基于径向基函数的机翼二维气动代理模型设计

采用多学科设计优化方法进行机翼气动/结构优化时,结构学科的优化需要气动学科提供机翼压力分布的代理模型。本文引入了等参单元形函数的几何变换思想,利用径向基函数,解决了复杂形状机翼的二维气动代理模型的构造问题,进行了某巡航导弹弹翼考虑结构变形的气动力代理模型的构建。算例结果表明,本文所用代理模型构造方法简单易行,预测结果的精度很好。     

基于径向基函数的频响函数模型修正

针对使用频响函数进行有限元模型修正存在的时间成本和精度问题,结合模态参与变异系数法和模态动能法分别优化激励点和测点,使获得的模态信息更完整;然后,引入径向基函数(RBF)模型减少原有限元模型计算次数,并根据均方根误差准则对所构建代理模型的参数(spread)进行优选,提高模型预测精度;最后,选用一个36自由度的二维桁架模型进行可行性验证,对比有限元法、Kriging模型及二阶响应面模型的修正精度和迭代时间,结果表明,本文方法具有较好的优势。     

基于更新径向基函数网络模型的广义Pareto分布函数拟合

广义Pareto分布函数GPD(Generalized Pareto Distribution)是一种针对随机参数尾部进行渐进插值的方法,能够对高可靠性问题进行评估。但这种方法要求样本空间较大,计算成本较高,尽管可以通过径向基函数网络RBFNN(Radial Basis Function Neural Network)辅助抽样的方法削减计算成本,但对于非线性程度较高的问题,RBFNN精度问题使得辅助抽样方法失效。针对这类问题,根据GPD的特点,提出了高效的更新RBFNN训练样本的方法,改善了RBFNN在功能函数分布尾部的精度,将RBFNN辅助抽样方法推广应用到非线性程度较高的问题,准确地得到了所有需要的尾部样本,基于该尾部样本集的GPD拟合结果与基于直接计算所有样本的GPD拟合结果完全一致。     

Computing non‐Newtonian fluid flow with radial basis function networks

This paper is concerned with the application of radial basis function networks (RBFNs) for solving non‐Newtonian fluid flow problems. Indirect RBFNs, which are based on an integration process, are employed to represent the solution variables; the governing differential equations are discretized by means of point collocation. To enhance numerical stability, stress‐splitting techniques are utilized. The proposed method is verified through the computation of the rectilinear and non‐rectilinear flows in a straight duct and the axisymmetric flow in an undulating tube using Newtonian, power‐law, Criminale–Ericksen–Filbey (CEF) and Oldroyd‐B models. The obtained results are in good agreement with the analytic and benchmark solutions. Copyright © 2005 John Wiley & Sons, Ltd.     

基于近似模型的非线性区间数优化方法及其应用

在不确定优化中,非线性区间数优化方法由于需要嵌套优化,造成计算效率低下而阻碍其应用于工程实际.本文提出了一种基于径向基函数近似模型的求解方法,以提高非线性区间数优化方法的计算效率.该方法利用拉丁超立方实验设计方法采样,建立目标函数和各约束的径向基函数近似模型.利用近似模型代替嵌套优化中的真实模型,再用非线性区间数优化方法进行求解,从而提高了非线性区间数优化方法的计算效率,使得该算法在工程应用方面成为可能.用一个测试函数验证了该方法的可行性,最后将方法应用于车身薄壁梁的耐撞性优化.     

CFD‐based optimization of aerofoils using radial basis functions for domain element parameterization and mesh deformation

A novel domain element shape parameterization method is presented for computational fluid dynamics‐based shape optimization. The method is to achieve two aims: (1) provide a generic ‘wrap‐around’ optimization tool that is independent of both flow solver and grid generation package and (2) provide a method that allows high‐fidelity aerodynamic optimization of two‐ and three‐dimensional bodies with a low number of design variables. The parameterization technique uses radial basis functions to transfer domain element movements into deformations of the design surface and corresponding aerodynamic mesh, thus allowing total independence from the grid generation package (structured or unstructured). Independence from the flow solver (either inviscid, viscous, aeroelastic) is achieved by obtaining sensitivity information for an advanced gradient‐based optimizer (feasible sequential quadratic programming) by finite‐differences. Results are presented for two‐dimensional aerofoil inverse design and drag optimization problems. Inverse design results demonstrate that a large proportion of the design space is feasible with a relatively low number of design variables using the domain element parameterization. Heavily constrained (in lift, volume, and moment) two‐dimensional aerofoil drag optimization has shown that significant improvements over existing designs can be achieved using this method, through the use of various objective functions. Copyright © 2008 John Wiley & Sons, Ltd.     

