多小波在一维有限元方法中的应用

有限元方法(FEM)是建立在变分原理基础上的一种频域数值计算方法.其基函数的选取相当重要,既影响到计算结果的精度也影响到计算效率.通常情况下,都是利用拉格朗日线性插值函数作为基函数.文中利用了多尺度函数.由于多尺度函数及它的偏导数的差值特性,可以快速逼近某个函数.同时这个新的基函数的一阶偏导数在相邻节点上是连续的.最后得到的数值结果显示:在保证一定计算效率的基础上,使得精度大幅度提高.因此采用多尺度函数作为基函数具有很多优势.     

关于高斯函数的小波性质的研究

本文基于小波理论框架，分析探讨了有关高斯函数的小波特性。根据多尺度微分算予理论和多分辨分析思想，证明了高斯函数构造了一个多分辨分析(MRA)，高斯函数的各阶导数均构成小波基函数。从滤波器组的角度，由高斯函数的导数构成的小波函数构造了低通滤波器的脉冲响应，也可视为一尺度函数。     

基于降基多尺度有限元的PGD方法及其在含参数椭圆方程中的应用

为了提高模拟多尺度模型的效率,提出基于降基多尺度有限元的广义特征分解方法.广义多尺度有限元方法是模拟多尺度模型的一种有效方法,在粗网格上构造局部基函数,不仅反映了细尺度上的信息,而且能减少大量的计算量.在广义多尺度有限元方法的框架下,通过交叉验证的思想将多尺度模型映射到降基多尺度有限元空间上,提出基于交叉验证的降基多尺度有限元方法.最后,结合广义特征分解方法和基于交叉验证的降基多尺度有限元方法,将其应用于带参数椭圆偏微分方程的计算.数值例子表明,广义特征分解方法和基于交叉验证的降基多尺度有限元方法相结合,不仅比广义多尺度有限元方法具有更高精度,而且能提高在线计算效率.     

前馈神经网络导数特性分析

为分析前馈神经网络输出量的一阶、二阶偏导数特性,从一层网络结构入手,推导网络输出量的一阶偏导数,应用链式求导法则,推导多层网络输出量的一阶、二阶偏导数的计算公式。在此基础上推导网络的三阶偏导数,并针对二层结构网络,在其输出层激活函数为线性函数时,推导出该网络对输入量的高阶偏导数计算公式。实例分析结果表明,前馈神经网络一阶、二阶偏导数值的精度比网络输出值的精度要低,尤其是在区间的边界上有时会出现较大的偏差。网络的一阶、二阶偏导数值的精度也会随着隐含层神经元数量的增加明显降低,在基本相同的网络训练精度下,隐含层神经元较多的网络比神经元少的网络导数特性差。     

偏微分方程的算子自定义小波解耦算法研究

利用小波算法求解偏微分方程最困难的问题是随着尺度的升高,系统方程的耦合度越来越高,极大降低了计算效率和精度.针对此问题提出了采用算子自定义小波的多尺度解耦算法,首先建立有限元多分辨空间和小波细化关系,提出偏微分方程的多尺度计算理论方法.在优化方案的基础上,提出算子自定义小波的构造方法及解耦条件.改进方法的突出优点在于根据工程问题的实际需要灵活构造具有期望特性的小波基.提出偏微分方程的多尺度算子自定义小波算法,充分利用算子自定义小波的嵌套逼近和尺度解耦特性,实现问题的高效求解.仿真结果表明,改进的算子自定义小波解耦算法具有计算效率高、精度高等特点.     

基于泰勒基函数的移动最小二乘法及误差分析

移动最小二乘法通常选用不超过m次单项式生成基函数空间,本文选用了以计算点(?)为平移点的泰勒基函数生成基空间.理论和数值试验发现:选用此种基函数后会降低形函数及导数计算的复杂性,并且有效减小广义逆矩阵的条件数,提高了计算效率同时增加了计算稳定性,并用该方法推导出移动最小二乘近似的收敛阶及误差主部.     

