Sie sind auf Seite 1von 56

Numerical Modelling Background

Numerical Modelling
Background

By Terry Wiles
© Mine Modelling Pty Ltd 2006

This presentation is aimed at providing an overview of numerical modelling and


how it applied to mine design. The objective is to set the theme for discussion
that will follow.

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 1


Numerical Modelling Background

Numerical Modelling
Background

What is numerical modelling?


How does numerical modelling work?
What is the difference between various numerical
modelling codes?
What are we trying to achieve using numerical modelling?
What is the conventional approach for applying numerical
modelling to mine design?
How well does this approach work?
What are the limitations of this approach?

Here let s try to address these issues in a cursory sense only, leaving the details
until later.

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 2


Numerical Modelling Background

Analysis Results

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 3


Numerical Modelling Background

How does numerical modelling work?

Equilibrium
Continuity
Elasticity
Non-linearity

Numerical modelling achieves this simulation by using certain physical


constraints on how the rock mass can respond.

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 4


Numerical Modelling Background

Equilibrium

xx/x+ xy/ y+ xz/ z+X=0


xy/ x + yy/ y + yz/ z + Y = 0

xz/ x + yz/ y + zz/ z + Z = 0

All loads must balance one another.

xy

xx

Equilibrium Applied forces must always balance one another at all locations in
the model. If you cut out a small cube of material and examine the stresses acting
on this cube these stresses must be in equilibrium.
Note that terms X, Y and Z, represent body forces that can be used to apply any
sort of external field loading. External field loading can be the result of heating,
fluid pressures material non-linearity etc. In addition, these can be determined
from many forms of in situ monitoring including for example fluid pressures (e.g.
well drawdown, dams, hydro-fracturing), heating (e.g. natural heating, nuclear
waste storage), deformations (e.g. monitored with extensometers), and seismic
activity (e.g. definition of fault slip or material non-linearity from seismic data).
It is these terms that will be used for the integration of numerical modelling with
seismic monitoring.

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 5


Numerical Modelling Background

Continuity
2 y2 + 2 x2 = 2
xx/ yy/ xy/x/ y
2 2 2 2 2
yy/ z + zz/ y = yz/ y/ z
2 2 2 2 2
zz/ x + xx/ z = xz/ x/ z

2 2
xx/ y/ z = / x( yz/ x xz/ y xy/ z)
2 2
yy/ x/ z = / y( yz/ x xz/ y xy/ z )
2 2
zz/ x/ y = / z( yz/ x xz/ y xy/ z )

You cannot have material disappearing or being


created.

Continuity In the rock mass continuum, the mass of material must be


maintained. You cannot have material disappearing or being created.

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 6


Numerical Modelling Background

Elasticity
= f + E/(1+ )/(1-2 ) [(1- )
xx xx xx + ( yy + zz)]
f
yy = yy + E/(1+ )/(1-2 ) [(1- ) yy + ( zz + xx)]
f
zz = zz + E/(1+ )/(1-2 ) [(1- ) zz + ( xx + yy)]

= f + E/(1 )
xy xy xy
= f + E/(1 ) 1 1
yz yz yx
f
xz = xz + E/(1 ) xz
Strength
envelope

Note that this is where the


q
pre-mining stresses ( f) UCS 1 f

are incorporated 3 e 1

Elasticity - At locations where the stresses do not exceed the strength, the rock
deforms in a linear elastic manner: stresses varying in direct proportion to the
strains.

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 7


Numerical Modelling Background

Non-linearity

1 = UCS + q 3 q = tan2(45+ /2)


= (1+sin )/(1 sin )
p = p1 + p + p3
vol 2
1 1

= ( p1- p ) tan( )
2 i
Strength
envelope

s = Coh + tan( ) q Elastic-perfectly


n UCS 1
plastic model
f predictions
p = p tan( )
s n i
3 e p 1

Non-linearity - At locations where the stresses are concentrated to the point


where they exceed the strength, the rock will yield to these loads and deform:
deformations are all allowed to proceed until the stresses relax down to the
strength.

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 8


Numerical Modelling Background

How does numerical modelling work?

Note that the equations of equilibrium and


continuity are differential equations. What
we need to do to solve these is to integrate
them over the rock mass volume such that
the appropriate boundary conditions are
satisfied.

There are many ways of accomplishing this. BEM integrate the equations
analytically, then use a numerical approximation to satisfy the boundary
conditions. FEM and FDM packages use a numerical integration scheme to
integrate the differential equations.
It must be emphasized that all numerical models solve these same equations.
Different methods use different methods to achieve this.

