An importance sampling procedure for estimating failure probabilities of non-linear dynamic systems subjected to random noise

The first passage problem for linear and non-linear oscillators excited by white and coloured noise are considered. An iterative variance reduction scheme is used in a framework of a measure change in the space of sample functions according to the Girsanov transformation, which is based on introducing a Markov control process. It is proved that a good approximation to the optimal stochastic control process can be obtained from an equivalent white noise excited linear oscillator. It is shown that this leads to very accurate estimates of the failure probability of the original system. The advantage of this procedure is that expressions for the parameters of the equivalent linear system and the design point oscillations, which are needed to find the control process, are available analytically. The number of samples, the variance of the failure probability estimates and the computational time are reduced significantly compared with direct Monte Carlo simulations.     

基于优化算法的串联体系可靠度分析

结构体系的失效概率数学上可以表示为结构体系失效域上联合概率密度函数的积分,一般情况下很难直接积分求解。近几十年来,结构体系可靠度分析一直是可靠度领域的一个研究热点,人们提出许多方法,如：Monte—Carlo法、重要性抽样法与界限法和概率网络估算技术等,这些算法在求解精度、计算效率、收敛性和易使用性等方面是不同的。本文采用优化算法(改进的可行方向法、序列线性规划和序列二次规划法)进行串联体系可靠度分析,并且与其他算法(HL—RF法、Monte—Carlo法和重要性抽样法)的结果以及一些精确解进行了比较。结果表明,相对于其他算法,基于优化算法的可靠度分析适用性广,在收敛性和健实性等方面具有明显的优势。     

Time-delayed stochastic optimal control of strongly non-linear systems with actuator saturation by using stochastic maximum principle

A time-delayed stochastic optimal bounded control strategy for strongly non-linear systems under wide-band random excitations with actuator saturation is proposed based on the stochastic averaging method and the stochastic maximum principle. First, the partially averaged Itô equation for the system amplitude is derived by using the stochastic averaging method for strongly non-linear systems. The time-delayed feedback control force is approximated by a control force without time delay based on the periodically random behavior of the displacement and velocity of the system. The partially averaged Itô equation for the system energy is derived from that for the system amplitude by using Itô formula and the relation between system amplitude and system energy. Then, the adjoint equation and maximum condition of the partially averaged control problem are derived based on the stochastic maximum principle. The saturated optimal control force is determined from maximum condition and solving the forward–backward stochastic differential equations (FBSDEs). For infinite time-interval ergodic control, the adjoint variable is stationary process and the FBSDE is reduced to a ordinary differential equation. Finally, the stationary probability density of the Hamiltonian and other response statistics of optimally controlled system are obtained from solving the Fokker–Plank–Kolmogorov (FPK) equation associated with the fully averaged Itô equation of the controlled system. For comparison, the optimal control forces obtained from the time-delayed bang–bang control and the control without considering time delay are also presented. An example is worked out to illustrate the proposed procedure and its advantages.     

Monte Carlo simulation in the stochastic analysis of non-linear systems under external stationary Poisson white noise input

A method for the evaluation of the probability density function (p.d.f.) of the response process of non-linear systems under external stationary Poisson white noise excitation is presented. The method takes advantage of the great accuracy of the Monte Carlo simulation (MCS) in evaluating the first two moments of the response process by considering just few samples. The quasi-moment neglect closure is used to close the infinite hierarchy of the moment differential equations of the response process. Moreover, in order to determine the higher order statistical moments of the response, the second-order probabilistic information given by MCS in conjunction with the quasi-moment neglect closure leads to a set of linear differential equations. The quasi-moments up to a given order are used as partial probabilistic information on the response process in order to find the p.d.f. by means of the C-type Gram-Charlier series expansion.     

First passage failure of quasi-partial integrable generalized Hamiltonian systems

The first passage failure of quasi-partial integrable generalized Hamiltonian systems is studied by using the stochastic averaging method. First, the stochastic averaging method for quasi-partial integrable generalized Hamiltonian systems is introduced briefly. Then, the backward Kolmogorov equation governing the conditional reliability function and the Pontryagin equation governing the conditional mean of first passage time are derived from the averaged Itô equations. The conditional reliability function, the conditional probability density and mean of the first passage time are obtained from solving these equations together with suitable initial condition and boundary conditions, respectively. Finally, one example is given to illustrate the proposed procedure in detail and the solutions are confirmed by using the results from Monte Carlo simulation of the original system.     

