Dr Marcelo Llano

Breaching the gap between geotechnical and

hydraulic engineering to improve tailings dam
breach studies through large deformation modelling
27 October 2020
Outline

Motivation

Flowability of tailings (Mobility)

Material Point Method (MPM) – large deformation modelling

The effective stress concept, strength and rheology

Case Studies

Summary and final thoughts

2
Motivation

CDA (2021) Technical Bulletin: Tailings Dam Breach Analysis

5.2 Tailings characterisation: “Flow liquefaction results in large deformations and directly impacts the volume of
tailings that can be released during a breach”.

5.3 Rheological behaviour: “The measurement of yield stress and viscosity are often conducted for tailings with a
volumetric solid concentration less than 45-55%. Yield strength and viscosity increase exponentially and become
very sensitive to the solids concentration when the solids concentration is close to a threshold, illustrating the
influence of water content on the rheological behaviour”….

Liquid limit, traditionally ranges between 1.7kPa to 2.3kPa. Soils at the plastic limit have around 100 times the
undrained shear strength measured at the liquid limit.

“Although the concepts of viscosity and yield stress are commonly used in conventional fluid mechanics to
characterise the shear properties of a fluid, they can be inadequate to describe some non-Newtonian flows with
high solid concentrations in which the particle-to-particle contacts become dominant and control the flow”.

“… at higher solids concentrations the mechanisms of compression and dilation may affect the tailings flows
and should be taken into account.”… “integration of the principles of advanced soil mechanics in combination with
fluid mechanics and rheology can be used”.

3
Mobility of flow events (modified from CDA, 2021)

Mud flood 5%
9%
18% 14% Percentages on
28% the plot added by
33%
the author are
43% 39% equivalent
54% gravimetric water
content assuming
117% Landslide 100% saturation.
150% Note that lines in
186% the plot for lower
water contents
Mudflow might be different
depending on the
shrinkage limit.
Water flood

4
Mobility of tailings in the context of modelling techniques

Hydraulic Modelling Geotechnical Modelling

Rheology Site Characterisation

Could be triggered by erosion/overtopping

Decreasing water content
Water flood Mud flood Mudflow Landslide Slope stability
Increasing flowability (deformability-large strains)
Run-out solving numerical capability
Inundation modelling Stress- Limit
(HEC-RAS, TUFLOW, FLOW-3D, FLOW-2D) Strain/Small Equilibrium/No
deformations deformation
Flow slide, debris flow, slurry flow, avalanche (FLAC, Plaxis) (Geostudio,
(MADFLOW) Slide, Plaxis LE)

Large deformation
modelling, MPM
(Anura3D)

Erosion modelling (EMBREA-Mud)

The horizontal size and relative position of white rectangles vs the red arrow (flowability) represents the approximate
5
numerical capability of techniques to model run-out events
Material Point Method (MPM) Background - Marcelo

▪ 2012, MSc in Geotechnics – University of Brasilia, Brazil

Thesis: Material Point Method Applications to Geotechnical Engineering
Problems

▪ 2016, PhD in Geotechnics – University of Brasilia, Brazil

(with an externship at the GEC in the University of Queensland)
Thesis: Experimental and Numerical Study of Geotechnical Engineering
Problems Using the Material Point Method

Around 30 scientific publications including two papers in Géotechnique and

>100 citations

6
MPM Background - Introduction

BACKGROUND: Finite differences (used in FLAC) was introduced by Taylor in 1715. FEM-
Finite Elements (used in Plaxis) was developed in the 1940s (Hrennikoff, 1941, Courant,
1942). Both tackle complex engineering problems for which there is no analytical solution
based on continuum mechanics approach.

PROBLEM: The traditional formulations are not well suited to tackle large deformation
problems.

ALTERNATIVE APPROACH: Meshfree or Meshless Methods (MPM) may solve the problem of
simulating large deformations.

FEM, modified from Vermeer et al., (2008)

MPM, Llano-Serna (2012)
7
MPM Background - Relevance

Large deformation phenomena are

very common in Geotechnical
Engineering

▪ STP, CPT, DMT, Fall cone test…

▪ Landslides, dam failure
▪ Piling

Classical numerical techniques fail to

deal with large deformations

8
MPM Background – The idea

Background Grid or Computational Mesh (Cells and

Vertices) used only for calculation purposes. Simple
and structured in general, exceeds the real
problem domain.

