Efficient reliability analysis of slopes integrating the random field method and a Gaussian process regression‐based surrogate model

Efficient evaluation of slope stability is a frontier in geo‐disaster prevention fields. While many slope stability evaluation methods, ranging from deterministic to probabilistic, have been proposed, reliability methods are particularly advantageous as they can account for uncertainties during slop...

Full description

Saved in:
Bibliographic Details
Published in:International Journal for Numerical and Analytical Methods in Geomechanics Vol. 45; no. 4; pp. 478 - 501
Main Authors: Zhu, Bin, Hiraishi, Tetsuya, Pei, Huafu, Yang, Qing
Format: Journal Article
Language:English
Published: Bognor Regis Wiley 01.03.2021
Wiley Subscription Services, Inc
Subjects:
Defects
Emergency preparedness
Evaluation
Feasibility studies
Fields (mathematics)
Gaussian process
Gaussian process regression
Karhunen‐Loève expansion
Monte Carlo simulation
Normal distribution
Procedures
random field
Regression analysis
Regression models
Reliability
Reliability analysis
response surface method
Response surface methodology
slope reliability
Slope stability
Soil layers
Soil stability
Soil strength
Spatial variations
Stability analysis
Statistical analysis
Statistical methods
surrogate model
ISSN:0363-9061, 1096-9853
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Efficient evaluation of slope stability is a frontier in geo‐disaster prevention fields. While many slope stability evaluation methods, ranging from deterministic to probabilistic, have been proposed, reliability methods are particularly advantageous as they can account for uncertainties during slope stability evaluation. To address the problem of the low efficiency of direct Monte Carlo simulations and overcome the defects of the traditional response surface method, in this work, a novel probabilistic procedure that integrates a Gaussian process regression‐based surrogate model and the random limit equilibrium method for slope stability evaluation while accounting for the spatial variability of soil strength is proposed. The novel Gaussian process regression‐based surrogate model is efficiently used in the Monte Carlo simulation to reduce the number of calls for direct stability analysis of a slope with spatially varied soil strength. To verify the accuracy and efficiency of the proposed procedure, it was applied to three case studies: the first one is a slope in saturated clay under undrained conditions; the second one is a slope for which the friction angle and cohesion values are cross‐correlated; and the third one is a real slope with multiple soil layers. The results obtained from the comparisons with other methods confirmed the precision and feasibility of the proposed procedure.
Bibliography:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:0363-9061
1096-9853
DOI:10.1002/nag.3169