Small Failure Probability Research Articles

Estimation of Small Failure Probability in High‐Dimensional Groundwater Contaminant Transport Modeling Using Subset Simulation Coupled With Preconditioned Crank‐Nicolson MCMC

AbstractThe accurate prediction of groundwater contamination is challenging due to uncertainties arising from the inherent heterogeneity of aquifers, inadequate site characterization, and limitations in conceptual mathematical models. These factors can result in an underestimation of contaminant concentrations. For effective contaminant prevention and control, it is important to estimate the probability of exceeding the allowed threshold for contaminant concentrations, known as the failure probability of groundwater contamination. Computing small failure probabilities using classical Monte Carlo simulation (MCS) requires computing a large number of samplers to converge to a stationary target value, which is time‐consuming. To address this, in this paper, we develop a novel approach for calculating small failure probabilities, known as subset simulation (SS) coupled with preconditioned Crank‐Nicolson Markov chain Monte Carlo (pCN‐SS), which combines subset simulation with preconditioned Crank‐Nicolson Markov chain Monte Carlo (pCN‐MCMC) to promote computational efficiency. We have tested the performance of the proposed algorithm in both a mathematical example and a numerical case study of groundwater contamination. The results demonstrate that pCN‐SS provides improved accuracy and efficiency for evaluating small failure probabilities for high‐dimensional groundwater contamination, specifically for hydraulic conductivity as a source of uncertainty. Compared to classical MCS and traditional SS, pCN‐SS requires fewer model evaluations but produces stable and accurate results.

A stratified beta-sphere sampling method combined with important sampling and active learning for rare event analysis

Accurate and efficient estimation of small failure probability subjected to high-dimensional and multiple failure domains is still a challenging task in structural reliability engineering. In this paper, we propose a stratified beta-spheres sampling method (SBSS) to tackle this task. Initially, the whole support space of random input variables is divided into a series of subdomains by using multiple specified beta-spheres, which is a hypersphere centered in the origin in standard normal space, then, the corresponding samples truncated by beta-spheres are generated explicitly and efficiently. Based on the truncated samples, the real failure probability can be estimated by the sum of failure probabilities of these subdomains. Next, we discuss and demonstrate the unbiasedness of the estimation of failure probability. The proposed method stands out for inheriting the advantages of Monte Carlo simulation (MCS) for highly nonlinear, high-dimensional problems, and problems with multiple failure domains, while overcoming the disadvantages of MCS for rare event. Furthermore, the SBSS method equipped with importance sampling technique (SBSS-IS) is also proposed to improve the robustness of estimation. Additionally, we combine the proposed SBSS and SBSS-IS methods with GPR model and active learning strategy so as to further substantially reduce the computational cost under the desired requirement of estimated accuracy. Finally, the superiorities of the proposed methods are demonstrated by six examples with different problem settings.

Dual Power Transformation and Yeo–Johnson Techniques for Static and Dynamic Reliability Assessments

This paper addresses key challenges in the static and dynamic reliability analysis of engineering structures, particularly the difficulty in accurately estimating large reliability indices and small failure probabilities. For static reliability problems, a dual power transformation is employed to transform the performance function into a form approaching a normal distribution. The high-order unscented transformation is then applied to compute the first four moments of the transformed performance function. Subsequently, the fourth-moment method is used to calculate large reliability indices, offering a novel improvement over traditional methods such as FORM and SORM. For dynamic reliability problems, the low-discrepancy sampling technique is integrated to efficiently compute structural responses under random seismic excitation, improving computational efficiency for complex dynamic systems. The Yeo–Johnson transformation is introduced to normalize the extreme values of dynamic responses, and the first four moments of the transformed extreme values are statistically evaluated. Additionally, a third-order polynomial transformation (TPT) is applied to approximate the probability density function, leading to the derivation of the probability of exceedance (POE) curve. The optimal transformation parameters for both the dual power and Yeo–Johnson transformations are determined using the Jarque–Bera (JB) test. Four numerical examples, coupled with Monte Carlo simulation, validate the proposed framework’s accuracy and efficiency, providing a robust tool for static and dynamic reliability analysis. This unified approach represents a significant advancement by integrating novel transformations and fourth-moment method, providing a powerful and efficient tool for static and dynamic reliability analysis of engineering structures.

