Determination of Consolidation Properties

Proceedings Tailings and Mine Waste 2011
Vancouver, BC, November 6 to 9, 2011
Determination of Consolidation Properties, Selection of

Computational Methods, and Estimation of Potential
Error in Mine Tailings Settlement Calculations
David Geier
Golder Associates Inc., USA
Gordan Gjerapic, Ph.D., PE

Golder Associates Inc., USA
Kimberly Morrison, PE, RG

AMEC Environmental & Infrastructure, USA
Abstract
Accurate estimates of tailings density and settlement are critical for successful design, operation, and closure of
tailings storage facilities (TSFs). Limitations of the computational methods employed to determine these
quantities are often overlooked, leading to inaccurate results and potentially inadequate engineering designs.
This paper provides a brief discussion of the methods that are commonly used to model tailings consolidation,
discusses relative accuracy of these methods, and outlines common pitfalls encountered in their use.
Engineering estimates of the commonly encountered errors are presented for selected methods. Applicability of
various methods and specific recommendations are discussed with regard to TSF geometries, tailings properties,
filling rates, and the calculation accuracy requirements. The present study compares traditional small strain and
large-strain consolidation analyses, and evaluates differences between one-dimensional and three-dimensional
calculation approaches. Also, a comparison between estimated (calculated) and actual tailings densities (based
on mill production data and bathymetric surveys) is provided for an existing operational tailings facility.
Comparison between Small Strain and Large Strain Methods

Small Strain Method
A traditional small strain method for calculating settlements utilizes the approach that was originally
developed by Terzaghi. This method often assumes that the compressibility is constant over the range
of stresses used in the analysis. The amount of compression is typically determined based on the
change in void ratio:
(1)
where
e
= void ratio
Cc
= compression index of the soil
= final vertical effective stress
= initial vertical effective stress

Cc is the slope of the e vs. log curve for normally consolidated soils. A typical e vs. log curve for
tailings material is shown in Figure 1. Generally, the slope between two arbitrary points on the e-log
curve is not a constant value, which needs to be accounted for in the consolidation analysis. Typically,
tailings are deposited at relatively high void ratios, corresponding to solid contents of 40% or less and
effective stress magnitude close to zero. Over time, the tailings material will consolidate under its own
weight with partial consolidation of tailings typically achieved while the height of the deposit is still
increasing. Tailings heights in excess of 100 m and ultimate effective stresses at the base of the
impoundment of over 1,500 kPa are not uncommon.
Typically, small strain consolidation analyses assume that the coefficient of consolidation (material
parameter required to calculate the time rate of consolidation) remains constant. The coefficient of
consolidation depends on the compressibility, void ratio, and permeability of a soil:
(2)
where
cv
= coefficient of consolidation
= permeability
= unit weight of water
mv
= coefficient of volume compressibility
av
e0
= coefficient of compressibility (change in void ratio with respect to change in stress,

which is assumed constant for small strains)
= initial void ratio
In reality, the coefficient of consolidation may display significant variation over the applicable stress
range. Consequently, the small strain consolidation approach is likely to result in errors between the
actual and predicted settlement behavior when applied to large strain problems.
Figure 1:
Example of a typical e vs. log curve

Large Strain Method

Both the compressibility and the permeability of tailings deposits are likely to exhibit significant
changes when subjected to stress increases caused by continuous tailings deposition.
Large strain analyses presented in this study are based on the consolidation method proposed by
Gibson et al. (1967). The large strain method removes the constraints imposed on the compressibility
and permeability relationships by the small strain method. For large strain analyses, permeability and
compressibility relationships are typically expressed as arbitrary functions of the void ratio.
The large strain equations are relatively complex, and are typically solved using computer programs
such as CONDES (Yao and Znidarcic 1997) or FSCONSOL (GWP Software 1999). In these computer
programs, the relationships between permeability, compressibility, void ratio, and effective stress are
often expressed in closed form. For example, the relationships used in FSCONSOL are:
(3)
(4)
where is the vertical effective stress and A, B, M, C, and D are material parameters determined from
appropriate laboratory tests.
Assigning and Interpreting Laboratory Tests