自动微分在隐式曲线绘制中的应用

自动微分是用于计算多变量函数的导数和偏导数的一种微分技术,在给定一个多变量光滑函数值的程序代码后,可以很容易地利用自动微分来实现有关导数和偏导数的精确计算。将自动微分技术与泰勒方法相结合应用到计算机图形学领域隐式函数曲线绘制的细分算法中,并与未使用自动微分技术前的隐式曲线绘制方法作比较和分析,展示了自动微分方法在绘制隐式曲线方面的优势。     

扩展的多尺度有限元法基本原理

阐述一种适用于非均质材料力学性能分析的扩展的多尺度有限元法（Extended Multiscale Finite Element Method,EMsFEM）的基本原理．该方法的基本思想是利用数值方法构造能反映胞体单元内部材料非均质影响的多尺度基函数,在此基础上求得粗网格层次的等效单元刚度阵,从而在粗网格尺度上对原问题进行求解,很大程度地减少计算量．以该方法进行的具有周期和随机微观结构的材料计算示例,通过与传统有限元法的结果比较,说明这一方法的有效性．EMsFEM的优势在于,能容易地进行降尺度计算,可较准确地求得单元内部的微观应力应变信息,在非均质材料强度和非线性分析中有很大的应用潜力．     

基于等几何分析基函数重用的积分方法

针对等几何分析方法计算过程中的积分效率严重制约计算效率的问题,提出一种基于基函数分类重用的积分方法.首先通过均匀B样条基函数的性质将基函数分类,然后通过支撑域的线性变换实现了基函数的重用,最后采用适合等几何分析的精确高斯积分方法,在保证计算精度的同时显著提高积分效率.数值算例结果表明,该积分方法是可行的和有效的.     

应用于地下水模拟的多尺度有限体积元方法

应用多尺度有限体积元方法模拟地下水流动问题,其中地下渗透场系数采用二维对数正态随机场.与传统的有限体积元法相比,多尺度有限体积元法的基函数具有能够反映单元内参数变化的优点,所以这种方法能在大尺度上捕捉解的小尺度特征获得较精确的解.文中算例分别对均匀、各向同性和各向异性对数正态随机场的二维地下水流动问题用传统数值模拟方法和多尺度有限体积元方法进行了计算.计算结果表明多尺度有限体积元方法收敛,且与传统数值模拟方法相比,多尺度有限体积元方法既节省计算量,又有较高的精度.     

Basis functions for an MoM solution of a corner‐truncated patch antenna

A new set of entire‐domain basis functions for the numerical solution of corner‐truncated patch antennas via a method of moments (MoM) algorithm is proposed in this article. The basis functions used here allow improvement of the computational efficiency and/or the accuracy of the numerical method, compared to the analysis techniques previously presented in the literature and commonly employed in practical designs. Some numerical results showing the capabilities of the new basis functions proposed here and good agreement with the measurements are also presented. © 2005 Wiley Periodicals, Inc. Int J RF and Microwave CAE, 2005.     

半解析高阶有限谱元法及其在波导介质层PBG结构滤波器优化设计中的应用

本文采用高阶有限谱单元对分层结构的横截面进行半解析离散．将结构中沿纵向均匀的区段视为子结构，运用基于Riccati方程的精细积分算法求出其出口刚度阵．网格拼装后即可对分层介质问题进行求解．半解析高阶谱单元的采用可以避免著名的龙格现象，该算法的数值精度能随着基函数的阶数的增加呈指数级提高．即高阶有限谱单元能够达到任意需要的精度．数值算例证明这种方法具有很高的精度与效率．在高精度高效率分析的基础上建立了滤波器的优化设计模型，利用遗传算法对优化模型进行全局优化，得到了PBG结构滤波性能全局最优的设计参数．     

A new radial basis functions method for pricing American options under Merton's jump-diffusion model

A new radial basis functions (RBFs) algorithm for pricing financial options under Merton's jump-diffusion model is described. The method is based on a differential quadrature approach, that allows the implementation of the boundary conditions in an efficient way. The semi-discrete equations obtained after approximation of the spatial derivatives, using RBFs based on differential quadrature are solved, using an exponential time integration scheme and we provide several numerical tests which show the superiority of this method over the popular Crank–Nicolson method. Various numerical results for the pricing of European, American and barrier options are given to illustrate the efficiency and accuracy of this new algorithm. We also show that the option Greeks such as the Delta and Gamma sensitivity measures are efficiently computed to high accuracy.     

