ASCE-ASME Journal of Risk and Uncertainty in Engineering Systems, Part A: Civil Engineering

In the context of sampling, monitoring, and sensing in infrastructures, there is an interest in algorithms to produce an observation plan that is cost effective, while maximizing the benefits of the new observations. This paper proposes a method to obtain an optimal sampling plan in terms of the number and placement of additional sampling points based on value of information (VoI). VoI can be computed easily through updating a Gaussian random field, i.e., kriging, which is a probabilistic interpolation method. Particle swarm optimization is introduced to optimize a set of sites for new observations with respect to VoI. In the paper, after presenting the basic concept and formulation, we describe applying the method to the placement of additional borings as a liquefaction countermeasure for an embankment along a river. The optimal sampling placement may be obtained at a feasible computational cost even when the number of additional sampling points is greater than 10. The optimal number of sampling points is also evaluated based on VoI.

Assessment of the reliability of lifelines under seismic loads requires particular attention, since the proper functioning of these systems during or after a destructive earthquake is very important. The main objective of this study is to investigate the system reliability of lifeline networks subjected to earthquake loads by concentrating on highway transportation systems consisting of viaducts, bridges, roads, and highways. Seismic reliability of a structural component is determined based on the seismic capacity of the component and earthquake loads acting on it. The future earthquake threat on the components is calculated by performing a probabilistic seismic hazard analysis. Failure reasons of bridges and asphalt roads are explained and behavior of these components under earthquake loads is discussed to determine the seismic capacities of them. In this respect, the system reliability concepts are used and bounds on the seismic reliability of a system are established. A case study based on real-life data is also presented in order to illustrate the implementation of the method.

The windstorm resilience of a 10-story steel frame building is examined by adapting procedures and software initially developed for performance-based seismic engineering. Building vulnerability to wind was predicted using a nonlinear finite element model of the structure subjected to wind loads based on loads measured in wind tunnel tests of a small-scale model of the building. The model of the building structure was idealized using a concentrated plasticity approach, including both the main wind force resisting and gravity systems. Wind tunnel load records were modified to emulate the nonstationary effects of windstorms. Nonlinear response history analyses were used to calculate story drifts and roof and floor accelerations. The calculated values were then correlated to probable building damage using empirical fragility data of structural and nonstructural building components. Monte Carlo simulations of possible building response scenarios were run and the simulation results indicated that during service-level windstorms buildings were generally habitable. During extreme-level sustained windstorms, damage to cladding and a need for the repair of structural components, especially nonductile beam connections, was predicted.

This paper examines the task of approximating or generating samples according to a target probability distribution when this distribution is expressed as a function of the response of an engineering system. Frequently such approximation is performed in a sequential manner, using a series of intermediate densities that converge to the target density and may require a large number of evaluations of the system response, which for applications involving complex numerical models creates a significant computational burden. To alleviate this burden an adaptive Kriging stochastic sampling and density approximation framework (AK-SSD) is developed in this work. The metamodel approximates the system response vector, whereas the adaptive characteristics are established through an iterative approach. At the end of each iteration, the target density, approximated through the current metamodel, is compared to the density established at the previous iteration, using the Hellinger distance as a comparison metric. If convergence has not been achieved, then additional simulation experiments are performed to inform the metamodel development, through a sample-based design of experiments that balances between the improvement of the metamodel accuracy and the addition of experiments in regions of importance for the stochastic sampling. These regions are defined by considering both the target density and any intermediate densities. The process then moves to the next iteration, with an improved metamodel developed using all the available simulation experiments. Although the theoretical discussions are general, the emphasis is placed on rare-event simulation. For this application, once the target density is approximated (first stage), it is used (second stage) as an importance sampling density for estimating the rare-event likelihood. For the second stage, use of either the metamodel or the exact numerical model is examined.

Most international design codes either prohibit construction or recommend alternative seismic analysis and design of vertically irregular building frames, depending on the degree of irregularity and site hazard. However, the irregularity quantifiers recommended by design codes and published literature do not appear to correlate well with the potential for seismic damage. A new parameter, the seismic vulnerability indicator (SVI), is proposed based on the interstory drift ratio damage parameter to indicate the expected seismic risk of any building with single or coupled vertical irregularity. The SVI is computationally simple and found to correlate well with the potential seismic risk of all categories of vertically irregular buildings.

