Metaheuristics in Reliability and Risk Analysis

The special collection on Metaheuristics in Reliability and Risk Analysis is available in the ASCE Library at https://ascelibrary.org/page/ajrua6/metaheuristics_analysis.

Reliability and risk analysis for civil engineering systems is a complex procedure including various design parameters and constraints. Metaheuristic algorithms are viable tools to deal with such complicated modeling and optimization problems (Gandomi et al. 2013). These techniques attempt to reproduce natural phenomena or social behavior, e.g., biological evolution, stellar evolution, thermal annealing, animal behavior, music improvisation, and so forth. In the engineering domain, a variety of modern and computationally efficient metaheuristic algorithms have been presented recently. However, the capabilities of metaheuristics have not been investigated sufficiently for problems related to the reliability and risk analysis. This special issue of ASCE-ASME Journal of Risk and Uncertainty in Engineering Systems, Part A: Civil Engineering strives to review the development and key applications of the metaheuristic algorithms in engineering systems reliability and risk analysis. We assembled a set of scientific contributions to provide a window to a successful implementation of the metaheuristics to challenging problems in this field.

In the opening paper of this issue, Machiani et al. (2016) present a physics-based metaheuristic called simulated annealing to solve a complex traffic congestion problem. Traffic congestion impacts safety, and aggressive driving leads to negative environmental impacts such as air pollution and excess fuel consumption. The authors formulate the problem to determine location, speed limit reduction, start time, and duration of a limited number of variable speed limit signs while maximizing travel time reliability on selected critical paths on a network. They use a bilevel optimization approach for this purpose in which the upper level is for travel time reliability optimization and the lower level assigns traffic to the network. The authors use a 20-dynamic traffic assignment simulation tool and choose the real network of Guam as a case study. The simulated annealing results showed that it can improve travel time reliability up to 0.6%.

Gholizadeh and Mohammadi (2016) propose a hybrid algorithm to tackle reliability-based seismic design optimization problems. Using the performance-based design concept, they try to optimize steel moment resisting frames. Three types of constraints are involved in these problems, including serviceability, ultimate limit-state, and 464 probabilistic checks. The authors propose a hybridized particle swarm optimization (PSO) and bat algorithm (BA) called the PSO-BA algorithm. In the PSO-BA algorithm, PSO explores the search space, and then BA continues the optimization process by exploiting the neighboring region around the best solution found by the PSO algorithm. They used Monte Carlo simulation (MCS) to evaluate the reliability constraints during the optimization process. Wavelet back-propagation neural network models are developed to predict the required deterministic and probabilistic seismic responses at different performance levels. Two case studies considered are a 3-story and a 10-story frame. The PSO-BA simulation results are compared with the standard PSO and BA algorithm. The results confirm that not only the proposed algorithm converge faster, but also it obtains a better final solution in comparison with the PSO and BA algorithms.

Sajedi et al. (2016) formulate a reliability-based design optimization of reinforced concrete bridges by two conflicting objectives—maximizing reliability of the structure and minimizing the construction material costs. This problem usually consists of several different variables, such as geometrical, mechanical, and environmental as well as deterministic and probabilistic constraints. The authors consider the effect of corrosion in this problem. To solve this constrained biobjective problem, they coupled the first-order reliability method with nondominated sorted genetic algorithm-II. A bridge with interior T-beams is studied and optimized in the study, and three levels of concrete materials are investigated here. The nondominated solution set is presented, and a decision on optimum design can be made based on this set. The authors found this procedure computationally efficient and reliable. Based on the results, they also made some recommendations about the type of the concrete that should be used to meet both objectives simultaneously.

Talatahari et al. (2018) proposed a new metaheuristic method for a risk-based optimization of dams via the hybridizing charged system search (CSS) and big bang–big crunch (BBBC). In this problem, the objective is defined as minimizing the expected costs (or maximizing the expected profit) of an arch dam that can be achieved by finding the optimum level of safety. Here, the total cost includes the cost of concrete volume of the dam body, and the life cycle, maintenance, and reconstruction costs. The considered case study is the well-known Morrow Point dam with full reservoir. To formulate this case study, interaction between dam, reservoir and foundation, hydrodynamic effects resulting from seismic loading, and the flexibility and damping effects of the foundation rock are considered. The results of the hybrid method on risk-based optimization of this arch dam are compared with basic CSS and BBBC, as well as some other methods proposed in the literature. The results show that the proposed algorithm can find a better solution (costs) in comparison with the other optimization methods.

In the closing paper of this special issue, Nasr and Saad (2016) coupled genetic algorithm (GA) and Kalman filter method to optimize the sensor placement for the purpose of structural identification. Therefore, the objective function is minimization of the error between the actual measurement data and their predicted values. A case study of a 10-story shear building subjected to El-Centro earthquake excitation is considered to evaluate the performance of the proposed hybrid method. The results indicate that the lowest and highest floors of the building are among the best sensor locations in all scenarios.

The papers in this special issue bring new findings and perspectives of metaheuristics applications in reliability and risk analysis. We appreciate Professor Bilal M. Ayyub, editor in chief of ASCE-ASME Journal of Risk and Uncertainty in Engineering Systems for his continuous support. We also thank other editors and managers from ASCE for their assistance during the review and publication process of the manuscripts. We are also grateful to the respected reviewers for providing constructive reviews and valuable comments for the submitted papers. Finally, we express our sincere thanks to all authors who contributed to our special issue.