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.     

钢筋混凝土柱抗弯承载力的概率模型与校准分析

为了克服传统确定性抗弯承载力模型和校准方法无法合理考虑不确定性所存在的缺陷,分别建立了钢筋混凝土(RC)柱的概率抗弯承载力模型与概率校准方法。首先,基于RC柱正截面受弯承载力的基本假定,结合偏心受压RC柱的截面内力平衡条件,分别建立了大(小)偏心受压RC柱的确定性抗弯承载力模型;然后,综合考虑固有不确定性和认知不确定性的影响,分别建立了大(小)偏心受压RC柱概率抗弯承载力模型的解析表达式,进而结合贝叶斯理论和MCMC法确定了概率模型参数的后验分布信息,从而建立了RC柱的概率抗弯承载力模型;最后,基于概率抗弯承载力模型所确定的概率密度函数、置信区间和置信水平,提出了传统确定性抗弯承载力模型的概率校准方法。研究结果表明,所建立的概率抗弯承载力模型不仅可以合理描述RC柱抗弯承载力的概率分布特性,而且可以校准传统确定性抗弯承载力模型的计算精度和置信水平。     

基于Bayes方法的多元正态应力强度可靠度分析

实际工程中,应力和强度是多维相关的,且相关信息很难获得,使用传统的单维应力-强度干涉模型很难进行可靠度分析,需要研究多元应力-强度干涉模型的可靠度计算问题。描述了两种多元应力-强度干涉模型,针对应力、强度服从正态分布的情形,使用Bayes方法,充分利用验前信息,由多元应力-强度干涉模型获得分布参数验后分布,并基于多元统计学知识,采用仿真抽样方法,进行可靠度验后预报。巡航弹可用过载的仿真算例验证Bayes方法在应力、强度试验信息较少的情况下,可高精度地预测可靠度的验后分布。     

基于随机加权的捷联惯组历次测试数据验前分布研究

针对捷联惯性测量组合(捷联惯组)历次测试数据小样本的特点,在总体分布参数形式未知的情况下,根据已有的先验信息,提出了通过随机加权法,获得捷联惯组历次测试数据总体参数的验前分布。结合当前样本信息,利用贝叶斯方法得到捷联惯组历次测试数据的验后分布。将统计推断建立在验后分布基础之上,可以减小小样本情况下的统计分析误差。     

A quadratic hierarchical Bayesian dynamic prediction model for infrastructure maintenance

The main objective of this study is the development of a correlation model in dynamic Bayesian belief networks (DBBNs) followed by an inverse economic analysis. This is based on a quadratic hierarchical Bayesian inference prediction method using Markov chain Monte Carlo simulations. The developed model is implemented to predict the future degradation and maintenance budget for a suspension bridge system. Bayesian inference is applied to find the posterior probability density function of the source parameters (damage indices and serviceability), given 10 years maintenance data. The simulated risk prediction under decreased serviceability conditions gives posterior distributions based on a prior distribution and likelihood data updated from annual maintenance tasks. Compared with a conventional linear prediction model, the proposed quadratic model provides highly improved convergence and closeness to the measured data. Finally, the developed inverse DBBN analysis method allows forecasts of future performance and the financial management of complex infrastructures by providing the sensitivity of serviceability and risky factors to the maintenance budgets of structural components and the overall system.     

Analytical rheology of blends of linear and star polymers using a Bayesian formulation

A Bayesian data analysis technique is presented as a general tool for inverting linear viscoelastic models of branched polymers. The proposed method takes rheological data of an unknown polymer sample as input and provides a distribution of compositions and structures consistent with the rheology, as its output. It does so by converting the inverse problem of analytical rheology into a sampling problem, using the idea of Bayesian inference. A Markov chain Monte Carlo method with delayed rejection is proposed to sample the resulting posterior distribution. As an example, the method is applied to pure linear and star polymers and linear–linear, star–star, and star–linear blends. It is able to (a) discriminate between pure and blend systems, (b) accurately predict the composition of the mixtures, in the absence of degenerate solutions, and (c) describe multiple solutions, when more than one possible combination of constituents is consistent with the rheology.     

