Sei sulla pagina 1di 13

SOLUTION MINING

RESEARCH INSTITUTE

Technical
Paper

105 Apple Valley Circle


Clarks Summit, Pennsylvania 18411
USA
Country Code: 1 Voice: 570.585.8092 Fax: 570.585.8091
E-mail: smri@solutionmining.org www.solutionmining.org

Subsidence Modeling Techniques


An Overview of Graphical, Analytical,
and Numerical Methods

Paul E. Nelson
RESPEC
Rapid City, South Dakota, USA

Fall 2005 Meeting


25 October 2005
Nancy, France

ABSTRACT
Ground subsidence is a consequential result of underground mining. Associated with ground
subsidence is surface tilt and extensional and compressional strains which may cause damage
to pipelines, roads, railways, buildings, and other structures. Therefore, it is essential to
understand and predict the effects that an underground opening may have at the ground
surface. The earliest subsidence modeling methods used empirical relations relating subsidence
measurements to mine geometry and graphical techniques to apply the empirical relation
toward making subsidence predictions. Other subsidence modeling methods, and most
currently used methods, are based on theoretical models of material response (e.g., elasticity).

INTRODUCTION
One result of creating an underground excavation is vertical subsidence at the ground
surface. Associated with subsidence is surface tilt and extensional and compressional strains
which may result in damage to pipelines, roads, railways, buildings, and other structures.
Therefore, it is essential to understand and predict the effects that an underground opening
may have at the ground surface.
The earliest methods used to predict subsidence were developed by the European coal mining
industry. What may be the earliest formula to predict ground subsidence is [Kratzsch, 1983]:
vz = M cos

(1)

where:
vz = vertical displacement at the ground surface
M = mine height
= dip angle of the mining seam

which states that the vertical subsidence is equal to the height of a vertical section through the
mining seam.
Subsidence prediction techniques have evolved significantly since Equation 1 was derived.
The modern discipline of subsidence engineering is based on the theories of solid mechanics and
typically incorporate numerical techniques. The increased sophistication has allowed modeling
of increasingly complex geometries and stratigraphies. Current numerical techniques also
allow the prediction of time-dependent subsidence above excavations created in creeping
materials.
The underlying assumptions used in the prediction of ground subsidence can be based on
empirical relations or material behavior models such as elastic, plastic, or stochastic theories,
1

while the technique used to make subsidence predictions may be graphical, analytical,
numerical, or use influence functions. For example, ground displacement predictions may be
made by creating a computer model of the underground opening and surrounding stratigraphy.
The computer model is then used to make subsidence predictions based on elastic rock behavior.
This paper is organized to first give a basic understanding of the assumptions upon which most
subsidence modeling techniques are based and then discusses the modeling techniques
themselves.
Ground subsidence, as discussed in this paper, should be differentiated from catastrophic
subsidence events such as sinkholes (catastrophic is used here in the sense of the magnitude of
surface displacements). Modern subsidence modeling techniques assume that the rock mass
behaves as a continuum; that is, the rock beds deform uniformly. Sinkholes result in
discontinuous displacements that are much more difficult to model.

EMPIRICAL METHODS
Empirical methods were developed in the coal mining regions of Europe and are some of the
earliest techniques used to predict ground subsidence. The empirical method uses data gained
from observations (i.e., subsidence measurements) to develop an equation or set of equations to
make predictions about the effect of future mining. In effect, an equation (the empirical
relation) is fit to past subsidence measurements. An example (one method of many) is the
Polish profile-curve method which uses the following exponential equation to match subsidence
data [Kratzsch, 1983]:
vz = vz Max e nr

(2)

In Equation 2, vz Max is the maximum measured subsidence, r is the distance from the mining
panel center to the point where subsidence is to be calculated, and n is expressed as:
n=

vz Max
R2c

(3)

where:
R = radius of the critical area
c = average roof displacement.

The example given above fits an exponential equation to measured subsidence data.
However, other empirical relations may incorporate logarithmic, quadratic, or trigonometric
functions.
Empirical methods are typically developed for a particular mining district with its own
unique mine geometry, stratigraphy, and rock properties and are continually modified as new
2

data are collected. Although an empirical relation may accurately predict subsidence for a
particular mining district, the empirical relations are developed to suit the ground response for
a particular set of data (and stratigraphy). Therefore, the empirical relations developed for one
mining district may not accurately predict subsidence at another region. Also, the empirical
relations cannot be developed until measurable subsidence has already occurred. Thus the
potential effects of subsidence as a result of mining at a new site cannot be predicted accurately
a priori.