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 9


Numerical Modelling Background

All numerical models solve


these same underlying equations

FEM/FDM BEM
FLAC Map3D
UDEC Ex3D
ABAQUS Besol
Phases

P = K[n n] u = + M[n n] D
surface far_field

Both of these equations constitute a mathematical description of how the rock


mass responds. All numerical models (Map3D, Flac, Udec, Minsim, Abaqus,
etc.) use some variation on this approach.
These equations need to be solved simultaneously such the boundary conditions
are satisfied. By solving these equations throughout the rock mass, along with the
loading conditions and geometry, you conduct a stress analysis .
No matter which method is used, you end up with a large set of equations that
describes how various parts of the rock mass interact with one another. A typical
mine wide model would have in the range of 100,000 to 1,000,000 equations
depending on how much detail we build into the model, and whether we use
elastic or non-linear modelling.

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 10


Numerical Modelling Background

Analysis Results

Numerical models provide predictions of stresses, strains and displacements.


How do they do this?

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 11


Numerical Modelling Background

Numerical Modelling
1. Geometry
Background
2. Geology
1.
3. f Numerical
2.
4. Model Type Model
3.
5. Material Properties
6. Numerical Error Safety factor, probability of
failure, stability, stand-up time
etc. are not predicted.

There are six inputs to a numerical model:


1. Geometry (approximation of actual mining shapes , 2D, 3D)
2. Geology (lithology)
3. Pre-mining stress state (magnitude, stress ratios, orientation)
4. Constitutive Model type (elastic, plastic, fault slip)
5. Material properties (strength, stiffness)
6. Numerical approximation (element sizes)
Each of these needs to be characterized and numerical values provided.

When the numerical model is run there are only three outputs:
1. Stresses
2. Strains
3. Displacements
No other values are predicted by the model. Safety factor, probability of failure,
stability, stand-up time etc. are not predicted.

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 12


Numerical Modelling Background

Numerical Modelling
Background
1. Geometry CGeometry
2. Geology CGeology
1.
3. f C f Numerical
2.
4. Model Type CModel Model
3.
5. Properties CProperties
COutput
6. Num Error CNumerical

2 2
CGeometry CGeology C2f 2
CModel 2
C Properties 2
C Numerical COutput

Uncertainty or variability (coefficient of variation) of the outputs COutput is


directly dependent on the uncertainty or variability (coefficient of variation) in
the inputs CGeometry CGeology C f CModel CProperties CNumerical

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 13


Numerical Modelling Background

Uncertainty of predictions?
2 2
COutput CGeometry CGeology C2f 2
C Model 2
C Properties 2
C Numerical

1. 1D Tributary area C ±100%?


2. 2D approximation C ±50%?
3. Quasi-3D tabular C ±20%?
4. 3D blocky model C ±10%?
5. 3D refined model C ±5%?

1. Tributary area calculations provide total pillar loads only, i.e. cross-section
only. There is no confinement information must use pillar strength formula
(height to width based formula). Only works for very regular repeating
geometries. Fails near abutments. No details of stress distributions.
2. 2D approximations only works where appropriate cross-sections can be taken.
3. Tabular approximations provide pillar loads only with some details of the
stress distributions. There is no confinement information must use pillar
strength formula (height to width based formula). Only works for tabular
mining shapes fails for bulky 3D shapes. Fine near abutments.
4. 3D blocky models give errors if blocks are too coarse. It is necessary to
represent the pillars sizes accurately so that the correct amount of stress get
diverted to the abutments. Poorly shaped pillars give inaccurate stress
concentrations.
5. 3D refined models require no compromise on accuracy if the model is built
well.

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 14


Numerical Modelling Background

Uncertainty of predictions?
2 2
COutput CGeometry CGeology C2f 2
C Model 2
C Properties 2
C Numerical

COutput
CGeometry ±5% to ±100%
depends on model

Clower limit CGeometry= 0

$
Tributary 2D Quasi Blocky Refined
Area 3D 3D 3D

The uncertainty contributed by the geometric approximation depends on what


type of model we use (e.g. 1-D, 2-D, tabular, 3-D) and how carefully we
represent the actual mined shapes. Although carefully built models should
contribute no more than ±5% error, in some cases the actual mined shapes
may be poorly known because of over-break and blast damage. Details of the
geometry are often sacrificed in order to reduce model building and run times
resulting in much larger contributions from this parameter.
There is a steep cost rise in going to refined 3D modelling. Note that at some
point there will be nothing to gain from improving the geometric accuracy as
other contributions will exceed errors introduced from CGeometry.
This cost is large enough for domain methods (finite element or finite difference
methods, e.g. ABAQUS, FLAC3D) that geometry is often severely
compromised. This raises the lower limit for these modelling methods.
Refined 3D geometry is much less expensive to achieve with boundary element
methods (e.g. Map3D) thus reducing the lower limit for these modelling
methods.
The amount of effort you can justify putting into refinement of the geometry
depends on how much uncertainty is introduced the other contributions

