基于预测模型的虚拟试验不确定性分析方法

 工程系统中不可避免地存在各种参数不确定性,利用数值计算模型对系统进行虚拟试验时应进行不确定性分析.大型耗时计算模型的不确定性分析将面临严重的的计算复杂性问题,为此,针对工程应用中耗时计算模型,提出一种基于贝叶斯预测模型的不确定性分析仿真方法,采用概率分布为参数不确定性建模,研究系统响应预测不确定性的概率特征.泰勒杆撞击实例验证了该方法的高效性.     

Nonlinear random vibration analysis: A Bayesian nonparametric approach

Random vibration analysis aims to estimate the response statistics of dynamical systems subject to stochastic excitations. Stochastic differential equations (SDEs) that govern the response of general nonlinear systems are often complicated, and their analytical solutions are scarce. Thus, a range of approximate methods and simulation techniques have been developed. This paper develops a hybrid approach that approximates the governing SDE of nonlinear systems using a small number of response simulations and information available a priori. The main idea is to identify a set of surrogate linear systems such that their response probability distributions collectively estimate the response probability distribution of the original nonlinear system. To identify the surrogate linear systems, the proposed method integrates the simulated responses of the original nonlinear system with information available a priori about the number and parameters of the surrogate linear systems. There will be epistemic uncertainty in the number and parameters of the surrogate linear systems because of the limited data. This paper proposes a Bayesian nonparametric approach, called a Dirichlet Process Mixture Model, to capture these uncertainties. The Dirichlet process models the uncertainty over an infinite-dimensional parameter space, representing an infinite number of potential surrogate linear systems. Specifically, the proposed method allows the number of surrogate linear systems to grow indefinitely as the nonlinear system observed dynamic unveil new patterns. The quantified uncertainty in the estimates of the unknown model parameters propagates into the response probability distribution. The paper then shows that, under some mild conditions, the estimated probability distribution approaches, as close as desired, to the original nonlinear system’s response probability distribution. As a measure of model accuracy, the paper provides the convergence rate of the response probability distribution. Because the posterior distribution of the unknown model parameters is often not analytically tractable, a Gibbs sampling algorithm is presented to draw samples from the posterior distribution. Variational Bayesian inference is also introduced to derive an approximate closed-form expression for the posterior distribution. The paper illustrates the proposed method through the random vibration analysis of a nonlinear elastic and a nonlinear hysteretic system.     

Integration of computation and testing for reliability estimation

This paper develops a methodology to integrate reliability testing and computational reliability analysis for product development. The presence of information uncertainty such as statistical uncertainty and modeling error is incorporated. The integration of testing and computation leads to a more cost-efficient estimation of failure probability and life distribution than the tests-only approach currently followed by the industry. A Bayesian procedure is proposed to quantify the modeling uncertainty using random parameters, including the uncertainty in mechanical and statistical model selection and the uncertainty in distribution parameters. An adaptive method is developed to determine the number of tests needed to achieve a desired confidence level in the reliability estimates, by combining prior computational prediction and test data. Two kinds of tests — failure probability estimation and life estimation — are considered. The prior distribution and confidence interval of failure probability in both cases are estimated using computational reliability methods, and are updated using the results of tests performed during the product development phase.     

Bayesian time–domain approach for modal updating using ambient data

The problem of identification of the modal parameters of a structural model using measured ambient response time histories is addressed. A Bayesian time–domain approach for modal updating is presented which is based on an approximation of a conditional probability expansion of the response. It allows one to obtain not only the optimal values of the updated modal parameters but also their associated uncertainties, calculated from their joint probability distribution. Calculation of the uncertainties of the identified modal parameters is very important if one plans to proceed in a subsequent step with the updating of a theoretical finite-element model based on modal estimates. The proposed approach requires only one set of response data. It is found that the updated PDF can be well approximated by a Gaussian distribution centered at the optimal parameters at which the updated PDF is maximized. Examples using simulated data are presented to illustrate the proposed method.     