THEORETICAL MODELS
Theoretical models involve the application of solid mechanics (in the case of elasticity or
plasticity) or statistics (in the case of a stochastic model) to derive a set of equations describing
the displacement of a surface over an underground opening. The theoretical models assume
that the rock mass behaves as a continuum. This assumption implies that the rock behaves as
if it is composed of infinitesimally small particles, all with the same properties. Thus
deformations are assumed continuous (i.e., no discontinuities) throughout the material.
However, inhomogeneous materials and anisotropic material properties can be simulated with
some numerical techniques; the finite element method is an example.

Elastic Theory
An elastic material will deform under a given load and return to its original shape upon
removal of the load. Although this often does not exactly describe the behavior of a typically
porous and fractured material such as rock, an elastic model can often reasonably simulate the
ground surface response to the excavation of an underground opening in many situations. In
addition to assuming that the rock behaves as a continuum, it is assumed that volume is
conserved. That is, the volume of the subsidence bowl equals the closed volume of the mine or
cavern.
Several elastic models exist dependent upon how the underground excavation is treated. For
example, Maruyama [1964] derives the following equation from the closing of a volume in a
semi-infinite half-space:

Zu ( x , y) =

UV
b
UVD
UVD

tan 1
2
2
2
2
2 U + D V + D
D

(4)

where:
Z u ( x , y ) = ultimate subsidence at any point (x , y) on the ground surface
b = vertical closure of the opening
D = depth to the top of the opening

and:
= U 2 + V 2 + D2

The notation

(5)

is used to represent an equation of the form:


f (U ,V ) = f (U 2 ,V2 ) f (U 2 ,V1 ) f (U1 ,V2 ) + f (U1 ,V1 )

(6)

where:
U1 = X c

W
x
2

U2 = X c +

W
x
2

(7)

L
V1 = Yc y
2
V2 = Yc +

L
y
2

where (Xc, Yc) are the coordinates of the center of the underground opening represented by a
parallelepiped where L and W are its length and width, respectively. Equations 4 through 7 are
the basis of the subsidence calculations made by the Solution Mining Research Institute (SMRI)
sponsored subsidence-modeling computer program SALT_SUBSID [Nieland, 1991].
The linear behavior of an elastic material allows the superposition of subsidence predictions
made by the closing of multiple excavations. It is important to note that elastic theory does not
follow the rule of the angle of draw. That is, elastic models predict finite displacements outside
the area defined by the angle of draw.

Plastic Theory
Because of discontinuities in the rock mass (e.g., fractures and void spaces), rock typically
does not behave as a perfectly elastic material. This is because the discontinuities present in
the rock absorb some of the movement (in addition to the elastic response) upon application of a
load. This movement is then not recovered upon removal of the load.

An example of an equation derived for the prediction of subsidence above a plastic rock body
is given by Kratzsch [1983]. This equation is:
vz =

aM
2

x 0.5l
x + 0.5l
tanh
tanh H
H

(8)

where:
a = ratio of vertical subsidence to mine closure
M = room height
H = depth to mine horizon
x = distance from the center of subsidence
l = opening width

and:
=

Vz m
4

(9)

where:
Vz m = rate of subsidence
= kinematic stiffness of the rock body.

(10)

Stochastic Theory
The stochastic theory is a statistical approach that treats the rock body as a cohesionless
medium. A rock mass that contains numerous and pervasive fractures may behave as a
material with zero cohesion, much like a sand.
The stochastic theory states that when a volume of rock is removed (an extraction element),
it is replaced by an equal volume of rock lying above. The resultant particle movement at the
ground surface is predicted by a bell-shaped probability curve representing the subsidence
profile. This is demonstrated by Figure 1. A probabilistic approach to predicting surface
subsidence is not a commonly used but may be the basis of certain particle flow computer codes.

SUBSIDENCE MODELING TECHNIQUES


Subsidence modeling is based on either an empirical relation or a theoretical model.
However, there are several techniques for utilizing the model or empirical relation to perform
predictive subsidence calculations. These include (1) graphicalthe use of graphs, nomograms,

Figure 1. Representation of Probabilistic Particle Movement in a Cohesionless Material (After


Kratzsch [1983]).
or gridded overlays to estimate subsidence; (2) influence functions; (3) analyticalthe direct
calculation of subsidence at a point; and (4) numericalthe use of numerical models (e.g., finite
element).

