基于赫林格?赖斯纳变分原理的一致高效无网格本质边界条件施加方法

无网格法具有高阶连续光滑的形函数, 在结构分析中呈现出显著的精度优势. 但无网格形函数在节点处一般没有插值性, 导致伽辽金无网格法难以直接施加本质边界条件. 采用变分一致尼兹法施加边界条件的数值解具有良好的收敛性和稳定性, 因而得到了非常广泛的应用, 然而该方法仍然需要引入人工参数来保证算法的稳定性. 本文以赫林格?赖斯纳变分原理为基础, 建立了一种变分一致的本质边界条件施加方法. 该方法采用混合离散近似赫林格?赖斯纳变分原理弱形式中的位移和应力, 其中位移采用传统无网格形函数进行离散, 而应力则在背景积分单元中近似为相应阶次的多项式. 此时的无网格离散方程可视为一种新型的尼兹法施加本质边界条件, 其中修正变分项采用再生光滑梯度和无网格形函数进行混合离散, 稳定项则内嵌于赫林格?赖斯纳变分原理弱形式中, 无需额外增加稳定项, 消除了对人工参数的依赖性. 该方法无需计算复杂耗时的形函数导数, 并满足积分约束条件, 保证了数值求解的精度. 数值结果表明, 所提方法能够保证伽辽金无网格法的计算精度最优误差收敛率, 与传统的尼兹法相比明显提高了计算效率.      

双重点最小二乘配点无网格法

配点类无网格法需要计算近似函数的二阶导数,因而在移动最小二乘(MLS)近似中至少要采用二次基函数。本文利用Voronoi图对双重点移动最小二乘近似法进行了改进,建立了基于Voronoi图的双重点移动最小二乘近似(VDG),并利用加权最小二乘法离散微分方程,导出了双重点最小二乘配点无网格法(MD GLS)。该方法将求解域用节点离散,并以节点为生成点建立Voronoi图,取Voronoi多边形的顶点为辅助点。近似函数及其二阶导数的计算过程可分解为两个步骤:首先用场函数节点值拟合辅助点处近似函数的一阶导数,再以辅助点处近似函数的一阶导数值拟合节点处近似函数的二阶导数。由于在每一步中只需计算MLS形函数及其一阶导数,这种近似方法需要较少的影响点和较小的影响域。同时借助于Voronoi结构的优良几何性质,可以快速地搜索影响点。研究表明,与基于MLS的加权最小二乘无网格法(MWLS)相比,这种方法可以显著提高计算效率,并且在精度和收敛性方面也有所改善。     

节点梯度光滑有限元配点法

配点法构造简单、计算高效, 但需要用到数值离散形函数的高阶梯度,而传统有限元形函数的梯度在单元边界处通常仅具有C$^{0}$连续性,因此无法直接用于配点法分析. 本文通过引入有限元形函数的光滑梯度,提出了节点梯度光滑有限元配点法. 首先基于广义梯度光滑方法,定义了有限元形函数在节点处的一阶光滑梯度值,然后以有限元形函数为核函数构造了有限元形函数的一阶光滑梯度,进而对一阶光滑梯度直接求导并用一阶光滑梯度替换有限元形函数的标准梯度,即完成了有限元形函数二阶光滑梯度的构造.文中以线性有限元形函数为基础的理论分析表明,其光滑梯度不仅满足传统线性有限元形函数梯度对应的一阶一致性条件,而且在均布网格假定下满足更高一阶的二阶一致性条件.因此与传统线性有限元法相比,基于线性形函数的节点梯度光滑有限元法的$L_{2}$和$H_{1}$误差均具有二次精度,即其$H_{1}$误差收敛阶次比传统有限元法高一阶, 呈现超收敛特性.文中通过典型算例验证了节点梯度光滑有限元配点法的精度和收敛性,特别是其$H_{1}$或能量误差的精度和收敛率都明显高于传统有限元法.      

无网格法的理论及应用

详细论述了近年来迅速发展的无网格法的理论基础及其在各个领域内的应 用. 无网格法网格依赖性弱, 避免了传统的有限元、边界元等基于网格的数值方法 中可能出现的网格畸变和扭曲, 在一些有限元、边界元等方法难以较好处理的领域体现 出独特的优势. 以加权余量法为主线归纳了已有的30多种无网格法, 各类 无网格法的主要区别在于使用了不同的加权余量法和近似函数. 详尽介绍 了各种无网格近似方案(包括移动最小二乘近似、核近似和重构核近似、单位分 解近似、径向基函数近似、点插值近似、自然邻接点插值近似等)和无网格法 中常用的各类加权余量法(伽辽金格式、配点格式、局部弱形式、加权最小二乘 格式和边界积分格式等), 并讨论了数值积分方法和边界条件的处理等问题. 在 此基础上较系统地总结了无网格法在冲击爆炸、裂纹传播、超大变形、结 构优化、流固耦合、生物力学和微纳米力学等领域的应用, 展示了无网格法相 对于传统数值方法的优势.     