Meshless method based on collocation with consistent compactly supported radial basis functions

Based on our previously study, the accuracy of derivatives of interpolating functions are usually very poor near the boundary of domain when Compactly Supported Radial Basis Functions (CSRBFs) are used, so that it could result in significant error in solving partial differential equations with Neumann boundary conditions. To overcome this drawback, the Consistent Compactly Supported Radial Basis Functions (CCSRBFs) are developed, which satisfy the predetermined consistency conditions. Meshless method based on point collocation with CCSRBFs is developed for solving partial differential equations. Numerical studies show that the proposed method improves the accuracy of approximation significantly. The project supported by the National Natural Science Foundation of China (10172052)     

A numerical method for KdV equation using collocation and radial basis functions

Recently, there has been an increasing interest in the study of initial boundary value problems for Korteweg–de Vries (KdV) equations. In this paper, we propose a numerical scheme to solve the third-order nonlinear KdV equation using collocation points and approximating the solution using multiquadric (MQ) radial basis function (RBF). The scheme works in a similar fashion as finite-difference methods. Numerical examples are given to confirm the good accuracy of the presented scheme.     

Numerical solution of Navier–Stokes equations using multiquadric radial basis function networks

A numerical method based on radial basis function networks (RBFNs) for solving steady incompressible viscous flow problems (including Boussinesq materials) is presented in this paper. The method uses a ‘universal approximator’ based on neural network methodology to represent the solutions. The method is easy to implement and does not require any kind of ‘finite element‐type’ discretization of the domain and its boundary. Instead, two sets of random points distributed throughout the domain and on the boundary are required. The first set defines the centres of the RBFNs and the second defines the collocation points. The two sets of points can be different; however, experience shows that if the two sets are the same better results are obtained. In this work the two sets are identical and hence commonly referred to as the set of centres. Planar Poiseuille, driven cavity and natural convection flows are simulated to verify the method. The numerical solutions obtained using only relatively low densities of centres are in good agreement with analytical and benchmark solutions available in the literature. With uniformly distributed centres, the method achieves Reynolds number Re = 100 000 for the Poiseuille flow (assuming that laminar flow can be maintained) using the density of , Re = 400 for the driven cavity flow with a density of and Rayleigh number Ra = 1 000 000 for the natural convection flow with a density of . Copyright © 2001 John Wiley & Sons, Ltd.     

非线性函数的混沌优化方法比较研究

已有的混沌优化方法几乎都是利用Logistic映射作为混沌序列发生器,而Logistic映射产生的混沌序列的概率密度函数服从两头多、中间少的切比雪夫型分布,不利于搜索的效率和能力。为此,首先根据Logistie映射混沌轨道点密度函数的特点,建立改进的混沌-BFGS混合优化算法。之后,考虑到Kent映射混沌轨道点密度为均匀分布,建立了基于Kent映射的混沌-BFGS混合优化算法。然后对五种混合优化方法——不加改进的和改进的基于Logistic映射的混沌-BFGS法,基于Kent映射的混沌-BFGS法,Monte Carlo试验-BFGS法,网格-BFGS法进行了研究,分别对3个低维和2个高维非线性复杂测试函数进行优化计算,对它们的全局优化计算效率和寻优能力做了比较,并探讨了混合优化方法全局优化性能差异的原因。结果表明,混沌优化方法是与Monte Carlo方法类似的一种随机性试验优化方法。而且,这类优化方法的计算性能至少与以下因素有关：混沌／随机序列的统计性质,优化问题全局最优点位置。     