The numerical solution of nonlinear generalized Benjamin–Bona–Mahony–Burgers and regularized long-wave equations via the meshless method of integrated radial basis functions

In this paper, a numerical technique is proposed for solving the nonlinear generalized Benjamin–Bona–Mahony–Burgers and regularized long-wave equations. The used numerical method is based on the integrated radial basis functions (IRBFs). First, the time derivative has been approximated using a finite difference scheme. Then, the IRBF method is developed to approximate the spatial derivatives. The two-dimensional version of these equations is solved using the presented method on different computational geometries such as the rectangular, triangular, circular and butterfly domains and also other irregular regions. The aim of this paper is to show that the integrated radial basis function method is also suitable for solving nonlinear partial differential equations. Numerical examples confirm the efficiency of the proposed scheme.

     

A meshless method for Asian style options pricing under the Merton jump-diffusion model

In this paper, we consider the partial integro-differential equation arising when a stock follows a Poisson distributed jump process, for the pricing of Asian options. We make use of the meshless radial basis functions with differential quadrature for approximating the spatial derivatives and demonstrate that the algorithm performs effectively well as compared to the commonly employed finite difference approximations. We also employ Strang splitting with the exponential time integration technique to improve temporal efficiency. Throughout the numerical experiments covered in the paper, we show how the proposed scheme can be efficiently employed for the pricing of American style Asian options under both the Black–Scholes and the Merton jump-diffusion models.     

Meshless local radial point interpolation to three-dimensional wave equation with Neumann's boundary conditions

In this article, the meshless local radial point interpolation (MLRPI) method is applied to simulate three-dimensional wave equation subject to given appropriate initial and Neumann's boundary conditions. The main drawback of methods in fully 3-D problems is the large computational costs. In the MLRPI method, all integrations are carried out locally over small quadrature domains of regular shapes such as a cube or a sphere. The point interpolation method with the help of radial basis functions is proposed to form shape functions in the frame of MLRPI. The local weak formulation using Heaviside step function converts the set of governing equations into local integral equations on local subdomains where Neumann's boundary condition is imposed naturally. A two-step time discretization technique with the help of the Crank-Nicolson technique is employed to approximate the time derivatives. Convergence studies in the numerical example show that the MLRPI method possesses reliable rates of convergence.     

More accurate results for two-dimensional heat equation with Neumann’s and non-classical boundary conditions

In this article, recently proposed spectral meshless radial point interpolation (SMRPI) method is applied to the two-dimensional diffusion equation with a mixed group of Dirichlet’s and Neumann’s and non-classical boundary conditions. The present method is based on meshless methods and benefits from spectral collocation ideas. The point interpolation method with the help of radial basis functions is proposed to construct shape functions which have Kronecker delta function property. Evaluation of high-order derivatives is possible by constructing and using operational matrices. The computational cost of the method is modest due to using strong form equation and collocation approach. A comparison study of the efficiency and accuracy of the present method and other meshless methods is given by applying on mentioned diffusion equation. Stability and convergence of this meshless approach are discussed and theoretically proven. Convergence studies in the numerical examples show that SMRPI method possesses excellent rates of convergence.     

Sorting-Based Calculation of Zeros and Extrema of Functions as Applied to Search and Recognition. II

A stable address sorting and an extrema localization condition based on it are applied for approximate calculation of complex roots of polynomials and zeros of derivatives of functions of one variable. This method is used to localize extrema in discrete numerical sequences, to identify chains of repeating figures, and, on the basis of all this, to search for and recognize patterns presented as numerical arrays.     

Vibration analysis of symmetrically laminated plates based on FSDT using the moving least squares differential quadrature method

In this paper, we adopt the first-order shear deformation theory in the moving least squares differential quadrature (MLSDQ) procedure for predicting the free vibration behavior of moderately thick symmetrically laminated composite plates. The transverse deflection and two rotations of the laminate are independently approximated with the moving least squares (MLS) approximation. The weighting coefficients used in the MLSDQ approximation are obtained through the fast computation of the MLS shape functions and their partial derivatives. The natural frequencies of vibration are computed for various laminated plates and compared with the available published results. Through numerical experiments, the capability and efficiency of the MLSDQ method for eigenvalue problems are demonstrated, and the numerical accuracy and convergence are thoughtfully examined. Effects of the size of support, order of completeness of the basis functions and node irregularity on the numerical accuracy are also investigated.