Beruflich Dokumente
Kultur Dokumente
1. Introduction
2. Definitions
3. Disaggregate Characterization
Parameter Description
Joint Size
Joints usually occupy only a part of the joint surface (joint plane), a
fact that is represented by persistence (Table 1). Complete information on
joint size is difficult to obtain and one generally uses joint trace lengths
formed by the intersection of a joint with an outcrop surface as indicators
(or descriptors) of joint size.
Joint sizes and trace lengths can range from centimeters to many
thousand meters. Given the stochastic character of joint trace lengths, a
number of distributions have been proposed: Exponential (e.g.
Robertson, 1970; C a l l et al., 1976), lognormal (e.g. M a c M a h o n ,
1971; B r i d g e s , 1976; B a r t o n , 1978; E i n s t e i n et al., 1980), hyperbolic
( S e g a l l and P o l l a r d , 1983). The Gamma-1 distribution could also be
used (see D e r s h o w i t z , 1984), its advantage being that the exponential
distribution is a special case of it and the hyperbolic distribution can be
approximated with it. When proposing particular distributions, the just-
mentioned authors do also provide geologic evidence; actually in most
cases these authors collected data and then determined the best fitting
distribution. The wide variety of observed distributions is to some extent
caused by the fact that one observes surface traces and not actual joint
sizes, an issue which will be addressed in the next paragraph. Also,
different mechanical processes lead to different distributions, e. g. uniform
processes to exponential distributions, multiplicatory processes as they
occur in breakage to lognormal distributions and the continuity of the
process from smallest to largest sizes to hyperbolic distributions.
What one is actually more interested in, but cannot obtain, is the
distribution of joint sizes (rather than trace lengths). Exercises with circular
disk shaped joints ( B a e c h e r et al., 1977) have shown that both expo-
nential and lognormal radius distribution lead to lognormal trace length
distributions. W a r b u r t o n (1980 a and b) who developed methods t o
infer joint size distribution from trace length distributions for parallelo-
grams showed, for example, that lognormal trace length distributions lead
to lognormal size distributions for parallelograms with edges parallel to the
sampling surface. One should note, however, that if joint sizes were deter-
ministic and only rock joint location were stochastic the trace lengths still
would be stochastic.
To conclude this section, one can state that a variety of joint trace
distributions occur, that this variety may to some extent be caused by the
fact that one observes surface traces and not actual sizes, but that the
variety of underlying mechanical processes and therefore the geologic
materials and history are probably most responsible for the occurence of
Characterizing Rock Joint Geometry with Joint System Models 25
Aperture
Aperture measurements have only recently become a part of joint
characterization, this mainly in the context of fluid flow through fractured
media. With regard to fluid flow, however, aperture per se may be an insuf-
ficient characterization in that roughness combined with aperture produces
channels. Some statistical studies on aperture have been conducted and
both lognormal (S n o w, 1965 ; G r o s s m a n, 1985 b) as well as exponential
distributions (B a r t o n, 1986) have been observed.
Joint Termination
In observations, two main categories of joint termination can be
distinguished:joints that terminate at the intersection with other joints and
joints that terminate against intact rock. Direct observations by the MIT
group in the Porter Square Station Cavern and evaluations of Stripa obser-
Characterizing Rock Joint Geometry with Joint System Models 27
4.10rthogonal Model
j\
\/
%
\/
\
Applicability
The most fundamental assumption of the Orthogonal model is that it
consists of three mutually perpendicular sets of parallel joints. As a conse-
quence of this assumption, rock blocks are all rectangular prisms. Actual
rock jointing patterns involve substantial scatter in joint orientation, and
are usually not exactly represented by the Orthogonal model. However, the
Orthogonal model provides a reasonable approximation for many jointing
patterns ( W e i s s , 1972) in which most blocks can be approximated by
rectangular prisms.
By definition of the orthogonal joint system model, only very minor
variations in joint orientation are permissible. As a result, this model is
only appropriate where joint formation processes are sufficiently regular as
to produce subparallel jointing. Complicated mechanisms such as complex
folding, subsequent shearing, or superposition of other jointing patterns
could introduce sufficient dispersion so as to make the orthogonal joint
system model inappropriate. Also the orthogonal joint system model
assumes planar joints; since many mechanisms can produce non-planar
jointing, this requirement is another significant limitation of the model.
ii! "Li
A j -..------ A'
11
t mela~
~-o.s
I
I J 1i
, I
L (
llI'/~ J
i f
i
B
) B'
i
/ :. r I ! ,/i I: i~ T i
i
i, i o-
I I ~ ~=
' o! ,~.oI !