基于Kriging模型的汽轮机基础动力优化设计

随着汽轮机容量的增加和核电站的迅速发展,汽轮机基础动力优化设计已经成为世界前沿的研究课题.本文提出一种基于Kriging模型的有效优化方法,用以求解上述动力优化设计问题.该问题的优化模型是在汽轮机基础框架重量约束条件下,优化汽轮机基础中柱的位置和粱、柱的截面面积,使基础振动的最大幅值最小化.Kriging模型用于建立基础振动的最大动位移幅值与设计变量间的近似函数关系,从而避免了优化迭代中灵敏度分析.开发了动力分析程序,作为黑箱用于动力响应分析.算例结果表明,本文方法在效率和稳定性上优于序列线性规划方法.     

基于Kriging代理模型的两类全局优化算法比较

代理模型在结构优化领域中的应用逐渐增多。相对传统优化方法,代理模型方法在处理带有噪音或仿真模拟十分耗时的问题时有明显优势。加点准则是代理模型技术的一个关键,为了避免陷入局部最优解,加点准则需要同时考虑局部搜索(exploitation)和全局搜索(exploration)两部分并加以平衡。本文在Kriging代理模型基础上提出一种基于几何全局搜索的全局优化算法MSG(Multi-start Local Search with Geometrical Exploration),通过数值算例将其与基于不确定性全局搜索的有效全局优化算法EGO(Efficient Global Optimization)进行比较,研究了MSG算法参数的影响,并讨论了MSG与EGO各自的特点和适用范围。     

Numerical study of radial basis function interpolation for data transfer across discontinuous mesh interfaces

Allowing discontinuous or non‐matching mesh spacing across zonal interfaces within a computational domain offers many advantages, particularly in terms of easing the mesh generation process, reduction of required mesh densities, and relative motion between mesh zones. This paper presents a numerical study of a universal method for interpolating solution data across such interfaces. The method utilises radial basis functions (RBFs) for n‐dimensional volume interpolation, and treats the available solution data points simply as arbitrary clouds of points, eliminating all connectivity requirements and making it applicable to a wide range of computational problems. Properties of the developed meshless interface interpolation are investigated using analytic functions, and three issues are considered: the achievable order of spatial accuracy of the RBF interpolation alone and comparison with a variable order polynomial; the effect of a combined RBF and polynomial interpolation; and the ability of the method to recover frequency content. RBF interpolation alone is shown to achieve fourth‐order to sixth‐order spatial accuracy in one and two dimensions, and in three dimensions, using a small number of data points, third‐order and above is achievable even for a 3 : 1 discontinuous cell spacing ratio, that is a 27 : 1 volume ratio, across the interface. Hence, it is inefficient to include polynomial terms, since improving on the RBF spatial accuracy results in a significant increase in the system size and deterioration in conditioning. It is also shown that only five points per wavelength are required to capture both frequency and amplitude content of periodic solutions to less than 0.01% error.Copyright © 2013 John Wiley & Sons, Ltd.     

基于局部DQ-RBF不可压缩N-S方程的数值求解

以RBF作为DQ方法的基函数,将迎风机制引入DQ-RBF中,建立了二维不可压缩黏性N-S方程数值求解模型,采用Levenberg-Marquardt算法求解非线性方程组.求解时分析了形状参数对求解精度的影响,改进了边界速度的处理方法.对平板Couette流及有限宽台阶绕流流动问题进行了数值求解.比较了本文方法和FLUE...     

Radial basis function neural network model for mean velocity and vorticity of capillary flow

The radial basis function neural network (RBFNN) simulation has been designed to simulate and predict the mean velocity of capillary flow in transition from laminar to turbulent flow and the root‐mean‐square vorticity as a function of wall‐normal position at different values of Reynolds number. The system was trained on the available data of the two cases. Therefore, we designed the system to work in automatic way for finding the best network that has the ability to have the best test and prediction. The proposed system shows an excellent agreement with that of an experimental data in these cases. The technique has been also designed to simulate the other distributions not presented in the training set and predicted them with effective matching. Copyright © 2010 John Wiley & Sons, Ltd.     