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 15


Numerical Modelling Background

Uncertainty of predictions?
2 2
COutput CGeometry CGeology C2f 2
C Model 2
C Properties 2
C Numerical

FEM/FDM BEM
FLAC Map3D
UDEC Ex3D
ABAQUS Besol
Phases

Almost impossible to Easy to build complex


build complex geometries
geometries
Any skew shapes or oblique
Symmetry is often stress states are easily
required limiting shapes
accommodated
and stress states

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 16


Numerical Modelling Background

Uncertainty of predictions?
2 2
COutput CGeometry CGeology C2f 2
C Model 2
C Properties 2
C Numerical

COutput
CGeology ±5% to ±100%
depends on model

Clower limit CGeology= 0

$
Homogeneous Broadly defined Detailed variations
Material Lithology from place to place

There may be considerable stiffness and strength contrasts from place to place
due to natural variation of the rock. Although broadly defined lithological
units can at times be identified and readily incorporated into the modelling,
often this is either unknown or simply ignored due to the high cost of
characterizing the properties of the alternate material zones.
There is a steep cost rise in attempting to model detailed variations of the geology
from place to place. Such variations will generally never be known anyway.
This cost is large enough that a compromise is always made. Note that at
some point there will be nothing to gain from improving the geometric
accuracy as other contributions will exceed errors introduced from CGeology .
Detailed geology is less expensive to achieve with domain methods (finite
element or finite difference methods, e.g. ABAQUS, FLAC3D) since the rock
mass is already subdivided in small zones.
Detailed geology is more expensive to achieve with boundary element methods
(e.g. Map3D).
The amount of effort you can justify putting into refinement of the geology
depends on how much uncertainty is introduced by the other contributions

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 17


Numerical Modelling Background

Uncertainty of predictions?
2 2
COutput CGeometry CGeology C2f 2
C Model 2
C Properties 2
C Numerical

COutput
C f ±5% to ±30%
depends on in situ variability and
calibration effort
Clower limit C f=0

$
Estimated In situ stress Minimization
measurements via modelling

Most often the stress state is measured in situ or simply estimated. It is generally
accepted that owing to geological non-homogeneity, the pre-mining stress
state varies on the order of ±20%. It is possible to refine this through back-
analysis.
There is a steep cost rise in attempting to accurately determine f. Variations
from place to place are almost never measured. Detailed variations will
generally never be known anyway. This cost is large enough that a
compromise is always made. Note that at some point there will be nothing to
gain from improving the accuracy as other contributions will exceed errors
introduced from C f.
Neither domain methods nor boundary element methods offer any advantage in
this situation.
Note that at some point there will be nothing to gain from conducting additional
measurements as there is natural variability in the rock mass. Additional
measurements simply confirm what this natural variability is.
The amount of effort you can justify putting into refinement of the pre-mining
stress state depends on how much uncertainty is introduced the other
contributions

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 18


Numerical Modelling Background

Natural variability is on the order of 20% to 30%.

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 19


Numerical Modelling Background

Natural variability is on the order of 20% to 30%.

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 20


Numerical Modelling Background

Natural variability is on the order of 20% to 30%.

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 21


Numerical Modelling Background

Variability of lab tests is on the order of 20% to 30%.

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 22


Numerical Modelling Background

Variability of lab tests is on the order of 20% to 30%.

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 23


Numerical Modelling Background

Uncertainty of predictions?
2 2
COutput CGeometry CGeology C2f 2
C Model 2
C Properties 2
C Numerical

COutput
CModel ±5% to ±100%
depends on mine

Clower limit Cmodel = 0

$
Equilibrium Elastic Elastic-perfectly Strain-softening
plastic plastic

In most hard rock mines, elastic modelling can provide sufficiently accurate
simulations (less than ±10% error). However, in cases where there is a significant
amount of yielding (fault slip or pillar yielding) inelastic modelling needs to be
considered.
The primary disadvantage of elastic modelling is that stresses are not re-
distributed when the strength is exceeded. This discussion applies to pillar
yielding as well as fault slip problems. As long as there is not a significant
amount of stress re-distribution, the errors introduced by this approximation are
minimal. The vital question here is how to assess how significant these inelastic
effects are! This can be tested during model calibration. If it is found that the
errors due to this effect are too large, then inelastic modelling needs to be
considered.
When there are many highly yielding pillars, stress transfer effects can become
very important. Elastic modelling can still be used in such cases. At the extreme
where pillars are obliterated upon failure, these situations can be dealt with
simply by excavating the failed pillars while sticking with elastic modelling.
Pillars can be softened by adjusting their stiffness according to some empirically
derived rule. However, this later situation is probably best dealt with using
inelastic modelling.