An active learning reliability method combining population Monte Carlo and Kriging model for small failure probability

This study investigates the analysis of small failure probability problems and proposes the AK-PMC method, which combines the Population Monte Carlo (PMC) method with an active learning Kriging model (AK). The PMC method, as a variant of importance sampling, iteratively improving the quality of particles by reweighting them and updating the proposal distributions. In AK-PMC, Kriging model and PMC collaborate with each other. In each iteration, PMC draws potential importance samples according to the prediction information of Kriging model and Kriging is updated by choosing an optimal training point among the importance samples of PMC. After several iterations, the importance samples of PMC will cover the important failure regions and the Kriging model will accurately predict the limit state surface. To avoid the waste of training points, the relative error of failure probability estimated by PMC is derived by considering the prediction information of Kriging model. Three numerical examples and one practical engineering example are investigated to verify the performance of the proposed method.

Variational Bayesian Monte Carlo simulation for rare event assessment

Estimating rare events in structural engineering demands intensive computational resources, and the complexity of the performance function poses challenges to accurate probability estimation. To efficiently conduct the structural reliability analysis on rare events, we propose a novel variational Bayesian Monte Carlo (VBMC)-based subset simulation (SS) method. The proposed approach can be divided into two aspects: the Monte Carlo simulation (MCS) is first applied to estimate the first-layer unconditional failure probability, then VBMC is employed to estimate subsequent conditional failure probabilities. Finally, the original small failure probability is obtained by the product of the first-layer unconditional failure probability and a series of conditional failure probabilities. The proposed approach inherits the merits of SS as well as VBMC, which converts the tricky rare event estimation into a series of high-frequency event estimations and leverages VBMC instead of the canonical Markov chain to produce the independent and informative next-generation failure-conditional samples. Four case studies, including high-dimensional and discrete state space scenarios, are performed to illustrate the feasibility and generality of the proposed method.

Tiny Pointers

This paper introduces a new data-structural object that we call the tiny pointer. In many applications, traditional \(\log n\) -bit pointers can be replaced with \(o(\log n)\) -bit tiny pointers at the cost of only a constant-factor time overhead and a small probability of failure. We develop a comprehensive theory of tiny pointers, and give optimal constructions for both fixed-size tiny pointers (i.e., settings in which all of the tiny pointers must be the same size) and variable-size tiny pointers (i.e., settings in which the average tiny-pointer size must be small, but some tiny pointers can be larger). If a tiny pointer references an item in an array filled to load factor \(1-\delta\) , then the optimal tiny-pointer size is \(\Theta(\log\log\log n+\log\delta^{-1})\) bits in the fixed-size case, and \(\Theta(\log\delta^{-1})\) expected bits in the variable-size case. Our tiny-pointer constructions also require us to revisit several classic problems having to do with balls and bins; these results may be of independent interest. Using tiny pointers, we apply tiny pointers to five classic data-structure problems. We show that: A data structure storing \(n\) \(v\) -bit values for \(n\) keys with constant-factor time modifications/queries can be implemented to take space \(nv+O(n\log^{(r)}n)\) bits, for any constant \(r\gt0\) , as long as the user stores a tiny pointer of expected size \(O(1)\) with each key—here, \(\log^{(r)}n\) is the \(r\) -th iterated logarithm. Any binary search tree can be made succinct, meaning that it achieves \((1+o(1))\) times the optimal space, with constant-factor time overhead, and can even be made to be within \(O(n)\) bits of optimal if we allow for \(O(\log^{*}n)\) -time modifications—this holds even for rotation-based trees such as the splay tree and the red-black tree. Any fixed-capacity key-value dictionary can be made stable (i.e., items do not move once inserted) with constant-factor time overhead and \((1+o(1))\) -factor space overhead. Any key-value dictionary that requires uniform-size values can be made to support arbitrary-size values with constant-factor time overhead and with an additional space consumption of \(\log^{(r)}n+O(\log j)\) bits per \(j\) -bit value for an arbitrary constant \(r\gt0\) of our choice. Given an external-memory array \(A\) of size \((1+\varepsilon)n\) containing a dynamic set of up to \(n\) key-value pairs, it is possible to maintain an internal-memory stash of size \(O(n\log\varepsilon^{-1})\) bits so that the location of any key-value pair in \(A\) can be computed in constant time (and with no IOs). In each case tiny pointers allow for us to take a natural space-inefficient solution that uses pointers and make it space-efficient for free.

