结构优化半解析灵敏度及误差修正改进算法

提出结构半解析灵敏度分析及其针对刚体位移的误差修正方法的改进算法,构建灵敏度分析与误差修正项可分离形式.该方法实现简便,数值精度不受摄动步长与单元数目的影响.首先从总体角度推得静力问题的误差修正半解析灵敏度分析方法,提出了位移误差修正灵敏度列式,并给出算法实施途径;然后将此思路推广于自振频率、屈曲临界载荷问题,提出了相应的计算步骤.随后,给出梁单元与壳单元误差修正项的具体推导方法,并分别使用两种单元构建有限元模型进行算例测试.结果表明,该方法适用于多种分析类型,数值精度不受单元数目与摄动步长的影响.由于灵敏度分析与误差修正项可以分开计算,该方法支持将误差修正项直接叠加于灵敏度求解结果进行误差修正,使已有灵敏度分析程序得到充分利用.尤其对于复杂工程结构的优化设计,特别是形状优化设计以及尺寸、形状混合优化设计,相比于原误差修正方法,实现更为简便,效率有所提升,能为半解析灵敏度分析方法及其程序实现提供新的思路.     

结构优化半解析灵敏度分析的改进算法

提出了结构半解析灵敏度分析的改进算法,该算法实现简便,对于设计变量摄动步长具有极佳的数值稳定特性。首先,从总体角度推导静力问题半解析法灵敏度分析新算法,提出了位移与应力灵敏度列式,并给出了算法实施途径;然后,将此思路推广于自振频率、屈曲临界荷载和瞬态响应等多种问题,提出了相应的计算步骤。以梁单元与壳单元等典型结构为例,开展了多个算例测试。测试表明,改进算法计算精度和效率均有提升,特别是设计变量步长有更大的数值稳定区域,为复杂工程结构形状优化的灵敏度分析提供了新途径。     

基于虚荷载变量的形状优化和灵敏度分析

基于选择施加在结构“控制点”上的虚荷载作为优化设计变量,针对一种新的承受约束的形状优化数值方法进行了研究。借助于节点位移与虚荷载之间的线性关系,提出了一种新的计算灵敏度系数的解析方法。利用节点移动速度域概念构造了优化新形状产生的计算公式,以结构中节点的最大应力最小化作为优化目标,通过控制网格结点的最大位移量,较好地解决了单元网格在形状优化中的扭曲问题。对三个不同的实例成功地完成了形状优化。     

风力发电塔的自振特性分析

将风力发电塔视为带有附加质量的变截面悬臂梁,进行横向振动的自振特性分析.采用直接模态摄动法建立风力发电塔自振特性的近似求解方法,与采用梁单元模型和壳单元模型的有限元法的计算结果相比较.数值计算结果表明直接模态摄动法具有较好的精度,形成了半解析解形式.     

热传导问题的非协调数值流形方法

数值流形方法通过引入数学与物理双重网格，将插值域与积分域分别定义在两个不同的覆盖上，其优点是网格划分随意，不受复杂边界形状和材料界面的限制，是较之于有限元方法更一般化的数值模拟方法。在计算精度方面，数值流形方法远远高于有限元法。但它的精度还是不够理想。为此本文在单元总体位移场上附加非协调位移基本项，使单元位移函数趋于完全，构造了非协调流形单元来改善流形单元的计算精度和计算效率，并将其应用于热传导问题，推导了势问题的非协调数值流形方法。     

等几何壳体分析与形状优化

首先基于Reissner-Mindlin理论进行了三维壳体等几何分析,而后基于此对三维壳体进行形状优化,提出了形状优化中灵敏度的全解析计算方法,包括位移应变阵、雅克比阵和刚度阵等相对控制顶点位置的灵敏度解析计算公式;通过实例验证了壳体等几何分析和灵敏度全解析计算方法的有效性。与传统的基于网格的灵敏度半解析计算方法相比,基于NURBS的灵敏度全解析计算具有精确、计算效率高的特点,且可以避免优化迭代中的网格畸变。     

等几何壳体分析与形状优化

首先基于Reissner-Mindlin理论进行了三维壳体等几何分析,而后基于此对三维壳体进行形状优化,提出了形状优化中灵敏度的全解析计算方法,包括位移应变阵、雅克比阵和刚度阵等相对控制顶点位置的灵敏度解析计算公式;通过实例验证了壳体等几何分析和灵敏度全解析计算方法的有效性。与传统的基于网格的灵敏度半解析计算方法相比,基于NURBS的灵敏度全解析计算具有精确、计算效率高的特点,且可以避免优化迭代中的网格畸变。     