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 24


Numerical Modelling Background

Uncertainty of predictions?
2 2
COutput CGeometry CGeology C2f 2
C Model 2
C Properties 2
C Numerical

COutput
CProperties ±5% to ±60%
depends on calibration effort

Clower limit CProperties = 0

0 1 2 3 4 5 6 7 8 9 10 # of params $
Equilibrium Elastic Simple-plastic Complex-plastic

Although full-scale in situ testing can be conducted, most often properties are
estimated by conducting laboratory tests then degrading the results to field
scale. Although poorly documented, it is not uncommon to find uncertainties
in the procedure for degradation of laboratory strength to field scale values to
add ±30% or more. It is possible to refine these values through back-analysis.
Uncertainty in material properties can add anywhere from ±5% to ±60% or
more depending on the calibration effort.

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 25


Numerical Modelling Background

Uncertainty of predictions?
2 2
COutput CGeometry CGeology C2f 2
C Model 2
C Properties 2
C Numerical

Laboratory Strength: c50 mi 20% to 30%


Field Scale Strength: 2
1= 3+ m c 3+s c

Reduction to Field Scale: m/mi=e[RMR-100)/28] 15% to 50%


s=e[RMR-100)/9]

(20%)2 + (15%)2 25% (30%)2 + (50%)2 60%

The least expensive approach to determine material properties is to use a


combination of laboratory strength measurements and a technique like Hoek-
Brown to degrade these to field scale values. The limitation of this approach is
the relatively large uncertainty ±25% to ±60% that results.

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 26


Numerical Modelling Background

Uncertainty of predictions?
2 2
COutput CGeometry CGeology C2f 2
C Model 2
C Properties 2
C Numerical

20000 Calibration
effort
Number of inputs:
15000 Equilibrium 1
10000 Elastic 3
5000
Perfectly Plastic 6
Strain Softening 9
0
# model
0
Equilibrium
2
Elastic
4 6
Simple
8 10
Complex
input
parameters
Time dependent >9
model model Plastic Plastic

An alternative is to determine the input parameters using a through back-analysis.


While this approach is more expensive, it has the distinct advantage of offering
much reduced uncertainty (±5%).
The various model types we can use (equilibrium, elastic, plastic etc.) provide a
potential improvement in accuracy of our model predictions. With increased
constitutive capability (I.e. plastic versus elastic), these models can simulate our
rock mass response more realistically by taking into account full three-
dimensional geometries, stiffness and or pre-mining stress state varying across
out mine site, stress redistribution due to plastic yielding of pillar or fault-slip,
etc. However, this increase in constitutive capability comes at the cost of having
to specify ever more material properties used as parameters in these equations.
For equilibrium (tributary area) models there is really only 1 parameter hence the
effort for calibration is quite small (on the order of 3). For elastic models, there
are 3 parameters hence the effort for calibration is on the order of 27. For elastic-
perfectly plastic models, there are 6 parameters hence the effort for calibration is
on the order of 729. For more complex plastic models the effort required is very
large, easily exceeding 19,683. Beyond this we could consider time-dependant
creep response and more.

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 27


Numerical Modelling Background

Uncertainty of predictions?
2 2
COutput CGeometry CGeology C2f 2
C Model 2
C Properties 2
C Numerical

CNumerical
L
20

15

10

D 0
0 1 2 3 4 D/L

The accuracy with which various numerical models solve the specified equations
is well known and readily quantifiable. Basically this boils down to ratio of the
distance from the nearest excavation surface, D, divided by the element width, L
(i.e. size of the elements used to approximate the equations). To obtain accurate
results near excavation surfaces (D is small) you need to ensure that L is also
small. The important point here is that these errors are easily controlled to any
desired magnitude.

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 28


Numerical Modelling Background

Uncertainty of predictions?
2 2
COutput CGeometry CGeology C2f 2
C Model 2
C Properties 2
C Numerical

COutput
CNumerical ±5% to ±20%
depends on D/L

Clower limit CNumerical = 0

0 1 2 3 4 D/L $

The accuracy with which various numerical models solve the specified equations
is well known and readily quantifiable. Basically this boils down to ratio of the
distance from the nearest excavation surface, D, divided by the element width, L
(i.e. size of the elements used to approximate the equations). To obtain accurate
results near excavation surfaces (D is small) you need to ensure that L is also
small. The important point here is that these errors are easily controlled to any
desired magnitude.

