Validation of computational liquefaction for tailings: Tar Island slump

Abstract

Finite-element analyses using critical state theory proved necessary to understand the development of static liquefaction during three recent large tailing dam failures at Fundao (in Brazil), Cadia (in Australia) and Brumadinho (in Brazil). However, the complexity of these events prevents these analyses being viewed as a complete validation of the methodology. Here the authors evaluate a far simpler case of static liquefaction: the 1974 Tar Island slump (in Canada). This upstream slump involved a rapid drop of 5 m during construction of a 12.5 m high upstream raise over loose tailings. While not a dam stability issue, the event has the attraction for validation of being load-induced, with simple geometry, and with known material properties and in situ state. The computed liquefaction develops from a prior drained condition before propagating rapidly undrained – there are similarities to the video record at Brumadinho (an animation is provided as online supplementary material to illustrate this). A range of scenarios are explored, with the base case of taking reported conditions at face value giving deformations close to those measured. An important aspect was using elastic shear moduli determined by geophysical methods. The analyses were carried out with commercial software (Plaxis) and used critical state theory with largely familiar soil properties measured by standard methods.

Discussion of “Evaluation of flow liquefaction and liquefied strength using the cone penetration test: an update”

“…The better agreement seen for the torsional simple shear testsplane strain tests, with rotation of principal stress angle during shearingis consistent with the well-documented cross-anisotropic nature of sands (Sivathayalan and Vaid 2002;Reid 2020), clays (Seah 1990), and tailings (Reid et al 2022) along with the lower values of instability stress ratio and critical friction ratio expected in plane strain conditions (Wanatowski and Chu 2007;Jefferies and Shuttle 2011;Shuttle et al 2022). Indeed, it has been observed in the past that torsional simple shear tests generally provide better agreement with back-analysis-inferred liquefied strengths (e.g., Yoshimine et al 1999) than triaxial compression.…”

Discussion of “Improved cone penetration test predictions of the state parameter of loose mine tailings”

“…E4-5 of (Morgenstern et al 2019). In the case of Tar Island (Shuttle et al 2021), there was a close match between the measured and computed displacement during the liquefaction slump--a displacement that is sensitive to assessed state (which was based on CPTwidget). More generally, there is a limited range of states at both Cadia and Tar Island that allow stable construction of what was built with the subsequent evolution of liquefaction-driven slumps--the in situ state parameter is quite constrained for these two case histories.…”