基于子集模拟和重要抽样的可靠性灵敏度分析方法

针对工程实际中大量存在的小失效概率问题,提出了基于子集模拟和重要抽样的可靠性灵敏度分析方法. 在子集模拟重要抽样可靠性分析方法中,通过引入合理的中间失效事件,将小的失效概率表达为一系列较大的条件失效概率的乘积,而较大的条件失效概率则可通过构造中间失效事件的重要抽样密度函数来高效求解. 基于子集模拟重要抽样可靠性分析的思想,论文将可靠性灵敏度转化为条件失效概率对基本变量分布参数的偏导数形式,推导了基于子集模拟和重要抽样的可靠性灵敏度估计值及估计值方差的计算公式,并采用算例对所提方法进行了验证. 算例结果表明所提方法具有较高的计算精度和效率,并且适用单个和多个失效模式系统.      

COMMENTS ON AMSAA-BISE RELIABILITY GROWTH MODEL (Ⅰ)

From 1986 to 1991, based on AMSAA model, Zhou Yuan-quan and Weng Zhao-xi presented AMSAA-BISE model to estimate reliability growth for multiple systems development, for the case that more than one system of the same type is put into reliability growth test, once a Type B failure mode is seen during test, corrective action will be taken to all systems. It is shown that there is something wrong with AMSAA-BISE model. According to AMSAA-BISE model, the maximum likelihood estimation of MTBF for multiple systems reliability growth test is much larger than that according to AMSAA model for a single system; The more systems is put into test, the larger the estimation of MTBF. An example is given, and an approximate method is presented. Biography: Mei Wen-hua (1965-), Senior Engineer, Doctor     

Generalized Second Moment Reliability Index

ABSTRACT

A crucial property of any measure of structural reliability should be comparativeness. With this point in mind this paper discusses some well-known versions of so-called reliability indices. Such reliability indices, defined by use of second moment information, have been used for the last decade, specifically within safety code committee work. This paper defines a generalized second moment reliability index that satisfies some few fundamental canonical rules and principles of simplicity. The reliability index is specifically defined to be used when no high quality information is available to the engineer other than the limit state surface and a second moment representation for the set of basic variables of the structural problem. In a parallel paper the author demonstrates the practical operability of this reliability index even for multimode failure systems.     

含概率与区间混合不确定性的系统可靠性分析方法

系统可靠性问题中通常存在大量的不确定参数,传统方法一般是基于概率模型对系统进行可靠性分析,但是实际工程中由于数据缺乏或试验条件的限制往往难以得到参数的精确概率分布.本文将结构体系一部分样本信息充足的不确定变量用随机变量进行描述,而另一部分样本缺乏的用区间表示,并提出了一种新的含概率与区间混合不确定性的系统可靠性分析方法.首先,基于一个高效求解方法获得单失效模式下结构的最小可靠度指标;再针对多失效模式下含概率与区间混合不确定性问题建立了系统可靠性分析模型;考虑各失效模式之间的相关性,通过线性相关度计算方法求得相关系数矩阵;最后提出了串联体系和并联体系可靠度求解方法.3个数值算例表明,该方法可以实现含概率与区间混合的多个非线性失效模式下系统可靠度的计算.通过对比传统的概率可靠性分析方法,本文方法只需要少量的不确定信息便可确保系统更加安全,更适合复杂结构系统可靠性的分析和设计.     

非正态变量可靠性分析的鞍点线抽样方法

结合鞍点概率分布估计和传统线抽样方法的优点,提出了非正态变量可靠性分析的鞍点线抽样方法。传统的线抽样方法对非正态变量问题进行可靠性分析时需将非正态变量等价转换为标准正态变量,非正态变量向标准正态变量的非线性转换将增加响应功能函数的非线性程度,进而加大了转换后响应函数失效概率估计的难度。所提鞍点线抽样方法则无需将非正态变量转化为标准正态变量,它利用鞍点概率分布估计方法可以直接估计非正态变量空间中线性响应函数概率分布的特点,并利用线抽样方法可以将非线性功能函数的失效概率转化为一系列线性功能函数失效概率平均值进行估计的优点,实现了非正态变量空间非线性功能函数失效概率的高精度估计。鞍点线抽样方法使用前需将变量进行标准化变换,这种变换是线性的,通过对变量的标准化变换可以消除变量的量纲,从而使得标准化变量空间概率分布更具规律性。理论推导可以证明：鞍点线抽样方法在基本变量服从正态分布时将退化为传统的线抽样方法。算例验证结果表明：针对非线性功能函数的可靠性问题,鞍点线抽样方法比传统的直接鞍点估计具有更高的精度。     

