Sie sind auf Seite 1von 20

Rock Fragmentation by Blasting Sanchidrin (ed)

2010 Taylor & Francis Group, London, ISBN 978-0-415-48296-7

Fragmentation characterization; the Swebrec function and its use


in blast engineering

F. Ouchterlony
Swebrec, Swedish Blasting Research Centre at Lule University of Technology, Sweden

ABSTRACT: This work summarizes recent research at Swebrec and SveBeFo. It has led to a new
fragment size distribution with an unexpected ability to reproduce sieving curves of all kinds of rock and
concrete/mortar that have been blasted or crushed in almost every conceivable way. The three-parameter
Swebrec function is described but its parameters are not independent. Examples of curve fits are given as
is some advice about the fitting procedure. Its relations to other comminution concepts such as t10 from
JKMRC are explored. How to use the Swebrec function for full-scale blasts is described, especially for
large blasts where the available information is very limited. The final section treats design curves, which
can predict the changes in fragmentation that are caused by changes in specific charge and other geo-
metrical parameters of a bench blast. Evidence to the effect of confining debris at the free face is finally
presented.

1 INTRODUCTION parameters, e.g. the median fragment size x50 and a


uniformity index n. With P(x) the fraction of mass
Fragmentation plus looseness and shape of the passing a sieve cloth with a quadratic mesh of side x,
muck pile are the primary outcomes of a produc- it can e.g. be written P(x) = 1 exp[ln2 (x/x50)n].
tion blast. They determine the effectiveness and One realizes that when x << x50 then P ln2 (x/x50)n
cost of digging, loading, hauling, crushing and and that in a diagram of logP vs. log(x/x50), this
other downstream operations in falling importance. reduces to a straight line of slope n.
Most mines and quarries produce a fragmentation Hardly any sieving data for blasted and crushed
that allows an efficient operation, yet they do not rock look like that though, not even Kuznetsovs
know nor can they accurately measure the frag- (1973) original data, see Figure 1.
ment size distribution they produce. Nor could The Rosin-Rammler function has no upper
they with good accuracy predict the fragmentation size limit xmax either but this may be incorporated
if they change their operating conditions substan-
tially or open production at another site. Passing, %
100
The fragment size distribution in comminution is 90 Kuznetsov table 6
usually determined by sieving and quarry products 80
Rosin-Rammler, r2 = 0,9827
are e.g. classified by their sieving curves. Sieving 70
60 Swebrec function, r2 = 0,9995
could in principle also be applied to the fragmen-
50
tation of a blast, but a complete sieving curve of
a full-size blast round hardly exists. The digital 40
image based technique to measure fragmentation
30
(Franklin & Katsabanis 1996) has many draw-
backs (Ouchterlony 2004, 2005b). A major one is
the limited range of fragment sizes that could be 20
resolved, from xmax down to xmax/10 or xmax/30 e.g.
Another drawback is its tendency to give an erro-
neous form of the size distribution (Ouchterlony
et al. 2006). Yet the method has found many practi- 10
cal applications. 100 1000
Mesh size, mm
The most commonly used fragmentation model .

is the Kuz-Ram (Cunningham 1983, 1987 & 2005). Figure 1. Plot of fragmentation data from Kuznetsov
To describe the sieving curve it uses a single Rosin- (1973) with curve fits. Note deviation from Rosin-
Rammler or Weibull distribution, which has two Rammler behavior.

3
by truncation of the function (King 2001) or a 2 THE SWEBREC FUNCTION
transformation of the variable x when needed. The
size range in which it gives an accurate description The basic Swebrec function is given by the expres-
of the sieving curve is 12 orders of magnitude in sion (Ouchterlony 2003)
fragment size.
The Kuz-Ram models poor ability to describe P(x) = 1/{1+[ln(xmax/x)/ln(xmax/x50)]b}, (1)
fines was one of the major reasons why the Two
Component Model (TCM; Djordjevic 1999) and when 0 < x xmax. This cumulative distribution
the Crush Zone Model (CZM; Kanchibotla et al. function may take any value in the range 01.
1999) were developed. Both combine two Rosin- Figure 2 gives an example of a fit to bench blast-
Rammler distributions or components, one for the ing data in which the range of fit is three orders of
coarse part of the curve and one mainly for the magnitude, 0.5500 mm.
fines. The ability to reproduce a sieving curve accu- The logarithm term becomes zero when x = xmax
rately improves at the price of introducing 2 or 3 and one when x = x50, the median fragment size.
extra parameters and the problem of a linear size Thus the necessary conditions P(xmax) = 1 and
distribution for the fines in log-log space remains. P(x50) = 0.5 are met. The function has 3 param-
Recently Blair (2004) has used both a one- and eters, xmax, x50 and the curve undulation exponent
two-component log-normal and sigmoidal func- b, about which more later. xmax limits the frag-
tion description of fragmentation to compare with ment size. The logarithm term dominates for small
the Rosin-Rammler fits on three sieved data sets. x-values.
He restricted his studies to up to 40 % passing, i.e. This function has an inflection point in log P
he focused on the fines part down to below 0.1 mm. versus log x space at
He found that the two-component lognormal func- 1/b 1/b
tion with 5 parameters provided the best fits, the x/xmax = (x50/xmax)(b1) or x/x50 = (x50/xmax)(b1) 1