Consolidation Testing
For small strain analyses, a traditional oedometer test (ASTM D2435) is often used to determine
material parameters. Oedometer tests are relatively inexpensive, and offered by most geotechnical
soils laboratories. A standard oedometer test is performed by applying a vertical load at the surface of
a confined sample and measuring deformation during the consolidation process. The applied load is
increased (typically doubled) in every loading increment until reaching the maximum desired stress
level. The oedometer test provides cv value as a function of the applied vertical stress (although
typically only one cv value is selected for small strain analysis) and an e vs. log curve, from which a
compression index, Cc value, can be derived.
For large strain analyses, sufficient laboratory data must be collected to fit material parameters A, B,
M, C, and D in equations (3) and (4). Typically, this requires either a staged slurry consolidation test
or a seepage induced consolidation test (SICTA). These tests are not standardized by ASTM, and may
require shipping samples to a specialized laboratory. While providing sufficient data to model the nonlinear tailings behavior, these tests also have an advantage over the standard oedometer test in
determining tailings parameters at low stresses. A SICTA test is particularly well suited for
determining soil parameters at effective stresses below 30 kPa.
A slurry consolidation test is typically performed on a loose/low density tailings sample. Ideally, the
initial sample placed in the cell should be close to its initial in-situ (deposition) density, but at the same
time prepared at a high enough density to avoid segregation. The first loading increment is typically
between 15 and 30 kPa. Once the sample has consolidated, a permeability test is performed. This
procedure is repeated, typically by doubling consolidation loads until reaching maximum desired
vertical effective stress.

A SICTA test is conducted similarly to the slurry consolidation test for effective stresses larger than
approximately 30 kPa. Consolidation parameters at lower effective stresses, however, are determined
by inducing seepage flows in order to consolidate the sample. Material parameters are typically
determined via inverse-solution modeling procedures. The SICTA test may be used to accurately
determine consolidation parameters for effective stresses below 10 kPa.
Material Sampling
Regardless of the selected laboratory test, it is important to obtain a representative sample of the
tailings material, and to prepare test specimens in accordance with the testing objectives.
Consolidation properties are a function of grain size distribution, particle shape, mineralogy, etc. In
addition, the initial tailings void ratio and the tendency of tailings to segregate depend on the selected
depositional method. Traditional dilute slurry deposition often produces highly segregated tailings,
resulting in sandy material with relatively high density near the discharge points (beach sands) and
fine-grained, slowly consolidating, low density tailings in the center of the TSF (tailings slimes).
Conversely, thickened tailings or paste have significantly lower tendency to segregate, and are more
likely to result in uniform tailings properties over the entire TSF footprint.
If tailings are prone to segregation, it is important to determine the range of consolidation properties for
different tailings fractions, from beach sands to tailings slimes, and estimate their distribution and
relative proportion within the impoundment. Ideally, tailings samples required to characterize
conditions at an existing facility are collected at different locations within the TSF (e.g. near beach, end
of beach, tailings pool). For a proposed or relatively new facility, this type of geotechnical
investigation may not be possible, while at an existing facility, the sampling program may be difficult
to execute considering financial and logistical constraints and/or conflicts with production.
Flume Testing
A representative sample of the tailings feed (composite tailings sample) is often obtained from a pilot
plant or from a tailings discharge point (for an existing facility). As noted previously, the segregation
potential for thickened tailings or paste samples may be minimal. For conventional slurry, however,
the segregation potential is often significant requiring further evaluation based on flume test results. A
flume test is performed by discharging composite tailings at a pre-defined solid content and flow rate,
based on the operating and design parameters. Following the test, segregated tailings are sampled from
various points along the flume at different distances from the discharge point. These samples are then
subjected to classification and consolidation testing to obtain the representative range of material
parameters.
A flume test example displaying the depth of tailings along a flume for various initial solid contents is
shown in Figure 2. In general, lower discharge densities (lower solid contents) typically yield greater
tailings segregation. For example, a sample deposited at 35.5% solid content (Figure 2) displays a
relatively large quantity of material at relatively high deposition angle near the discharge point, a clear
indication of segregating tailings. Grain size distributions for the flume deposits sampled near the
discharge point and at the distal end of the flume (Figure 3) demonstrate a relatively high segregation
potential. In this particular case, a composite feed sample with 73% fines produced material containing
43% fines near the discharge point and material with 100% fines at the end of the flume.