Seismic reliability analysis using Subset Simulation enhanced with an explorative adaptive conditional sampling algorithm

Reliability analysis of structures under earthquake loading represents a significant engineering challenge. This is due to the required and rather numerically involving non-linear dynamic analysis, the large computational burden when targeting small failure probabilities, and the synthetic earthquake model representation that may contain thousands of random variables. Subset Simulation is an efficient reliability analysis technique that can handle the challenge of a high-dimensional space with a reduced number of structural analysis calls compared to crude Monte Carlo Simulation. In this contribution, firstly, we investigate the conditions for which Subset Simulation performs efficiently. Thereafter we propose an enhancement to the existing Subset Simulation schemes that shows significant potentials for enhancing the strategy for the starting of the Markov Chain Monte Carlo simulations whenever a new level is reached in the Subset Simulation. Finally, the information gathered from the simulations is investigated to verify that Subset Simulation provides meaningful results from a physical point of view.

An improved sequential importance sampling method for structural reliability analysis of high dimensional problems

Reliability analysis for rare events has proven to be a significant challenge, particularly in non-linear and complex scenarios. Despite the effectiveness of Sequential Importance Sampling (SIS) and Subset Simulation (SUS) in addressing high-dimensional small failure probability problems, the computational burden associated with mechanical simulations for complex structures remains substantial. To address this issue, this paper introduces a novel approach referred to as SIS and Kriging Metamodel Integration (SISAK), designed for computing small failure probabilities in high dimensions. This method involves generating a small number of samples from the original distribution to construct an adaptive metamodel, which is then iteratively optimized using a series of intermediate distributions introduced by SIS. By substituting the function for intermediate distributions with a continuously optimized metamodel, the enhanced approach significantly reduces the computational load associated with mechanical model evaluations compared to the direct use of SIS. This approach does not depend on determining the most probable failure point and does not require consideration of the failure domain shape. Additionally, it inherits the advantages of SIS in addressing high-dimensional small failure probabilities, making it well-suited for structural reliability analysis of nonlinear systems, discontinuous failure domains, cascading functions, and high-dimensional problems.

A study on the small failure probabilities of cylindrical composite hydrogen storage tanks using subset simulation

Although composite hydrogen tanks are becoming increasingly intriguing, a detailed structural reliability analysis is still lacking. The current landscape of analysis for pressurised multilayer cylinders, while rich in research, lacks tools to deal with the low probabilities of failure that govern the design of hydrogen tanks. This study presents a computational framework integrating the Subset Simulation method (SS) for assessing the failure probability of composite tanks used for hydrogen storage. To explore how randomness impacts design parameters, this study utilises a limit state equation derived from the thin-walled circumferential model of composite pressure tanks. Five random variables, each with varying coefficients of variation (COVs), are incorporated into the analysis. For comparison and validation purposes, two methods Monte Carlo simulations and FORM have been used. The analysis revealed that SS excels at estimating the low failure probabilities of composite hydrogen tanks, and showed also the feasibility and accuracy of SS in the prediction of burst pressure since good agreement was obtained between the probabilistic approach and the experimental results. Furthermore, the uncertainties related to internal pressure and composite thickness affect significantly the reliability of the structure and can lead to the shrinkage of the safety margin and the failure of the vessel.

A single-loop reliability sensitivity analysis strategy for time-dependent rare events with both random variables and stochastic processes

To deal with the time-dependent reliability sensitivity (TDRS) analysis of rare events with both random variables and stochastic processes, a single-loop sampling procedure combined with the cross entropy-based importance sampling strategy (CEIS), namely SCEIS, is derived to lighten the computational burden without loss of precision. Firstly, the TDRS indices are derived to make them implementable by a single-loop importance sampling (IS) procedure. Then, the quasi-optimal IS density is obtained through the extension of the cross-entropy method, which is an adaptive procedure that overcomes the challenges of determining the position and number of design points in traditional IS. Furthermore, the SCEIS can be easily implemented to estimate the TDRS indices for problems with small failure probability. Three examples involving numerical and engineering problems are analyzed to demonstrate the performance of the proposed method. The results obtained from comparing with other existing methods reveal that the proposed method provides satisfactory TDRS analysis for rare events while significantly reducing computational costs.