Rosin-Rammler the worst ones. Unfortunately he (2)


also found that the solutions obtained were rela-
tively sensitive to the initial guesses for the input when b 1, the inflection point tends to x = xmax.
parameters, i.e. the solutions were not robust. For increasing values it moves towards x = x50,
This paper presents a new distribution called which is reached when b = 2. When b increases
the Swebrec function (Ouchterlony 2003) discov- further, the inflection point moves to smaller val-
ered during the Less Fines project (Moser 2005, ues of x and then moves back towards x = x50. The
Sanchidrin et al. 2009). It is robust and has just inflection point and hence the undulating charac-
three parameters, with physical meaning, in its ter of the Swebrec function is always there and this
basic form. Yet it covers a size range of 23 orders makes it possible to pick up the fines behaviour
in magnitude of fragment size and even more in already from coarse fractions data near x50.
an extended version. It has been tested on several Common expressions for the variance of size dis-
hundreds of sets of sieving data from different tributions like x80x30 or x80/x30 may be easily derived
kinds of blasting and crushing of many different
kinds of rock. The coefficient of determination
r2 has in 95 % of the cases been better than 0.995, 2 2
Residuals

often substantially better. 0.5 0.5

This paper gives examples of such curve fits -1 -1

and it points out the relation of the Swebrec func-


tion to traditional comminution concepts like the
Mass passing, %

NBC concept (Steiner 1991, 1998), the t10 concept 10 10


(Napier-Munn et al. 1996) and the GGS-slope.
It gives an experimental relation between the three
curve fit parameters, whereby the number of inde- 1 1

pendent parameters in a way reduces to two.


Applications like an extrapolation from the
0.1 0.1
coarse range to the fines and vice versa will be 0.02 0.2 2 20 200 2000
Mesh size, mm
shown and as will how the Swebrec function could
be used to construct the fragment size distribution
Figure 2. Sieving curve for bench blasting with an emul-
for muck piles from full-scale blasts, based on lim- sion explosive, Brarp round 4, with best fit Swebrec func-
ited data. tion. Function parameters: x50 = 459 mm, xmax = 1497 mm
A final remark on how to implement the Swebrec and b = 2.238. r2 = 0.9973. Data range 0.5500 mm.
distribution into fragmentation models will be x = 1000 mm is based on oversize counting, not a sieved
given. value. Moser et al. (2003b).

4
from the Swebrec function. The mesh size corre- 2 2

Residuals
sponding to a percentage passing Pp is given by 0.5 0.5

-1 -1
1/b
xp/xmax = (x50/xmax)(1/Pp1) (3)

Mass passing, %
By choosing p = 30 or 80 %, i.e. Pp = 0.3 or 0.8, 10 10
the correct expression is obtained. As can be seen
the dependence on b and xmax is relatively com-
plicated. Another way to express the variance of 1 1

the distribution is the slope at a central point. We


choose to give the expression at the median frag- 0.1 0.1
ment size, P(x50) or s50, which is especially simple, 0.02 0.2 2 20 200 2000
Mesh size, mm

s50 = P(x50) = b/[4x50 ln(xmax/x50)] (4) Figure 3. Brarp round 4 data with best fit extended
Swebrec function. Function parameters: x50 = 459 mm,
Its use will become apparent below. xmax = 1480 mm, b = 2.224, a = 0.99999812 and c = 2.0.
Even if P(x) fits most sieving data for fragmented r2 = 0.9976. Data range 0.075500 mm. Note small mag-
rock quite well, it can not be used to estimate the nitude of prefactor (1a) in Equation 5.
surface area of the fragments in a muck pile. A sim-
ple conversion from volume to area in the probabil-
Residuals

1.5 1.5
ity density function dA/dV = 6/x gives an estimate
of the cumulative area as being proportional to -0.5 -0.5

P(x) dx/x. This integral doesnt converge however.