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 29


Numerical Modelling Background

Uncertainty of predictions?
2 2
COutput CGeometry CGeology C2f 2
C Model 2
C Properties 2
C Numerical

??% ??% 10% ??% 25% 5%


to to to
30% 60% 20%

(10%)2 + (25%)2 + (5%)2


Hoek-Brown
to 30% to 70%
(no minimization)
(30%)2 + (60%)2 + (20%)2

We can now estimate the total combined uncertainty COutput. If we use in situ
stress measurements and the Hoek-Brown technique, COutput is estimated to be on
the order of 30% to 70%. Although this approach has the advantage of low
cost, the uncertainty is quite large. One can hardly justify spending a great deal
on money on geometric refinement or complex inelastic modelling since any
improvement in accuracy obtained from inelastic modelling will be swamped
from inaccuracies in the stress and strength parameters.
Note that given a stress prediction for example of 1=50 MPa, 30% means
50 15MPa, 70% means 50 35MPa.

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 30


Numerical Modelling Background

Uncertainty of predictions?
2 2
COutput CGeometry CGeology C2f 2
C Model 2
C Properties 2
C Numerical

??% ??% 5% ??% 5% 5%

(5%)2 + (5%)2 + (5%)2 10% (with minimization)

(10%)2 + (25%)2 + (5%)2


Hoek-Brown
to 30% to 70%
(no minimization)
(30%)2 + (60%)2 + (20%)2

If we minimize the errors in the input parameters through detailed back-analysis,


COutput is estimated to be on the order of 10%. Although this approach has the
disadvantage of high cost, the uncertainty is quite low.
Note that given a stress prediction for example of 1=50 MPa, 10% means
50 5MPa.

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 31


Numerical Modelling Background

What are we trying to achieve


using numerical modelling?

The simple goal of numerical modelling is


to predict how the rock mass will respond to
mining.

PREDICTION OF ROCK MASS RESPONSE


TO MINING

We specify the loading conditions, specify the geometry (this can include dykes,
faults, bedding etc.), assume that elasticity and/or non-linearity applies, then
solve these equations for the stresses, strains, displacement etc. throughout the
rock mass.
By examining the results from the stress analysis we can identify locations that
are for example over-stressed or yielding. We can determine what locations need
support and what type of support is appropriate, etc.

Knowing what problems to expect, and the time and location of these problems is
the ultimate goal.
Compare alternative mine layouts.
Develop strategies to deal with problems.
Avoid unexpected production interruptions.
Lay down hard design numbers for support, pillar widths, stope heights, etc.

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 32


Numerical Modelling Background

What are we trying to achieve


using numerical modelling?

Numerical modelling provides a tool that


helps you design your mine.

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 33


Numerical Modelling Background

Analysis Results

While modelling by itself is reasonably straight forward, it is the interpretation of


the modelling results that is the real challenge. Somehow we must relate the
model predictions of stress, strain or displacement to the response we see in situ.

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 34


Numerical Modelling Background

What is the conventional approach for


applying numerical modelling to mine
design?

Initially (i.e. pre-mining) you can conduct


parametric studies.
compare alternate designs
identify important factors affecting stability
trade off the cost-benefit of increased
production versus increased maintenance
and support requirements.

Parametric studies certainly provide useful information as they allow you to


develop strategies to deal with problems if and when they occur.

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 35


Numerical Modelling Background

In the end it is fundamentally


necessary to have knowledge of the
reliability of a prediction to assess
risks and base cost benefit
decisions on.

But if you want to use modelling to lay down hard design numbers for support
requirements, specific pillar widths, acceptable stope heights and actual costs,
you need to have confidence in the accuracy of your predictions.
For example, there would be no point in specifying that a pillar be made 1.2m
wide based on a model prediction if you had no idea how reliable this prediction
was. You would never allow personnel to walk under a brow based on modelling
results that indicated the brow was safe unless you also were confident that the
predictions were reliable.

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 36


Numerical Modelling Background

Reliability

You may know that the reliability of your


predictions is very low and that they could
have large errors, alternatively you may
have evidence to demonstrate that the
reliability of your predictions is very high
and that they are probably very accurate.
Anyone in a decision making position needs
to know this so they can weigh this into
there cost benefit considerations.

It is worth spending a moment to reflect on this statement. I can think of no


examples where predictions of unknown reliability are of any value.
Let s ask these two questions:
If somehow you could show that your model predictions were 99% reliable (i.e.
only ±1% error) why wouldn t you use them?
If somehow you could show that your model predictions were only 1% reliable
(i.e. ±99% error) why would you use them?
The real crux of the matter here is reliability. We must know the reliability of our
predictions or we have nothing.

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 37


Numerical Modelling Background

Terzaghi s Observational
Approach to Design

Decide on some sort of initial mine layout -


parametric studies.
Begin mining.
Monitor the rock mass response normally
visual.
Redesign based on the observed behaviour -
model calibration

The procedure used to establish the reliability is called Terzaghi s Observational


Approach to Design .

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 38


Numerical Modelling Background

This process constitutes model


calibration

Redesign

Monitor

Mine

Terzaghi s Observational Approach to Design

Whether we use numerical modelling or not, this is the basic approach we all use
to design in our mines.