不确定度的Bayes表征与分析

本给出了测量不确定度的Ｂａｙｅｓ表征方法，该方法是通过对待测对象的测量结果验后分布的分析，获得不确定度表征所需参数值。与古典统计方法相比，Ｂａｙｅｓ方法具有估计精度高，不确定度的表征更为客观的特点。     

A Bayesian approach to diagnosis and prognosis using built-in test

Accounting for the effects of test uncertainty is a significant problem in test and diagnosis, especially within the context of built-in test. Of interest here, how does one assess the level of uncertainty and then utilize that assessment to improve diagnostics? One approach, based on measurement science, is to treat the probability of a false indication [e.g., built-in-test (BIT) false alarm or missed detection] as the measure of uncertainty. Given the ability to determine such probabilities, a Bayesian approach to diagnosis, and by extension, prognosis suggests itself. In the following, we present a mathematical derivation for false indication and apply it to the specification of Bayesian diagnosis. We draw from measurement science, reliability theory, signal detection theory, and Bayesian decision theory to provide an end-to-end probabilistic treatment of the fault diagnosis and prognosis problem.     

基于最大熵区间分析的测量不确定度评定

为了提高测量不确定度评定的精度,采用最大熵区间分析方法。首先通过贝叶斯模型结合最大熵算法建立模型;接着对输入量样本信息下限和上限区间的不对称性进行分析,引入Jaynes熵以及引入拉格朗日量得出最短区间;考虑了输入量的不确定度随概率分布的传递过程,最后对输入量样本信息通过划分区间比值来确定被测量的不确定度评定。实验仿真显示该算法计算测量不确定度的区间较小,评定结果更为精确。     

基于贝叶斯信息融合的测量不确定度评定与实时更新

基于贝叶斯信息融合与统计推断原理,建立不确定度动态评定模型,对测量不确定度进行实时更新。引入最大熵原理和爬山搜索优化算法,确定先验分布概率密度函数及样本信息似然函数,结合贝叶斯公式求出后验分布概率密度函数,实现不确定度的优化估计。仿真及实例分析表明,基于贝叶斯和最大熵方法评定及更新的测量不确定度更加接近理论值。     

基于多分辨率分析的结构物理参数识别贝叶斯估计方法:方法推导与验证

在观测噪声和模型误差等不确定性因素的影响下,结构物理参数识别问题是一个不确定性问题.针对此问题,该文从结构运动微分方程出发,利用小波多分辨率分析原理,建立结构多尺度动力方程,由该方程以结构激励和响应信息在多尺度上的细节信号和最大尺度上的概貌信号为观测量推得物理参数线性回归模型,对该模型应用贝叶斯估计理论得到物理参数后验...     

Model validation under epistemic uncertainty

This paper develops a methodology to assess the validity of computational models when some quantities may be affected by epistemic uncertainty. Three types of epistemic uncertainty regarding input random variables - interval data, sparse point data, and probability distributions with parameter uncertainty - are considered. When the model inputs are described using sparse point data and/or interval data, a likelihood-based methodology is used to represent these variables as probability distributions. Two approaches - a parametric approach and a non-parametric approach - are pursued for this purpose. While the parametric approach leads to a family of distributions due to distribution parameter uncertainty, the principles of conditional probability and total probability can be used to integrate the family of distributions into a single distribution. The non-parametric approach directly yields a single probability distribution. The probabilistic model predictions are compared against experimental observations, which may again be point data or interval data. A generalized likelihood function is constructed for Bayesian updating, and the posterior distribution of the model output is estimated. The Bayes factor metric is extended to assess the validity of the model under both aleatory and epistemic uncertainty and to estimate the confidence in the model prediction. The proposed method is illustrated using a numerical example.     

NPUA: A new approach for the analysis of computer experiments

An important problem in the analysis of computer experiments is the specification of the uncertainty of the prediction according to a meta-model. The Bayesian approach, developed for the uncertainty analysis of deterministic computer models, expresses uncertainty by the use of a Gaussian process. There are several versions of the Bayesian approach, which are different in many regards but all of them lead to time consuming computations for large data sets.In the present paper we introduce a new approach in which the distribution of uncertainty is obtained in a general nonparametric form. The proposed approach is called non-parametric uncertainty analysis (NPUA), which is computationally simple since it combines generic sampling and regression techniques. We compare NPUA with the Bayesian and Kriging approaches and show the advantages of NPUA for finding points for the next runs by reanalyzing the ASET model.     