This paper introduces a probabilistic formulation for the estimation of the spatial distribution of multiseverity casualties due to earthquakes. The formulation assesses the number, severity, and distribution of injuries in the affected region. The model is an essential component of resilience formulations of health care systems in a community because it represents the demand on the system. Moreover, the model extends the performance-based earthquake engineering (PBEE) framework from single-building analysis to multiple-building analysis. The paper gives a full description of both the underlying statistical interdependencies among the model’s variables and the extension of the formulation of the PBEE integral to a regional context with multiple buildings. Thus, the formulation advances current methodologies that focus only on single casualty types (e.g., Prompt Assessment of Global Earthquakes for Response [PAGER]) or on the mean number of casualties (e.g., Hazus) rather than their joint probability distribution. Two numerical algorithms are presented in this paper to solve for the casualty model: one based on traditional forward Monte Carlo and another based on the central limit theorem (CLT). It is shown that the latter model is highly computationally efficient while providing accurate results.

This study obtains a three-dimensional unified peak strength expression for confined concrete from the reported experimental results of circular concrete columns confined with different fiber-reinforced polymer (FRP) materials under axial compressive load for application in pile design. The reported experimental results are divided into two groups: FRP-wrapped and FRP tube–encased specimens. An assessment using various statistical indicators is carried out to evaluate the performance of existing predictive equations for peak strength. A reliability analysis of the models is conducted using the advanced first-order second-moment approach. Based on the assessment, it is inferred that most of the models conform to the reliability index criteria suggested by current requirements for structural concrete.

This paper considers the quantification of system reliability in scenarios in which data, that is, failures or the absence of failures, occurring from the system’s use over time, are considered surprising from the perspective of prior information. A generalized, or imprecise, Bayesian approach is presented for general system structures in which the component lifetimes have Weibull distributions with a known shape parameter. For the scale parameter, a specific set of prior distributions is assumed that enables the prior-data conflict to be reflected through the increased imprecision in the posterior reliability bounds.

Particular scenarios of emergent and future conditions are more disruptive than others in terms of project schedules and the evolving priorities of project managers. It is essential to associate the scenarios with cost and schedule impacts to guide corrective or precautionary actions. The innovation of this paper is to use scenarios to investigate two modes of project disruption: (1) scenario-based preferences that change the relative emphasis placed on project management success criteria, and (2) scenarios that may change the structure of the project network and parameters of the constituent activities. These disruptions are used to reprioritize the activities of projects across scenarios subject to precedence constraints. The developed methodology is demonstrated with applications to strategic planning, capacity expansion, and operations of a shipping container port with billions of dollars of planned investment and millions of container transactions per year. The results characterize the scenarios that are most and least disruptive.

The impacts of extreme weather events on railway operations are complex and in the most severe cases can cause significant disruption to the rail services, leading to delays for passengers and financial penalties to the industry. This paper presents a prototype data model with logistic regression analysis, which enables exploration of the underlying causal factors impacting on weather-related incidents on the rail network. The methodology is demonstrated by using wind-related delay data gathered from the Anglia Route of Great Britain’s rail network between financial year 2006–2007 and 2014–2015. The work presented draws on a diverse group of data resources, including climatic, geographical, and vegetation data sets, in order to include a wide range of potential contributing factors in the initial analysis. It investigates ways in which these data may be used to predict when and where wind-related disruptions would be likely to occur, thus enabling us to gain a deeper understanding of the conditions that prevail in sites at risk of disruption events, pointing to possible mitigation in the design of the infrastructure, and their relationship to the local environment.

Tolerance analysis is a key issue in proving the compatibility of manufacturing uncertainties with the quality level of mechanical systems. For rigid and isostatic systems, multiple methods (worst case, statistical, or probabilistic approaches) are applicable and well established. Recent scientific developments have brought enhancements for rigid overconstrained systems using probabilistic and optimization-based methods. The consideration of nonrigid systems is more complex since a large-scale numerical model must be taken into account for an accurate prediction of the quality. The aim of the present paper is the illustration of the probabilistic tolerance analysis approach for an industrial application involving deformable parts. The distributions associated with the dimensions of the components were identified using real components collected from the assembly lines. A nonlinear finite-element model was used to predict the mechanical behavior. A reliability analysis was performed in order to compute the defect probability and estimate the quality of the products. A kriging-based surrogate model was used to reduce the numerical efforts required for the reliability analysis.

