【作者】 许林；

【导师】 王振国；

【作者基本信息】 国防科学技术大学 ， 航空宇航科学与技术， 2009， 博士

【摘要】 飞行器设计方法在日益激烈的空天竞争中占据重要地位。传统设计方法在应对飞行器设计中存在的复杂耦合交互时效率低下,无法达到“快,好,省”的研制目标。多学科设计优化(Multidisciplinary Design Optimization,MDO)方法研究耦合多学科系统的设计最优化问题,对于提高飞行器设计水平具有非常现实与重要的意义。以提高飞行器设计水平为目的,本文以MDO方法的核心——MDO过程为主要研究对象,在系统研究飞行器设计问题、多种优化过程及相关技术的基础上,对优化过程进行了改进,并结合相关技术,使之适应飞行器设计的计算与工程实际需要,进一步考虑飞行器发展趋势,将其应用于亚声速近空间飞行器、MEMS(Micro Electro-Mechanical System)微梁的设计问题中。首先,研究了基于协同优化过程的高超声速飞行器设计问题。在系统研究协同优化过程(Collaborative Optimization, CO)的基础上,完成了高超声速飞行器概念设计CO表述,并结合变复杂度技术进行高超声速飞行器概念设计协同优化,确认了MDO过程的效益与进一步研究方向。其次,开展了BLISS过程(Bi-Level Integrated System Synthesis)的研究与改进。系统研究了多种BLISS过程,并在BLISS 2000过程基础上,发展了EBLISS 2000过程,该过程在优化目标、灵敏度分析、约束处理、移动限制等方面均有改进,算例测试表明,EBLISS 2000优化效率明显优于BLISS 2000过程;基于EBLISS 2000过程,还给出了三级系统的ETLISS过程。然后,研究了EBLISS 2000过程与飞行器设计计算实际的结合。证明了EBLISS 2000过程分解前后的最优解一致性;研究了确定性全局优化方法,发展了基于广义简约梯度法的填充函数法GRGFF;给出了MDO问题的广义高维拓扑优化本质与可采用的研究模式,并针对其中的“黑箱”模式,给出了基于拉丁超立方取样的多二次径向基序贯近似方法LHS-MQ;基于LHS-MQ与GRGFF方法,发展了伪数论网格序贯优化方法QSNTO;在EBLISS 2000过程的基础上,结合QSNTO,发展了具有全局优化能力的GBLISS 2000过程,该过程更适合于飞行器设计中存在大量计算的实际情况;算例测试表明了该过程的有效性。进一步,研究了GBLISS 2000过程与飞行器设计不确定性工程实际需求的结合。系统研究了飞行器设计不确定性;分析了系统不确定性分析方法SUA(System Uncertainty Analysis, SUA),给出了修正的系统不确定性分析方法RSUA;分析并确定了RSUA与MDO过程结合的方式;在GBLISS 2000的基础上,结合RSUA方法,发展了UGBLISS 2000过程,该过程更适应飞行器研制工程实际中存在不确定性的情况,算例测试表明了该过程的有效性。最后,利用以上研究的成果,结合飞行器的发展趋势,研究了亚声速近空间飞行器机翼气动弹性设计问题与微机电系统微梁的设计问题。在非线性气动弹性的基本计算方法研究的基础上,建立了基于CFD-GEOM与Python语言和MSC-PATRAN与PCL语言的亚声速近空间飞行器静气动弹性分析数值模型;以航程为目标,完成了该设计问题的GBLISS 2000表述与求解,最优设计的机翼性能相对于初始设计的机翼性能有了大幅度的优化;在系统研究MEMS系统设计与MDO技术结合的基础上,给出了基于MDO的MEMS CAD“top-down”设计流程,建立了基于ANSYS与APDL语言的微梁结构-电磁分析模型,并以微梁某综合性指标为目标,完成了该设计问题的UGBLISS 2000表述与求解,其结果在不确定性方面优于确定性最优设计。总之,论文研究以提高飞行器设计水平为目的,发展了性能优良的EBLISS 2000,GBLISS 2000与UGBLISS 2000过程,并应用于先进飞行器设计中。对MDO过程理论与先进飞行器设计应用做了大量探索性的工作,为进一步开展飞行器MDO奠定了良好的基础。更多还原