Figure 2:
Sediment height vs distance from discharge for flume tests at varying solid contents
Figure 3:
Grain size distribution curves for segregated tailings samples taken from flume test
sediment. Flume test performed using a 35.5% solids feed.
A SICTA test was performed on the feed sample, with slurry consolidation tests conducted on the neardischarge and end-of-flume (distal) samples. Material parameters for these samples are summarized in
Table 1 with the corresponding compressibility and permeability curves shown in Figures 4 and 5,
respectively.
Table 1:
Material parameters for FSCONSOL input

Sample
C (m/s)
Discharge Point
0.85
-0.13
0.19
8.6E-07
4.7
End of Flume
2.42
-0.16
0.041
5.4E-09
3.1
Feed
2.78
-0.07
-1.15
6.7E-08
3.2

Figure 4:
Best fit void ratio vs. effective stress curves from SICTA and slurry consolidation
tests. X marks points from laboratory tests. Lines are best fit curves utilizing material
parameters shown in Table 1
Figure 5:
Best fit void ratio vs. permeability curves from SICTA and slurry consolidation
tests. X marks points from laboratory tests. Lines are best fit curves utilizing material
parameters shown in Table 1
Figures 4 and 5 indicate that tailings segregation may lead to large differences in material properties
between beach (near-discharge) and distal (end-of-flume) tailings. The distal tailings void ratio at the
vertical effective stress of 10 kPa is approximately two times larger than the beach tailings void ratio:
edistal=1.72, ebeach=0.82 (Figure 4). The difference in permeability between beach and distal tailings
samples at the same stress level exceeds one order of magnitude (Figure 5). Finer grained distal
tailings are likely to exhibit higher compressibility (higher settlements), higher void ratio (lower
density), and lower permeability (slower consolidation rates).
Column Settling
At low solids contents, the tailings behavior is often influenced by sedimentation. Only after the
tailings attain sufficient consistency, preventing relative movement between different size particles (i.e.
preventing segregation), may tailings settlements be modeled using large strain consolidation theory.

Column settling tests are commonly performed to determine an appropriate initial void ratio for the
consolidation analysis. A column settling test utilizes settlement of a dilute tailings sample poured in a
graduated cylinder to estimate segregation potential and estimate consolidation characteristics at low
effective stresses. The height of the column is monitored and recorded until the settlement is
completed. A grain size distribution of the tailings samples collected from the top and the bottom of
the settling column may be used as a rough estimate of the segregation potential.
Figure 6 shows the results of a column settling test. As before, different tailings types behave
significantly different. The near-discharge (beach) sample reached a higher equilibrium density (void
ratio of 1.2) than the end-of-flume (distal) sample (void ratio of 2.9). In addition, the consolidation of
the near-discharge sample was significantly faster. Column settling tests for both samples were initiated
with approximately the same initial height and using the same boundary conditions.
Comparison of Calculated Densities and Consolidation Time

The largest disadvantage of small strain methods is that it is relatively cumbersome to account for the
non-constant material properties. To illustrate this point, a small strain model was developed to
determine settlement of a 1 m thick layer of tailings when gradually loaded to 500 kPa.
The ranges of cv and Cc values for the feed and distal (end-of-flume) samples were estimated from the
SICTA and slurry consolidation tests, and are presented in Table 2. Table 3 illustrates the variance in
calculated settlements for the 1 m thick tailings layer at the base of the TSF.
Figure 6:
Example of column settling test results
Table 2:
Range of cv and Cc values derived from SICTA and slurry consolidation tests
Sample
Max Cc
Min Cc
Feed Sample
0.53
End of Flume
0.41
Max cv
2
Min cv
(cm /sec)
(cm2/sec)
0.34
2.0E-02
9.5E-04
0.22
2.0E-01
5.1E-03

