An active learning method using adaptive Bayesian compressive sensing and Monte Carlo simulation (ABCS-MCS) for slope reliability analysis considering soil spatial variability

Research output: Chapters, Conference Papers, Creative and Literary WorksRGC 32 - Refereed conference paper (with host publication)peer-review

View graph of relations

Author(s)

Detail(s)

Original languageEnglish
Title of host publicationProceedings of the 9th International Symposium on Reliability Engineering and Risk Management (ISRERM 2024)
PublisherTongji University Press
Publication statusPresented - 18 Oct 2024

Conference

Title9th International Symposium on Reliability Engineering and Risk Management (ISRERM)
Location
PlaceChina
CityHefei
Period18 - 21 October 2024

Abstract

Reliability analysis of slope systems considering the soil spatial variability is a challenging problem, because such slope systems often contain a large number of input variables generated by random field discretization and exist highly nonlinear system responses. For the slope systems with computationally time-consuming deterministic model, directly using Monte Carlo simulation (MCS) requires a huge computational cost, especially for the small failure probability event. To tackle this difficulty, this study develops an innovative active learning method using adaptive Bayesian compressive sensing (ABCS) and MCS, called ABCS-MCS, for reliability analysis of highly nonlinear slope systems considering soil spatial variability. In ABCS-MCS, Bayesian compressive sensing (BCS) is adopted to construct a response surface which can self-evaluate the uncertainty of response predictions, and it combines a learning function to adaptively choose additional sampling points to enrich the training set and improve the reliability analysis accuracy until a target accuracy is achieved. Moreover, ABCS-MCS is directly applicable to highly nonlinear or non-stationary slope systems because BCS is non-parametric. Performance of ABCS-MCS is investigated using a two-layered slope reliability analysis problem with consideration of spatial variability in soil properties, and results show that ABCS-MCS performs well in terms of accuracy and efficiency of reliability analysis.

Research Area(s)

  • Active learning, reliability analysis, soil spatial variability, Bayesian compressive sensing, non-parametric method

Citation Format(s)

An active learning method using adaptive Bayesian compressive sensing and Monte Carlo simulation (ABCS-MCS) for slope reliability analysis considering soil spatial variability. / LI, Peiping; WANG, Yu.
Proceedings of the 9th International Symposium on Reliability Engineering and Risk Management (ISRERM 2024). Tongji University Press, 2024.

Research output: Chapters, Conference Papers, Creative and Literary WorksRGC 32 - Refereed conference paper (with host publication)peer-review