基于随机响应面法的结构可靠度分析

提出了考虑变量间相关性的随机响应面法,采用正交变换解决随机响应面法输入随机变量间相关性问题。推导了4阶和5阶Hermite随机多项式展开的解析表达式。编写了基于C#语言的随机响应面法计算程序。最后采用算例证明了随机响应面法在结构可靠度分析中的有效性。结果表明,提出的随机响应面法能够有效地分析输入随机变量间相关性的可靠度问题。3阶随机响应面法的计算精度在大多数情况下可以满足结构可靠度分析的需要,而且计算效率较高。但是随着变量间相关性的增加,4阶或5阶随机响应面法才能保证足够的计算精度。概率配点数目为随机多项式待定系数数目的两倍并不总能保证足够的计算精度,一般来说,配点数目要大于两倍待定系数的个数才能保证随机响应面法足够的计算精度。     

考虑交叉项的自适应响应面法

“精度”和“效率”是近似方法的重要评价指标。传统的二次多项式响应面法,无论是不含交叉项的二次多项式还是完全二次多项式均不能兼顾“效率”和“精度”。为此,该文中提出了一类可在两者之间达到较好平衡的自适应响应面法。一方面,为确保响应面形式更具合理性,通过严格的数学推导给出了极限状态曲面中交叉项是否存在的判断准则,将该准则与完全二次多项式相结合即可确定合理的、自适应的响应面形式;另一方面,针对该判断准则,构造了与之对应的实现算法,并结合可靠度问题的特点,将算法进一步完善;为克服此算法选点中心位于均值点的特性,引入了样本点选取的迭代方案对其改进。最后,该文中通过一个数学算例和一个工程算例分别对建议方法及算法进行验证,结果表明:1) 交叉项存在的判断准则准确、有效;2) 对于较为简单的二次极限状态曲面,建议方法可以真实还原;3) 对于涉及一般极限状态曲面的可靠度问题,建议方法具有颇为理想的精度和较高的效率。     

二次序列响应面法分析重力坝的动力可靠度

提出了用二次序列响应面法分析重力坝可靠度的方法。该方法用界面元法对重力坝进行确定性计算,在适当的区间内用统计的方法确定各变量的样本点,经过若干次重复计算,由界面元法的计算结果可以得到该结构的响应面函数,然后由这一函数算出重力坝可靠度的近似值。用同样方法可得第二、第三个响应面函数,并使可靠度计算结果逐渐达到良好的精度要求。这一方法与统计可靠度分析方法相比可以减少大量的计算工作量。对于包含混凝土特性和坝上游水位等随机变量的重力坝,本文用上述方法计算了一些关键时刻重力坝抗裂的动力可靠度,并据此分析了在整个地震过程中该坝抗裂可靠指标β（t）随时间t的变化规律。     

基于神经网络响应面法的随机结构动力可靠度分析

在对神经网络响应面法的原理和算法进行研究的基础上，建立了基于神经网络响应面法的随机结构动力可靠度分析方法。首先，基于首次超越破坏准则，参照静力可靠度的功能函数模式，建立了随机结构的动力可靠度功能函数；然后引入响应面法，以三层BP神经网络作为拟合函数，推导了功能函数的拟合表达式；最后结合一次二阶矩方法求解可靠指标。算例分析表明了本文方法有较好的计算精度和计算效率，在复杂结构的动力可靠度分析中有较强的实用价值。     

偏最小二乘回归在响应面法可靠度分析中的应用

针对目前多维变量可靠度问题中广泛应用的均匀设计响应面法,分析了采用最小二乘法拟合样本数据回归模型时存在的局限性,并在已有方法的基础上提出了一种改进的方法。该方法将均匀设计与偏最小二乘回归技术相结合来回归响应面模型,从而计算结构的失效概率,有效的解决了变量间多重相关性及小样本条件下建立回归模型的问题。通过算例验证了该方法的适用性,尤其对于高维变量的可靠度问题,与最小二乘拟合响应面相比,计算结果更加精确。     

一种高效的结构可靠度近似分析方法

本文简要讨论了结构的可靠度计算的数值模拟方法,如Monte-Carlo法、响应面和两点近似法;结合结构分析的确定性有限元法及可靠度计算的一次二阶矩法,提出了一种结构可靠度分析的近似方法;算例表明,当荷载的不确定性远大于结构的不确定性时,本文提出的方法满足工程精度要求,从而大大简化了结构可靠度分析过程。     

基于等概率近似变换的 向量型层递响应面法分析结构可靠度

基于正交变换和等概率近似变换,研究建立了随机变量为非高斯互相关的工程结构可靠度分析的向量型层递响应面法。首先利用正交变换将非高斯互相关随机变量变换为互不相关的非高斯标准随机变量,建立结构总体刚度矩阵和荷载列阵,据此定义预处理器并形成预处理随机Krylov子空间,进而利用该空间的层递基向量将结构总体节点位移向量近似展开,建立关于互不相关非高斯标准随机变量的层递响应面;然后根据等概率近似变换,将独立标准正态空间的样本点转换为层递响应面在非高斯空间中的概率配点;最后通过回归分析确定层递响应面待定系数,并利用层递响应面建立极限状态方程求解结构可靠度。分析表明:该文提出的等概率近似变换方法不仅成功地将层递响应面法拓展到非高斯互相关随机变量下的结构可靠度分析,而且方法简便、适用范围广、计算精度和效率较高,具有良好的全域性。     