枚举模糊随机主要失效模式的准则法

基于模糊不确定性向随机不确定性的等价转换和有限元分析计算，建立了以失效模式相对重要度为依据的广义主要失效模式枚举的工程准则法．该方法可以综合考虑材料参数、结构几何以及外载荷等基本变量的随机和模糊两种不确定性因素对失效模式相对重要度的影响，通过合理选择门槛值，可以在保证计算精度的基础上尽可能的减小计算工作量．以平尾转轴的模糊随机可靠性分析为算例，给出了门槛值的经验值，并说明了所建工程准则法的合理性．     

多模式自适应重要抽样法及其应用

针对多模式的可靠性分析，研究了其失效概率计算的自适应重要抽样法，该方法用模拟退火 算法来自动调整每个失效模式的重要抽样函数，使其逐渐趋近于估计方差最小的重要抽样 函数. 对于多个模式系统失效概率的计算，采用混合加权自适应重要抽样的方法, 反映了每个 失效模式对系统失效概率的贡献；对于系统失效模式所含基本变量不全相同的情况，提出了 扩展自适应重要抽样法, 来统一所有失效模式中的基本变量，从而使得混合自适应 重要抽样, 可以方便地求解变量不全相同时的系统失效概率. 对估计值方差和变异系数的计算公 式进行了推导. 验证算例结果, 充分说明方法的合理性与可行性.     

Combined interval analysis - Monte Carlo simulation approach for the analysis of uncertainties in parallel manipulators

The quantification of the impact of uncertainties may increase the reliability and robustness of parallel manipulators. Monte Carlo simulation (MCS) and interval analysis are among the most common techniques used in uncertainty quantification. Interval analysis provides guaranteed performance since the interval evaluation of a function always contains the exact result. Nevertheless, interval analysis estimations are very conservative, frequently yielding overestimated results. Conversely, Monte Carlo Simulation avoids overestimation, but does not provide guaranteed performance. This paper proposes a novel hybrid algorithm combining the best features of interval analysis and Monte Carlo simulation for estimating probabilities of failure in the positioning error of parallel manipulators. A 3RRR manipulator is employed as case-study. The hybrid approach provides information on the bounds (minimum and maximum values) and estimated values of failure probabilities. The simulations herein compare the hybrid approach with pure interval analysis and pure Monte Carlo simulation. The results reveal that the hybrid technique may estimate not only the probability of failure in a continuous region, but also the worst and best case probabilities, much faster than interval analysis based approaches.

     

结构体系的非概率可靠性分析方法

结构体系的可靠性与结构的失效模式有关。在非概率条件下 ,结构体系的可靠性取决于非概率可靠性指标最小的最危险失效模式。最危险失效模式的识别及相应极限状态方程的建立是结构体系非概率可靠性计算的关键问题。文中考虑了结构参数及强度、外载荷等参量的不确定性 ,基于随机可靠性方法中常用的优化准则法 ,提出了非概率结构体系主要失效模式的枚举准则。可在只需枚举出少量主要失效模式的情况下 ,不漏掉最危险失效模式。并提出用区间增量载荷法生成主要失效模式的极限状态方程。算例分析表明文中方法是实用和有效的。     

基于BP神经网络和Laplace渐进积分法的结构可靠性计算

传统基于代理模型的可靠性研究大多将抽样方法与代理模型相结合,并假定随机变量相互独立,且没有考虑到代理模型的不确定性对失效概率的影响。本文将反向传播(BP)神经网络和Laplace渐进积分法相结合,提出一种结合代理模型和高次阶矩的可靠性计算方法,称之为BP-Lap法。采用Latin超立方抽样技术,结合学习函数选取样本点,基于函数逼近原理,利用BP网络代理极限状态方程及其梯度向量和Hessian矩阵。利用训练好的BP网络通过Laplace渐进积分法求解失效概率,基于十折交叉验证思想,得到失效概率取值区间。通过四个算例,分别在随机变量相关和不相关的条件下,验证了BP-Lap法的有效性。研究表明：BP-Lap法可以衡量代理模型的不确定性对失效概率的影响,得到失效概率的上、下界;BP-Lap法同时适用于显示和隐式的极限状态方程,对相关随机变量的可靠性问题具有较高精度。     