] f
,i t/
/
i,! o~
I
i
,J
I
I
< CONTiNUiTY
UNCERTAIN
Fig. 4. Joint trace map, Sierra Nevada, California (from S e g a 11 and P o I I a r d, 1983)
34 W.S. Dershowitz and H. H. Einstein:
Applicability
The applicability of this model hinges on the existence of circular or
elliptical joint shapes. As discussed in Section 3 such shapes have been
observed and many documented cases exist and can be explained from a
fracture mechanics point of view. Direct observation of joint shapes (and
sizes for that matter) whether elliptical, circular or polygonal will always be
difficult because of the limited exposure. Indirect inferences can be drawn,
however. Disk shaped joints can only form blocks if they are large
compared to the region under consideration; for instance in Fig. 3 the disk
dimensions would have to be equal or greater than the cubic region. Thus
the often observed blocky appearance of rock masses implies either that
the joints are not disk shaped or, if they are disks, that they are (or were at
one time) greater than the visible part of the rock mass. In cases where
joints terminate in intact rock, disks of any size are possible. Fig. 4 from the
Sierra Nevada is an example of an outcrop in which almost all joint traces
are planar, and terminate to a point (rather than against another joint).
Such traces could be produced by elliptical or circular joints; the evidence
is, however, not sufficient to draw definite conclusions. Another restriction
is the fact that joints have to be planar which eliminates a number of joint
formation mechanisms which can result in non-planar jointing.
(B.) Marking of
Polygonal Joints
However, the simple Poisson plane joint model is based on the assumption
of infinite extent of joints, which makes it unsuitable for modelling of
bounded joints. V e n e z i a n o (1978) introduced a method for adaptation
of the concept of Poisson plane joints to bounded joints. Fig. 6 illustrates
the generation of a Veneziano rock joint system model. The model requires
three consecutive stochastic processes. First, joint planes are generated as
Poisson planes. These joint planes are located in space by a uniform
distribution, but may have any desired distribution of orientation. Second,
a Poisson line process on each joint plane divides joint planes into poly-
gonal regions. Finally, a portion "PA" of these polygons is randomly
marked as jointed, while the remainder is defined as intact rock where
"PA" corresponds to persistence as defined in Table 1. With this model,
joint shapes are polygonal, and joint sizes are defined by the intensity of
the Poisson line process and the proportion of polygons marked as joints.
The use of joint planes as a first step in the generation of joints results in a
tendency toward coplanar jointing.
In a two dimensional trace plane, the Veneziano model resembles the
Baecher model, except that joints are represented by coplanar line
segments or fibers (Fig. 7) rather than independent fibers. In addition,
V e n e z i a n o (1978) demonstrated that his model leads to an exponential
distribution of joint trace lengths, which contrasts with the lognormal
distribution found with the Baecher model. As a matter of fact by selection
of extreme values of characteristics, the Veneziano model can be trans-
formed to the Priest and Hudson and orthogonal models, and approximate
the Baecher model (see D e r s h o w i t z, 1984).
l ,
Fig. 7. Two Dimensional Veneziano Model
Applicability
Polygonal shapes as defined by the Veneziano model can often be
observed in nature. As long as the polygons are rectangles (or squares) and
the orientation of joint sets is approximately constant and perpendicular
between sets the Orthogonal model is applicable. In the many cases where
other polygonal shapes and orientation distributions occur the Veneziano
model is more appropriate, particularly if the joints are coplanar (Fig. 8).
Fig. 8. Rock blocks, Butte Mountain Formation, California. Many coplanar joints but not
orthogonally oriented sets. The Veneziano model can represent such geometries.)
Rock blocks can be created with the Veneziano models if the joints are
100 % persistent and unbounded; in such a case joints will cross each other
and not terminate (like the pure Poisson model, Fig. 5). In the case of non-
persistent, bounded joints the Veneziano model can, but will usually not,
produce blocks. This is so because joints are defined by Poisson lines on
precedingly defined Poisson planes; intersections between joints on
different joint planes, therefore, do not often match joint edges.
38 W.S. Dershowitz and H. H. Einstein:
define a process of lines on each joint plane, which divides each plane into
polygons. The second process is the marking of a persistent portion of
polygons on each plane as joints, and the remainder as intact rock. As in
the Veneziano model, this is done by a stochastic process in which each
potential joint has an equal probability of being marked as an open joint.
Thus joint edges are defined by joint plane intersections, and as a result all
joint intersections occur at joint edges. Also, joints correspond directly to
the faces of the polyhedra defined by joint planes. As a result, polyhedra
faces are either completely intact or completely jointed and rock blocks
can be modelled relatively easily.
Applicability
Since joint edges correspond to joint intersections in the Dershowitz
model, distinct rock blocks can be defined at any scale, regardless of the
proportion of each joint plane which is defined as jointed. Thus the
~- 4 /Subsequent joint
I L__ i~i------i I - - - J s e t , 50% open
i I
planes do not intersect within the scale of the problem. The virtual joint set
does not add additional jointing, but does add the required joint plane
intersections. This provides the Dershowitz model with the flexibility to
approximate the properties of the Veneziano model. Difficulties are caused
by the fact that polygon (block face) sizes are controlled by the intensity of
intersecting joint plane processes. As the intensity of the plane process
increases, the number of intersecting lines on each plane increases, and
therefore the size of polygons defined by the lines decreases. If joints are
defined as a constant percentage of each plane, the increase in plane
process intensity results in a larger number of smaller polygons.
