挠性膜片联轴器优化设计

本文改进了Ｐｏｗｅｌｌ的直接优化方法，提出了有效的约束函数法，并用此法研究了膜片联轴器在强度刚度振动条件下，以重量轻，刚度最小为目标的优化设计计算，基于膜片的强度，刚度是在有限元分析基础上的，必须要很好地解决有限元前置处理（自动分单元、分载荷）和后置处理（绘等值线）。开发了一套比较完善的膜片外形优化设计软件系统，简称ＯＤＦＣ系统。     

Multi-objective optimization of frequency and damping of vertical stabilizer skin structure placed with variable-angle tows

The vibration characteristics of composite vertical stabilizer skin structures play a critical role in damping effects designed for overcoming the air disturbances experienced by aircraft structural components during flight. The first-order fundamental frequencies and their corresponding damping characteristics of the vertical stabilizer skin structure tow-steered by automatic fiber placement technique were optimized with the parameterized trajectories and plies as design variables. Firstly, the vibration and damping numerical models were derived based on Kirchhoff laminate theory, the Rayleigh-Ritz method, and the Strain Energy Method. Then the optimization model was developed by adopting the self-adaptive Differential Evolution Multi-objective optimization algorithm and incorporating the solution method of Pareto Front. The constraints of this optimization model considered the experimentally obtained minimum turning radius of prepregs tow-steered in automatic fiber placement process obtained from experimental tests. Finally, the comparison of numerical simulation results was conducted for the optimized trajectories and the conventional straight trajectories under various boundary conditions, and the numerical results were partially validated through damping and frequency tests. The results indicate the vibration characteristics of the composite vertical stabilizer skin structure can be enhanced to a large extent by optimizing fiber trajectories, and the enhancement percentage is affected by the boundary conditions of the actual structure.     

Geometry and size optimization of stiffener layout for three-dimensional box structures with maximization of natural frequencies

Based on the growth mechanism of natural biological branching systems and inspiration from the morphology of plant root tips, a bionic design method called Improved Adaptive Growth Method (IAGM) has been proposed in the authors’ previous research and successfully applied to the reinforcement optimization of three-dimensional box structures with respect to natural frequencies. However, as a kind of ground structure methods, the final layout patterns of stiffeners obtained by using the IAGM are highly subjected to their ground structures, which restricts the optimization effect and freedom to further improve the dynamic performance of structures. To solve this problem, a novel post-processing geometry and size optimization approach is proposed in this article. This method takes the former layout optimization result as start, and iteratively finds the optimal layout angles, locations, and lengths of stiffeners with a few design variables by optimizing the positions of some specific node lines called active node lines. At the same time, thicknesses of stiffeners are also optimized to further improve natural frequencies of three-dimensional box structures. Using this method, stiffeners can be successfully separated from their ground structures and further effectively improve natural frequencies of three-dimensional box structures with less material consumption. Typical numerical examples are illustrated to validate the effectiveness and advantages of the suggested method.     

Variational Bayesian Kalman filter using natural gradient

We propose a technique based on the natural gradient method for variational lower bound maximization for a variational Bayesian Kalman filter. The natural gradient approach is applied to the Kullback-Leibler divergence between the parameterized variational distribution and the posterior density of interest. Using a Gaussian assumption for the parametrized variational distribution, we obtain a closed-form iterative procedure for the Kullback-Leibler divergence minimization, producing estimates of...     

基于直二元喷管形面尺寸的涡扇发动机参数优化设计研究

为了在设计阶段提升带直二元喷管涡扇发动机的总体性能,开展了基于直二元喷管形面尺寸的涡扇发动机参数优化的研究。首先,建立了带直二元喷管的涡扇发动机模型,提出了发动机正后向排气系统红外辐射特征的计算方法,分析了直二元喷管尺寸对发动机性能参数的影响;其次,提出了基于序列二次规划算法的设计参数多目标优化方法,优化的目标包括高单位推力、低油耗和低红外辐射强度;最后,基于以上模型,利用序列二次规划算法对在设计点非加力情况下的涡扇发动机设计参数进行多目标优化。仿真结果表明：在设计点上,相较于不带直二元喷管的涡扇发动机,带直二元喷管的涡扇发动机具有更好的红外性能,并且通过算法优化后,带直二元喷管的涡扇发动机具有更好的性能参数。     

