丰富的计算结果后处理及结果图形显示能力。多个二维叶型和三维叶片的优化算例验证知:本文软件用来设计高气动性能叶片,设计结果较可靠,设计效率较高,能节约设计成本、缩短设计周期。更多还原

【Abstract】 Numercial aerodynamic optimization is an interdiscipline automatic design method via computational fluid dynamics and optimization theory. It can implement multi-objective optimization and multi-parameter combinatorial optimization with less dependency on designer’s experience. However, numercial aerodynamic optimization design requires considerable flow field calculation accuracy and computing speed, it has not been developing so fast until recent years with the rapid advancement of computer technology.Fan/compressor design using numercial aerodynamic optimization focus on four key technologies: Fan/compressor blade passage flow field numerical method; Numerical optimization method with a good search capability; Parameterization description method of three-dimensional blade shape using Computer Aided Geometric Design Technology; The construction methodology of multi-objective function for aerodynamic performance. Estimated 30K LOC have been coded to implement the aforementioned critical technologies and based on these, the following research is carried out in detail:Firstly, fan/compressor flow field calculation procedure based on the distributed body force proposed by Denton is carried out in this thesis, and this method can shorten flow field computing time effectively compared to that based on Reynolds-averaged NS equations. The law of experience parameter setting is explored by NASA Rotor67 and Rotor37 examples. The accuracy of this procedure is verified by comparing the procedure calculation result to the NUMECA calculation results and experiment data.Secondly, an advanced genetic algorithm using Real-coded, adaptive operators, niching techniquesis at the basis of Simple Genetic Algorithm has improved the search efficiency and global optimization ability. Then this algorithm is paralleled on the LAN and server through winsock, multi-thread and Client-Server architecture to reduce the time-consuming of optimization.Thirdly, a fan/compressor blade parameterization method based on modification is studied including curved/swept stacking line, profile, meridional channel and chord length. In order to improve the optimization convergence rate, a multi-level parameterization algorithm is involved following Multi-Grid Method and Bezier recursive algorithm in this thesis. And corresponding relationships are set up between blade stacking line curved/swept, chord length and translation operation, rotating operation, zoom operation in CAGD. Fourthly, in the process of fan/compressor blade numercial aerodynamic optimization, the initial blade is necessary. Analysis and comparisons of revolving surface cascade and plane cascade give their differences, and then the revolving surface cascade is selected to design blade profile. At last, all profiles are stacked together on the radial to form initial blade.Finally, the software called OTMBPGA (Optimization of Turbomachinery Blades based on Parallel Genetic Algorithm) is integrated with Genetic Algorithm module, blade parameterization module, grid generation and flow field calculation module and objective fuction setting module. The data exchanges among those modules are convenient by using public memery. And every module is also independent, so it’s easy to maintain, extend software function and program debugging. OTMBPGA has friendly user interface, the functionality of two-dimensional and three-dimensional blade parallel optimization, the capacity of post-processing and graphical display of optimization result. 2D and 3D blade optimization examples verify that the design of the blade with high aerodynamic performance using OTMBPGA are reliable and high-efficiency, it can save design cost and shorten design period.更多还原