Eine Plattform für die Wissenschaft: Bauingenieurwesen, Architektur und Urbanistik
Eine Plattform für die Wissenschaft: Bauingenieurwesen, Architektur und Urbanistik
Reliability-Based Geotechnical Design Method Using the Gaussian Process Regression–Based Differential Evolution Algorithm
A fundamental challenge in reliability-based geotechnical design is optimizing cost-effectiveness while adhering to a predefined failure probability target. Traditionally, failure probability is assessed via Monte Carlo simulation (MCS), which, despite its accuracy, is often prohibitively time-consuming and expensive. This scenario frames reliability-based geotechnical design as a constrained optimization problem (COP) characterized by a low-cost objective and high-cost constraints. Recent advancements have seen the application of Gaussian process regression (GPR) enhanced evolutionary algorithms (EAs) for managing COPs where both objectives and constraints incur significant expenses. However, methodologies adept at handling COPs with inexpensive objectives yet costly constraints remain underexplored. This paper introduces a novel approach utilizing a GPR–based differential evolution (DE) algorithm designed specifically for COPs with this cost disparity. Here, GPR serves as a surrogate model to estimate the actual performance derived from MCS assessments. The innovative use of expected improvement (EI) as a selection criterion for potential solutions is a key feature of this method. EI quantitatively evaluates each candidate’s potential to enhance economic efficiency and safety reliability, effectively converting the COP into a single-objective optimization problem (SOOP). We demonstrate the efficacy of our proposed GPR–based DE algorithm through a case study of the Sau Mau Ping rock slope in Hong Kong, highlighting the method’s ability to achieve superior accuracy and substantial computational savings.
Reliability-Based Geotechnical Design Method Using the Gaussian Process Regression–Based Differential Evolution Algorithm
A fundamental challenge in reliability-based geotechnical design is optimizing cost-effectiveness while adhering to a predefined failure probability target. Traditionally, failure probability is assessed via Monte Carlo simulation (MCS), which, despite its accuracy, is often prohibitively time-consuming and expensive. This scenario frames reliability-based geotechnical design as a constrained optimization problem (COP) characterized by a low-cost objective and high-cost constraints. Recent advancements have seen the application of Gaussian process regression (GPR) enhanced evolutionary algorithms (EAs) for managing COPs where both objectives and constraints incur significant expenses. However, methodologies adept at handling COPs with inexpensive objectives yet costly constraints remain underexplored. This paper introduces a novel approach utilizing a GPR–based differential evolution (DE) algorithm designed specifically for COPs with this cost disparity. Here, GPR serves as a surrogate model to estimate the actual performance derived from MCS assessments. The innovative use of expected improvement (EI) as a selection criterion for potential solutions is a key feature of this method. EI quantitatively evaluates each candidate’s potential to enhance economic efficiency and safety reliability, effectively converting the COP into a single-objective optimization problem (SOOP). We demonstrate the efficacy of our proposed GPR–based DE algorithm through a case study of the Sau Mau Ping rock slope in Hong Kong, highlighting the method’s ability to achieve superior accuracy and substantial computational savings.
Reliability-Based Geotechnical Design Method Using the Gaussian Process Regression–Based Differential Evolution Algorithm
Kai Wen (Autor:in) / Zemin Kuang (Autor:in) / Wei Zeng (Autor:in) / Sanyou Zeng (Autor:in) / Yue Qiu (Autor:in)
2024
Aufsatz (Zeitschrift)
Elektronische Ressource
Unbekannt
Metadata by DOAJ is licensed under CC BY-SA 1.0
Reliability-Based Geotechnical Design Code Development
| ASCE | 2014
Geotechnical Uncertainties and Reliability-Based Design
| ASCE | 2013