Material “Points” (with finite dimensions) represent

the actual discretization. MP points are free to
move across the background grid.

The MPs carry all relevant information as they move.

Henry Tan’s blog

(iMechanica.org)

9
MPM development with emphasis on tailings dams- timeline

▪ 1964 - Particle in Cell (PIC) Method formulation (fluids) - Harlow

▪ 1994 – Material Point Method (PIC extension for solids) - Sulsky
▪ 1999 – First application in Geotechnical Engineering- Zhou et al.
(Embankment settlement and geotextiles)
▪ 2010 – First application in slope stability and tailings dam failure
(Aznalcollar) - Zabala
▪ 2017 – Rankine lecture on triggering and motion of landslides,
Prof. Eduardo Alonso –UPC, Spain (Emphasis on MPM)
▪ 2017-2018 – Conceptual models of various run-out examples –
The Author, while working at the GEC, UQ
▪ 2019 – MPM modelling of Brumadinho failure (Results included
in the expert panel report) – The Author
▪ 2021 – MPM modelling of Cadia – Ian Pierce, Alba Yerro
(Virginia Tech)
▪ 2021 - Contributions to the development of Tailings Dam
Breach Assessments for various projects in Australia – The 1
0
Author
MPM examples – Geotechnical models, conceptual capabilities

Active/passive earth pressure

5.5m tall, 11.0m wide

Wieckowski (2004)

Wall movement

1.0m tall wall

Beuth et al. (2011)

11
MPM examples – Geotechnical models, conceptual capabilities

Slope instability / erodible barriers

20.0m tall slope

Andersen (2010)

Shin et al. (2011)

1
2
Software available

Commercially available software do not exist to date. Several research packages exist.

Anura3D - http://www.mpm-dredge.eu/ (Recommended)

CB-Geo - https://www.cb-geo.com/research/mpm/
Uintah - http://uintah.utah.edu/
NairnMPM - http://people.oregonstate.edu/~nairnj/

1
3
Partial summary 1

Large deformations in the context of geotechnical engineering

Large deformations are a common occurrence in the practice of geotechnical engineering. Unfortunately, tools to
model and further understand some of the mechanisms when large deformation take place are limited (but exist).

In slope stability/dam engineering, engineers have trained and learnt to perform calculations to “avoid failure”,
which in this field is understood as deformation (large or small) of a given slope. Despite significant progress in
computational capacity, limit equilibrium techniques with no deformation involved are the most common
approach.

Limit equilibrium will be challenged more often (but unlikely to disappear) as earthen structures grow due to
commercial demand, international standards and guidelines are becoming more refined. An obvious step forward
is a wider adoption of small deformation approaches for which the availability of commercial packages is vast.

1
4
The effective stress concept, strength and rheology

Meaning of effective stress (1925)

Trenchlesspedia.com

“Effective stress can be defined as the stress that keeps particles together. In soil, it is the combined effect of pore
water pressure and total stress that keeps it together”.

Wikipedia.org

“Can be defined as the stress, depending on the applied tension and pore pressure, which controls the strain or
strength behaviour of soil and rock (or a generic porous body)”.

𝜎𝑣′ = 𝜎𝑣 − 𝑢 Most common (Easy but does not take

horizontal stress into account)
′ ′
𝜎1 + 2𝜎3
𝑝′ = Cambridge notation
3
′ ′
𝜎1 + 𝜎3
Environment.uwe.ac.uk 𝑠′ = MIT notation, some authors use p’ to refer to
2
s’ which makes it confusing
1
5
The effective stress concept, strength and rheology

Rheological chracterisation
Conventionally includes shear yield stress vs solids mass fraction, and

Shear stress vs shear rate.

Usually expressed in terms of total stresses. Is there an influence of effective stresses (i.e. different rheological
behaviour varying with depth)?

1
6
The effective stress concept, strength and rheology

“In the first type of aggregate,

interfiber friction is not very
strong and nanofibers could
reorganise as the aggregate is
tilted by the shear flow, given
rise to a “paramagnetic”
aggregate with changeable
magnetisation. In the second
type of aggregate, interfiber
friction is so strong that
nanofiber orientations are fixed
and, thus, the magnitude and
directions (along the aggregate
main axis) of the magnetisation
of the aggregates are fixed too.”

