Markov processes for tailings dams – Part 2: Applications
Australian Geomechanics Journal · 01.09.2026 · 61 (3) · 147-159
DOI: 10.56295/agj6139 ↗Kurz gefasst
Fachartikel in der Zeitschrift Australian Geomechanics Journal, online erschienen am 01.09.2026. Verfasst von Franco Oboni.
Abstract (Originalsprache)
This paper describes various evaluations carried out using an unorthodox Markovian process on single tailings storage facilities dams (TSFs) and a small portfolio of those structures. The Markovian model was described in the first part of this discussion (Markov Processes For Tailings Dams. Part 1: The Model, Oboni, 2026 in this Journal) and the application described in this paper uses a proven quantitative risk assessment (QRA) procedure (Oboni, Oboni, 2025) for the prior estimates of the probabilities of failure (PoFs), but any other validated, equally proven QRA method delivering the initial estimates of the various PoFs needed for the Markovian model could be substituted. Thus, the Markovian model proposed in this paper is agnostic to the QRA procedure, under the condition that it can deliver the required data. As described in the companion paper, the driving idea is to offer users a way to optimize costly QRA updates frequency and to guide the selection of adequate TARP types along the life of the considered dam, while showing the model capabilities and offer insight on possible practical uses. The five real-life dams of the considered portfolio have initial probabilities of failure covering almost three orders of magnitude, from the low bound of the world-wide TSF benchmark up to what is considered pre-catastrophic conditions. The comparison of the Markovian results with the a priori QRA is very encouraging and in the rare case of apparent discrepancies, intuitive convincing explanations delivered. The Markovian approach shows that better dams require longer steps sequences to attain the so-called Markovian steady state, i.e. a state where the probabilities of each individual failure mode and causalities remain constant over time. The global PoF of the dams varies during the path to steady state, as the probabilities of the hazardous conditions vary at each step. As none of the considered dams exhibits an absorbing state, i.e. a certainty of a single failure mode (FM)/causality within the studied path to steady state, several probable failure modes (PFM) remain jointly possible through the “life” of the dam including at steady state. However, the fundamental difference between dams is that better-quality ones sport a larger number of dominant PFMs at steady state whereas lesser quality dams concentrate toward one or two PFMs. That means that better dams need a combination of multiple driving conditions to collapse as their resilience is higher whereas “weaker” dams just require a small trigger to collapse following one of the PFMs. A second phase of the development led to analyse the decay of a dam through a progressive relaxation of the standard of care. Seven levels of reduction were simulated varying the causalities, but not the initial QRA. This is a very theoretical case, as it is obvious that past a certain level of decay the QRA should be updated. It is inferred that the Markovian model would guide users to optimize updating frequency. On an initially good quality dam some cracks, leaks etc. are not necessarily fatal conditions as shown by numerous real-life cases of damaged dams surviving for years, sometimes even without repairs (not a recommended practice, of course). However, as expected, the number of steps to steady state reduces significantly as the standard of care decays, which is an indicator of the need to re-evaluate specific TARPs. Also, the initial post QRA observations significantly increase the probability of failure, shift the PoFs to a higher level, almost half order of magnitude. As the decay continues the PoF evolves during the considered life (up to the steady state).
Zitieren
APA
Oboni, F. (2026). Markov processes for tailings dams – Part 2: Applications. Australian Geomechanics Journal, 61(3), 147-159. https://doi.org/10.56295/agj6139
Vancouver
Oboni F. Markov processes for tailings dams – Part 2: Applications. Australian Geomechanics Journal. 2026;61(3):147-159. doi:10.56295/agj6139
BibTeX
@article{Oboni2026,
author = {Oboni, Franco},
title = {Markov processes for tailings dams – Part 2: Applications},
journal = {Australian Geomechanics Journal},
year = {2026},
volume = {61},
number = {3},
pages = {147--159},
doi = {10.56295/agj6139}
}
EUVIMED