等几何修正准凸无网格法

采用等几何B样条基函数的多项式再生条件对无网格形函数的多项式再生条件进行了修正,使得无网格形函数的负值部分明显减少,在域内趋于非负函数,即等几何修正准凸无网格形函数。该准凸无网格形函数仍然具有与传统再生核无网格形函数相似的构造形式,数值实现比较便捷,同时该准凸无网格形函数的多项式再生条件具有准确的修正系数,无需引入额外的人工节点松弛参数。更重要的是,等几何修正准凸无网格形函数可在确保形函数高阶光滑的前提下减小相对支持域,提高计算效率。最后,基于等几何修正准凸无网格形函数对杆梁和膜板结构进行了伽辽金无网格振动分析。结果表明,与标准再生核无网格法相比,等几何修正准凸无网格法具有更优的计算精度。     

薄板分析的线性基梯度光滑伽辽金无网格法

薄板问题的控制方程为四阶微分方程,因而当采用伽辽金法进行分析时,形函数需要满足C$^{1}$连续性要求,且至少使用二次基函数才能保证方法的收敛性.无网格形函数虽然易于满足C$^{1}$连续性要求,但由于不是多项式,其二阶导数的计算较为复杂耗时,同时也对刚度矩阵的数值积分提出了更高的要求.本文提出了一种薄板分析的线性基梯度光滑伽辽金无网格法,该方法的基础是线性基无网格形函数的光滑梯度.在梯度光滑构造的理论框架内,无网格形函数的二阶光滑梯度可以表示为形函数一阶梯度的线性组合,因而可以提高形函数二阶梯度的计算效率.分析表明,线性基无网格形函数的光滑梯度不仅满足其固有的线性梯度一致性条件,还满足本属于二次基函数对应的额外高阶一致性条件,因此能够恰当地运用到薄板结构的伽辽金分析.此外,插值误差分析也很好地验证了线性基无网格光滑梯度的收敛特性.算例结果进一步表明,线性基梯度光滑伽辽金无网格法的收敛率与传统二次基伽辽金无网格法相当,但精度更高,同时刚度矩阵所需的高斯积分点数明显减少.      

伽辽金型无网格法的数值积分方法

无网格法直接通过节点信息构造形函数,不依赖于节点之间的有序单元连接,能够建立任意高阶连续的整体协调形函数.与传统的有限元法相比,无网格法对大变形问题、移动边界问题和高阶问题的求解有比较明显的优势.伽辽金型无网格法是目前应用最为广泛的一类无网格法.虽然无网格形函数本身不依赖于单元,但伽辽金型无网格法需要采取合适的方法进行弱形式的数值积分.由于无网格形函数一般不是多项式,具有非插值性且影响域与背景积分网格通常不重合,伽辽金型无网格法通常需要采用高阶的高斯积分进行数值积分,导致了计算效率低下,难于求解大型实际问题.因此,如何通过建立高效积分方法提高无网格法的计算效率成为无网格法研究领域的一个核心问题.论文总结了伽辽金型无网格法中若干常用的数值积分方法,并对伽辽金型无网格法的数值积分方法领域存在的一些问题进行了探讨.     

无网格全局介点法

提出了一种新型无网格法,即无网格全局介点(MGIP)法。该方法采用移动最小二乘核近似来构造形函数,有利于提高数值方法的计算稳定性,而且算法更为简单。该方法需要引入全局介点进行数值离散,基于有限点的广义变分法导出求解系统方程,并采用罚系数法来保证边界条件,数值实现较为简洁。数值算例结果表明:MGIP法的计算耗时不到无网格局部彼得洛夫-伽辽金法的1%,具有较高的计算效率;相比于一般配点法,本文方法的计算稳定性更好,计算精度更高。     

用无网格LRPIM分析中厚板弯曲问题

用径向基函数构造无网格点插值法的形函数,插值函数具有Kronecker delta函数性质,因此可以很方便地施加本质边界条件.利用无网格局部径向点插值方法分别对一个对边固支另对边简支中厚板和一个悬臂中厚板的弯曲进行了分析计算.该方法不需要任何形式的网格划分,所有的积分都在规则形状的子域及其边界上进行,是一种真正的无网格方法.算例表明:将无网格局部径向点插值法应用于计算中厚板的弯曲问题,所求得的位移场和应力场都是光滑的;在径向基函数的基础上,附加多项式大大提高了插值精度;所得结果与弹性力学理论解以及有限元解都十分吻合.     