Stochastic averaging of energy harvesting systems

A stochastic averaging method is proposed for nonlinear energy harvesters subjected to external white Gaussian noise and parametric excitations. The Fokker–Planck–Kolmogorov equation of the coupled electromechanical system of energy harvesting is a three variables nonlinear parabolic partial differential equation whose exact stationary solutions are generally hard to find. In order to overcome difficulties in solving higher dimensional nonlinear partial differential equations, a transformation scheme is applied to decouple the electromechanical equations. The averaged Itô equations are derived via the standard stochastic averaging method, then the FPK equations of the decoupled system are obtained. The exact stationary solution of the averaged FPK equation is used to determine the probability densities of the displacement, the velocity, the amplitude, the joint probability densities of the displacement and velocity, and the power of the stationary response. The effects of the system parameters on the output power are examined. The approximate analytical outcomes are qualitatively and quantitatively supported by the Monte Carlo simulations.     

Importance sampling for elasto-plastic systems using adapted process with deterministic control

A new importance sampling method is presented for computing the first passage probability of elasto-plastic systems under white noise excitations. The importance sampling distribution corresponds to shifting the mean of the excitation to an ‘adapted’ stochastic process whose future is determined based on information only up to the present. This causal property is responsible for maintaining the theoretical rigor of the resulting importance sampling procedure. It opens up great possibilities in the design of adapted process for variance reduction using importance sampling. In this work the adapted process is designed using concepts of deterministic optimal control. Numerical results show that the use of adapted process is particularly useful for hysteretic systems where hysteretic effects undermine the effectiveness of conventional importance sampling method based on fixed design points.     

Nonlinear stochastic analysis of subharmonic response of a shallow cable

The paper deals with the subharmonic response of a shallow cable due to time variations of the chord length of the equilibrium suspension, caused by time varying support point motions. Initially, the capability of a simple nonlinear two-degree-of-freedom model for the prediction of chaotic and stochastic subharmonic response is demonstrated upon comparison with a more involved model based on a spatial finite difference discretization of the full nonlinear partial differential equations of the cable. Since the stochastic response quantities are obtained by Monte Carlo simulation, which is extremely time-consuming for the finite difference model, most of the results are next based on the reduced model. Under harmonical varying support point motions the stable subharmonic motion consists of a harmonically varying component in the equilibrium plane and a large subharmonic out-of-plane component, producing a trajectory at the mid-point of shape as an infinity sign. However, when the harmonical variation of the chordwise elongation is replaced by a narrow-banded Gaussian excitation with the same standard deviation and a centre frequency equal to the circular frequency of the harmonic excitation, the slowly varying phase of the excitation implies that the phase difference between the in-plane and out-of-plane displacement components is not locked at a fixed value. In turn this implies that the trajectory of the displacement components is slowly rotating around the chord line. Hence, a large subharmonic response component is also present in the static equilibrium plane. Further, the time variation of the envelope process of the narrow-banded chordwise elongation process tends to enhance chaotic behaviour of the subharmonic response, which is detectable via extreme sensitivity on the initial conditions, or via the sign of a numerical calculated Lyapunov exponent. These effects have been further investigated based on periodic varying chord elongations with the same frequency and standard deviation as the harmonic excitation, for which the amplitude varies in a well-defined way between two levels within each period. Depending on the relative magnitude of the high and low amplitude phase and their relative duration the onset of chaotic vibrations has been verified.     