基于自适应点估计和最大熵原理的结构体系多构件可靠度分析

准确而高效地求解结构体系中多个构件的可靠度水准对结构维护和优化具有重要意义,目前已有学者将蒙特卡洛法和响应面法用于此类可靠度分析。然而,蒙特卡洛法所需结构分析次数取决于失效概率的量级,通常计算成本较高。而响应面法的所需结构分析次数取决于杆件数量,当其数量较多时同样有较高的成本。鉴于此,该文提出了一种基于自适应点估计和最大熵原理的结构体系多构件可靠度分析方法,其所需的结构重分析次数上限与杆件数量无关,计算过程简便无需迭代。首先,通过引入自适应交叉项判定和双变量降维近似模型求解各杆件的前四阶矩;然后,根据各杆件的前四阶矩,采用最大熵原理求解各杆件的可靠度指标;最后,通过多个算例对比了蒙特卡洛法、响应面法和建议方法的精度和效率。结果表明建议方法所需的结构重分析次数远少于蒙特卡洛法和响应面法,实现过程简便,且精度能够满足工程要求。     

基于逆响应面法的有限元模型修正

基于响应面法（RSM）的有限元模型修正是以若干设计参数（如密度、弹性模量等）为自变量,以若干特征参数（如固有频率、振型等）为因变量,通过回归分析方法来拟合特征参数关于修正参数的显式表达式。提出的逆响应面法（IRSM）则是以特征参数作为自变量,设计参数作为因变量。利用此法的有限元模型修正可直接根据特征参数的目标值得到设计参数的修正量,而不需要经过迭代计算,有效地提高计算速度和精度。介绍逆响应面法及其应用,讨论使用响应面法和逆响应面法进行有限元模型修正的优缺点和适用范围,分析适合于逆响应面法的逆响应面函数、实验设计方案和回归精度检验的方法。利用逆响应面法对一简支梁进行有限元模型修正的结果表明,逆响应面法能高效准确地修正设计参数,对于输出变量少于输入变量的情况更能显著减少有限元计算次数,适用于复杂的工程结构。     

基于自适应响应面法的输电塔线体系腐蚀疲劳可靠度研究

环境腐蚀和风振疲劳耦合作用下输电塔体的结构性能逐渐退化,满足预定设计功能的概率减小。然而,传统响应面法计算结构可靠度时均不能兼顾"效率"和"精度"。为此,首先通过严格的数学推导给出了交叉项是否存在的判断准则,将该判断准则与传统二次响应面法相结合建立了考虑部分交叉项的自适应响应面法。然后,通过Q345等边角钢腐蚀疲劳试验结果给出了构件腐蚀疲劳t-P-S-N曲线方程,再与概率论相结合建立了随机疲劳曲线方程。最后,通过工程算例采用建议自适应响应面法以风速、腐蚀时间和随机腐蚀疲劳S-N曲线方程为随机变量对输电塔线体系进行了腐蚀疲劳可靠度研究,结果表明:①交叉项判断准则能有效地保留相互影响随机变量之间的交叉项;②建议自适应响应面法在满足精度的同时能有效减少计算量;③构件随机腐蚀疲劳S-N曲线模型在结构可靠性分析中简单易行。     

Application of the response surface methods to solve inverse reliability problems with implicit response functions

The inverse first-order reliability method (FORM) is considered to be one of the most widely used methods in inverse reliability analysis. It has been recognized that there are shortcomings of the inverse FORM in solving inverse reliability problems with implicit response functions, primarily inefficiency and difficulties involved in evaluating derivatives of the implicit response functions with respect to random variables. In order to apply the inverse FORM to structural inverse reliability analysis, response surface methods can be used to overcome the shortcomings. In the present paper, two different response surface methods, namely the polynomial-based response surface method and the artificial neural network-based response surface method, are developed to solve the inverse reliability problems with implicit response functions, and the accuracy and efficiency of the two response surface methods are demonstrated through two numerical examples of steel structures. It is found that the polynomial-based response surface method is more efficient and accurate than the artificial neural network-based response surface method. Recommendations are made regarding the suitability of the two response surface methods to solve the inverse reliability problems with implicit response functions.     