Geometrical theory of fluid flows and dynamical systems

Various dynamical systems have often common geometrical structures and can be formulated on the basis of Riemannian geometry and Lie group theory, provided that a dynamical system has a group symmetry, namely it is invariant under group transformations, and further that the group manifold is endowed with a Riemannian metric. The basic ideas and tools are described, and their applications are presented for the following five problems: (a) free rotation of a rigid body, which is a well-known system in mechanics and presented as an illustrative example of the geometrical theory; (b) geodesic equation and KdV equation on the group of diffeomorphisms of a circle and its extended group; (c) a self-gravitating system of a finite number of points masses and a geometrical interpretation of chaos of Hénon–Heiles system; (d) geometrical formulation of hydrodynamics of an incompressible ideal fluid on the group of volume preserving diffeomorphisms, where an interpretation of the origin of Riemannian curvatures of the fluid flow is given; (e) geodesic equation on a loop group and the local induction equation for the motion of a vortex filament, where the geodesic equation on its extended group is found to be equivalent to the equation for a vortex filament with an axial flow along it.

It is remarkable that the present geometrical formulations are successful for all the problems considered here and give an insight into the deep background common to the diverse physical systems. Furthermore, the geometrical formulation opens a new approach to various dynamical systems, which is rewarded with new results.     

可靠性敏度分析方法及其在非线性蠕变疲劳失效模型中的应用

针对非线性极限状态方程,发展了两种基本随机变量为非正态情况下的可靠性敏度分析方法:基于改进一次二阶矩的近似解析法和基于Monte-Carlo的数字模拟法.近似解析法中非正态变量首先被等价变换为正态变量,然后用正态变量的敏度分析法和隐函数求导法则来得到失效概率对非正态变量分布参数的灵敏度,求解敏度的数字模拟法是从计算失效概率的所有样本点中选取合适的抽样点,利用回归分析和隐函数求导法则来求取可靠性灵敏度的.所提方法被用于非线性蠕变疲劳失效模式的可靠性灵敏度分析,近似解析法和数字模拟法结果的一致说明了所提方法的合理可行.蠕变疲劳失效的可靠性灵敏度随参数的变化趋势分析为工程设计提供了有益指导.     

COMMENTS ON AMSAA-BISE RELIABILITY GROWTH MODEL (Ⅰ)

ＩｎｔｒｏｄｕｃｔｉｏｎＴｈｅＡＭＳＡＡｒｅｌｉａｂｉｌｉｔｙｇｒｏｗｔｈｍｏｄｅｌｐｒｏｐｏｓｅｄｂｙＬａｒｒｙＨ .Ｃｒｏｗ[1,2 ]ｉｎ 1 972ｗａｓｆｉｒｓｔｅｍｐｌｏｙｅｄｂｙｔｈｅＵ .Ｓ .ＡｒｍｙＭａｔｅｒｉｅｌＳｙｓｔｅｍｓＡｎａｌｙｓｉｓＡｃｔｉｖｉｔｙ (ＡＭＳＡＡ) .ＩｔｉｓａｌｓｏｒｅｆｅｒｒｅｄｔｏｉｎｔｈｅｌｉｔｅｒａｔｕｒｅａｓｔｈｅＣｒｏｗＭｏｄｅｌ.ＡＭＳＡＡｍｏｄｅｌｉｓａｗｉｄｅｌｙｕｓｅｄｒｅｌｉａｂｉｌｉｔｙｇｒｏｗｔｈｍｏｄｅｌ.ＩｔｗａｓａｃｃｅｐｔｅｄｂｙＭＩＬ…     

多失效模式下的模式重要性测度及解法

针对工程中普遍存在的多失效模式系统可靠性问题,为提高或改进设计,需要研究失效模式对系统不确定性的影响因素。基于前人提出的基本变量重要性测度,提出了三种模式重要性测度:模式对系统失效概率的重要性测度、模式对系统响应量概率分布的重要性测度和模式功能函数与系统功能函数的相关系数。这三种模式重要性测度从不同的角度反映了失效模式对多模式系统不确定性的贡献。初步讨论了三种模式重要性测度的性质以及其所服务的工程目的,同时给出了模式重要性测度的标准数字模拟求解方法。根据本文提出的模式重要性测度,可以得到模式对系统不确定性的贡献的排序,从而为可靠性设计提供指导。给出了数值算例和工程算例,以说明所提出的模式重要性测度的合理性和可行性。     

On the stability of systems having several equilibrium states

A frequency domain theorem is given for the stability analysis of systems that have several equilibrium states. Under consideration are closed loop systems containing a single instantaneous nonlinearity in an otherwise linear system. A condition is derived, involving the frequency response of the linear part of the system, which guarantees that all bounded solutions approach an equilibrium state, and a geometrical interpretation of the results is found.     