Internal dosimetry of uranium isotopes using Bayesian inference methods

A group of personnel at Los Alamos National Laboratory is routinely monitored for the presence of uranium isotopes by urine bioassay. Samples are analysed by alpha spectroscopy, and the results are examined for evidence of an intake of uranium. Because the measurement uncertainties are often comparable to the quantities of material we wish to detect, statistical considerations are crucial for the proper interpretation of the data. The problem is further complicated by the significant, but highly non-uniform, presence of uranium in local drinking water and, in some cases, food supply. Software originally developed for internal dosimetry of plutonium has been adapted to the problem of uranium dosimetry. The software uses an unfolding algorithm to calculate an approximate Bayesian solution to the problem of characterising any intakes which may have occurred, given the history of urine bioassay results for each individual in the monitored population. The program uses biokinetic models from ICRP Publications 68 and later, and a prior probability distribution derived empirically from the body of uranium bioassay data collected at Los Alamos over the operating history of the laboratory. For each individual, the software creates a posterior probability distribution of intake quantity and solubility type as a function of time. From this distribution, estimates are made of the cumulative committed dose (CEDE) to each individual. Results of the method are compared with those obtained using an earlier classical (non-Bayesian) algorithm for uranium dosimetry. We also discuss the problem of distinguishing occupational intakes from intake of environmental uranium, within a Bayesian framework.     

Improving optical measurement uncertainty with combined multitool metrology using a Bayesian approach

Recently, there has been significant research investigating new optical technologies for dimensional metrology of features 22?nm in critical dimension and smaller. When modeling optical measurements, a library of curves is assembled through the simulation of a multidimensional parameter space. A nonlinear regression routine described in this paper is then used to identify an optimum set of parameters that yields the closest experiment-to-theory agreement. However, parametric correlation, measurement noise, and model inaccuracy all lead to measurement uncertainty in the fitting process for optical critical dimension measurements. To improve the optical measurements, other techniques such as atomic force microscopy and scanning electronic microscopy can also be used to provide supplemental a priori information. In this paper, a Bayesian statistical approach is proposed to allow the combination of different measurement techniques that are based on different physical measurements. The effect of this hybrid metrology approach will be shown to reduce the uncertainties of the parameter estimators.     

基于分位数函数和Bootstrap分布的不确定度评定研究

最大信息熵方法是基于概率分布评定测量不确度的主要方法之一。其所依赖的高阶矩需要较大样本的测量数据,而校准/检测实验室的测量一般为小样本,故用最大熵方法评定小样本测量不确定度缺乏一定的可靠性。提出了基于分位数函数和概率权重矩作为约束条件的最大信息熵不确定度评定法,把矩的计算从高次降为一次,并结合遗传算法求解概率分布,用Bootstrap分布估计扩展不确定度和包含区间,解决了由分位数区间估计分布不对称所致的复杂计算问题。     

Computer arithmetic for probability distribution variables

The uncertainty in the variables and functions in computer simulations can be quantified by probability distributions and the correlations between the variables. We augment the standard computer arithmetic operations and the interval arithmetic approach to include probability distribution variable (PDV) as a basic data type. Probability distribution variable is a random variable that is usually characterized by generalized probabilistic discretization. The correlations or dependencies between PDVs that arise in a computation are automatically calculated and tracked. These correlations are used by the computer arithmetic rules to achieve the convergent approximation of the probability distribution function of a PDV and to guarantee that the derived bounds include the true solution. In many calculations, the calculated uncertainty bounds for PDVs are much tighter than they would have been had the dependencies been ignored. We describe the new PDV Arithmetic and verify the effectiveness of the approach to account for the creation and propagation of uncertainties in a computer program due to uncertainties in the initial data.     