基于卷积神经网络的无网格形函数影响域优化研究

在无网格法中,离散节点之间的相互联系由节点形函数影响域的大小确定,因此形函数影响域的大小对无网格法的计算精度有着直接和重要的影响。但由于无网格形函数的形式较为复杂,目前形函数影响域大小的选择仍然缺乏系统的理论依据,通常在实际计算中仍凭借经验进行选取,难以保证计算精度。卷积神经网络是一类机器学习方法,其感受野与无网格形函数的影响域具有内在相似性,因此在形函数影响域选择方面有很好的适用性。基于该特性,本文通过引入卷积神经网络对无网格形函数的影响域进行了优化选择。首先,针对感受野和影响域的匹配关系,分析了卷积神经网络的结构设计和超参数选择,提出了一种无网格法内禀卷积神经网络结构的设计方法;然后依托该网络结构设计方法,建立了对无网格形函数影响域和数值解分别优化或同时优化的卷积神经网络。文中通过算例系统验证了所提无网格法内禀卷积神经网络对形函数影响域选择和计算结果的优化效应。     

A MESH-FREE ALGORITHM FOR DYNAMIC IMPACT ANALYSIS OF HYPERELASTICITY

A mesh-free method based on local Petrov-Galerkin formulation is presented to solve dynamic impact problems of hyperelastic material.In the present method,a simple Heaviside test function is chosen for simplifying domain integrals.Trial function is constructed by using a radial basis function(RBF)coupled with a polynomial basis function,in which the shape function possesses the kronecker delta function property.So,additional treatment is not required for imposing essential boundary conditions.Governing equations of impact problems are established and solved node by node by using an explicit time integration algorithm in a local domain,which is very similar to that of the collocation method except that numerical integration can be implemented over local domain in the present method.Numerical results for several examples show that the present method performs well in dealing with the dynamic impact problem of hyperelastic material.     

平面广义四节点等参元GQ4及其性能探讨

广义节点有限元是将传统有限元方法中的节点广义化,在不增加节点个数的前提下,仅通过提高广义节点的插值函数的阶次,从而达到提高有限元解精度的目的.与现有的p型和hp型有限元不同,在这种新的有限元中,节点自由度全部定义在节点处,在理论与程序实现上与传统有限元方法具有很好的相容性,传统有限元方法是这种新方法的广义节点退化为0阶时的特殊情形.文中主要讨论了这一新方法的四节点等参元（记为GQ4）的形式.对GQ4进行的各种数值试验表明,所发展的广义四节点等参单元具有精度高且无剪切自锁与体积自锁等的特点.     

A mesh-free radial basis function–based semi-Lagrangian lattice Boltzmann method for incompressible flows

In this paper, a local radial basis function–based semi-Lagrangian lattice Boltzmann method (RBF-SL-LBM) is proposed. This is a mesh-free method that can be used for the simulation of incompressible flows. In this method, the collision step is performed locally, which is the same as in the standard LBM. In the meanwhile, the steaming step is solved in a semi-Lagrangian framework. The distribution functions at the departure points, which may be not the grid points in general, are computed by the local radial basis function interpolation. Several numerical tests are conducted to validate the present method, including the lid-driven cavity flow, the steady and unsteady flow past a circular cylinder, and the flow past an NACA0012 airfoil. The present results are in good agreement with those published in the previous literature, which demonstrates the capability of RBF-SL-LBM for the simulation of incompressible flows.     

NEW TREATMENT OF ESSENTIAL BOUNDARY CONDITIONS IN EFG METHOD BY COUPLING WITH RPIM

One of major difficulties in the implementation of meshfree methods using the moving least square (MLS) approximation, such as element-free Galerkin method (EFG), is the imposition of essential boundary conditions as the approximations do not pass through the nodal parameter values. Another class of meshfree methods based on the radial basis point interpolation can satisfy the essential boundary conditions exactly since its approximation function passes through each node in an influence domain and thus its shape functions possess the properties of delta function. In this paper, a coupled element-free Galerkin(EFG)-radial point interpolation method (RPIM) is proposed to enhance their advantages and avoid their disadvantages. Discretized equations of equilibrium are obtained in the RPIM region and the EFG region, respectively. Then a collocation approach is introduced to couple the RPIM and the EFG method. This method satisfies the linear consistency exactly and can maintain the stiffness matrix symmetric. Numerical tests show that this method gives reasonably accurate results consistent with the theory.     

ELASTO－PLASTICITY ANALYSIS BASED ON COLLOCATION WITH THE MOVING LEAST SQUARE METHOD

A meshless approach based on the moving least square method is developed for elasto-plasticity analysis, in which the incremental formulation is used. In this approach, the displacement shape functions are constructed by using the moving least square approximation, and the discrete governing equations for elasto-plastic material are constructed with the direct collocation method. The boundary conditions are also imposed by collocation. The method established is a truly meshless one, as it does not need any mesh, either for the purpose of interpolation of the solution variables, or for the purpose of construction of the discrete equations. It is simply formulated and very efficient, and no post-processing procedure is required to compute the derivatives of the unknown variables, since the solution from this method based on the moving least square approximation is already smooth enough. Numerical examples are given to verify the accuracy of the meshless method proposed for elasto-rdasticity analysis.     

