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

Uncertainty inherent in the operations of water distribution systems (WDSs) poses risks to public health due to the likely insufficient or excessive disinfectants caused by inevitable changes in WDS operations. A methodology that integrates flexibility in WDS operations and location of booster stations was formulated. The flexibility incorporated into the multiobjective problem facilitates more effective chlorine disinfection of WDSs by (1) minimizing the mass injection rate of both the treatment plant and booster stations and (2) minimizing the risk associated with chlorine disinfection in WDSs. The problem was solved using a nondominated sorting genetic algorithm. The decision variables are the locations of booster stations and the mass injection rates of both the treatment plant and booster stations. The results suggest that flexibility adds value in the chlorine booster disinfection of WDSs compared to deterministic solutions if WDSs are subject to uncertain water supply paths, operational interventions, and water demands.

The resilience of a system is generally defined in terms of its ability to withstand external perturbations, adapt, and rapidly recover. This paper introduces a probabilistic formulation to predict the recovery process of a system given past recovery data and to estimate the probability of reaching or exceeding a target value of functionality at any time. A Bayesian inference is used to capture the changes over time of model parameters as recovery data become available during the work progress. The proposed formulation is general and can be applied to continuous recovery processes such as those of economic or natural systems, as well as to discrete recovery processes typical of engineering systems. As an illustration of the proposed formulation, two examples are provided. The paper models the recovery of a reinforced concrete bridge following seismic damage, as well as the population relocation after the occurrence of a seismic event when no data on the duration of the recovery are available a priori.

The dynamic nature of construction projects triggers uncertainty in activity durations. Several approaches such as interval analysis, probabilistic models, and fuzzy set theory have been proposed in the literature to address this uncertainty and provide more realistic duration estimates. However, difficulties in defining the boundaries of duration intervals in the presence of uncertainty and unavailability of historical data make the fuzzy set theory an effective alternative. Despite the potency of fuzzy set theory in providing a systematic tool to address the uncertainty caused by incomplete or ill-defined data, its application in activity duration estimation is very limited. This limitation mainly stems from the complexity of fuzzy calculations and lack of a precise analytical solution for the fuzzy weighted average. To address this gap, this paper presents a fuzzy analytical solution with a concise structure and efficient computation process to estimate the activity durations under uncertainty. The developed model generates a set of risk-related duration modifiers by considering the dependency and combinatory effect of different risk factors. Experts’ confidence level has been modeled using interval-valued fuzzy numbers and an analytical solution based on the Karnik-Mendel (KM) algorithm has been employed to calculate the fuzzy weighted average. The advantages of this approach compared to existing models are (1) considering the dependency and combinatory effect of risk factors on activity durations, (2) accounting for uncertainty in experts’ evaluations by employing the interval-valued fuzzy numbers, and (3) avoiding the creation of incorrect membership functions by using an analytical solution rather than a discrete algorithm to calculate the fuzzy weighted average. To elaborate the model, a simulated bridge project and four risk factors affecting activity durations have been presented. Applying the proposed methodology, a set of duration modifiers to calculate the optimistic, most likely, and pessimistic duration values for each activity under uncertainty are calculated. Compared to similar approaches, the resultant modified durations are more reliable since the combinatory effect of risk factors and the uncertainty of decision makers are taken into consideration.

Crashes involving heavy trucks and vulnerable road users are a major road safety concern because of the higher likelihood of fatal and serious injury. This research identified the socioeconomic characteristics affecting injury severity in collisions involving trucks and vulnerable road users. Specifically, the influences of the sociodemographic characteristics of road users, the neighborhoods where the road users live, and where the crashes occur were explored using the random parameter ordered logit model. This study found that the age and gender of vulnerable road users involved; the age, language, occupation, religion, and income of the person’s residency neighborhood; and the age, occupation, religion, and income of the crash neighborhood had a significant influence on injury severity.

Machine learning has taken a front seat in railway big data analysis. This is partly due to perpetual data collection and the need for automated systems to expedite maintenance decisions. A case for track defect prognosis in rail track engineering is presented in this paper. Fatigue defects are very common and are influential on rail maintenance. Understanding such defects is essential for optimized maintenance scheduling. The literature is replete with machine learning models developed for defect prediction. Because no single machine learning model is guaranteed to surpass others with every kind of data, each model has its inherent deficiencies. Classifier ensembles such as bagging or boosting aggregate strengths from different models to enhance prediction. The outcome is very effective, although highly correlated. This work proposes a stacking method of combining average learners into powerful learning machines while considering memory, time, computational, structural complexities, and bias-variance trade-offs. Because of the large scale of rail infrastructure considered in this work (35,406 km), this study shows that classical Weibull analysis underestimates annual fatigue defects by at least 25% throughout rail life. The proposed stacking ensemble compensates for this shortfall by aggregating the probability predictions of diverse learners. These predictions were combined from a binary classification ensemble of 0.783 receiver operating characteristic area under curve (ROC-AUC) score with significant room for improvement in computation time and curve fitting.

Currently, frequent crash occurrences significantly influence traffic operation conditions and travel reliability for urban expressway systems. Therefore, it is vital to understand the crash occurrence mechanisms and then introduce safety improvement countermeasures. Emerging studies have been conducted to unveil the relationships between traffic operation conditions and crash occurrence with advanced traffic-sensing data. However, the majority of previous studies have only identified correlation relationships, which are insufficient for traffic-safety improvement. On the other hand, existing crash causal investigations have limitations of utilizing aggregated traffic-flow data and considering the crash occurrence mechanisms only in a reflective way (in contrast to the formative way). In this study, the confounding impacts among crash risk contributing factors and the crash causal relationships were revealed through the partial least-squares path modeling (PLS-PM) analysis approach. Data from the Shanghai urban expressway system in China were utilized for the empirical analyses. First, random forest models were adopted to rank the variable importance, and a total of six contributing factors were selected as inputs that feed into the PLS path models. Then, two different causal relationship structures (formative and reflective) were established, and the best-fitted model structures were identified. The results showed that average operation speed has negative impacts on crash occurrence, and the variables indicated that disturbed traffic flows have positive causal relationships. Finally, the analysis results shed some light on proactive safety management strategies.