Numerical implementation of a neural network based material model in finite element analysis

Neural network (NN) based constitutive models can capture non‐linear material behaviour. These models are versatile and have the capacity to continuously learn as additional material response data becomes available. NN constitutive models are increasingly used within the finite element (FE) method for the solution of boundary value problems. NN constitutive models, unlike commonly used plasticity models, do not require special integration procedures for implementation in FE analysis. NN constitutive model formulation does not use a material stiffness matrix concept in contrast to the elasto‐plastic matrix central to conventional plasticity based models. This paper addresses numerical implementation issues related to the use of NN constitutive models in FE analysis. A consistent material stiffness matrix is derived for the NN constitutive model that leads to efficient convergence of the FE Newton iterations. The proposed stiffness matrix is general and valid regardless of the material behaviour represented by the NN constitutive model. Two examples demonstrate the performance of the proposed NN constitutive model implementation. Copyright © 2004 John Wiley & Sons, Ltd.     

Finite layer analysis of layered elastic materials using a flexibility approach. Part 1—Strip loadings

It is well known that the analysis of a horizontally layered elastic material can be considerably simplified by the introduction of a Fourier or Hankel transform and the application of the finite layer approach. The conventional finite layer (and finite element) stiffness approach breaks down when applied to incompressible materials. In this paper these difficulties are overcome by the introduction of an exact finite layer flexibility matrix. This flexibility matrix can be assembled in much the same way as the stiffness matrix and does not suffer from the disadvantage of becoming infinite for an incompressible material. The method is illustrated by a series of examples drawn from the geotechnical area, where it is observed that many natural and man-made deposits are horizontally layered and where it is necessary to consider incompressible behaviour for undrained conditions.     

Development of a finite dynamic element for free vibration analysis of two-dimensional structures

The paper develops an efficient free-vibration analysis procedure of two-dimensional structures. This is achieved by employing a discretization technique based on a recently developed concept of finite dynamic elements, involving higher order dynamic correction terms in the associated stiffness and inertia matrices. A plane rectangular dynamic element is developed in detail. Numerical solution results of free-vibration analysis presented herein clearly indicate that these dynamic element combined with a suitable quadratic matrix eigenproblem solution technique effect a most economical and efficient solution for such an analysis when compared with the usual finite element method.     

An assumed strain finite element model for large deflection composite shells

A nine node finite element model has been developed for analysis of geometrically non-linear laminated composite shells. The formulation is based on the degenerate solid shell concept and utilizes a set of assumed strain fields as well as assumed displacement Two different local orthogonal co-ordinate systems were used to maintain invariance of the element stiffness matrix. The formulation assumes strain and the determinant of the Jacobian matrix to be linear in the thickness direction. This allows analytical integration in the thickness direction regardless of ply layups. The formulation also allows the reference plane to be different from the shell midsurface. The results of numerical tests demonstrate the validity and the effectiveness of the present approach.     

基于应用程序交互访问技术的桥梁有限元模型修正研究

传统的有限元模型修正方法对于具有复杂多自由体系的大型桥梁结构显得束手无策,其中一个重要的原因是大型桥梁结构体系很难用有限元编程精确地表达,难以建立完整的或者缩聚的质量矩阵和刚度矩阵。该文开发了基于应用程序交互访问的有限元模型修正模式,首先在Strand7软件中建立初始的有限元模型,然后利用MATLAB建立迭代程序并调用Strand7软件,通过读写Strand7中的物理参数来更新模型,实现了大型桥梁结构有限元模型修正。该文对一座实桥进行多参考点脉冲锤击法模态试验和静载试验,基于获得的静动力试验数据和Strand7有限元模型分析结果,引入损伤函数的概念识别得到了该桥各梁的分段刚度,成功地实现了单元层次的参数识别。     

Analysis of cracked anisotropic plates using the hybrid finite element method

This paper presents a convenient and efficient method to obtain accurate stress intensity factors for cracked anisotropic plates. In this method, a complex variable formulation in conjunction with a hybrid displacement finite element scheme is used to carry out the stiffness and stress calculations of finite cracked plates subjected to general boundary and loading conditions. Unlike other numerical methods used for local analysis such as the boundary element method, the present method results in a symmetric stiffness matrix, which can be directly incorporated into the stiffness matrix representing other structural parts modeled by conventional finite elements. Therefore, the present method is ideally suited for modeling cracked plates in a large complex structure.     

