Multi‐objective optimization of mechanically stabilized earth retaining wall using evolutionary algorithms

• Published in: International journal for numerical and analytical methods in geomechanics Vol. 46; no. 8; pp. 1433 - 1465
• Main Authors: Kashani, Ali R., Camp, Charles V., Azizi, Koorosh, Rostamian, Mehdi
• Format: Journal Article
• Language: English
• Published: Bognor Regis Wiley Subscription Services, Inc 01.06.2022
• Subjects: Algorithms; Coefficient of variation; Confidence intervals; Design; Earthquakes; Evolutionary algorithms; Genetic algorithms; mechanically stabilized wall; Minimum cost; multi‐objective optimization; Optimality criteria; Optimization algorithms; Parameter sensitivity; Particle swarm optimization; retaining wall; Retaining walls; Safety factors; Seismic activity; Sensitivity analysis; Sorting algorithms; Statistical analysis; Structural engineering
• ISSN: 0363-9061; 1096-9853
• DOI: 10.1002/nag.3352

This paper uses evolutionary optimization algorithms to study the multi‐objective optimization of mechanically stabilized earth (MSE) retaining walls. Five multi‐objective optimization algorithms, including the non‐dominated sorting genetic algorithm II (NSGA‐II), strength Pareto evolutionary algorithm II (SPEA‐II), multi‐objective particle swarm optimization (MOPSO), multi‐objective multi‐verse optimization (MVO), and Pareto envelope‐based selection algorithm II (PESA‐II), are applied to the design procedure. MSE wall design requires two major requirements: external stability and internal stability. In this study, the optimality criterion is to minimize cost and its trade‐off with the factor of safety (FOS). To this end, two objectives are defined: (1) minimum cost, (2) maximum FOS. Three different strategies are considered for reinforcement combinations in the numerical simulations. Moreover, a sensitivity analysis was conducted on the variation of significant parameters, including backfill slope, wall height, horizontal earthquake coefficient, and surcharge load. The efficiency of the utilized algorithms was assessed through three well‐known coverage set measures, diversity, and hypervolume. These measures were further examined using basic statistical measures (i.e., min, max, standard deviation) and the Friedman test with a 95% confidence level.