Bayesian Statistic Based Multivariate Gaussian Process Approach for Offline/Online Fatigue Crack Growth Prediction

Offline and online fatigue crack growth prediction of Aluminum 2024 compact-tension (CT) specimens under variable loading has been modeled, using multivariate Gaussian Process (GP) technique. The GP model is a Bayesian statistic stochastic model that projects the input space to an output space by probabilistically inferring the underlying nonlinear function. For the offline prediction, the input space of the model is trained with parameters that affect fatigue crack growth, such as the number of fatigue cycles, minimum load, maximum load, and load ratio. For the online prediction, the model input space is trained using piezoelectric sensor signal features rather than training the input space with loading parameters, which are difficult to measure in a real time scenario. Principal Component Analysis (PCA) is used to extract the principal features from sensor signals. In both the offline and online case, the output space is trained with known associated crack lengths or crack growth rates. Once the GP model is trained, a new output space for which the corresponding crack length or crack growth rate is not known, is predicted using the trained GP model. The models are validated through several numerical examples.     

Advanced statistics to improve the physical interpretation of atomization processes

This paper reports an analysis of the physics of atomization processes using advanced statistical tools. Namely, finite mixtures of probability density functions, which best fitting is found using a Bayesian approach based on a Markov chain Monte Carlo (MCMC) algorithm. This approach takes into account eventual multimodality and heterogeneities in drop size distributions. Therefore, it provides information about the complete probability density function of multimodal drop size distributions and allows the identification of subgroups in the heterogeneous data. This allows improving the physical interpretation of atomization processes. Moreover, it also overcomes the limitations induced by analyzing the spray droplets characteristics through moments alone, particularly, the hindering of different natures of droplet formation. Finally, the method is applied to physically interpret a case-study based on multijet atomization processes.     

预测结构性能退化的混合粒子滤波方法

由于载荷,环境以及材料内部因素的作用,结构的性能一般随时间而逐渐退化. 为了评估结构服役期间的状态,常采用随机变量模型来描述结构性能的退化规律. 即,采用含不确定性模型参数的物理模型来逼近结构响应特性. 利用同类型结构的先知数据集信息可确定模型参数的先验分布. 结合结构服役期间的检测信息和贝叶斯原理,对模型参数进行更新,从而提高物理模型的准确性. 本文提出一种混合粒子滤波方法(particle filter-differential evolution adaptive Metropolis,PF-DREAM)用于模型更新,即:在确定参数先验分布时,采用证据理论(Dempster-shafer theory, DST)初始化模型参数;结合差分进化自适应 Metropolis 算法(differential evolution adaptive Metropolis, DREAM)和粒子滤波(particle filter, PF)算法,来计算更新公式中的复杂的高维积分. 相比于传统的 PF 算法,混合 PF-DREAM 方法可以有效提高样本粒子的多样性,解决重采样算法中粒子多样性匮乏的问题,从而得到更加合理的物理模型. 为了证明该方法的有效性,将提出的方法分别应用于电池性能退化和裂纹扩展规律预测. 算例表明采用本文提出的模型参数确定方法,使得物理模型更加合理,性能预测更加准确. 用于更新的数据越多,模型参数的分散性越小. 本文方法应用于高维问题或隐式函数问题时,计算原理和步骤不发生改变,但函数评价次数和计算时间会随之增大.      

The reduced‐order hybrid Monte Carlo sampling smoother

Hybrid Monte Carlo sampling smoother is a fully non‐Gaussian four‐dimensional data assimilation algorithm that works by directly sampling the posterior distribution formulated in the Bayesian framework. The smoother in its original formulation is computationally expensive owing to the intrinsic requirement of running the forward and adjoint models repeatedly. Here we present computationally efficient versions of the hybrid Monte Carlo sampling smoother based on reduced‐order approximations of the underlying model dynamics. The schemes developed herein are tested numerically using the shallow‐water equations model on Cartesian coordinates. The results reveal that the reduced‐order versions of the smoother are capable of accurately capturing the posterior probability density, while being significantly faster than the original full‐order formulation. Copyright © 2016 John Wiley & Sons, Ltd.     

