Augmented-Space Hamiltonian Monte Carlo Subset Simulation Method for Estimating the Failure Probability Function
Abstract In structural reliability analysis, the estimation of the failure probability function (FPF) is often computationally prohibitive and numerically unstable. To address this issue, this paper proposes an augmented-space integral–based Hamiltonian Monte Carlo subset simulation (ASI-HMC-SS) method for efficient FPF estimation. Based on the Bayesian formulation, the distribution parameters are treated as auxiliary random variables with instrumental prior distributions. Accordingly, the FPF is formulated based on the failure probability in the augmented space and the parameter posterior density conditioned on failure, which supports efficient global FPF estimation through a single-run calculation. In the conditional sampling stage of subset simulation, a Hamiltonian Monte Carlo (HMC) scheme with an adaptive trajectory-length adjustment strategy is incorporated to significantly improve sampling efficiency and numerical stability in failure domains. Several static and dynamic engineering examples are presented to validate the proposed method. The results suggest that, in terms of efficiency and accuracy, the proposed method outperforms conventional augmented-space subset simulation methods. This superiority highlights its significant potential for engineering applications.
Authors
- Xiukai Yuan (ORCID: https://orcid.org/0000-0003-0332-2614)
- Yiwei Dong
- Chaoxin Lu (ORCID: https://orcid.org/0009-0002-6850-8102)
- Yuhan Wang
Institutions
- Xiamen University (CN)
Publication Details
- Journal
- ASCE-ASME Journal of Risk and Uncertainty in Engineering Systems Part A Civil Engineering
- Published
- 2026-11-14
- DOI
- https://doi.org/10.1061/ajrua6.rueng-2075
- Primary Topic
- Probabilistic and Robust Engineering Design
- Type
- article
- Field-Weighted Citation Impact
- 0.00