弹性力学中的一种非协调数值流形方法

通过引入数学和物理双重网格，将插值域与 积分域分别定义在不同的覆盖上，即在数学网格上进行插值函数的构造，物理网格上完成 系统能量泛函积分运算，最后通过覆盖权函数将二者联结在一起. 它的优点是单元网格划 分随意，不受复杂边界形状和二相材料界面的限制，单元可以是任意形状，是较之于有限 元方法更一般的数值模拟方法. 在4节点四边形数值流形方法中，由于单元总体位移函数 包含的完全多项式不完全，使得计算精度不够精确，为此，在单元总体位移函数上附 加非协调位移基本项，使之趋于完全，提出了弹性力学问题的一种改进的数值流形 方法------非协调数值流形方法. 通过内部自由度静力凝聚处理，导出了消除内参后的单元应变矩阵 和单元刚度矩阵，使得在不增加广义节点自由度的前提下，大大提高了数值流形方法的计 算精度和计算效率. 同时对非协调项进行了显式处理，可以对工程实践起到更切实的帮助. 数值试验表明，它们能够保证收敛，有较高的精度，对畸变不敏感，从而证明了该方法的 可行性.     

基于摄动法的不确定性有限元模型修正方法研究

开展了考虑不确定性的有限元模型修正方法的研究。基于摄动法推导了待修正参数均值和协方差矩阵的迭代格式,其中协方差的迭代格式包括是否考虑试验数据与修正参数之间相关性的两种形式。在理论研究基础上开展数值仿真研究,实现了不确定性有限元模型修正的摄动法,并研究了试验数据样本数量对修正误差的影响。仿真结果表明,该方法适用于解决系统参数与试验数据存在不确定性的模型修正问题,试验样本数量对待修正参数标准差的修正精度影响较大;忽略试验模态参数与待修正参数不确定性之间的相关性,能够避免计算二阶灵敏度矩阵,在保证修正结果准确性的前提下减少计算量。     

比例边界有限元二阶灵敏度设计及断裂力学分析

比例边界有限元是一种只需在边界上划分网格且无需基本解的半解析方法,能有效处理应力奇异性和无边界问题.论文提出了一种比例边界有限元的二阶灵敏度分析方法,可以准确而高效地求解响应关于参数的二阶梯度.首先通过建立仅需右特征向量的哈密顿矩阵特征灵敏度分析方程,发展了一种改进的比例边界有限元一阶灵敏度分析方法;其次,进一步通过构建二阶哈密顿矩阵特征灵敏度分析方程,并对比例边界有限元系统方程进行一系列二次直接微分,提出了一种半解析形式的比例边界有限元二阶灵敏度分析方法.该方法被应用于线弹性裂纹结构的形状灵敏度分析和不确定性传播分析.最后,给出了两个数值算例验证论文方法的有效性.     

Accuracy Analysis of the Semi-Analytical Method for Shape Sensitivity Calculation∗

ABSTRACT

The semi-analytical method of design sensitivity analysis that is widely used for calculating derivatives of static response with respect to design variables for structures modeled by finite elements is studied in this paper. It is shown that the method can have serious accuracy problems for shape design variables in structures modeled by beam, plate, truss, frame, and solid elements. Errors are shown to be associated with an incompatibility of the sensitivity field with the structure. An error index is developed to test the accuracy of the semi-analytical method. It characterizes the difference in errors between a general finite difference method and the semi-analytical method. A method for improving the accuracy of the semi-analytical method (when possible) is provided. Examples are presented to demonstrate the use of the error index.     

Binary discrete method of topology optimization

The numerical non-stability of a discrete algorithm of topology optimization can result from the inaccurate evaluation of element sensitivities. Especially, when material is added to elements, the estimation of element sensitivities is very inaccurate, even their signs are also estimated wrong. In order to overcome the problem, a new incremental sensitivity analysis formula is constructed based on the perturbation analysis of the elastic equilibrium increment equation, which can provide us a good estimate of the change of the objective function whether material is removed from or added to elements, meanwhile it can also be considered as the conventional sensitivity formula modified by a non-local element stiffness matrix. As a consequence, a binary discrete method of topology optimization is established, in which each element is assigned either a stiffness value of solid material or a small value indicating no material, and the optimization process can remove material from elements or add material to elements so as to make the objective function decrease. And a main advantage of the method is simple and no need of much mathematics, particularly interesting in engineering application.     

On Accuracy Problems for Semi-Analytical Sensitivity Analyses

Abstract

The semi-analytical method of sensitivity analysis combines ease of implementation with computational efficiency. A major drawback to this method, however, is that severe accuracy problems have recently been reported. A complete error analysis for a beam problem with changing length is carried out in this paper. It is shown that the sensitivity error is proportional to the relative length difference, but in agreement with Eq. 3.8. The approximate pseudo loads thus violate moment equilibrium for rigid body motion while the correct pseudo loads do not.