【Abstract】 Aerospace vehicle design method plays an important role in increasingly intensive aerospace competition. Traditional design method handles the complex coupled interaction in aerospace vehicle design ineffectively and can not fulfill the target of“fast, good and economical”. The MDO (Multidisciplinary Design Optimization) method handles the optimal design problem of coupled multidisciplinary system. It has very practical and important meaning to increase the aerospace vehicle design level. Aimed at increasing the aerospace vehicle design level, this thesis takes MDO process, the core of MDO, as main research object. Based on the systematic study of aerospace vehicle design problem, several optimization processes and related technologies, improvements are made to some optimization process and adapted to the requirements of computation and engineering practice of aerospace vehicle design with related technologies. According to the further consideration of aerospace vehicle development trend, these research productions are utilized in the subsonic near-space vehicle design and MEMS (Micro Electro-Mechanical System) micro beam design.At first, the hypersonic vehicle design problem is studied based on collaborative optimization. On the basis of systematic study of collaborative optimization, the hypersonic vehicle preliminary design collaborative optimization is brought forward and performed with variable complexity model. The effectiveness of MDO process and further study direction are confirmed.Secondly, the BLISS 2000 process is studied and improved. Several BLISS process are studied, and EBLISS 2000 is developed based on BLISS 2000, which is improved on the aspects of object functions, sensitivity analysis, constraints handling and moving limit. The test case shows the efficiency of EBLISS 2000 is much higher than BLISS 2000. The ETLISS process is also proposed for the three level system based on EBLISS 2000Thirdly, the combination of EBLISS 2000 and aerospace vehicle design computation is studied. The consistency of EBLISS 2000 optimal before and after decomposition is proved. The deterministic global optimization is studied and the GRGFF (Generalized Reduced Gradient Method based Filled Function) method is developed. The general high dimensional topology optimization nature of MDO problem is pointed out with several research modes. For the“black box”mode, the sequential approximation method LHS-MQ based on Latin hyper cubic sampling and multiquadric radial basis function is provided. Quasi sequential number-theoretic optimization method QSNTO is developed based on LHS-MQ and GRGFF. Combined with QSNTO, GBLISS 2000(Global Bi-Level Integrated System Synthesis 2000) is developed, which fits the existence of much computation in aerospace vehicle design better. The test case shows the effectiveness of GBLISS 2000.Fourthly, the combination of GBLISS 2000 and aerospace vehicle design uncertainty engineering practical demands is studied. A systematic research of aerospace vehicle design uncertainty is done. The system uncertainty analysis method SUA is analyzed and the revised system uncertainty analysis method RSUA is given. The combination method of RSUA and MDO process is analyzed and the UGBLISS 2000 (Uncertainty based Global Bi-Level Integrated System Synthesis 2000) process is developed based on the combination of RSUA and GBLISS 2000. The UGBLISS 2000 fits the existence of uncertainty in the aerospace vehicle design better. Test case shows the effectiveness of UGBLISS 2000.Finally, applied these research results and combined with the development trend of aerospace vehicle, the design problems of subsonic near space vehicle wing and MEMS micro beam are studied. Based on the nonlinear aeroelastic computation, the static aeroelastic numerical analysis model of subsonic near space vehicle is built with CFD-FASTRAN, Python, MSC-PATRAN and PCL. The GBLISS 2000 formulation of this design problem is given and solved with the object of range. The optimal wing performance is much better than the original wing performance. Based on the systematic MEMS system and MDO technology, the MEMS CAD“top-down”design process based on MDO is provided. The micro beam electromechanical model is built on ANSYS and APDL. The UGBLISS 2000 formulation of this problem is given and solved with some comprehensive indicator object. The result is better than deterministic optimal design.To sum up, the thesis aims at improving aerospace vehicle design level, develops excellent EBLISS 2000, GBLISS 2000 and UGBLISS 2000 processes and applies them in advanced aerospace design. Much exploring work of MDO process theory and advanced aerospace design application are done, which establishes a good base for aerospace vehicle MDO.更多还原