剪切型钢筋混凝土柱临界斜裂缝倾角的概率模型

对于剪切型钢筋混凝土(RC)柱,传统的确定性临界斜裂缝倾角模型难以有效考虑其材料参数、几何尺寸和边界约束条件等方面不确定性因素的影响,导致计算精度有限,且离散性较大。鉴于此,本文研究建立了剪切型RC柱临界斜裂缝倾角的概率模型。首先,基于变角桁架模型理论,建立了剪切型RC柱临界裂缝倾角的确定性模型;然后,考虑主观不确定性和客观不确定性因素的影响,结合贝叶斯理论和马尔可夫链蒙特卡洛(MCMC)法,建立了剪切型RC柱临界斜裂缝倾角的概率模型,并推导了均值和方差的解析表达式,从而能够合理描述临界斜裂缝倾角的概率分布特性;最后,利用试验数据对比验证了该模型的有效性,并校准了现有确定性模型的置信水平,进而确定了不同置信水平下剪切型RC柱临界斜裂缝倾角的概率特征值。     

Inference of polymer structure by simultaneous analysis of chromatographic and rheological measurements

A Bayesian framework that integrates chromatographic and rheological measurements to infer the structure of an unknown binary linear blend, with a high molar mass tail, is proposed and explored. A visualization method based on clustering of multidimensional data is introduced to facilitate the comprehension of the joint probability distribution that results from the Bayesian analysis. The relationship between the rheological and chromatographic data is found to be synergetic: the simultaneous analysis of both sets of data circumvents problems of degeneracy and insensitivity associated with rheology and chromatography, respectively.     

钢筋混凝土梁临界斜裂缝倾角计算的概率模型

综合考虑剪跨比、混凝土强度、配筋率和配箍率等重要因素的影响,研究建立了RC梁临界斜裂缝倾角计算的概率模型。首先,基于修正压力场理论,建立了RC梁临界斜裂缝倾角的确定性计算模型;然后,引入剪跨比修正系数,并综合考虑主观不确定性和客观不确定性因素的影响,结合贝叶斯理论和马尔科夫链蒙特卡洛法（MCMC）,建立了RC梁临界斜裂缝倾角计算的概率模型;最后,通过与试验数据和传统确定性计算模型的对比分析,验证了该模型的有效性和适用性。分析结果表明,随着剪跨比的增大,RC梁的临界斜裂缝倾角逐渐减小;临界斜裂缝倾角的试验测试值的范围为23°～41°,具有显著的离散性;本文概率模型不仅可以合理描述临界斜裂缝倾角的概率分布特性,而且可以校准传统确定性计算模型的计算精度和置信水平,以及根据预定的置信水平确定临界斜裂缝倾角的概率特征值,具有良好的计算精度和适用性。     

Finite element solutions of axisymmetric Euler equations for an incompressible and inviscid fluid

In this paper we present a finite element method for the numerical solution of axisymmetric flows. The governing equations of the flow are the axisymmetric Euler equations. We use a streamfunction angular velocity and vorticity formulation of these equations, and we consider the non-stationary and the stationary problems. For industrial applications we have developed a general model which computes the flow past an annular aerofoil and a duct propeller. It is able to take into account jumps of angular velocity and vorticiy in order to model the flow in the presence of a propeller. Moreover, we compute the complete flow around the after-body of a ship and the interaction between a ducted propeller and the stern. In the stationary case we have developed a simple and efficient version of the characteristics/finite element method. Numerical tests have shown that this last method leads to a very fast solver for the Euler equations. The numerical results are in good agreement with experimental data.     