Reliability Against Defect Generated Fracture

ABSTRACT

The subject of this study is the reliability determination for structures that may fail due to fracture generated from some material defect among a random number of small random defects distributed in some way throughout the body of the structure. The considered failure mechanism may be characterized as a weakest link mechanism with a random number of links. The calculation technique is based on some previously published general narrow probability bounds for unions of small probability events. In the case of high reliability structures, as are usually of interest in structural engineering, some very useful conclusions may be drawn from the study. Given only the defect fracture probability corresponding to the actual stress field and the mean number of defects per unit volume both as functions of the space coordinates, a sufficiently accurate assessment of the total reliability may be calculated. With rare exceptions, no dependency information is needed and no information about the defect point process is needed except its mean function. The reliability is insensitive to very large variations of these properties. If there is doubt about the dependency level being too high to neglect it, conditioning followed by fast probability integration may be useful. This is illustrated in the last section of the paper.     

一种基于凸集-概率混合模型的结构可靠性分析法

本文建立一种凸集-概率混合模型下的结构可靠性分析方法。考虑椭球模型和区间模型两种不同类型的凸集模型,在标准空间内通过拉丁超立方抽样生成样本点,通过矩阵变换将其转换到凸集空间内,得到凸集变量的样本点;将凸集变量样本点代入极限状态方程,从而混合可靠性模型转变为纯概率可靠性模型;利用Laplace渐进积分法对每个极限状态方程进行失效概率求解,统计结果的最大值和最小值,得到失效概率的上、下界,研究了凸集变量的个数对结果的影响,通过失效概率的变异系数描述计算结果的稳定性;通过三个算例验证了计算结果的精度,并采用混合模型的蒙特卡洛法进行对比计算。研究表明：本文所提方法计算精度和效率均较高;凸集模型的类型会对结果产生较大影响;为使结果趋于稳定,椭球模型所需的凸集样本点个数多于区间模型;若凸集样本点数目相同,椭球模型的失效概率计算结果变异系数较小,稳定性高于区间模型。     

一种基于凸集-概率混合模型的结构可靠性分析法

本文建立一种凸集-概率混合模型下的结构可靠性分析方法。考虑椭球模型和区间模型两种不同类型的凸集模型,在标准空间内通过拉丁超立方抽样生成样本点,通过矩阵变换将其转换到凸集空间内,得到凸集变量的样本点;将凸集变量样本点代入极限状态方程,从而混合可靠性模型转变为纯概率可靠性模型;利用Laplace渐进积分法对每个极限状态方程进行失效概率求解,统计结果的最大值和最小值,得到失效概率的上、下界,研究了凸集变量的个数对结果的影响,通过失效概率的变异系数描述计算结果的稳定性;通过三个算例验证了计算结果的精度,并采用混合模型的蒙特卡洛法进行对比计算。研究表明：本文所提方法计算精度和效率均较高;凸集模型的类型会对结果产生较大影响;为使结果趋于稳定,椭球模型所需的凸集样本点个数多于区间模型;若凸集样本点数目相同,椭球模型的失效概率计算结果变异系数较小,稳定性高于区间模型。     

具有多种失效模式的抗震结构的模糊可靠性分析

由于地震荷载和结构抗力具有明显的模糊性,在文献[1,2]中我们提出了具有单一失效模式的抗震结构的模糊可靠性分析.本文将上述概念和方法推广到具有多种失效模式的抗震结构,为此,又提出了结构的“模糊有效域”和“模糊失效域”的概念。     

COMMENTS ON AMSAA-BISE RELIABILITY GROWTH MODEL (Ⅱ)

The AMSAA-BISE reliability growth model for multiple systems development derived by the second approach is incorrect as well. The practical failure data, it included 35 systems’ reliability growth test data and 80 systems’ field data, cannot show its correctness of AMSAA-BISE model, it shows that there is something wrong with AMSAA-BISE model.