论文标题:变可信度模型管理结构优化方法应用研究
Investigation on the Temperature Distribution in a Reactor and Design of Measure System by Optical Fiber Sensor System
论文作者 左英桃
论文导师 高正红,论文学位 硕士,论文专业 飞行器设计
论文单位 西北工业大学,点击次数 37,论文页数 68页File Size3296k
2004-10-01论文网 http://www.lw23.com/lunwen_216867657/ 气动;优化设计;变可信度模型管理结构;NURBS;参数化
Aerodynamic; Design Optimization; Approximation Management Framework; NURBS ; parametrization
飞机的气动特性是影响飞机性能的一项重要指标,气动外形设计是飞行器设计过程最重要的方面之一。本文采用变可信度模型管理结构(ApproximationManagement Framework,AMF)进行了翼型、全机气动外形优化设计方法研究。变可信度模型管理结构是在信赖域方法的基础上,为建立基于不同分析模型的优化方法提出的模型管理框架。由于该框架可构成对高可信度模型和低可信度模型的有效管理,因此利用该方法建立的优化方法可以充分发挥不同可信度模型的优势,从而在保证一定设计精度要求的前提下,提高了优化效率。本论文在分析了变可信度模型管理结构的数学基础信赖域方法的基础上,根据高可信度模型和近似模型一致性要求,分别建立了一阶近似模型和近似一阶模型两种低可信度模型的修正方法;探讨了高可信度模型和低可信度模型以及近似模型之间的相互关系与选取原则,给出了AMF方法的工作流程;将基于一阶近模型的AMF方法与梯度搜索方法结合,建立了梯度AMF优化方法;将基于近似一阶模型的AMF方法与遗传算法结合,建立了遗传AMF优化方法。针对具有代表性的飞行器布局特点,分别从构成飞行器外形的基本元素翼面类(翼平面形状和翼型剖面形状)、机身类(机身外形)出发,建立了描述飞行器外形的参数化几何外形方法,并总结了设计变量提取准则,研究了NURBS方法在几何外形参数化中的应用。根据上述方法建立了变可信度模型管理结构的优化设计系统,将N-S方程与Euler方程分别作为高可信度模型与低可信度模型,采用梯度AMF优化方法进行了翼型优化设计,结果表明,梯度AMF方法能将优化速度提高接近一倍;将Euler方程与格林函数法分别作为高可信度模型和低可信度模型,采用遗传AMF优化方法进行了无人机全机气动外形优化设计,取得了很好的效果。通过这些数值实验考察了该设计系统的适用性,检验了优化前后气动性能的改善程度,获得了一些初步的结论。
In aircraft design aerodynamic characteristic is an important aspect that affecting aircraft performance. Aerodynamic configuration is one of the most important aspects of aircraft design. This thesis discusses the application of Approximation Management Frame-work (AMF) in airfoil optimization design and aircraft aerodynamic configuration optimization design.Approximation Management Framework is a model management framework that based on trust region method, and it aims to construct an optimization method that based on different models. Because this framework provides an effective management on variable-fidelity models, it tries to utilize the advantages of variable-fidelity models . It improves the optimization efficiency under certain design precision request.This thesis analyses the mathematic basis of Approximation Management Framework, the trust region method. According to the consistency condition of low-fidelity model and high-fidelity model, this thesis constructs two low-fidelity model correcting method: first order model and approximate first order model; It discusses the relationship of high-fidelity model, low-fidelity model and approximate model, and it gives a flow chart of AMF method. It combines the AMF that based on first order mod-el with gradient searching algorithm and constructs gradient-based AMF method; It combines the AMF that based on approximate first order model with genetic algorith-m and constructs genetic-based AMF method.Aiming at characteristics of representative aircraft configurations, the commonness and speciality of different kinds of aircraft configuration are investigated. According to the basic elements constitute the aircraft configuration, wing (planform and airfoil) and body (fuselage shape), the mathematic model of parametrization aircraft geometry which described the aircraft configuration is built and the rule of deciding geometry control parameters and design variables is established. It also studies the application of NURBS method in parametrization of aircraft geometry.According to forenamed methods, an optimization design system that based on AMF is built. Gradient-based AMF method is demonstrated on a 2D foil optimization problem. In this case, N-S equations are taken as high-fidelity model and Euler equations are taken as low-fidelity model. The results indicate that one time of time saving is achieved. Genetic-based AMF method is demonstrated on the aerodynamic configuration optimization of an uninhabited aerial vehicle. In this case, Euler equations are taken as high-fidelity model and Green method is taken as low-fidelity model. A very good result is achieved. These numerical experiments validate the practicability and applicability of the system. Some preliminary results are obtained.
|