Efficient estimation procedure for failure probability function by an augmented directional sampling

AbstractFailure probability function (FPF) can reflect quantitative effects of random input distribution parameter (DP) on failure probability, and it is significant for decoupling reliability‐based design optimization (RBDO). But the FPF estimation is time‐consuming since it generally requires repeated reliability analyses at different DPs. For efficiently estimating FPF, an augmented directional sampling (A‐DS) is proposed in this paper. By using the property that the limit state surface (LSS) in physical input space is independent of DP, the A‐DS establishes transformation of LSS samples in standard normal spaces corresponding to different DPs. By the established transformation in different standard normal spaces, the LSS samples obtained by DS at a given DP can be transformed to those at other DPs. After simple interpolation post‐processing on those transformed samples, the failure probability at other DPs can be estimated by DS simultaneously. The main novelty of A‐DS is that a strategy of sharing DS samples is designed for estimating the failure probability at different DPs. The A‐DS avoids repeated reliability analyses and inherits merit of DS suitable for solving problems with multiple failure modes and small failure probability. Compared with other FPF estimation methods, the examples sufficiently verify the accuracy and efficiency of A‐DS.

A method of combined metamodel and subset simulation for reliability analysis of rare events

Reliability analysis of large, complex structures with high reliability has always been a challenging task. The Subset Simulation (SUS) has proven highly effective for high-dimensional and small failure probability problems. However, the computational demands of time-consuming numerical simulations in the field of mechanical engineering remain substantial. Metamodeling techniques provide an effective means to reduce simulation computational costs. This paper introduces a novel approach, termed SUSAK, which combines the Subset Simulation with Kriging metamodels to address the challenge of small failure probability calculations. By using the original distribution of input variables representing the structural system, the SUSAK method establishes an adaptive metamodel and iteratively optimizes it with intermediate failure domain samples corresponding to the intermediate events. By replacing the performance function in subsequent calculations of the probability associated with intermediate events with metamodels, the enhanced method significantly reduces the computational burden compared to using the SUS method. This approach does not rely on solving for the design point, is not constrained by the shape of the failure domain, and retains the advantages of the SUS method in high-dimensional small failure probability calculations. As a result, it is well-suited for calculating small failure probabilities in nonlinear, discontinuous failure domains, multiple failure domains, and high-dimensional problems.

A double-loop adaptive relevant vector machine combined with Harris Hawks optimization-based importance sampling

PurposeEnsuring the safety of structures is important. However, when a structure possesses both an implicit performance function and an extremely small failure probability, traditional methods struggle to conduct a reliability analysis. Therefore, this paper proposes a reliability analysis method aimed at enhancing the efficiency of rare event analysis, using the widely recognized Relevant Vector Machine (RVM).Design/methodology/approachDrawing from the principles of importance sampling (IS), this paper employs Harris Hawks Optimization (HHO) to ascertain the optimal design point. This approach not only guarantees precision but also facilitates the RVM in approximating the limit state surface. When the U learning function, designed for Kriging, is applied to RVM, it results in sample clustering in the design of experiment (DoE). Therefore, this paper proposes a FU learning function, which is more suitable for RVM.FindingsThree numerical examples and two engineering problem demonstrate the effectiveness of the proposed method.Originality/valueBy employing the HHO algorithm, this paper innovatively applies RVM in IS reliability analysis, proposing a novel method termed RVM-HIS. The RVM-HIS demonstrates exceptional computational efficiency, making it eminently suitable for rare events reliability analysis with implicit performance function. Moreover, the computational efficiency of RVM-HIS has been significantly enhanced through the improvement of the U learning function.