MESHLESS METHOD OF DUAL RECIPROCITY HYBRID RADIAL BOUNDARY NODE METHOD FOR ELASTICITY

Combining the radial point interpolation method （RPIM）, the dual reciprocity method （DRM） and the hybrid boundary node method （HBNM）, a dual reciprocity hybrid radial boundary node method （DHRBNM） is proposed for linear elasticity. Compared to DHBNM, RPIM is exploited to replace the moving least square （MLS） in DHRBNM, and it gets rid of the deficiency of MLS approximation, in which shape functions lack the delta function property, the boundary condition can not be applied easily and directly and it＇s computational expense is high. Besides, different approximate functions are discussed in DRM to get the interpolation property, in which the accuracy and efficiency for different basis functions are compared. Then RPIM is also applied in DRM to replace the conical function interpolation, which can greatly improve the accuracy of the present method. To demonstrate the effectiveness of the present method, DHBNM is applied for comparison, and some numerical examples of 2-D elasticity problems show that the present method is much more effective than DHBNM.     

RBF collocation method and stability analysis for phononic crystals

The mesh-free radial basis function (RBF) collocation method is explored to calculate band structures of periodic composite structures. The inverse multi-quadric (MQ), Gaussian, and MQ RBFs are used to test the stability of the RBF collocation method in periodic structures. Much useful information is obtained. Due to the merits of the RBF collocation method, the general form in this paper can easily be applied in the high dimensional problems analysis. The stability is fully discussed with different RBFs. The choice of the shape parameter and the effects of the knot number are presented.     

Non‐intrusive reduced‐order modelling of the Navier–Stokes equations based on RBF interpolation

We present a new non‐intrusive model reduction method for the Navier–Stokes equations. The method replaces the traditional approach of projecting the equations onto the reduced space with a radial basis function (RBF) multi‐dimensional interpolation. The main point of this method is to construct a number of multi‐dimensional interpolation functions using the RBF scatter multi‐dimensional interpolation method. The interpolation functions are used to calculate POD coefficients at each time step from POD coefficients at earlier time steps. The advantage of this method is that it does not require modifications to the source code (which would otherwise be very cumbersome), as it is independent of the governing equations of the system. Another advantage of this method is that it avoids the stability problem of POD/Galerkin. The novelty of this work lies in the application of RBF interpolation and POD to construct the reduced‐order model for the Navier–Stokes equations. Another novelty is the verification and validation of numerical examples (a lock exchange problem and a flow past a cylinder problem) using unstructured adaptive finite element ocean model. The results obtained show that CPU times are reduced by several orders of magnitude whilst the accuracy is maintained in comparison with the corresponding high‐fidelity models. Copyright © 2015 John Wiley & Sons, Ltd.     

The improvement of numerical modeling in the solution of incompressible viscous flow problems using finite element method based on spherical Hankel shape functions

In this paper, the finite element method with new spherical Hankel shape functions is developed for simulating 2‐dimensional incompressible viscous fluid problems. In order to approximate the hydrodynamic variables, the finite element method based on new shape functions is reformulated. The governing equations are the Navier‐Stokes equations solved by the finite element method with the classic Lagrange and spherical Hankel shape functions. The new shape functions are derived using the first and second kinds of Bessel functions. In addition, these functions have properties such as piecewise continuity. For the enrichment of Hankel radial basis functions, polynomial terms are added to the functional expansion that only employs spherical Hankel radial basis functions in the approximation. In addition, the participation of spherical Bessel function fields has enhanced the robustness and efficiency of the interpolation. To demonstrate the efficiency and accuracy of these shape functions, 4 benchmark tests in fluid mechanics are considered. Then, the present model results are compared with the classic finite element results and available analytical and numerical solutions. The results show that the proposed method, even with less number of elements, is more accurate than the classic finite element method.     

A smoothed Hermite radial point interpolation method for thin plate analysis

A smoothed Hermite radial point interpolation method using gradient smoothing operation is formulated for thin plate analysis. The radial basis functions augmented with polynomial basis are used to construct the shape functions that have the important Delta function property. The smoothed Galerkin weakform is adopted to discretize the governing partial differential equations, and a curvature smoothed operation is developed to relax the continuity requirement and achieve accurate bending solutions. The approximation based on both deflection and rotation variables make the proposed method very effective in enforcing the essential boundary conditions. The effects of different numbers of sub-smoothing-domains created based on the triangular background cell are investigated in detail. A number of numerical examples have been studied and the results show that the present method is very stable and accurate even for extremely irregular background cells.