基于Kriging代理模型的结构形状优化方法研究

建立了一种进行结构形状优化的工程方法。该方法利用试验设计法选取位置变量样本点,并对各个样本点对应的结构进行尺寸优化,然后根据样本点和尺寸优化结果建立Kriging代理模型,再利用遗传算法求出位置变量最优解,进而得到结构最优解。最后进行了15杆平面拱式桁架、悬臂梁和25杆空间桁架三个形状优化算例,算例分析结果表明:该方法思路清晰、适应性强、优化效果明显,具有较大的工程应用价值。     

基于Kriging模型的频响函数有限元模型修正方法

针对使用频响函数进行有限元模型修正的问题,提出了一种基于Kriging模型的修正方法,用于检测结构由损伤引起的在单元刚度特性上的衰减。本文方法可以在不需要推导修正参数与频响函数残差代数关系的前提下,通过少量测点提供的有效数据快速求解;还可以通过控制算法的终止准则来提高对未知区域的探索程度,降低结果收敛到局部解上的可能。使用Kriging模型可以有效地减少原有限元模型的计算次数,保证计算效率的同时,为对结构进行更准确精密的有限元建模提供了便利。     

可靠性优化问题中遗传算法适应值函数的建立

应用遗传算法进行优化,约束的处理成为建立适应值函数和算法进行的关键.可靠性优化是以系统可拿性指标作为优化问题的约束条件.首先结合外罚函数法建立数学模型,处理约束的惩罚因子时根据种群情况白适应取值,构造适应值函数的映射公式.随后采用拉格朗日秉子法建立了新的约束与目标函数向适应值函数的映射公式,该公式可以避免因罚函数病态所导致的搜索终止,收敛更加快速,使遗传算法得以成功应用于可靠性优化问题中.分析计算结果表明乘子法具有更好地收敛效果,两个公式构造合理.     

遗传算法优化的神经网络在SINS/GPS中的应用

在不增加辅助系统的情况下,针对捷联惯性导航/全球定位组合导航系统(SINS/GPS)在GPS信号不可用时,其定位精度产生较大退化的问题,提出了遗传算法优化的径向基函数神经网络辅助组合导航系统定位的方法。当GPS信号可用时,采用遗传算法对径向基函数神经网络进行优化训练;当GPS信号不可用时,利用遗传算法优化后的径向基函数神经网络预测卡尔曼滤波器量测输入中的速度误差信息,使得卡尔曼滤波器能够继续工作并提供速度校正量。跑车实验表明,通过对速度进行误差补偿能够有效地修正位置误差,以GPS信号断开180 s的结果作分析,纯SINS模式的东向和北向位置误差分别为35.1 m和38.8 m,而本文所提方法的误差分别为10.5 m和7.2 m,其定位精度提高较为显著。     

Advances in viscous vortex methods—meshless spatial adaption based on radial basis function interpolation

Vortex methods have a history as old as finite differences. They have since faced difficulties stemming from the numerical complexity of the Biot–Savart law, the inconvenience of adding viscous effects in a Lagrangian formulation, and the loss of accuracy due to Lagrangian distortion of the computational elements. The first two issues have been successfully addressed, respectively, by the application of the fast multipole method, and by a variety of viscous schemes which will be briefly reviewed in this article. The standard method to deal with the third problem is the use of remeshing schemes consisting of tensor product interpolation with high‐order kernels. In this work, a numerical study of the errors due to remeshing has been performed, as well as of the errors implied in the discretization itself using vortex blobs. In addition, an alternative method of controlling Lagrangian distortion is proposed, based on ideas of radial basis function (RBF) interpolation (briefly reviewed here). This alternative is formulated grid‐free, and is shown to be more accurate than standard remeshing. In addition to high‐accuracy, RBF interpolation allows core size control, either for correcting the core spreading viscous scheme or for providing a variable resolution in the physical domain. This formulation will allow in theory the application of error estimates to produce a truly adaptive spatial refinement technique. Proof‐of‐concept is provided by calculations of the relaxation of a perturbed monopole to a tripole attractor. Copyright © 2005 John Wiley & Sons, Ltd.