Graphical Techniques
Graphical methods and empirical relations are some of the earliest techniques developed to
predict surface subsidence. However, graphical techniques tend to be quite labor intensive.
One graphical approach employs a set of graphs from which key points on the subsidence
surface are located (Figure 2). Another graphical technique uses a circular, sectored overlay to
calculate the amount of subsidence at a point by measuring the percentage of the mine area
that falls within each sector (Figure 3).

Influence Functions
Influence functions are based on several assumptionsone of which is superposition which
states that the subsidence at a point can be calculated by summing the elementary subsidence
troughs for each and every extraction volume located within that points cone of influence
(defined by the angle of draw). The previous statement follows another assumption which
states that a subsidence trough is delimited by outward extending lines which make some angle
with vertical (the angle of draw). Influence functions are also assumed to possess rotational

(a)

(b)

(c)

Figure 2. Graph (a) Developed for the Angle-of-Intersection Method Used to Locate Key Points
of the Subsidence Trough in the Dip Direction (b) and in the Strike Direction
(c) After Kratzsch [1983].
symmetry and equivalence (all extraction elements of the same size and at the same depth have
an equal amount of influence at the ground surface). Lastly, influence functions assume that
volume is conserved, which states that a specified volume of closure at depth is expressed at the
surface with an equal subsidence volume.
Influence functions derive their name from the concept that the closure of an extraction
element results in a finite amount of subsidence (influence) at a given point on the ground
surface. The magnitude of the influence (i.e., subsidence), kz, may be given as a function of the
solid angle ( ) , the radial distance from the point of measurement to the center of the
subsidence trough (d), or as the distance from the extraction element to the point of
measurement (f). These functions are represented in Figure 4 where the angle noted by
represents the angle of draw. Influence functions are derived from either empirical relations or
a theoretical material model (e.g., elasticity).

Figure 3. Sectured Overlay Used to Predict Subsidence at Point P (After Kratzsch [1983]).

dA

Figure 4. Key Parameters Commonly Used to Describe the Influence (Subsidence) at Point P
as a Result of Excavating Volume dA.

An example of an influence function (which is a function of the distance from the cavern axis,
d) is provided by Reitze [2000] for the prediction of subsidence vz ( x , y, z ) , at any point above a
solution-mined cavern.
vz ( x , y, z ) = vz Max e

d2
o u tan 2
Z Z

(11)

and
vz Max = a

kV E
tan 2
o u
Z Z

(12)

where:
vz Max = maximum subsidence
d = distance from the cavern axis
Z o = depth of the cavern roof
Z u = depth of the cavern floor
= limit angle
a = transmission factor (0 a 1)
k = convergence factor (0 k 1)
V E = final volume.

The advantage of using an influence function is that the extraction volume is discretized into
small extraction elements. Therefore, irregular mine or cavern geometries and multiple mining
levels are easily managed. A disadvantage of influence functions is that they ignore the
interaction between extraction elements unless accounted for in the function itself.

Analytical Techniques
Analytical techniques simply solve for subsidence at a point by substituting the appropriate
values into an equation derived from a material behavior model. Analytical techniques are
similar to influence functions, but violate one of the assumptions made by influence functions;
specifically, the assumption regarding the angle of draw. Analytical techniques based on an
elastic material model predict subsidence outside the area delineated by the angle of draw.
Similar to influence functions, when superimposing subsidence any interaction between
excavations is ignored.
Analytical techniques are best used in conjunction with a computer when calculating
multiple subsidence valuesat several hundred grid points while developing a contour plot, for
example. SMRIs subsidence modeling program SALT_SUBSID uses this approach.
9

Numerical Techniques
Numerical techniques include finite element, finite difference, and boundary element
methods. The numerical techniques noted above involve subdividing (discretizing) the area of
interest (the model region) into discrete areas or volumes (elements or zones) to create a mesh
or grid. Suitable material properties are then assigned to each of the elements (or zones) along
with appropriate boundary and initial conditions. Mines or caverns are represented as void(s)
in the model.
A typical numerical model involves the following tasks:
1. Discretization of the model region
2. Assign constitutive models and material properties to the elements
3. Apply appropriate boundary conditions (e.g., fixed or rollered)
4. Apply appropriate initial conditions (i.e., in situ stress and temperature fields)
5. Step to a solution
6. Extract the results.
Calculating the subsidence is a simple task that involves extracting the displacements of the
points lying on the plane of the model that represents the ground surface. Horizontal strains
can also be directly output from the computer model.
The advantage of numerical techniques is that complex geometries and stratigraphies can be
accurately represented. Most numerical programs include elastic, plastic, and viscoelastic
(creep) constitutive models that can be applied to individual elements. Thus material
inhomogeneities can be modeled. Directional anisotropy can also be modeled as well as
discontinuities such as faults (given appropriate properties). Also, the interaction between
caverns can be simulated which is not possible with influence functions or analytical
techniques.
A disadvantage of numerical techniques is the effort required to generate a model, with some
complex geometries requiring a significant amount of time to generate. Also, estimates of the
stratigraphy and rock properties may be absent or expensive to obtain.
Of course,
simplifications to the model and assumptions of material properties can be made at the expense
of accuracy.
However, it is possible to calibrate a numerical model to subsidence
measurements by adjusting model parameters until a reasonable match to the measured data is
obtained.
Advances in modeling software have made many numerical packages very user friendly,
making it possible for most anyone to generate a model and perform a subsidence simulation.
However, a significant amount of experience is required to judge whether the modeling results
10