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 39


Numerical Modelling Background

How well does this approach


work?

As mining progresses, you learn how


reliable (or unreliable) your model is.

The value of this approach is that as mining progresses, you numerical model
becomes better and better calibrated: you learn how reliable or unreliable your
model is. Note that traditionally observations of the rock mass response are made
visually and often supplemented with sparsely located instruments. By making
observations of rock mass response, over time you literally get to see when the
model works and when it does not. You learn what features need to be included
in the model (e.g. fault planes, lithology, loading conditions etc.). You not only
learn how to use the model to predict rock mass response, but also gain
confidence in the predictions and recognize situations where the model
predictions are suspect.

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 40


Numerical Modelling Background

In short you learn how reliable


(or unreliable) your predictions
are.

Once you have established the reliability, you are in a position to fine tune your
design. A well calibrated model allows you a glimpse into the future: to predict
how the rock mass will respond.
When properly applied, this procedure is extremely valuable. You can trim pillars
and modify the design with confidence, leaving ore only where you need to. You
avoid unexpected interruptions in production. You can mine with confidence.

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 41


Numerical Modelling Background

Using back-analysis to determine


field scale strength parameters
300
Blk 27 - 6800
250 Blk 71 - 6800
200 Blk 72 - 6800
Sigma 1

Blk 73 - 6800
150 Blk 22 - 7000
100 Blk 34 - 7000
Blk 33 - 7200
50
Blk 42 - 7200
0 Blk 72 - 7200
0 5 10 15 20 25 30 Blk 74 - 7200
Sigma 3 Best Fit

To reinforce this concept let s consider a real example. Consider several pillar
bursts that occurred at Inco s Creighton Mine over several years of mining. For
each rockburst, a numerical model was run to determine the stress state at the
time and location of the burst. If we plot all of these stress predictions on a set of
1 versus 3 axes we see that there is a strong correlation between the stress state
at the time of each burst and a strength criterion. The coefficient of correlation of
this data is 0.90.
Let s state this in a different way. If you calculate the difference between the
stress state for each pillar failure and the best fit line, then take the mean of these
errors, you will find that the mean error in prediction of 1 is approximately 14
MPa. This gives a coefficient of variation of only ±7%.

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 42


Numerical Modelling Background

Reliability
300
250
(a) (a)
200
Sigma 1

150
(b) (b)
100
50
0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
Sigma 3 Sigma 3

It is evident that for this particular example, the stress state predicted from
numerical modelling is a very reliable predictor for the the time and location of
pillar bursts.
Considering the figure on the left, certainly you would not expect pillars with
stress states (a) or (b) to burst at this point in the mining sequence (i.e. (a) should
have burst by now, (b) should have burst yet).
This is because they are a lot more than 14 MPa (the mean error in prediction)
from the best fit line. We can see the reliability of our predictions from the other
back-analysis results, and it is obvious that pillars with stress states (a) or (b) are
well out of the range of uncertainty.
Alternatively, considering the figure on the right, you could not say with
confidence that pillars with stress states (a) or (b) would or would not burst.
While pillar (a) is above the best fit line and hence should have burst by now, it is
well within the range of variability: other pillars reached this stress level without
bursting. Similarly, although pillar (b) is below the best fit line and hence should
not burst yet, it is well within the range of variability: other pillars reached this
stress level and had already burst. We can see the reliability of our predictions
from the other back-analysis results, and it is obvious that pillars with stress
states (a) or (b) are well within the range of uncertainty.

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 43


Numerical Modelling Background

Reliability
300
250
(a) (a)
200
Sigma 1

150
(b) (b)
100
50
0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
Sigma 3 Sigma 3

Now consider the case where no back-analyses have been done (shown on the
right). Can you say with confidence that pillars with stress states (a) or (b) would
or would not burst? While pillar (a) is clearly above the failure criterion and
hence should have burst by now, we have no idea what the variability is, nor do
we have any idea if this line is even representative of pillar failure. In other
words, we do not know the reliability!

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 44


Numerical Modelling Background

What are the shortcomings of this


approach?

It takes many passes through the


mine/monitor/redesign loop to learn how
reliable the model predictions are.
It may take several years before you
establish the reliability with any confidence.

Shortcomings is probably the wrong term to use here. It would be more


appropriate to call these problems or costs associated with this approach.
This takes time, manpower and continued dedication (a lot of trips underground).
Monitoring the rock mass response is no trivial task.
While I firmly believe that predictions of unknown reliability are of no value,
only you can decide if the effort required to determine the reliability can be
justified ( i.e. it is worth following through with this approach).

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 45


Numerical Modelling Background

What are the shortcomings of this


approach?