Table 3:
Range of calculated settlements for 1 m thick layer using small strain calculations
Sample
Maximum
Compression
(m)
Minimum
Compression
(m)
Feed Sample
0.35
0.19
End of Flume
0.37
0.24
Comparison of 1D and 3D Methods

Theoretical Preliminaries
Programs used to model large strain consolidation typically provide solution to a non-linear second
order partial differential equation (Gibson et al. 1967). These programs provide one-dimensional, timedependent solutions of void ratio distribution (solid content distributions), layer thickness, pore
pressures, and degree of consolidation. For 1D analyses, the TSF capacity can be calculated using the
following procedure (e.g. Gjerapic et al. 2008):
Input material parameters and TSF geometry into the modeling software.
Discretize the TSF into several columns of varying height. Each column has a base area selected
such that the sum of the base areas of each column multiplied by the height of each respective
column will produce a volume closely approximating that of the actual TSF. A schematic
showing a simplified TSF discretization is shown in Figure 7.
Use mine planning data to determine the average tailings inflow (typically equal to production
rate expressed in dry tonnes per day). Divide the production rate (in cubic meters per day) by the
area of the first column, A1, to determine the first filling rate, q1, as illustrated in Figure 7. Fill
the TSF at this rate until the elevation of tailings reaches the base of the second column, H1.
Increase the area to include the first and second columns, A2. Recalculate filling rate using the
larger area (i.e. determine the filling rate, q2), and continue filling until reaching the base of the
third column at elevation H2. Repeat this step, increasing the area (i.e. apply filling rates q3 and
q4 over areas A3 and A4) until the top of the TSF is reached.
Calculate the TSF capacity by multiplying the tailings production rate with the filling time.
Calculate the average density of tailings by dividing the TSF capacity determined in the previous
step with the TSF volume.
Other outputs (e.g. pore pressure, degree of consolidation) can typically be imported into a
spreadsheet from the modeling software for further analysis.
Figure 7:
Example of discretized TSF

The 1D method to determine TSF capacity is also referred to as an upper bound method (Gjerapic et al.
2008) because it implicitly assumes that the sides of the TSF undergo the same deformation as the
tailings material in the center of the TSF at the same elevation. Typically, foundation soils are much
less compressible than the tailings. As a result, the 1D model typically over-estimates foundation soil
settlements. Consequently, the time required to fill the TSF is also overestimated, resulting in
potentially unrealistic estimates of TSF capacity and average tailings dry density. Figure 8 illustrates
compression of the TSF foundation soils introduced by the 1D method.
Figure 8:
TSF foundation settlements in 1D model
To eliminate the error introduced by compressible TSF boundaries, a 3D approach can be implemented.
An approach for eliminating calculation errors caused by compressible TSF boundaries has been
developed by Gjerapic et al. (2008). In summary, a series of 1D large strain models is developed for
individual columns (from deep to shallow TSF areas) enforcing incompressible boundaries at the base
of each column. In addition, adjustments are made to filling rates and filling times of individual
columns in order to compensate for settlements occurring during filling in adjacent columns.
The error caused by compressible boundaries is a function of the TSF geometry, material properties,
boundary conditions, and filling rate. Table 4 compares the results of analyses performed using both
1D and 3D methods for two different TSFs. Figures 9 and 10 show filling curves for these two
impoundments. Figure 10 demonstrates that the error produced by using the 1D method (i.e. by
implicitly assuming compressible boundaries) may be significant. The difference between 1D and 3D
method results in an error of 94.6% for TSF B scenario.
Table 4:
Error due to compressible boundaries (1D analysis)

TSF
1D Capacity 3D
Capacity
% Error
(years)
(years)
4.8
4.4
8.6
30.0
15.4
94.6

Figure 9:
Filling curve for TSF A (see Table 4)
Figure 10:
Filling curve for TSF B (see Table 4)
Errors in Mass Conservation