组合梁界面滑移的计算分析

钢-混凝土组合梁(以下简称组合梁)的界面滑移总是存在的,滑移的存在会降低组合梁的组合作用和刚度,增大挠度,要计算组合梁界面的滑移及挠度,对于简支梁在简单荷载情况下,还可得到解析解,但对于连续梁要得到解析解是十分困难的,另外简支梁的解析解十分冗长,实际运用十分不便。用有限元法计算组合梁的滑移和挠度将是很有效的,不受荷载及支撑条件限制,而有限元法的关键是单元刚度矩阵,该文用虚功原理推导了组合梁的单元刚度矩阵,并用自编的有限元程序对组合梁的滑移和挠度进行了计算,在简支情况下与解析解进行了对比和验证,误差很小,在1%以内。该文推导的单元刚度矩阵可用于小型的自编有限元软件,为快速经济地解决相关的大量实际工程问题奠定了基础。     

Three-dimensional finite element for analysing thin plate/shell structures

A three-dimensional (3-D) hexahedron finite element is presented for the analysis of thin plate/shell structures. The element employs an explicit algebraic definition of six uniform (continuum) strains, six rigid body modes and classical Lagrange-Germain-Kirchhoff thin plate bending modes. Nine additional stiffness factors are used to control higher-order hourglass modes. The element may be used for plate/shell analyses where the flat plate assumptions are appropriate. Also it can easily be adapted to form transition elements to lower order 2-D elements, or to higher-order 3-D continuum elements. The stiffness matrix satisfies the geometric isotropy requirement, passes the patch test, and gives essentially identical response to either applied transverse corner forces or to twisting moments applied on the corner, a requirement of Kirchhoff's corner conditions for a classical thin plate. Several examples are presented to demonstrate the performance of this finite element.     

Direct-iterative solution of ill-conditioned finite element stiffness matrices

In this paper we discuss and expand the direct-iterative method proposed originally by Wilson.¹ First we introduce several simple numerical examples to illustrate the basic idea of this method before we proceed to prove the convergence of the direct-iterative method. We then discuss the methods for selecting the transformative matrix (Q) to be used in transforming an ill-conditioned matrix into a well-conditioned matrix in the direct-iterative method. There are two methods used to choose the matrix Q, namely the rigid body movement method and the imaginary element method. From examples 1-3 we can see that the imaginary element mesh is optional, and the finite element mesh is not necessary. The imaginary element method is a generalization of the mesh refinement method development in Reference 3. Because instead of local rotation angle we only choose displacements of nodes to represent rigid body movement, the rigid body movement method is an improvement of the method in Reference 2. The advantage of these two methods is that, in order to obtain well-conditioned matrices, only a few changes in the stiffness matrices are required even with general ill-conditioned stiffness matrices, and then convergency is achieved rapidly under SOR iteration. Finally, the examples for computing each type of the ill-conditioned matrix in three-dimensional finite element analysis are presented to demonstrate the effectiveness of the direct-iterative method in solving the large bandwidth problems.     

Free mesh method: A new meshless finite element method

A new meshless finite element method, named as the Free Mesh Method, is proposed in this paper. Once nodes are arranged in the domain to be analyzed, some temporary triangular elements are set around a node, i.e. a current central node. The contributions from the element matrices of the above temporary elements are assemebled to the total stiffness matrix. The above processes are performed on all the nodes in the domain. Finally, the solution is obtained by solving the total stiffness equation system as the usual finite element method. To demonstrate the effectiveness of the method, a simple two-dimensional heat conduction problem is solved.     

Elastic buckling analysis of 3-D trusses and frames with thin-walled members

 A finite element method is presented for the determination of the elastic buckling load of three-dimensional trusses and frames with rigid joints. The beam element stiffness matrix is constructed on the basis of the exact solution of the governing equations describing the coupled flexural-torsional buckling behaviour of a three-dimensional beam with an open thin-walled section in the framework of a small deformation theory. Large deformation effects are taken into account approximately through consideration of P−Δ effects. The structural stiffness matrix is obtained by an appropriate superposition of the various element stiffness matrices. The axial force distribution in the members is obtained iteratively for every value of the externally applied loading and the vanishing of the determinant of the structural stiffness matrix is the criterion used to numerically determine the elastic buckling load of the structure. The effect of initial member imperfections is also included in the formulation. Comparisons of accuracy and efficiency of the present exact finite element method against the conventional approximate finite element method are made. Cases where the axial force distribution determination can be done without iterations are also identified. The effect of neglecting the warping stiffness of some mono-symmetric sections is also investigated. Numerical examples involving simple and complex three-dimensional trusses and frames are presented to illustrate the method and demonstrate its merits. Received: 2 May 2000 / Accepted: 15 July 2002     

A formulation for the 4-node quadrilateral element

A formulation for the plane 4-node quadrilateral finite element is developed based on the principle of virtual displacements for a deformable body. Incompatible modes are added to the standard displacement field. Then expressions for gradient operators are obtained from an expansion of the basis functions into a second-order Taylor series in the physical co-ordinates. The internal degrees of freedom of the incompatible modes are eliminated on the element level. A modified change of variables is used to integrate the element matrices. For a linear elastic material, the element stiffness matrix can be separated into two parts. These are equivalent to a stiffness matrix obtained from underintegration and a stabilization matrix. The formulation includes the cases of plane stress and plane strain as well as the analysis of incompressible materials. Further, the approach is suitable for non-linear analysis. There, an application is given for the calculation of inelastic problems in physically non-linear elasticity. The element is efficient to implement and it is frame invariant. Locking effects and zero-energy modes are avoided as well as singularities of the stiffness matrix due to geometric distortion. A high accuracy is obtained for numerical solutions in displacements and stresses.     