There may be geologic features that cannot


be adequately characterized for modelling
Exclusion of such features can invalidate
the model results

There are other less obvious problems here. For example, often the details of the
location, shape, extent and condition of various faults are not well enough known
to be included in a numerical model with any accuracy.
Through proper application of Terzaghi s Observational Approach to Design you
can establish the level of reliability in the model predictions. It could be that you
find out that they are very unreliable and be left with an unpredictable situation
(at least unpredictable as far as your existing model is concerned). Even in this
unfortunate case we need to know this, so we can make intelligent decisions (i.e.
ignore the numerical modelling results)..

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 46


Numerical Modelling Background

What are the shortcomings of this


approach?

If local conditions such as lithology, pre-


mining stress or jointing vary rapidly from
place to place it may be impractical to
monitor with sufficient resolution to
identify the effects of such changing
features.

There are modelling examples where practitioners (Pariseau, 19??) have mapped
drives in extreme detail and assigned appropriate sets of material properties to
each bit of the rock according to the mapping. The resulting stress analysis
produced predictions with remarkable agreement to the observed deformations.
In another example (Wiles, 1988) showed that by measuring the in situ stresses
before and after pillar mining, that the stress analysis results agreed very well
with these measurements. This latter case demonstrated that it was our lack of
detailed knowledge of the pre-mining stress state that limited prediction
accuracy, not the model itself.
Normally we do not have this type of information, and hence it is not taken into
account. This shows up as large scatter in model predictions thus reducing
reliability.

Wiles, T.D., Accuracy and Applicability of Elastic Stress Analysis Methods in


Mining , 15th Canadian Rock Mechanics Symposium, Toronto, 1988.

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 47


Numerical Modelling Background

What are the shortcomings of this


approach?

This technique can only be applied where


we are actively mining.
We cannot apply this method to Greenfield
sites or feasibility studies.

Laboratory testing and degradation to rock mass strength should only be used as a
deficient alternative when back-analysis results are not available.

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 48


Numerical Modelling Background

What are the shortcomings of this


approach?

We would like to shorten the calibration


loop.
We would also like to be able to
characterize geologic features well enough
for incorporation into numerical modelling.
Finally we would like to reduce scatter and
develop a model with high reliability.

Clearly we would like to avoid these problems.

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 49


Numerical Modelling Background

How are we going to improve on


these shortcomings?

There is no substitute for back-analysis!

We must always correlate our results with observed rock mass response. This is
the only way to determine how well our model is working.

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 50


Numerical Modelling Background

Model
Poorly optimized
Mining
costs
mine design
Predictions

Well optimized
mine design

Level 1 Level 2 Level 4 Level 4


Modelling costs

Trade off mining costs versus modelling


costs requires meaningful cost projections
for various alternatives

The only way to decide on how to proceed is to trade off the cost of poorly
optimized unreliable designs associated with large values of s, versus the high
cost of minimizing s through calibration. Larger values of s cost less to determine
but require using larger safety factors and hence less well optimized, more
expensive designs. Smaller values of s cost more to determine but allow for
smaller safety factors and hence better optimized, more economic designs.
This trade-off can be more easily made if real cost figures have been associated
with each listed response category. This will allow you to present meaningful
cost projections for various alternatives when you make your predictions. It will
also allow you to justify expenditures for conducting the various components of
the model calibration exercise.

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 51


Numerical Modelling Background

Poorly optimized
Mining
costs
mine design
Model
Well optimized
mine design
Predictions
Level 1 Level 2 Level 3 Level 4
Modelling costs

Level 1 Parametric/demonstration modelling


Demonstrate realistic behaviour
Little in the way of real design numbers
Modelling can be completed inexpensively with little
or no site specific information
Predictions are vague and ambiguous
Impossible to assess safety or to make informed cost-
benefit trade-offs

At this level, no serious attempt is made to quantify site specific values for input
parameters such as pre-mining stress state or field scale strength. Reasonable
values are estimated based on depth, rock type and parameters determined for
nearby mines. Results of alternative designs are compared purely on the basis of
relative values. For example, span interrupting pillars may be demonstrated to
significantly reduce the size of tension zones compared to open spans. Pillar
confinement may be shown to drop for large height to width ratios. Realistic
fracture patterns may appear under loading in a random assortment of particles.
Results from such modelling cannot be used quantitatively as they have not been
calibrated to local site conditions.
While such modelling can demonstrate realistic behaviour, this approach is not
deterministic. This level of modelling is conducted for demonstration purposes
only and can provide very little in the way of real design numbers. This level of
modelling is often referred to a parametric study using common-sense
interpretations.
The appeal of this approach is that modelling can be completed inexpensively
with little or no site specific information. The problem with this approach is that
predictions are vague and ambiguous. It is near impossible to assess safety or to
make informed cost-benefit trade-offs. This does not allow for very much
optimization of mine designs as shown in the figure.