Accounting for Parameter Uncertainty in Simulation Input Modeling

We formulate and evaluate a Bayesian approach to probabilistic input modeling for simulation experiments that accounts for the parameter and stochastic uncertainties inherent in most simulations and that yields valid predictive inferences about outputs of interest. We use prior information to construct prior distributions on the parameters of the input processes driving the simulation. Using Bayes' rule, we combine this prior information with the likelihood function of sample data observed on the input processes to compute the posterior parameter distributions. In our Bayesian simulation replication algorithm, we estimate parameter uncertainty by independently sampling new values of the input-model parameters from their posterior distributions on selected simulation runs; and we estimate stochastic uncertainty by performing multiple (conditionally) independent runs with each set of parameter values. We formulate performance measures relevant to both Bayesian and frequentist input-modeling techniques, and we summarize an experimental performance evaluation demonstrating the advantages of the Bayesian approach.     

Surrogate-assisted Bayesian inference inverse material identification method and application to advanced high strength steel

Proper definition of certain material properties is a paramount issue for accurate simulation. However, the values of a material parameter are commonly uncertain due to multiple factors in practice. To obtain reliable material parameters, parameter identification via Bayesian theory has become an attractive framework and received more attention recently. Based on this frame, the determination of likelihood function is critical for posterior probability. Unfortunately, it is commonly difficult to be determined directly, especially for complex engineering problems. In this study, Bayesian formulas for material parameter identification are given. To make it feasible for real engineering problems, the least square-support vector regression surrogate and Monte Carlo Simulation are integrated to obtain the maximum likelihood estimation of likelihood function. The uncertainty of parameter identification is quantified via the Bayesian method. In two benchmarks, two cases with single and multiple uncertainty sources are used to propagate and quantify uncertainties in material parameters based on Bayesian approach. Moreover, the proposed method is used to identify the material parameters of advanced high strength steel used in vehicle successfully.     

An uncertainty importance measure using a distance metric for the change in a cumulative distribution function

A simple measure of uncertainty importance using the entire change of cumulative distribution functions (CDFs) has been developed for use in probability safety assessments (PSAs). The entire change of CDFs is quantified in terms of the metric distance between two CDFs. The metric distance measure developed in this study reflects the relative impact of distributional changes of inputs on the change of an output distribution, while most of the existing uncertainty importance measures reflect the magnitude of relative contribution of input uncertainties to the output uncertainty. The present measure has been evaluated analytically for various analytical distributions to examine its characteristics. To illustrate the applicability and strength of the present measure, two examples are provided. The first example is an application of the present measure to a typical problem of a system fault tree analysis and the second one is for a hypothetical non-linear model. Comparisons of the present result with those obtained by existing uncertainty importance measures show that the metric distance measure is a useful tool to express the measure of uncertainty importance in terms of the relative impact of distributional changes of inputs on the change of an output distribution.     

Evaluating systematic differences between laboratories in interlaboratory comparisons

A Bayesian analysis of the data from interlaboratory comparisons involving a single stable traveling standard is presented. The approach is based on the assumption that each participating laboratory provides an estimate of the value of the measurand with zero estimated bias. In addition, it is assumed that each of the reported uncertainties is given in the form of two separate components, one associated with random effects and the other associated with systematic effects. It is finally assumed that all information is consistent. Using Gaussian probability density functions, simple formulas for the joint estimate of the value of the measurand and for the a posteriori estimates of the biases and of their differences are derived. Formulas for the uncertainties of all these estimates are also given.     

GUM与基于观测方程的测量不确定度评估

比较了GUM系列文件与贝叶斯方法评估不确定度的过程,说明GUM是以测量方程为基础的前向分析方法,而贝叶斯分析是以观测方程为基础、以数据分析为主的一种后向不确定度评估方法。运用概率模型的描述方法对上述两种分析过程进行了分析与比较,说明了两种方法在线性测量模型的条件下,不考虑被测量先验分布时两种结果基本一致,对于非线性的测量模型,GUM S1得到的结果与一定先验分布条件下的贝叶斯分析的结果也基本一致。最后通过实例说明了该结论。