P-CS不确定性量化模型与其性能数据驱动更新方法

在实际工程中, 广泛存在大量的不确定性信息, 直接或间接影响着工程结构形式设计、结构性能评估与预测以及在役结构损伤识别等工作的开展与决策. 这些多源不确定性信息往往需要用多种不同的不确定性量化模型加以描述; 与此同时, 不确定性变量在使用过程中可能随时间变化且难以直接测量, 需要间接根据性能测试信息在使用工程中更新不确定性量化模型. 为兼顾上述两个问题, 本文基于等概率变换原则提出了一种P-CS (probability-convex set) 不确定性量化模型, 该模型将不确定性变量用概率随机变量与非概率凸集变量组合表征, 可统一表达概率模型、非概率模型以及非精确概率模型, 实现多源、多类型不确定性的统一量化. 本文进一步基于贝叶斯理论提出了一种针对该P-CS不确定性量化模型的性能数据驱动更新方法. 该更新方法根据性能测试数据信息更新P-CS不确定性量化模型参数取值的信度分布, 从而根据后验信度分布计算得出当前P-CS不确定性量化模型参数集合. 通过数值算例详述了P-CS不确定性量化模型的构建方法与其概率、非概率特性, 并验证了性能数据驱动更新P-CS模型方法的适用性.      

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.     

基于显示连通贝叶斯网络的机场塔台地震易损性分析

将基于性能的多维易损性分析方法,结合显示连通贝叶斯网络,应用于机场塔台的多维易损性分析。考虑地震激励的不确定性,通过非线性时程分析获得结构响应数据;将塔台结构分为三个层次,每个层次按包含的层数分为相应的子层次。根据功能特性确定子层次的评价指标和极限状态,建立服从多元对数正态分布的概率地震需求模型;考虑各种极限状态之间的相关性,建立极限状态方程,确定失效域,通过蒙特卡洛法求得构件的超越概率;建立塔台结构的显示连通贝叶斯网络模型,利用层次分析法获得中间节点的条件概率表,利用MATLAB进行贝叶斯网络的推理计算,实现从单一层次的易损性到整体易损性的推理。     

Uncertainty quantification of a graphite nitridation experiment using a Bayesian approach

In this paper, a stochastic system based Bayesian approach is applied to estimate different model parameters and hence quantify the uncertainty of a graphite nitridation experiment. The Bayesian approach is robust due to its ability to characterize modeling uncertainties associated with the underlying system and is rigorous due to its exclusive foundation on the axioms of probability theory. We choose an experiment by Zhang et al. [1] whose main objective is to measure the reaction efficiency for the active nitridation of graphite by atomic nitrogen. To obtain the primary physical quantity of interest, we need to model and estimate the uncertainty of a number of other physical processes associated with the experimental setup. We use the Bayesian method to obtain posterior probability distributions of all the parameters relevant to the experiment while taking into account uncertainties in the inputs and the modeling errors. We use a recently developed stochastic simulation algorithm which allows for efficient sampling in the high-dimensional parameter space. We show that the predicted reaction efficiency of the graphite nitridation and its uncertainty is ∼3.1 ± 1.0 × 10⁻³ that is slightly larger than the ones deterministically obtained by Zhang et al. [1].     

Global Continuous Riemann Solver for Nonlinear Elasticity

We consider in this paper an isothermal model of nonlinear elasticity. This model is described by two conservation laws that define a problem of mixed type, both elliptic and hyperbolic. We restrict ourselves to the linearly degenerate case, and consider Riemann data that lies in the hyperbolic regions. The lack of uniqueness of the Riemann problem is solved by the introduction of a so-called kinetic relation, used to narrow the set of admissible subsonic phase transitions. In this situation, we consider the Riemann problem for any data lying in the hyperbolic region, using either explicit computations or geometric arguments. This construction allows us to give sufficient conditions on the kinetic relation in order that the generated Riemann solver possesses properties of uniqueness, globality, and continuous dependence on the initial data in the L ¹ distance. Accepted October 1, 2000?Published online January 22, 2001