1
7
 The effective stress concept, strength and rheology
450 CIU 150 kPa
400 Tailings tested under conventional triaxial undrained
CIU 75 kPa
Deviator Stress (kPa)

350 compression for different effective stresses can also show N-like
300 CIU 300 kPa
stress-strain curves. There is evidence of different responses
250 CIU 600kPa
200 with varying confinement stress.
Max pore pressure
150
100 Max dev. Stress One might have aggregate type one or two (In Zubarev’s
50 Max stress ratio (s1/s3) framework) dominating the response of tailings based on
0
Instability line material characteristics. See Cadia and Brumadinho examples
0 20 40
Axial Strain (%)
below.

CIU triaxial at 50kPa, from Appendix E, loose CIU triaxial at 2000kPa, test certificate from Appendix E, loose
tailings properties. Cadia, Expert Review Panel. tailings properties. Feijao (Brumadinho), Expert Panel Report

1
8
The effective stress concept, strength and rheology

Model recommended to simulate strain softening

Brittleness may be simulated by introducing a strain-dependent parameters that controls the transition between
peak frictional angle or peak cohesion to residual frictional angle or cohesion. The softening behaviour is
accounted for by reducing the strength parameters with the accumulated plastic deviatoric strain.

Evolution of the shear

stress with shear strain for
several brittle responses

1
9
Partial summary 2

Constitutive behaviour of tailings during a breach event

The effective stress concept has been known to influence the mechanical response of granular arrangements
during loading. Specifically, a MPM model can simulate the “initiation” of the failure and the transition between a
static (solid) state trough a deformation process that can result in liquefied tailings.

Limited rheological testing under confining stresses exist, available results suggest that confining pressures have
an influence in the rheological response of materials, specially during the solid to fluid transition. See also Chen et
al., (2008) and Jia et al., (2021).

A detailed study about the role of different constitutive models in MPM simulations of granular column collapses is
presented by Fern and Soga (2016).

2
0
Case studies

Influence of deposit length and undrained shear strength on run -

out distance (The Author, software:NairnMPM)

Rockfill

2
1
Case studies

Pore water pressures

Aznalcollar (Los Frailes) – Zabala (2010)
prevailing in the
foundation clay at the time
of the failure, with strength
behaviour of brittle clay
with low permeability was
attributed as the main
cause of the event.

2
2
Case studies

Brumadinho – The Author (2019)

Using Anura3D

(adm)

(adm)

2
3
Case studies

Cadia (Pierce 2021)

Using Anura3D

2
4
Case studies

Cadia (Pierce 2021)

Using Anura3D

Key drivers of the observed behaviour were residual properties of the foundation, rate at which tailings
liquified, the initial extent of liquefied tailings, the presence of the excavation at the toe of the dam. The
softening characteristics of the foundation played a large role allowing for the failure to occur, without the
progressive softening of the foundation the dam does not move without the liquefaction of tailings.

2
5
Summary and final thoughts

We have covered relevant aspects of the numerical modelling of tailings dam breaches from a soil's mechanics
perspective. The review covered mobility of tailings, the role of large deformation modelling in TDBA, the relevance
of the effective stress concept in tailings dam breaches and several case studies.

MPM is a growing technique. Although no commercial software is yet available, there is a vast technical resource
on the web.

Despite all the advantages mentioned there are still limitations. Applicability is still limited to 2D problems, we
offset this limitation by coupling the MPM outputs with hydraulic engineering software using the advantages
offered by both techniques, which results in a better product for the clients. The publication of the recent GISTM
with emphasis on ALARP concepts and the new CDA bulletin will likely accelerate the wider implementation of this
(and other similar) type of approach in the industry.

2
6
Thanks,

[email protected]

2
7
Phase transformation and work required during dilation

The input energy per unit volume of soils was calculated to investigate its effect on the mechanical
behaviour of sand samples prepared at loose and dense states using air pluvation.

Phase transformation occurs at axial strains

of less than 5% and can consume a minimum
of 95% of the energy required to shear the
soil to axial strains of around 20%

2
8
Shahnazari et al. (2016)