子集模拟的混合采样算法

子集模拟是当前可靠度计算领域常用的估计算法,相比于直接蒙特卡罗积分法,极大减少了函数调用的次数。子集模拟法一般使用单一采样器,然而不同采样器适用范围不同。如使用单一椭圆切片采样器,其遍历性较好但函数调用次数较多;而使用单一自适应条件采样时,其采样效率较高但样本容易陷入局部极值。单一采样器由于本身特性面对不同问题时失效概率的积分结果可能出现偏差,模拟效果不稳定。本文首次提出了一种混合采样子集模拟法,在子集模拟的前几层使用椭圆切片采样,此时失效区域收缩程度有限,函数调用次数在可接受的范围内,样本经过采样扩充后能充分探索参数空间,更有效地检测出所有失效区域。当失效区域收缩至一定限度后,使用自适应条件采样,此时种子样本继承了前几层样本较低的相关性,并在此基础上通过自适应条件采样更高效地增殖样本。本文通过四个模拟算例多种维度下的数值积分验证了该算法具有椭圆切片采样器较好的遍历性,同时采样效率位于椭圆切片采样与自适应条件采样之间,对于不同问题拥有良好的通用性。     

Improvement of parameter estimation for non-linear hysteretic systems with slip by a fast Bayesian bootstrap filter

Modeling and identification of non-linear hysteretic systems are widely encountered in the structural dynamics field, especially for the hysteresis with slip. A model, called SL model, which can describe the pinching of most practical hysteresis loops perfectly was proposed by Baber and Noori (J. Eng. Mech. 111 (1985) 1010). A method of estimating the parameters of SL model on the basis of input-output data based on bootstrap filter was proposed by the writers. Bootstrap filter is a filtering method based on Bayesian state estimation and Monte Carlo method, which has the great advantage of being able to handle any functional non-linearity and system and/or measurement noise of any distribution. The standard bootstrap filter, however, is not time efficient, i.e., it is very time consuming and is not suitable for real-time applications. In this paper, previous work by the writers is extended to do the parameter estimation of SL model by a fast Bayesian bootstrap filtering technique. Simulation results are presented to demonstrate the performance of the algorithm.     

Stochastic stability of non-linear SDOF systems

A semi-analytical procedure for obtaining stability conditions for strongly non-linear single degree of freedom system (SDOF) subjected to random excitations is presented using stochastic averaging technique. The method is useful for finding stability conditions for systems having highly irregular non-linear functions which cannot be integrated in closed form to yield analytical expressions for averaged drift and diffusion coefficients. In spite of numerical methods available for finding stability of SDOF system by determining Lyapunov exponent, the proposed technique may have to be adopted (i) when the excitation is non-white; and (ii) when numerical integration fails due to convergence problem. The method is developed in such a way that it lends itself to a numerical computational scheme using FFT for obtaining numerical values of drift and diffusion coefficients of Its differential equation and the corresponding FPK equation for the system. These values of averaged drift and diffusion coefficients are then fit into polynomial form using curve fitting technique so that polynomials can be used for stability analysis. Two example problems are solved as illustrations. The first one is the Van der Pol oscillator having non-linearities which can be treated purely analytically. The example is considered for the validation of the proposed method. The second one involves non-linearities in the form of signum function for which purely analytical solution is not possible. The results of the study show that the proposed method is useful and efficient for performing stability analysis of dynamic systems having any type of non-linearities.     

一种用于蠕变损伤失效概率计算的分层抽样方法

根据蠕变损伤试验和相关的研究资料,确定出影响材料蠕变损伤的随机参数;建立了蠕变损伤失效的概率模型.其次根据分层抽样法的基本思想,建立了一种考虑对蠕变损伤失效概率区间分情况抽样的蒙特卡洛分层抽样方法.并且在蠕变损伤模型中选择等效应力作为一个分层抽样变量.随后根据工程实际情况,提出了一种在统计意义下确定抽样变量的安全区域和失效区域的分层抽样的策略,以减少抽样次数.并推导出确定分层抽样区间的计算方法.然后对选择区间大小进行了讨论,可知如果所选择的分层抽样区间太小,会漏掉许多导致蠕变损伤失效的等效应力抽样值,从而使计算结果误差增大;但是如果分层抽样区间选择太大,会增加无效抽样次数,因而降低抽样效率.最后的算例表明了该方法在同样的计算精度下,其计算效率比直接抽样法的效率要高.     