Packaging Parameters Analysis for the Fatigue Reliability of Stacked Chip Ball Grid Array by Using the Optimal Equivalent Solder Balls

A global/local method with the modified sub‐modeling approach of the two‐parametric optimal equivalent volume solder balls is introduced to predict the deformation and reliability of the package. The equivalent solder balls as exhibit in this method can obviously reduce the required elements/nodes quantities to enhance computing efficiency. A package model of wire‐bonded stacked chip ball grid array under cyclic thermal loading is used as a test vehicle to verify the influences of design factors by fatigue life indicator. Comparing the proposed method with the global fine mesh model, it is found that the difference in the accumulated strain energy density is merely 5.77%, but the optimal equivalent model has highly saved 90% finite element analysis required elements, which means the adopted method can effectively replace the global fine mesh model because both results are in accordance with each other. Using design of experiments to efficiently verify each factor influence with their cross‐coupling effects, this paper adopts two kinds of response surface methods that confirm the fatigue life of the proposed approach can be improved by as much as 123.7% for the dual response surface method and 126.3% for the mixed response surface method when comparing with the baseline model. In addition, the optimization of generic algorithm for both response surface methods is demonstrated in this study. From the reviews of factor coupling effects, it is concluded that the response surface method is eligible to achieve the optimum design for package reliability improving. Copyright © 2013 John Wiley & Sons, Ltd.     

分片响应面法在随机振动响应分析中的应用

在分析结构的随机振动响应时,响应面法(Response Surface Method)可有效地降低随机仿真的计算代价。然而,当随机变量存在大变异系数时,传统的响应面法无法满足所要求的精度。分片响应面基于对随机变量变异系数进行合理分块的原则,缩小响应面的近似范围,并对分割后的响应面进行独立分析,从而提高了响应面在该空间的近似精度。首先采用分块响应面法结合蒙特卡洛MC抽样技术,以单质点振子模型的随机振动响应为算例,对分块响应面法的正确性进行验证。计算结果表明,在随机变量的变异系数不大时,分片响应面法的计算精度不低于传统响应面法,而当随机变量具有大变异系数时,分片响应面法的近似精度远高于传统响应面法。此外,以随机地震动作用下的桥墩随机振动为背景,将该方法进行了进一步地推广及应用。     

An efficient response surface method using moving least squares approximation for structural reliability analysis

The response surface method (RSM) is widely adopted for structural reliability analysis because of its numerical efficiency. However, the RSM is time consuming for large-scale applications and sometimes shows large errors in the calculation of the sensitivity of the reliability index with respect to random variables. In order to overcome these problems, this study proposes an efficient RSM applying a moving least squares (MLS) approximation instead of the traditional least squares approximation generally used in the RSM. The MLS approximation gives higher weight to the experimental points closer to the most probable failure point (MPFP), which allows the response surface function (RSF) to be closer to the limit state function at the MPFP. In the proposed method, a linear RSF is constructed at first and a quadratic RSF is formed using the axial experimental points selected from the reduced region where the MPFP is likely to exist. The RSF is updated successively by adding one new experimental point to the previous set of experimental points. Numerical examples are presented to demonstrate the improved accuracy and computational efficiency of the proposed method compared to the conventional RSM.     

Reliability analysis of structures using neural network method

In order to predict the failure probability of a complicated structure, the structural responses usually need to be estimated by a numerical procedure, such as finite element method. To reduce the computational effort required for reliability analysis, response surface method could be used. However the conventional response surface method is still time consuming especially when the number of random variables is large. In this paper, an artificial neural network (ANN)-based response surface method is proposed. In this method, the relation between the random variables (input) and structural responses is established using ANN models. ANN model is then connected to a reliability method, such as first order and second moment (FORM), or Monte Carlo simulation method (MCS), to predict the failure probability. The proposed method is applied to four examples to validate its accuracy and efficiency. The obtained results show that the ANN-based response surface method is more efficient and accurate than the conventional response surface method.     

Combination of finite element and reliability methods in nonlinear fracture mechanics

This paper presents a probabilistic methodology for nonlinear fracture analysis in order to get decisive help for the reparation and functioning optimization of general cracked structures. It involves nonlinear finite element analysis. Two methods are studied for the coupling of finite element with reliability software: the direct method and the quadratic response surface method. To ensure the response surface efficiency, we introduce new quality measures in the convergence scheme. An example of a cracked pipe is presented to illustrate the proposed methodology. The results show that the methodology is able to give accurate probabilistic characterization of the J-integral in elastic–plastic fracture mechanics without obvious time consumption. By introducing an “analysis re-using” technique, we show how the response surface method becomes cost attractive in case of incremental finite element analysis.     

Full long-term extreme response analysis of marine structures using inverse FORM

An exact and an approximate formulation for the long-term extreme response of marine structures are discussed and compared. It is well known that the approximate formulation can be evaluated in a simplified way by using the first order reliability method (FORM), known for its computational efficiency. In this paper it is shown how this can be done for the exact formulation as well. Characteristic values of the long-term extreme response are calculated using inverse FORM (IFORM) for both formulations. A new method is proposed for the numerical solution of the IFORM problem, resolving some convergence issues of a well-established iteration algorithm. The proposed method is demonstrated for a single-degree-of-freedom (SDOF) example and the accuracy of the long-term extreme response approximations is investigated, revealing that the IFORM methods provide good estimates in a very efficient manner. The reduced number of required short-term response calculations provided by the IFORM methods is expected to make full long-term extreme response analysis feasible also for more complex systems.