Probabilistic analysis of slopes designed by the partial material factor approach

TY - JOUR

T1 - Probabilistic analysis of slopes designed by the partial material factor approach

AU - Li, Yuting

AU - Chen, Xiaobin

AU - He, Pengpeng

AU - Fenton, Gordon A.

AU - Griffiths, D. V.

N1 - Copyright: © 2025 Elsevier Ltd

PY - 2025/12

Y1 - 2025/12

N2 - Traditional slope stability design typically relies on a single global factor of safety to account for various sources of uncertainties. However, the same global factor of safety may lead to different levels of reliability for different site conditions and slope configurations. This study probabilistically calibrated the partial material factors used for slope stability design within the partial factor method design framework using the Random Finite Element Method (RFEM). Drained and undrained soil conditions were separately considered. The slope height was first designed using the partial factor method, followed by RFEM to estimate the failure probability of the design slope. The results show that a worst-case correlation length exists for all of the cases, suggesting that the worst-case correlation length can be used for conservative design when insufficient data is available for accurate estimates of the soil correlation length. In addition, the sampling location significantly influences the estimated failure probability and the required partial material factor. For drained soil conditions, lower cross correlation coefficients between c ′ and ϕ ′ result in lower estimated failure probabilities, indicating that an assumption of independence between c ′ and ϕ ′ is conservative. Overall, this study provides probabilistic insights into slope design practices and can guide the calibration of partial material factors for more reliable slope design.

AB - Traditional slope stability design typically relies on a single global factor of safety to account for various sources of uncertainties. However, the same global factor of safety may lead to different levels of reliability for different site conditions and slope configurations. This study probabilistically calibrated the partial material factors used for slope stability design within the partial factor method design framework using the Random Finite Element Method (RFEM). Drained and undrained soil conditions were separately considered. The slope height was first designed using the partial factor method, followed by RFEM to estimate the failure probability of the design slope. The results show that a worst-case correlation length exists for all of the cases, suggesting that the worst-case correlation length can be used for conservative design when insufficient data is available for accurate estimates of the soil correlation length. In addition, the sampling location significantly influences the estimated failure probability and the required partial material factor. For drained soil conditions, lower cross correlation coefficients between c ′ and ϕ ′ result in lower estimated failure probabilities, indicating that an assumption of independence between c ′ and ϕ ′ is conservative. Overall, this study provides probabilistic insights into slope design practices and can guide the calibration of partial material factors for more reliable slope design.

KW - slope stability

KW - Partial factor design

KW - Random finite element method

KW - Failure probability

KW - Cross correlation

KW - Slope stability

UR - https://www.scopus.com/pages/publications/105013479502

U2 - 10.1016/j.compgeo.2025.107572

DO - 10.1016/j.compgeo.2025.107572

M3 - Article

SN - 0266-352X

VL - 188

JO - Computers and Geotechnics

JF - Computers and Geotechnics

M1 - 107572

ER -