考虑性能及成本的固体火箭发动机多学科设计优化

为综合考虑固体火箭发动机的燃烧室、药柱、内弹道、喷管及成本等学科影响,梳理学科间耦合关系,并建立了以总冲最大、成本最小为优化目标的固体火箭发动机多学科设计优化（MDO）模型。为降低MDO问题的计算成本,提出一种基于Kriging代理模型的多目标自适应优化方法（KRG-MAOM）。优化过程中,分别对目标与约束构建Kriging模型,并采用多目标优化算法求解,在伪Pareto解中综合考虑支配关系与分布特性选取新增样本点,引导优化快速收敛。算例结果表明,KRG-MAOM算法在全局收敛性与优化效率方面具有显著优势。最后,采用KRG-MAOM算法求解该MDO问题,得到可行的Pareto解集方案,与初始方案相比,同性能情况下成本节省约3.36%;同成本情况下性能提升约10.93%,从而验证MDO模型合理性与KRG-MAOM算法有效性。     

基于变精度遗传算法的翼型快速优化设计方法

低碳环保的电动飞机在要求较高升阻比的同时,需要尽量降低成本、缩短研制周期。但高精度的数值模拟时间代价很大,因此针对电动飞机翼型设计中初始翼型较难选取、优化速度较慢的问题,提出了一种基于变精度遗传算法的翼型多点快速优化方法。以常用的 Hicks-Henne 型函数为基础,改进了其对翼型后缘描述不精确的问题。在数值模拟阶段,介绍了一种快速气动参数计算软件XFOIL,并分析了该软件的适用性与局限。之后给出了使用XFOIL 与 Matlab 进行联合求解的方法,在无人干预的情况下完全实现了翼型设计与优化的自动化,提高了设计效率。在翼型优化阶段,为保持较高的精度和寻优效率,设计了翼型参数的实数编码方法。针对传统遗传优化算法了改进,设计了染色体变精度杂交方法以及动态惩罚方法。最后,给出了基于遗传算法的多点优化方案,以及翼型多目标快速优化一体化设计方案。仿真分成两部分进行,首先改进的 Hicks-Henne 型函数能够有效实现参数化翼型的后缘夹角改变。通过与 NSGA-II 方法的优化结果对比,本文的方法在一定迭代次数范围内获得的升阻比更高,失速特性更加缓和,特别是在综合提高翼型优化效率方面表现较好。仿真结果表明,该方法能够快速获得多种工况下具有较高升阻比的翼型,也可以作为进一步优化的初始翼型,能提高翼型优化效率。     

基于DMOM算法的航空发动机性能寻优控制

提出一种分散迁移优化算法(DMOM),可实现多峰值优化问题的全局最优解搜索.该算法通过随机选择参考粒子,不断迁移搜索自身所处区域峰值点,再通过分散操作排除局部最优点,重新生成新个体,可快速搜索到全局最优区域.将DMOM应用于航空发动机性能寻优控制仿真,结果表明:在最小油耗和最低涡轮温度模式下, DMOM的寻优速度相比遗传算法(GA)和粒子群算法(PSO)提高了2倍以上;同时DMOM的优化精度相比自组织迁移算法(SOMA)提高了60%以上,相比可行性序列二次规划(FSQP)算法提高了20%以上.验证了DMOM相比其他优化算法有更强的跳出局部最优的能力,在航空发动机最小油耗和最低涡轮温度这类多峰值寻优问题中具有明显的优势.      

二维气动问题中Kriging代理模型精度影响因素

以二维跨声速临界翼型的阻力特性为对象,探讨样本点数目、Kriging代理模型参数及其类型等对模型精度的影响.阻力系数采用计算流体力学（CFD）方法得到.模型精度的验证采用交叉验证方法,采用平均误差、最大误差和标准交叉验证残差来衡量Kriging代理模型的精度.研究结果表明:①Kriging代理模型预测气动阻力效果较好.②模型精度随样本点的增多而提高,剔除与样本点响应趋势不相符的“奇异点”后,模型精度显著提高,平均误差减小5%~38%,最大误差减小13%~77%.③核函数类型对模型精度的影响最大,相关参数次之,回归模型的影响最小.采用高斯相关函数、2阶多项式回归模型,以及合适的相关参数值时,Kriging代理模型的精度最高.      

推力室结构力学分析与优化设计

针对某液体火箭发动机推力室建立了参数化模型,使用有限元方法分析结构固有频率和振型,并进行了模态试验验证.采用最优拉丁超立方法进行计算机数值试验设计,得到了结构质量和1阶固有频率对于设计变量的敏感性,对于优化设计过程具有指导意义.使用NSGA-Ⅱ算法开展多目标优化设计,得到了同时满足静力学与动力学设计要求的推力室结构最优设计方案.最优设计方案的推力室质量减轻了9.1%,1阶固有频率值提高了22.6%,结果表明该方法能有效提高推力室的力学性能.