As in the Veneziano model, joints in the Dershowitz model are
coplanar, and the model is not accurate for models with bounded, non
coplanar joints. However, by a reduction in the proportion of each joint
which is defined as jointed, the effect of coplanarity can be reduced, and
the Dershowitz model can be used as an approximation. As this proportion
is reduced, however, rock block formation and the proportion of joints
terminating at joint intersections is also reduced, such that the model will
then have the same strengths and weaknesses as the other non-coplanar
model, the Baecher model.
~x\ / "x\ / z~ I \ t \ ,
, / , /
-4---
--4- -k._
I
t ~ ...................... :'
Orthogonal Hexagonal Triangular
i j I~
j
/^
k / ,~ / J 17-. 3
/ ;'-. --7, \ ] .
b) Stochastic Tessellation
/R--/K
b. Stochastic Tessellations
Fig. 13. Three-dimensionalmosaic tessellations
\ /
\ /
\ ' 9 I
J \ /
9 F--- 9
~ 26 b :;.j
~A 46 A ~7 - \ 7 ~ ~r.o J. 2~ ,p:,-
5g
64
~" el s 14 =c . . . . ~.-~/////
so ~ 56
r 63
(~r SO 13
57
57
62
61
o4 12
c 56
55
74 ~a 11 2
e/ ~ NO X 54 . .
73
i \
Fig. 15. Columnar Jointing at Burg Stolpen, taken from Stoyan 1980, after Koch, 1974
46 W.S. Dershowitz and H. H. Einstein:
Applicability
Mosaic Tessellation, and specifically Mosaic Block Tessellation, intro-
duces a class of rock joint system models which has not been utilized
before in rock mechanics. Mosaic Tessellation models feature polygonal
joints, with terminations at the edges of rock blocks, and systematic joint
intersections. These two features make mosaic tessellation similar to the
Dershowitz model. However, unlike the Dershowitz model, Mosaic Tessel-
lation offers the capability to model joints which are not coplanar. This
provides a new level of flexibility in joint system modeling.
Another interesting but somewhat problematic feature of Mosaic
Block Tessellation is the generation of rock blocks first, followed by rock
joints derived from the faces of the blocks. This feature is distinctly
different from the process used in models discussed previously. Since
joints are modeled only indirectly, joint sizes, shapes, locations, and orien-
tations are also modeled indirectly. It is, therefore, not possible to directly
use sampled distributions of joint characteristics to construct a particular
block tessellation model. Specification of distributions for these properties
must be done indirectly, as part of the definition of the process of block
formation. For example, in order to obtain a specific distribution of joint
shapes, an appropriate combination of processes for location of the seeds
and for block growth must be determined. Similarly, in order to obtain a
desired orientation distribution, the process of block seed location and
block growth must be controlled simultaneously.
Since jointing in Mosaic Block Tessellation models is defined by the
faces of a process of non-overlapping blocks completely containing the
rock mass, they are appropriate for joint systems which are actually the
result of a process of block formation in a rock mass. One example of such
a joint system is jointing in columnar basalts. However, columnar jointing
can in most cases only be modelled with Stoyan's (1980) approach and not
with Voronoi or Delauney tessellations.
The primary limitations on the application of Mosaic Block Tessel-
lation models are requirements for joint and block shapes. Mosaic Block
Tessellation model joints are the faces of polyhedral blocks, and as a result
must always be polygonal. Ample geological evidence has been presented
for polyhedral rock blocks and polygonal jointing. For cases which do not
display polyhedral blocks, and polygonal joints, Mosaic Block Tessellation
models may not be appropriate.
To our knowledge, only the Stoyans (1980) have applied mosaic Block
Tessellation to the characterization of rock mass geometry. The other
models have so far not been used.
Joint system models capture the fact that geometric rock mass prop-
erties are related to each other and that some relationships predominate.
Observations such as the blocky appearance of many rock masses or, in
Characterizing Rock Joint Geometry with Joint System Models 47
5. Conclusions
Acknowledgements
References
A g t e r b e r g, E. P. (1976): Geomathematics, Elsevier, New York.
A m b a r c u mj a n, R. V. (1974): Convex Polygons and Random Tessellations.
In: H a r d i n g , E. F., K e n d a l l , D. G. (eds.), Stochastic Geometry, Johan Wiley &
Sons, New York, 176--191.
B a e c h e r , G.B., L a n n e y , N.A. (1978): Trace Length Biases in Joint
Surveys. Proceedings of the 19th U. S. Symposium on Rock Mechanics, Vol. 1,
56--65.
B a e c h e r , G.B., L a n n e y , N.A., E i n s t e i n , H . H . (1977): Statistical
Description of Rock Properties and Sampling. Proceedings of the 18th U.S.
Symposium on Rock Mechanics, 5C1-8. . .