A layer assigned probability space partition method for structural small failure probability problem

The Physical Synthesis Method (PSM) stands out as a robust framework for conducting structural reliability analyses due to its clear conceptual foundation. However, this approach often necessitates significant computational resources when addressing scenarios with small failure probabilities. In response to this challenge, this study introduces a layer assigned probability space partition method designed to identify pivotal points based on the ultimate bearing capacity failure criterion of structural components within the PSM framework. Drawing inspiration from Harbitz's β-sphere, this method effectively utilizes the minimum reliability index of components to discern essential representative points within the probability space, thus streamlining computations. The efficacy of this approach is showcased through two case studies: a simply supported beam and a six-story reinforced concrete frame. The outcomes demonstrate that the proposed method, when integrated with PSM, exhibits a substantial enhancement in efficiency compared to the conventional Monte Carlo method. Besides, under equivalent computational resources, it achieves superior computational accuracy compared to the importance sampling method, particularly in scenarios with small failure probabilities. Furthermore, by introducing the notion of a common safe domain, this method addresses challenges in structural reliability analyses involving multiple failure surfaces.

Reliability analysis of complex systems using subset simulations with Hamiltonian Neural Networks

We present a new Subset Simulation approach using Hamiltonian neural network-based Monte Carlo sampling for reliability analysis. The proposed strategy combines the superior sampling of the Hamiltonian Monte Carlo method with computationally efficient gradient evaluations using Hamiltonian neural networks. This combination is especially advantageous because the neural network architecture conserves the Hamiltonian, which defines the acceptance criteria of the Hamiltonian Monte Carlo sampler. Hence, this strategy achieves high acceptance rates at low computational cost. Our approach estimates small failure probabilities using Subset Simulations. However, in low-probability sample regions, the gradient evaluation is particularly challenging. The remarkable accuracy of the proposed strategy is demonstrated on different reliability problems, and its efficiency is compared to the traditional Hamiltonian Monte Carlo method. We note that this approach can reach its limitations for gradient estimations in low-probability regions of complex and high-dimensional distributions. Thus, we propose techniques to improve gradient prediction in these particular situations and enable accurate estimations of the probability of failure. The highlight of this study is the reliability analysis of a system whose parameter distributions must be inferred with Bayesian inference problems. In such a case, the Hamiltonian Monte Carlo method requires a full model evaluation for each gradient evaluation and, therefore, comes at a very high cost. However, using Hamiltonian neural networks in this framework replaces the expensive model evaluation, resulting in tremendous improvements in computational efficiency.

System reliability analysis with small failure probability based on relevant vector machine and Meta-IS idea

Structural failure is a complex system problem and traditional methods face difficulties in solving system reliability problems with small failure probability. This paper proposes an efficient sampling simulation method termed SDMI, which integrates metamodel-based importance sampling (Meta-IS) with spherical decomposition (SD). SDMI transforms the failure probability into a product of spherical augmented failure probability and a correction factor. The implementation of SDMI for each component further facilitates the proposition of a system probability classification function. The integration of the relevant vector machine (RVM) leads to the proposal of a more efficient methodology, termed RVM-SDMI. The efficacy and precision of the proposed methods have been validated through four numerical examples and two engineering examples. The results demonstrate that SDMI successfully addresses the convergence challenges inherent in Meta-IS. Moreover, RVM-SDMI exhibits remarkable accuracy and computational efficiency in the system reliability analysis with small failure probabilities.

Meta-model based sequential importance sampling method for structural reliability analysis under high dimensional small failure probability

Reliability analysis poses a significant challenge for complex structures with stringent reliability requirements. While Sequential Importance Sampling (SIS) and Subset Simulation (SUS) have proven highly effective in addressing high-dimensional problems with small failure probabilities, the computational burden of mechanical simulations remains substantial due to the time-consuming nature of numerical simulation processes. Consequently, this paper introduces a novel approach, denoted as AK-SIS, which combines SIS with Kriging metamodeling specifically designed to address computational challenges associated with small failure probabilities. The fundamental principle of this approach involves utilizing AK-MCS technology (Echard et al., 2011) [1] as a precursor to the SIS approach to initially generate metamodels. These metamodels are then employed in lieu of performance functions in subsequent steps, significantly reducing the number of function calls required to simulate complex engineering problems when applying SIS techniques directly. By inheriting the advantages of SIS, AK-SIS has demonstrated its suitability for reliability analysis in scenarios involving high-dimensional spaces and small fault probabilities. Furthermore, AK-SIS is not limited by the shape of the failure domain, eliminates the need to solve the design point, and is particularly well-suited for analyzing reliability in cases of discontinuous failure domains, multiple failure domains, as well as complex failure domains and rare events. The efficacy of AK-SIS is substantiated through rigorous evaluation encompassing nonlinear, high-dimensional examples, and an engineering application. These empirical validations collectively contribute to a robust methodological framework for reliability analysis of intricate structures characterized by stringent reliability requirements.