accurately represent reality. It is easy to get results from a numerical model; the difficulty is
determining whether the results are correct or not.

TIME DEPENDENCY
Up until now, little mention has been made of time-dependent surface subsidence with the
exception of numerical models which can simulate the ground response as a result of salt creep.
All equations noted in regard to influence functions and analytical techniques predict final or
ultimate subsidence.
Influence functions and numerical techniques can be used to predict time-dependent
subsidence with certain assumptions. A key assumption associated with making timedependent subsidence predictions is that the spatial distribution of surface subsidence rates is
proportional to the spatial distribution of surface subsidence. This states that the amount of
subsidence at specific points of the ground surface is, at any time, proportional to the ultimate
subsidence. This approach allows the surface subsidence to be predicted at any time by
multiplying the ultimate subsidence by a time-dependent function. This approach will be
explained more fully in the description of SALT_SUBSID that follows.

SALT_SUBSID
Equation 4 is used in SALT_SUBSID to predict the ultimate surface displacement at any point.
However, an opening created in salt closes as some function of time. Therefore, Equation 4 is
modified as follows to account for this time-dependency:
Z ( x , y,t ) = Z u ( x , y ) G (t )

(13)

where:

N
1
; if ysst + y0 1 e E t > 1

Gt =
y t + y 1 e E N t ; otherwise
0
ss

where:
yss , y0 , , N = model parameters
E = extraction ratio ( = 1 if a solution mine)
t = time since excavation.

The model parameters are modified to match existing subsidence measurements or


estimated when no subsidence measurements are available. One method of estimating the

11

SALT_SUBSID parameters is with the use of a numerical model, such as a finite element

simulation. Numerous case studies have been given testifying to the effectiveness of using
SALT_SUBSID to model surface subsidence rates (see Ratigan [2000]; Van Sambeek [2000];
Cartwright and Ratigan [2005].

SUMMARY
Subsidence modeling has evolved from the use of empirical relations and graphical
techniques to predict surface subsidence unique to a particular mining district to the use of
sophisticated numerical models allowing for the prediction of horizontal and vertical surface
displacements above complex geometries and stratigraphies. However, the use of less
sophisticated modeling software, such as SALT_SUBSID, has also been shown to be widely
adaptable to a variety of conditions while providing reasonable subsidence predictions.

REFERENCES
Cartwright, M. J. and J. L. Ratigan, 2005. Case HistorySolution Mining a Cavern That
Intersects a Plane of Preferred Dissolution, Solution Mining Research Institute Fall Meeting,
Nancy, France, October 25.
Kratzsch, H., 1983. Mining Subsidence Engineering, Springer-Verlag Berlin Heidelberg, New
York, NY.
Maruyama, T., 1964. Statistical Elastic Dislocations in an Infinite and Semi-Infinite
Medium, Bulletin of the Earthquake Research Institute, Tokyo University, Vol. 42, pp. 289368.
Nieland, J. D., 1991. SALT_SUBSID: A PC-Based Subsidence Model, Users Manual, RSI-0389,
prepared by RE/SPEC Inc., Rapid City, SD, for the Solution Mining Research Institute,
Woodstock, IL.
Ratigan, J. L., 2000. Anomalous Subsidence at Mont Belvieu, Texas, Solution Mining
Research Institute Fall Meeting, San Antonio, TX, October 1518.
Reitze, A., 2000. Prediction of Ground Movement Above Salt Caverns Using Influence
Functions, Solution Mining Research Institute Fall Meeting, San Antonio, TX, October 1518.
Van Sambeek, L. L., 2000. Subsidence Modeling and Use of Solution Mining Research
Institute SALT_SUBSID Software, Solution Mining Research Institute Fall Meeting, San
Antonio, TX, October 1518.

12

Potrebbero piacerti anche