One source of errors that is often overlooked is the computer program itself. In some cases, the
computer program may exhibit difficulties in converging to a correct solution potentially resulting in a
mass conservation error, i.e. the mass conservation may not be maintained (e.g. mass may be lost)
throughout the calculation process. The convergence and mass conservation errors appear to vary
between different programs. Preliminary studies indicate that higher rates of rise, more compressible
tailings, and slower consolidating may increase the magnitude of the mass conservation error.
To illustrate the magnitude of this error, a 1D large strain calculation was performed on two theoretical
TSFs. One of the TSFs was relatively flat-bottomed, with an average rate of rise of 0.3 m/yr. The
second TSF was cone-shaped, with an average rate of rise of 2.7 m/yr.
To estimate the mass conservation error, one may calculate the average dry density of tailings in two
different ways.
Method 1: The average impoundment density is calculated using the 1D procedure described in the
previous section. In summary, the time required to fill the impoundment is multiplied by the filling
rate to determine total mass of solids residing in the impoundment. The average dry density of tailings
is determined by dividing the total mass of solids with the TSF volume at the end of deposition.

Method 2: The second method is based on integrating the void ratio profile generated by the modeling
software using the stage curve (elevation-area-volume relationship) for a given tailings impoundment.
In effect, the TSF is divided into horizontal slices (e.g. 1 m high). Model output is then used to
determine tailings density within each slice. Finally, the average density is determined by averaging
the density over all slices, the average tailings density effectively calculated as a weighted average
(weighted by volume of the individual slices).
If the compressible boundary effect discussed in the previous section is relatively negligible, both
Method 1 and Method 2 should yield the same average density. The difference between two methods,
however, is an indication that a 3D analysis may be necessary, or that there is a mass conservation error
caused by mesh discretization (see e.g. Gjerapic and Znidarcic 2007), or some other error.
Table 5 compares mass conservation error estimates based on the above methods. As expected, the
mass conservation error is smaller for the TSF exhibiting the lower rate of rise.
Generally, Method 2 produces more realistic results, e.g. one would expect lower tailings density for
the TSF exhibiting the higher rate of rise because tailings have less time to consolidate (before the TSF
capacity is reached). While the densities calculated based on void ratio profile integration (Method 2)
are consistent with the expected trend, the densities calculated using filling times (Method 1) are not.
In addition, Method 2 is more likely to result in conservative average density estimates.
Table 5:
Difference in calculated tailings density due to software error

TSF
Shape
Average Rate Average Dry Density Average Dry Density

Method 2
of Rise
Method 1
% Error
3
3
(m/yr)
(tonnes/m )
(tonnes/m )
Flat
0.3
1.44
1.34
7.6
Cone
2.7
1.63
1.23
27.7
Comparison between Calculated and Measured Densities Field Study

In order to verify the accuracy of the consolidation analyses, the authors compared calculated and insitu average densities for an operating TSF. The in-situ tailings volume was calculated in AutoCAD
using the as-built topography for the empty TSF and a bathymetric survey defining the top of tailings.
The mass of deposited tailings was estimated using the daily production rates provided by the TSF
operator. The average TSF in-situ densities were then calculated by dividing the estimated mass with
the in-situ tailings volume. Three comparisons were made at 5, 7, and 12 months after the TSF
commenced operation.
Calculated densities based on a 3D large strain model were used for comparison. The 3D model
incorporated two different tailings materials (segregated and un-segregated tailings). The volume
distribution of the two selected tailings types was estimated using the flume tests (see Figure 2) and
best-guess estimates for the operational percent solids of the tailings at the discharge points. A
comparison between average densities based on in-situ measurements and calculated values based on
the 3D consolidation model is shown in Table 6.
The calculated densities were lower than the measured densities by 11.9 to 12.5%. The consistency of
the calculated difference indicates the analysis methods were well suited for the tailings type and TSF
site under consideration. The difference between the in-situ data and values predicted by a numerical
model is likely due to engineering conservatism applied to the analysis. Specifically, the division

between segregated and un-segregated tailings was uncertain. The uncertainty was partially due to
difficulty extrapolating flume tests to the full-scale TSF, and partially due to uncertainty regarding the
solids content of the operational tailings discharge. Hence, the prediction in Table 6 can be relatively
easily corrected by assigning larger percentage of the impoundment to un-segregated (coarser) tailings.
Table 6:
Comparison of calculated and measured average dry density of tailings