Structural reliability analysis with extremely small failure probabilities: A quasi-Bayesian active learning method

The concept of Bayesian active learning has recently been introduced from machine learning to structural reliability analysis. Although several specific methods have been successfully developed, significant efforts are still needed to fully exploit their potential and to address existing challenges. This work proposes a quasi-Bayesian active learning method, called ‘Quasi-Bayesian Active Learning Cubature’, for structural reliability analysis with extremely small failure probabilities. The method is established based on a cleaver use of the Bayesian failure probability inference framework. To reduce the computational burden associated with the exact posterior variance of the failure probability, we propose a quasi posterior variance instead. Then, two critical elements for Bayesian active learning, namely the stopping criterion and the learning function, are developed subsequently. The stopping criterion is defined based on the quasi posterior coefficient of variation of the failure probability, whose numerical solution scheme is also tailored. The learning function is extracted from the quasi posterior variance, with the introduction of an additional parameter that allows multi-point selection and hence parallel distributed processing. By testing on four numerical examples, it is empirically shown that the proposed method can assess extremely small failure probabilities with desired accuracy and efficiency.

Small failure probability analysis of stochastic structures based on a new hybrid approach

The small failure probability problem of stochastic structures is investigated by using two types of surrogate models and the subset simulation method in conjunction with parallel computation. To achieve high computational efficiency, the explicit expression of dynamic responses of stochastic structures is first derived in the form based on the explicit time-domain method. Then, the small failure probability analysis of stochastic structures is efficiently carried out through the Monte Carlo simulation method utilizing explicit expressions. To avoid the repeated calculation for the coefficient matrices or vectors of the explicit expression of stochastic structures, two types of surrogate models, e.g., the backpropagation neural network model and the Kriging model, are introduced to obtain these matrices or vectors for each parameter sample of the stochastic structures. The computational cost is further reduced by using the subset simulation method to generate conditional samples which follow the rule of Metropolis-Hastings. Furthermore, in virtue of the independence of the surrogate models for each time instant and the independence of dynamic analysis for each sample, parallel computation is embedded in the proposed approach, which can fully exploit the characteristics of the proposed approach and further improve the computational efficiency of dynamic reliability analysis. Numerical examples are given to illustrate the validity of the proposed hybrid approach.

Seismic collaborative reliability analysis for a slope considering spatial variability base on optimized subset simulation

Seismic reliability analysis of actual slopes considering the spatial variability of soil materials is crucial. However, for the discretization of large-scale random fields, high-precision finite element analysis and analysis of small failure probability events, the random analysis of slopes under seismic loads is inefficient. To address this situation, this study proposes a collaborative reliability analysis framework based on the modified linear estimation method (MLEM) and optimized subset simulation (OSS). First, the random field of the uncertain parameters of the Jinping-I left bank slope model is efficiently discretized by the MLEM, and a sensitivity analysis is carried out. Then, considering the adoption of different degrees of cross-correlation of the sensitive random parameters, the OSS method is used to perform random finite element analysis on the coarse mesh model. Finally, the fine mesh samples are obtained according to the response conditioning method (RCM). The MLEM is used to ensure the consistency of the two sets of random fields, and the seismic failure probability and reliability index of the slope under different cross-correlation coefficients of uncertain parameters are obtained. The results suggest that the degree of cross-correlation of parameters has a great influence on the seismic reliability of the slope. Considering that the shear strength parameters of geotechnical materials are often negatively correlated, the fine analysis based on a fine model is necessary.