Extreme precipitation is one of the most important climate hazards that pose a significant threat to human property and life. Understanding extreme precipitation events helps to manage their risk to society and hence reduce potential losses. This paper provides two new stochastic methods to analyze and predict various extreme precipitation events based on nonstationary models with or without the consideration of serial dependency associated with different days. These methods, together with Monte Carlo simulation and dynamic optimization, bridge nonextreme precipitation and extreme precipitation so that abundant nonextreme precipitation data can be used for extreme precipitation analysis. On an annual basis, the analysis produces distributions for the maximum daily precipitation, number of days with heavy rainfall, and maximum number of consecutive days with heavy rainfall. The accuracy of the new methods is examined, using 10 decades of empirical data in the Washington, DC metropolitan area. Based on the new methods, predictions of various extreme events are provided under different assumptions. Finally, the impact of serial dependency on results is also discussed. The result shows that for the area studied, serial dependency can further improve the analysis result.

This paper compiles a database consisting of 16 tunnel excavation projects in clayey soils using the earth pressure balance (EPB) tunneling method. Statistical analysis of the data is performed and the probability distributions for ground loss ratio $\eta$ and fitting parameter $n$, which are two major parameters in Peck’s formula for transverse surface settlement, are identified based on 78 and 29 points of field data, respectively. The correlation between $\eta$ and $n$ is fitted based on the data and its influences on the failure probability is evaluated. A serviceability performance function is modeled based on Peck’s settlement formula, and the serviceability failure probability is defined and analyzed with Monte Carlo simulations. A series of design charts, in terms of tunnel diameter, cover/diameter ratio, $\eta$, and $n$, are generated for practical uses without running Monte Carlo simulations.

Monte Carlo simulation is the most versatile solution method for problems in stochastic computational mechanics but suffers from a slow convergence rate. The number of simulations required to produce an acceptable accuracy is often impractical for complex and time-consuming numerical models. In this paper, an element-based control variate approach is developed to improve the efficiency of Monte Carlo simulation in stochastic finite-element analysis, with particular reference to high-dimensional and nonlinear geotechnical problems. The method uses a low-order element to form an inexpensive approximation to the output of an expensive, high-order model. By keeping the mesh constant, a high correlation between low-order and high-order models is ensured, enabling a large variance reduction to be achieved. The approach is demonstrated by application to the bearing capacity of a strip footing on a spatially variable soil. The problem requires 300 input random variables to represent the spatial variability by random fields, and would be difficult to solve by methods other than Monte Carlo simulation. Using an element-based control variate reduces the standard deviation of the mean bearing capacity by approximately half. In addition, two methods for estimating the cumulative distribution function as a complement to the improved mean estimator are presented.

The Houston Ship Channel (HSC) is one of the busiest waterway corridors in the United States. Since the channel’s expansion in 2005, concerns have arisen about design deficiencies in the HSC in the area of the Bayport Ship Channel (BSC), especially north of the turn at Five Mile Cut. A mental models expert elicitation exercise was conducted in order to identify safety concerns arising from these design deficiencies and provide qualitative data that can structure analysis of technical data like those from automatic identification system (AIS) databases, which can better connect possible design deficiencies to incident outcomes. The elicitation produced an influence diagram to enable later causal reasoning and Bayesian analysis for the HSC and BSC confluence and nearby areas on the HSC, and helped to prime a comprehensive study of the feasibility of safety and performance modifications on this reach of the HSC.

Linear engineering skeletal structures with spatially varying random stiffness and mass properties, acted upon by a set of random vibratory loads, are considered in this paper. The focus of the study is on characterizing the global–local response sensitivity indexes for this class of structures by taking into account dependency among system parameters and/or forcing functions. This is achieved by extending the recently developed tools for modeling the sensitivity indexes based on various definitions of probability distance measures to problems involving stochastic inhomogeneities. The study allows for the non-Gaussian nature of stochastic variations and employs Monte Carlo simulation strategies and stochastic finite-element modeling tools. The spatially varying random properties are discretized using the optimal linear expansion methods. Illustrative examples include studies on single-span beams under the combined action of a train of random moving masses and earthquake support motions, and randomly parameterized skeletal structures under biaxial earthquake support motions.