Sequential design of computer experiments for the estimation of a probability of failure (original) (raw)
Related papers
A sequential Bayesian algorithm to estimate a probability of failure
2009
This paper deals with the problem of estimating the probability of failure of a system, in the challenging case where only an expensive-to-simulate model is available. In this context, the budget for simulations is usually severely limited and therefore classical Monte Carlo methods ought to be avoided. We present a new strategy to address this problem, in the framework of sequential Bayesian planning. The method uses kriging to compute an approximation of the probability of failure, and selects the next simulation to be conducted so as to reduce the mean square error of estimation. By way of illustration, we estimate the probability of failure of a control strategy in the presence of uncertainty about the parameters of the plant.
Computers & Structures, 2012
Estimation of small failure probabilities is one of the most important and challenging computational problems in reliability engineering. The failure probability is usually given by an integral over a high-dimensional uncertain parameter space that is difficult to evaluate numerically. This paper focuses on enhancements to Subset Simulation (SS), proposed by Au and Beck, which provides an efficient algorithm based on MCMC (Markov chain Monte Carlo) simulation for computing small failure probabilities for general high-dimensional reliability problems. First, we analyze the Modified Metropolis algorithm (MMA), an MCMC technique, which is used in SS for sampling from high-dimensional conditional distributions. The efficiency and accuracy of SS directly depends on the ergodic properties of the Markov chains generated by MMA, which control how fast the chain explores the parameter space. We present some observations on the optimal scaling of MMA for efficient exploration, and develop an optimal scaling strategy for this algorithm when it is employed within SS. Next, we provide a theoretical basis for the optimal value of the conditional failure probability p 0 , an important parameter one has to choose when using SS. We demonstrate that choosing any p 0 ∈ [0.1, 0.3] will give similar efficiency as the optimal value of p 0 . Finally, a Bayesian post-processor SS+ for the original SS method is developed where the uncertain failure probability that one is estimating is modeled as a stochastic variable whose possible values belong to the unit interval. Simulated samples from SS are viewed as informative data relevant to the system's reliability. Instead of a single real number as an estimate, SS+ produces the posterior PDF of the failure probability, which takes into account both prior information and the information in the sampled data. This PDF quantifies the uncertainty in the value of the failure probability and it may be further used in risk analyses to incorporate this uncertainty. To demonstrate SS+, we consider its application to two different reliability problems: a linear reliability problem and reliability analysis of an elasto-plastic structure subjected to strong seismic ground motion. The relationship between the original SS and SS+ is also discussed.
Estimating a probability of failure with the convex order in computer experiments
arXiv (Cornell University), 2019
This paper deals with the estimation of a failure probability of an industrial product. To be more specific, it is defined as the probability that the output of a physical model, with random input variables, exceeds a threshold. The model corresponds with an expensive to evaluate black-box function, so that classical Monte Carlo simulation methods cannot be applied. Bayesian principles of the Kriging method are then used to design an estimator of the failure probability. From a numerical point of view, the practical use of this estimator is restricted. An alternative estimator is proposed, which is equivalent in term of bias. The main result of this paper concerns the existence of a convex order inequality between these two estimators. This inequality allows to compare their efficiency and to quantify the uncertainty on the results that these estimators provide. A sequential procedure for the construction of a design of computer experiments, based on the principle of the Stepwise Uncertainty Reduction strategies, also results of the convex order inequality. The interest of this approach is highlighted through the study of a real case from the company STMicroelectronics.