A new finite element scheme for instability analysis of thin shells

This paper describes a new finite element scheme for the analysis of instability phenomena of arbitrary thin shells. A computationally efficient procedure is proposed for calculating the non-linear stiffness and tangential stiffness matrices for a doubly-curved quadrilateral element defined by co-ordinate lines. The essential feature is the explicit addition of the non-linear terms into the rigid-body motion of the element. Thus the non-linear and tangential element stiffness matrices can easily be generated by transforming the generalized element stiffness matrix for linear analysis, and the non-linear terms of these matrices are separated into a number of component terms multiplied by the rigid-body rotations. These component terms can be stored permanently and used to calculate efficiently the non-linear and tangential stiffness matrices at each iteration. Illustrative examples are presented which confirm the validity of the present approach in the analysis of instability phenomena of thin plates and shells.     

钢管混凝土结构材料非线性的一种有限元分析方法

为了更简单地考虑梁单元的材料非线性受力性能,把断面广义力和广义应变的概念运用于单元分析中,将单元的弹塑性刚度矩阵分离为弹性刚度矩阵和塑性刚度矩阵。这样,梁单元的变形可以由弹性变形和塑性变形简单地迭加,结构内力可通过弹性应变能的斜率(弹性刚度矩阵)与位移的乘积求得,从而在增量-迭代计算时可较准确且较快地计算出结构变形后的不平衡力。应用这一计算方法,推导了基于纤维模型的三维梁单元的钢管混凝土结构的有限元基本公式,并将其植入能考虑几何非线性的三维梁单元非线性计算程序NL_Beam3D中以计算结构的双重非线性问题。算例分析表明该方法和程序能较准确地反映钢管混凝土结构的双重非线性特性。     

考虑剪滞变形时箱形梁广义力矩的数值分析

为了简化变截面箱梁等复杂结构的剪滞效应分析,在明确定义相应于剪滞位移的广义力矩和有关几何特性的基础上,提出一种梁段有限元数值分析方法。选取控制微分方程的齐次解作为单元位移函数,以各积分常数为中间转换变量,推导梁段单元刚度矩阵和等效节点力向量的具体表达式,并给出用单元节点力直接计算应力的一般公式。编制了箱梁梁段有限元程序,对简支、悬臂、连续箱梁3个有机玻璃模型进行计算并与实测结果对比,验证了该文方法及公式的正确性。用所编程序对箱梁的剪滞广义力矩进行数值分析,并揭示了其变化规律。研究表明,在竖向荷载作用下,剪滞力矩与弯矩具有相似的分布规律,而且数值大小也接近。     

基于模态试验与优化的静动力学模型转换

在结构强度分析中,将刚度分布较为准确的静力学模型转换为动力学模型可以大大提高建模效率。提出了一种基于模态试验和优化算法的静、动力学模型转换方法。在调整静力学模型的刚度矩阵基础上,再按照质量、质心、惯矩、单元体积进行节点质量的预分配,最后根据模态试验识别出的模态参数优化节点质量的修正量,如此便可得到其动力学模型。以某飞机翼身组合结构模型的转换为例,证明了该方法的实用性。     

低张力缆索有限元模型及其应用

为准确计算缆索在低张力时的运动状态,该文考虑缆索低张力状态时的拉伸刚度、弯曲刚度和扭转刚度,建立了一种适用于低张力缆索的三维有限元模型。首先基于细长杆理论推导了缆索的动力学微分方程,接着以三次样条曲线为试函数,运用Galerkin加权残值法导出了单元刚度矩阵,最后对其进行组合建立了缆索的整体矩阵方程,并采用Matlab编写了求解程序。将其应用于实例中,所得结果与实验结果相一致。研究成果为拖曳缆、系泊缆、潜行器脐带缆等海洋缆索的运动分析与设计提供了理论依据。     

任意截面预应力混凝土细长柱的非线性分析

提出了轴力和双向弯曲作用下任意截面混凝土和预应力混凝土细长柱的非线性有限元计算模型。分析时既考虑了由单元变形和轴力二次矩引起的几何非线性效应,也考虑了由材料非线性应力应变关系和截面刚度矩阵引起的材料非线性效应。推导了非线性全过程分析的标准有限元公式,得到的单元刚度矩阵可分割成三个子矩阵,分别反映了材料非线性、材料非线性和单元大位移的耦合、轴力二次矩等三种不同的非线性作用效应。计算分析结果和试验结果吻合较好。     

Finite elements based on energy orthogonal functions

It is shown how the convergence requirements for a finite element may be written as a set of linear constraints on the stiffness matrix. It is then attempted to construct a best possible stiffness matrix. The constraint equations restrict the way in which these stiffness terms may be chosen; however, there is normally still room for improving or optimizing an element. It is demonstrated how an element stiffness matrix may be found using rigid body, constant strain and higher order deformation modes. Further, it is shown how the constraint equations may be exploited in deriving an ‘energy orthogonality theorem’. This theorem opens the door to a whole new class of simple finite elements which automatically satisfy the convergence requirements. Examples of deriving plane stress and plate bending elements are given.