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 52


Numerical Modelling Background

Poorly optimized
Mining
costs
mine design
Model
Well optimized
mine design
Predictions
Level 1 Level 2 Level 3 Level 4
Modelling costs

Level 2 - Qualitative modelling


Applied in feasibility studies/Green-field sites
Inexpensive since no back-analyses required
Predictions are relatively vague and ambiguous
s is quite large or unknown
Poorly optimized designs

At this level, attempts are made to measure site specific values for input
parameters including pre-mining stress state and laboratory strength. Field scale
strength values are estimated using a strength degradation technique such as
Hoek-Brown. However, since this later technique does not include a provision to
estimate s, these results generally cannot be used for quantitative design
purposes.
Results from this level should only be used qualitatively since this method does
not provide a clear statement of exactly what is supposed to happen in your mine
upon failure (i.e. when the stresses reach this strength envelope). For example,
do we expect to see cracking of the ground, pillar yielding, ground support
failure, blast-hole offsetting, ground stability problems, micro-seismic activity,
limited stand-up time, fault slip, excessive joint-set or bedding plane movement,
over-break depth? If so, to what level of damage do we expect to see these events
occur?
The major advantage of level 2 modelling is that it can be applied in feasibility
studies for Green-field sites where no observations are available. The method is
relatively inexpensive since no observations of field behaviour need to be made,
and no back-analyses are required. The major disadvantages of this approach are
that predictions are relatively vague and ambiguous. We must either assume that
C (and hence s) is quite large or unknown. Although it is possible to assess safety
and make cost-benefit trade-offs, the large safety factors required result in poorly
optimized designs. With this approach we can never have very much confidence
in the reliability of our modelling predictions.

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 53


Numerical Modelling Background

Poorly optimized
Mining
costs
mine design
Model
Well optimized
mine design
Predictions
Level 1 Level 2 Level 3 Level 4
Modelling costs

Level 3 - Quantitative modelling


Predictions are very specific giving the exact rock
response to be expected
s is determined as part of the back-analysis
Assess safety and make cost-benefit trade-offs
Cannot be applied in Green-field sites
More expensive than previous levels

At this level, correlations between observations of actual mine behaviour and


back-analysis results are evaluated providing site-specific strength criteria. Clear
statements are made regarding exactly what is supposed to happen when the
stresses reach specific strength levels. The goodness of fit of these correlations is
evaluated in terms of s. However, there is no systematic effort to reduce the value
of s by trialling a matrix of possible model input values.
The major advantage of level 3 modelling is that predictions are very specific
giving the exact rock response to be expected. It is possible to assess safety and
make cost-benefit trade-offs. Another advantage of this approach is that the
magnitude of s is determined as part of the back-analysis. When small values of s
are found, relatively small safety factors can be used resulting in well optimized
designs as shown in the figure. In cases where large values of s result from the
back-analysis, we will know that there is little or no basis for forward prediction.
The major disadvantage of this approach is that it cannot be applied in feasibility
studies for Green-field sites where no observations are available. The method is
more expensive than previous levels and requires considerable engineering effort.

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 54


Numerical Modelling Background

Poorly optimized
Mining
costs
mine design
Model
Well optimized
mine design
Predictions
Level 1 Level 2 Level 3 Level 4
Modelling costs

Level 4 Optimized quantitative modelling


Best reliability well optimized designs
Even though the cost of the extra modelling effort
required to minimize s can be traded off with the less
well optimized more expensive mine designs, this is
rarely done
Most expensive considered here, and can require a
large amount of engineering effort.

At this level, in addition to establishing correlations between observations of


actual mine behaviour and back-analysis results, there is also a systematic effort
to reduce the value of s by trialling a matrix of possible model input values. This
can include all model input parameters (geometry, geology, pre-mining stress
state, model type, model properties, numerical approximation).
The major advantage of level 4 modelling is that predictions are made with the
best possible reliability. By trialling a matrix of model input parameters, the
values that give the best match between observations and back-analysis are
found. This allow us to make the most well optimized designs where we use the
smallest possible safety factors will ensuring a specified level of reliability. This
can have substantial economic benefits in mine design as shown in the figure.
The major disadvantage of this approach is that the cost can be significant.
Owing to the relatively large effort required to minimize s (i.e. the large number
of trials that must be completed), this is often not carried out. Even though the
cost of the extra modelling effort required to minimize s can be traded off with
the economics of less well optimized more expensive mine designs, this is rarely
done. The method is the most expensive considered here, and can require a large
amount of engineering effort.

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 55


Numerical Modelling Background

Numerical Modelling
Background

By Terry Wiles
© Mine Modelling Pty Ltd 2006

Map3D Course Notes (c) Mine Modelling Pty Ltd 2006 56

Das könnte Ihnen auch gefallen