Cumulative TSF Average In-Situ Dry Calculated
Average
Filling Time
Density
Dry Density
% Difference
3
3
(months)
(tonnes/m )
(tonnes/m )
5
1.22
1.08
-11.9
1.21
1.06
-12.5
12
1.17
1.03
-12.1
Conclusions and Recommendations

The basis of a defensible settlement model for a TSF starts with performing laboratory tests on
representative tailings materials. Laboratory tests need to be designed and conducted in a manner to
provide the range of the consolidation parameters governing the in-situ tailings behavior.
One of the primary purposes of a TSF settlement model is to estimate TSF capacity. The assumption
that the tailings will not segregate may lead to unrealistic TSF capacity estimates. A justification for the
assumption that tailings will not segregate should be confirmed by laboratory testing.
While non-segregating tailings are typically beneficial from a TSF capacity standpoint, such tailings
may not be desirable if coarser material is required for constructing future raises of the TSF dam.
Small strain analyses may be cost-effective and reasonably accurate for old desiccated tailings deposits
that are being subjected to additional loads (e.g. during closure). However, small strain analyses are
not recommended for new, operational, or recently closed TSFs.
1D large strain analyses are often the most appropriate modeling tool for conceptual planning or tradeoff studies. The 1D large strain analyses are more accurate than small strain methods, while requiring
relatively small additional effort to perform.
3D large strain analyses are preferred for feasibility and detailed level engineering work due to
potentially large errors caused by compressible foundation boundaries, the assumption that is implicitly
applied when using 1D large strain analyses.
The mass conservation error should be evaluated for both 1D and 3D large strain consolidation
analyses. If survey data is available, settlement models can be calibrated to provide better prediction of
future tailings behavior and ultimate TSF capacity. By incorporating operational details into the
modeling process, the predicted tailings behavior is likely to exhibit favorable agreement with field
observations.
References
Gibson, R.E., England, G.L. & Hussey, M.J.L., 1967. The Theory of One-Dimensional Consolidation of
Saturated Clays. Geotechnique, 17, 261-273.
Gjerapic, G., Johnson, J., Coffin, J. & Znidarcic, D., 2008. Determination of Tailings Impoundment Capacity
via Finite-Strain Consolidation Models. In: Alshawabkeh, A.N., Reddy, K.R. & Khire, M.V. eds. GeoCongress

2008: Characterization, Monitoring, and Modeling of GeoSystems (GSP 179) conference proceedings. New
Orleans, LA, 798-805.
Gjerapic, G. & Znidarcic, D., 2007. A Mass-Conservative Numerical Solution for Finite-Strain Consolidation
during Continuous Soil Deposition. In: Siegel, T.C., Luna, R., Hueckel, T. & Laloui. L., eds. Geo-Denver 2007:
Computer Applications in Geotechnical Engineering: New Peaks in Geotechnics (GSP 157) conference
proceedings. Denver, CO, CD-ROM.
GWP Software 1999. FSConsol Users Manual.
Yao, D.T.C. & Znidarcic, D., 1997. Crust Formation and Desiccation Characteristics for Phosphatic Clays Users Manual for Computer Program CONDES, Florida Institute of Phosphate Research (FIPR) Publication
No. 04-055-134.

Determination of Consolidation Properties

Uploaded by

Copyright:

Available Formats

Determination of Consolidation Properties

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Determination of Consolidation Properties

Uploaded by

Copyright:

Available Formats

Proceedings Tailings and Mine Waste 2011

Vancouver, BC, November 6 to 9, 2011