Use of Evolutionary Computation to Improve Rock Slope Back Analysis

• Main Authors: An-Jui Li, Abdoulie Fatty, I-Tung Yang
• Format: Article
• Language: English
• Published: MDPI AG 2020-03-01
• Series: Applied Sciences
• Subjects: slope parameters; back analysis; genetic algorithm; particle swarm optimization
• Online Access: https://www.mdpi.com/2076-3417/10/6/2012

Generally, in geotechnical engineering, back analyses are used to investigate uncertain parameters. Back analyses can be undertaken by considering known conditions, such as failure surfaces, displacements, and structural performances. Many geotechnical problems have irregular solution domains, with the objective function being non-convex, and may not be continuous functions. As such, a complex non-linear optimization function is typically required for most geotechnical problems to attain a better understanding of these uncertainties. Therefore, particle swarm optimization (<i>PSO</i>) and a genetic algorithm (<i>GA</i>) are utilized in this study to facilitate in back analyses mainly based on upper bound finite element limit analysis method. These approaches are part of evolutionary computation, which is appropriate for solving non-linear global optimization problems. By using these techniques with upper-bound finite element limit analysis (<i>UB-FELA</i>), two case studies showed that the results obtained are reasonable and reliable while maintaining a balance between computational time and accuracy.