Adaptive Design of Experiments for Conservative Estimation of Excursion Sets
Technometrics, 2019
We consider the problem of estimating the set of all inputs that leads a system to some particular behavior. The system is modeled by an expensive-to-evaluate function, such as a computer experiment, and we are interested in its excursion set, i.e. the set of points where the function takes values above or below some prescribed threshold. The objective function is emulated with a Gaussian Process (GP) model based on an initial design of experiments enriched with evaluation results at (batch-) sequentially determined input points. The GP model provides conservative estimates for the excursion set, which control false positives while minimizing false negatives. We introduce adaptive strategies that sequentially select new evaluations of the function by reducing the uncertainty on conservative estimates. Following the Stepwise Uncertainty Reduction approach we obtain new evaluations by minimizing adapted criteria. Tractable formulae for the conservative criteria are derived, which allow more convenient optimization. The method is benchmarked on random functions generated under the model assumptions in different scenarios of noise and batch size. We then apply it to a reliability engineering test case. Overall, the proposed strategy of minimizing false negatives in conservative estimation achieves competitive performance both in terms of model-based and model-free indicators.
Practical reliability and uncertainty quantification in complex systems : final report
2009
The purpose of this project was to investigate the use of Bayesian methods for the estimation of the reliability of complex systems. The goals were to find methods for dealing with continuous data, rather than simple pass/fail data; to avoid assumptions of specific probability distributions, especially Gaussian, or normal, distributions; to compute not only an estimate of the reliability of the system, but also a measure of the confidence in that estimate; to develop procedures to address time-dependent or aging aspects in such systems, and to use these models and results to derive optimal testing strategies. The system is assumed to be a system of systems, i.e., a system with discrete components that are themselves systems. Furthermore, the system is "engineered" in the sense that each node is designed to do something and that we have a mathematical description of that process. In the time-dependent case, the assumption is that we have a general, nonlinear, time-dependent function describing the process.
Cost optimization of reliability testing by a bayesian approach
Mechanics & Industry, 2014
The Bayesian approach is a stochastic method, allowing to establish trend studies on the behavior of materials between two periods or after a break in the life of these materials. It naturally integrates the inclusion of the information partially uncertain to support in modeling problem. The method is therefore particularly suitable for the analysis of the reliability tests, especially for equipment and organs whose different tests are costly. Bayesian techniques are used to reduce the size of estimation tests, improving the evaluation of the parameters of product reliability by the integration of the past (data available on the product concerned) and process, the case "zero" failure observed, difficult to treat with conventional statistical approach. This study will concern the reduction in the number of tests on electronic or mechanical components installed in a mechanical lift knowing their a priori behavior in order to determine their a posteriori behavior.
A multilevel Monte Carlo method for computing failure probabilities
We propose and analyze a method for computing failure probabilities of systems modeled as numerical deterministic models (e.g., PDEs) with uncertain input data. A failure occurs when a functional of the solution to the model is below (or above) some critical value. By combining recent results on quantile estimation and the multilevel Monte Carlo method we develop a method which reduces computational cost without loss of accuracy. We show how the computational cost of the method relates to error tolerance of the failure probability. For a wide and common class of problems, the computational cost is asymptotically proportional to solving a single accurate realization of the numerical model, i.e., independent of the number of samples. Significant reductions in computational cost are also observed in numerical experiments.
Brazilian Journal of Physics, 2006
In this work, we present a new procedure, called sub-sampling, to obtain data concerning time of failure in trials without replacement, (NRT). With this data it is possible to determine the prediction interval (PI) for the future number of failures. We also present an alternative way to evaluate the coverage probability of the prediction interval (PI). The results presented show that the method proposed is reliable and can be useful for the statistical analyses of quality control of processes.
A New Monte Carlo Based Algorithm for the Gaussian Process Classification Problem
Gaussian process is a very promising novel technology that has been applied to both the regression problem and the classification problem. While for the regression problem it yields simple exact solutions, this is not the case for the classification problem, because we encounter intractable integrals. In this paper we develop a new derivation that transforms the problem into that of evaluating the ratio of multivariate Gaussian orthant integrals. Moreover, we develop a new Monte Carlo procedure that evaluates these integrals. It is based on some aspects of bootstrap sampling and acceptancerejection. The proposed approach has beneficial properties compared to the existing Markov Chain Monte Carlo approach, such as simplicity, reliability, and speed.