结构动力可靠度的重要抽样法

重要抽样法是蒙特卡洛数值模拟方法中的一种重要的方差缩减技术,目前重要抽样法在工程结构可靠度计算中的应用主要集中于静力问题。本文分析了动力可靠度蒙特卡洛方法的特点,提出了在结构动力可靠度问题中应用重要抽样法的方法,并针对白噪声荷载,给出了选择重要抽样函数的方法和重要抽样函数的具体表达式。理论和数值分析表明,本文所提出的重要抽样法应用于结构动力可靠度计算是可行的。     

重要性抽样法在计算球罐断裂失效概率中的应用

本文结合实例,在探讨了原分布函数平移一定距离后作为重要性函数的基础上,提出采用优化理论中的直接搜索法,以原分布函数与重要性函数相似比的变异系数最小为优化目标求得最佳平移量。同时提出迭代计算程序,即依据前次迭代计算得到的失效点分布确定计算时原分布函数的平移量,经多次迭代也能求出最佳平移量。多次迭代后诸失效点的中值与一阶二次矩法的设计点值接近。在本文算例中,由这二种方法求得的重要性函数减少了模拟计算总次数和提高了计算精度。     

Stochastic averaging method for estimating first-passage statistics of stochastically excited Duffing–Rayleigh–Mathieu system

The first-passage statistics of Duffing-Rayleigh- Mathieu system under wide-band colored noise excitations is studied by using stochastic averaging method. The motion equation of the original system is transformed into two time homogeneous diffusion Markovian processes of amplitude and phase after stochastic averaging. The diffusion process method for first-passage problem is used and the corresponding backward Kolmogorov equation and Pontryagin equation are constructed and solved to yield the conditional reliability function and mean first-passage time with suitable initial and boundary conditions. The analytical results are confirmed by Monte Carlo simulation.     

随机过程激励下随机结构系统可靠度分析的一种方法

提出了随机过程荷载激励下，具有随机参数的结构系统可靠度分析的一种方法，该方法基于首次超越破坏机制，分析随机过程荷载激励下，结构参数（随机变量）取某一确定向量时的条件失效概率，采用Ｍｏｎｔｅ Ｃａｒｌｏ技术模拟结构参数的随机性，由条件失效概率给出随机结构的无条件失效概率，最后对中方法和程序作了检验，并进行了实际计算。     

基于分层抽样法的可靠性灵敏度分析

提出基于分层抽样的可靠性灵敏度分析方法。该方法将变量空间分割成多个互斥子空间,根据子空间对可靠性灵敏度估计值贡献率的大小来分配子空间的抽取的样本量,这种抽样策略使得更多的样本来源于失效域,从而达到降低估计方差、提高了收敛速度的目的。本文还推导了基于分层抽样法的可靠性灵敏度估计值的方差和变异系数。三个算例显示出分层抽样在单模式、多模式并联、多模式串联结构可靠性灵敏度分析上的可行性和优越性。     

用于结构可靠性分析中的系统描述性抽样法

在用蒙特卡罗法计算结构系统可靠性时,动用了系统描述性抽样法。用系统描述性抽样代替蒙特卡罗的随即抽样过程,从而使得计算效率与降低方差都得到了明显地改善。     

Bayesian System Identification using auxiliary stochastic dynamical systems

Bayesian approaches to statistical inference and system identification became practical with the development of effective sampling methods like Markov Chain Monte Carlo (MCMC). However, because the size and complexity of inference problems has dramatically increased, improved MCMC methods are required. Dynamical systems based samplers are an effective extension of traditional MCMC methods. These samplers treat the posterior probability distribution as the potential energy function of a dynamical system, enabling them to better exploit the structure of the inference problem. We present an algorithm, Second-Order Langevin MCMC (SOL-MC), a stochastic dynamical system based MCMC algorithm, which uses the damped second-order Langevin stochastic differential equation (SDE) to sample a posterior distribution. We design the SDE such that the desired posterior probability distribution is its stationary distribution. Since this method is based upon an underlying dynamical system, we can utilize existing work to develop, implement, and optimize the sampler's performance. As such, we can choose parameters which speed up the convergence to the stationary distribution and reduce temporal state and energy correlations in the samples. We then apply this sampler to a system identification problem for a non-linear hysteretic structure model to investigate this method under globally identifiable and unidentifiable conditions.