It might be a good idea to modify the approximate pseudo loads in order to obtain general load equilibrium with rigid body motions. Such a method would be readily applicable for any element type, whether analytical expressions for the element stiffnesses are available or not. This topic is postponed for a future study.     

Free vibration analysis of open thin deep shells with variable radii of curvature

In the present paper, free vibration of a thin open curved shell with parabolic curvature was studied. This shell has a curvature with variable radius in one direction. The equations of motion of this shell were inferred by first order shell theory. According to perpendicular nature of loading on shell of marine structures, the assumptions of Donnell–Mushtari–Vlasov can be used with an acceptable level of accuracy and the in-plane displacement along shell straight direction “x” can be neglected as compared to the displacement in two other directions. The natural frequencies and mode shapes related to the first five vibrational modes were extracted using semi-analytical methods including power series method, Galerkin method and beam function method. The results of the semi-analytical methods were validated against those obtained by using the finite element method. Out of the studied semi-analytical methods, Galerkin method was found to have an appropriate convergence in both natural frequency and mode shape. Adopting eight terms of the response series, Galerkin method has an appropriate convergence compared with the results of finite element.     

基于改进灵敏度过滤策略的SIMP方法

在传统变密度法SIMP(Solid Isotropic Microstructures with Penalization)中,Sigmund灵敏度过滤策略是解决拓扑优化结果数值不稳定问题的重要方法之一。但在SIMP方法中采用Sigmund灵敏度过滤策略时,有时优化结果会有较多的灰度单元。为了减少拓扑优化结果中灰度单元的数量,提出了一种基于改进灵敏度过滤策略的SIMP方法。在过滤后单元灵敏度的计算中,增加了一个与中心单元过滤前灵敏度有关的部分,该部分在过滤后灵敏度中所占的比重需要通过过滤权重来调节。通过对灵敏度过滤需要的分析,确定了过滤半径和中心单元密度值是影响过滤权重的因素。根据拓扑优化实例的优化结果给出了过滤半径影响过滤权重的情况,并基于过滤半径对过滤权重的影响情况构建了过滤权重的函数。在改进灵敏度过滤策略的基础上,结合双重SIMP算法给出了本文SIMP方法的流程。最后,以悬臂梁结构和简支梁结构为例,验证了本文方法的有效性。与Sigmund灵敏度过滤策略相比,改进的灵敏度过滤策略能有效地减少灰度单元。与双重SIMP方法相比,本文SIMP方法能有效地减少数值不稳定现象。     

基于单元密度进化步长控制的双向渐进结构优化方法

在渐进结构优化方法中,单元密度的进化步长是获得全局最优解的关键因素之一。为了提高渐进结构优化方法的全局寻优能力,提出一种基于单元密度进化步长控制的双向渐进结构优化方法。该方法根据各单元对结构性能影响的权重系数,建立单元密度进化步长的控制模型以控制主/次要单元的删除速率和添加速率,减小灵敏度误差并抑制灰度单元的产生。在控制单元密度进化步长的基础上结合双向渐进结构优化方法中添加单元的特点,以避免由于误删单元导致优化失败。同时,采用灵敏度再分配技术抑制棋盘格式以获得更平滑的优化构形。最后,通过两个算例验证了本文方法能有效地通过控制单元密度进化步长提高全局寻优能力。     

结构优化中处理约束的统一公式

本文将杆、膜结构优化设计中处理位移和应力约束的公式推广为后者适用于频率等多种约束以及梁、板等多种单元构成的混合结构的优化设计,数值例子表明公式具有较高的精度,从而提高了结构优化设计的效率,方便了程序的编制。     

Generalized-order perturbation with explicit coefficient for damage detection of modular beam

A general method is formulated to estimate damage location and extent from the explicit perturbation terms in specific set of eigenvectors and eigenvalues. At first, perturbed orthonormal equation is generated from the perturbation of eigenvectors and eigenvalues to obtain the k-th explicit perturbation coefficients. At second, perturbed eigenvalue equation is generated from the perturbation of eigenvector and eigenvalue, and first-order expansion of the stiffness matrix to obtain other explicit perturbation coefficients. Stiffness parameters are computed from these equations using an optimization method. The algorithm is iterative and terminates under certain criteria. A fixed–fixed modular beam with various numbers of elements is used as test structure to investigate the applicability of the developed approach. By comparison with the Euler–Bernoulli beam, discretization errors are analyzed. In six elements beam, first-order algorithm converges faster for small percentage damage. Second-order algorithm is more efficient for medium percentage damage. For large percentage damage, the second-order algorithm converges more effectively. Meanwhile, for eight elements large percentage damage and ten elements small percentage damage, second-order algorithm converges faster to the termination criterion.