国际结构工程与计算学术会议纪要

1.概况国际结构工程与计算学术会议(ICSEC′ 90)于1990年4月25日至28日在中国北京举行.会议由中国力学学会主办,中国土木工程学会、中国航空学会、澳大利亚工程师协会、国际土力学与基础工程学     

The construction of large element in finite element method

In the usual finite element method, the order of the interpolation in an element is kept unchanged, and the accuracy is raised by subdividing the grid denser and denser. Alternatively, in the large element method, the grid is kept unchanged, and the terms of approximate series in the element are increased to raise the accuracy.In this paper, a method for constructing large elements is presented. When using this method, two sets of variables, one set defined inside the element, and the other defined on the boundary of the element, are adopted. Then, these two sets of variables are combined by the hybrid-penalty function method. This method can be applied to any elliptic equations in a domain with arbitrary shape and arbitrary complex boundary condition. It is proved with strict mathematical method in this paper, that in general cases, the accuracy of this method is much higher than that of the usual element and the large element method presented in [7]. Therefore, the degrees of freedom needed in this method are much fewer than those in the two methods if the same accuracy is preserved.     

THE CONSTRUCTION OF LARGE ELEMENT IN FINITE ELEMENT METHOD

In the usual finite element method.the order of the inter-polation in an element is kept unchanged.and the accuracyis raised by subdividing the grid denser and denser.Alter-natively.in the large element method.the grid is kept un-changed.and the terms of approximate series in the elementare increased to raise the accuracy.In this paper,a method for constructing large elementsis presented.when using this method,two sets of variables.one set defined inside the element.and the other defined onthe boundary of the element.are adopted.Then,these twosets of variables are combined by the hybrid-penalty functionmethod.This method can be applied to any elliptic equationsin a domain With arbitrary shape and arbitrary complex boun-dary condition.It is proved with strict mathematical methodin this paper.that in general cases,the accuracy of this me-thod is much higher than that of the usual element and thelarge element method presented in[7].Therefore.the deqreesof freedom needed in this method are much fewer than t     

程序元件化方法及元件程序组装语言

本文提出一种元件程序组装语言,能够支持结构程序设计和模块化方法,并能直接对程序进行分解和组合。     

有限元法中大单元的构造

在通常的有限元法中,单元内的插值多项式的阶数固定不变,通过加密剖分网格来提高精度.大单元法则剖分的网格固定不变而通过增加单元内逼近级数的项数来提高精度. 本文提出采用两套变量的办法来构造大单元,即单元内采用一套变量,单元的边界上采用另一套变量,然后用杂交罚函数法把两者联系起来.这种方法能适用于任何椭圆型方程,任意几何形状区域以及任何复杂的边界条件.本文用严密的数学方法证明了:在一般情况下,这种方法的精度比通常的有限元法和文[7]的大单元法高得多.即在达到相同的精度时,本文方法所需要的自由度(即未知数数目)比上述两种方法少得多.     

Mixed hybrid penalty finite element method and its application

The penalty and hybrid methods are being much used in dealing with the general incompatible element, With the penalty method convergence can always be assured, but comparatively speaking its accuracy is lower, and the condition number and sparsity are not so good. With the hybrid method, convergence can be assured only when the rank condition is satisfied. So the construction of the element is extremely limited. This paper presents the mixed hybrid penalty element method, which combines the two methods together. And it is proved theoretically that this new method is convergent, and it has the same accuracy, condition number and sparsity as the compatible element. That is to say, they are optimal to each other.Finally, a new triangle element for plate bending with nine freedom degrees is constructed with this method (three degreesof freedom are given on each corner——one displacement and tworotations), the calculating formula of the element stiffness matrix is almost the same as that of the old triangle element for plate bending with nine degrees of freedom But it is converged to true solution with arbitrary irregrlar triangle subdivision. If the true solution u∈H³ with this method the linear and quadratic rates of convergence are obtianed for three bending moments and for the displacement and two rotations respectively.     

MIXED HYBRID PENALTY FINITE ELEMENT METHOD AND ITS APPLICATION

The penalty and hybrid methods are being much used in dealing with the general incompatible element. With the penalty method convergence can always be assured, but comparatively speaking its accuracy is lower; and the condition number and sparsity are not so good. With the hybrid method, convergence can be assured only when the rank condition is satisfied. So the construction of the element isextremely limited. This paper presents the mixed hybrid penalty element method, which combines the two methods together.And it is proved theoretically that this new method is convergent, and it has the same accuracy, condition number and sparsity as the compatible element.That is to say, they are optimal to each other. Finally, a new triangle element for plate bending with nine freedom degrees is constructed with this method (three degrees of freedom are given on each corner——one displacement and two rotations), the calculating formula of the element stiffness matrix is almost the same as that of the old triangle     

混合杂交罚函数有限元方法及其应用

对一般非协调有限元,目前采用最多的两种方法是罚函数法和混合、杂交法.前一种方法总能保证收敛,但精度差,条件数和稀疏性不好;后一种方法则要满足“秩条件”才能保证收敛,故单元的构造受到很大的限制.本文提出把这两种方法结合一起的有限元方法——混合杂交罚函数法.从理论上严格证明了(在非常一般的条件下)这种新方法总是收敛的,并且其精度、条件数以及稀疏性等皆与协调元相同,也就是说都是最优的. 最后应用这一方法具体构造了一个新的九自由度任意三角形弯板单元(每个顶点给三个自由度——一个位移和两个转角),其单元刚度矩阵计算公式与旧的九自由度三角形弯板单元的计算公式相差不多.但它对任意几何形状的平板都收敛于真解,如果真解u∈H3的话,它的三个弯矩具有一阶精度,位移及两个转角均具有二阶精度.