In other words, the logarithmic behavior of P(x)
when x 0 is in all probability physically unsound.
Mass passing, %

An extended version of the Swebrec function


takes care of this anomaly, 10 10

P2(x) = 1/{1 + a [ln(xmax/x)/ln(xmax/x50)]b


+(1 a) [(xmax/x 1)/(xmax/x50 1)]c} (5)
1 1
0.5 5 50 500
The new term has the effect that P2(x) behaves Mesh size, mm

like a Gaudin-Schuhmann function (x/xmax)c for


Figure 4. Tunnel blasting with ANFO in Bondkall syenite.
small values of x. Its form is chosen so as to retain Function parameters: x50 = 132 mm, xmax = 500 mm and
the properties P2(xmax) = 1 and P2(x50) = 0.5. For b = 1.965. r2 = 0.9985. Data range 2500 mm. Ouchterlony
reasonable values of c, c > 0, the area integral (2003).
P(x) dx/x converges and the range of fit now
includes all points in the set of sieving data behind
Figure 1, see Figure 3.
Residuals

Figures 47 give examples of curve fits to the 2 2


sieving curves from different types of full scale -2 -2
blasts. Figures 810 give examples for crushing.
Ouchterlony (2003, 2005b) enumerates many
different applications and more are added all the
Mass passing, %

time. To date more than 600 sets of sieving data


from blasting in full-scale and model-scale have
been fitted and more than 300 sets from crushing 10 10

with many different types of machines and crush-


ing type tests.
The materials covered include concrete and
many types of rocks; amphibolite, andesite, dol- 1 1
erite, dolomite, gabbro, gneiss, granite, greywacke, 0.05 0.5 5 50 500
Mesh size, mm
hornfels, hyperite, leptite, limestone, magnesite,
magnetite, metabasalt, quartzite, sandstone, silt- Figure 5. Reef blasting in quartzite with Gelatine.
stone, syenite and tonalite etc. Function parameters: x50 = 37.8 mm, xmax = 215 mm
The b-values usually lie in a relatively narrow and b = 2.067. r2 = 0.9962. Data range 0.6215 mm.
range, 14. For the blasting curves the average lies Ouchterlony (2003).

5
Residuals

Residuals
1 1 1 1

-0.5 -0.5 -0.5 -0.5


Mass passing, %

Mass passing, %
10 10

10 10

1 1

0.1 0.1
1 1 0.05 0.5 5 50 500
5 50 500 Mesh size, mm
Screen size, mm

Figure 9. Toothed roller crusher product of limestone.


Figure 6. Bench blasting in dolomite with dynamite. Function parameters: x50 = 48.8 mm, xmax = 107 mm
Function parameters: x50 = 147 mm, xmax = 571 mm (fixed) and b = 2.451. r2 = 0.9994. Range 0.25100 mm.
and b = 2.150. r2 = 0.9990. Data range 9.5457 mm. Ouchterlony (2003).
Ouchterlony (2003).
Residuals [7]

Residuals [6]
1 1
1 1
-2 -2 0 0
-1 -1
Mass passing, %

Passing, %

10 10

1 1 10 10
0.1 1 10 100 1000 4 40 400
Mesh size, mm Mesh size, mm

Figure 7. Blasting of sublevel caving panels with ANFO. Figure 10. Gyratory crusher product of biotite
Function parameters: x50 = 85.8 mm, xmax = 1110 mm gneiss. Swebrec function parameters: x50 = 89.4 mm,
and b = 2.931. r2 = 0.9988. Data range 0.074800 mm. xmax = 300 mm and b = 1.535. r2 = 0.9993. Data range
Maripuu (1968) and Wimmer et al. (2008). 4300 mm.
Residuals [5]

4
2
4
2
The b-values tend to be relatively constant for a
0 0 given material but they are not true material prop-
-2 -2
erties as they e.g. depend on explosive type, type of
crusher and crusher settings etc.
Mass passing, %

Most curve fitting has been done with a commercial


10 10
program, TableCurve2D, which uses non-linear least
squares fitting based on the Levenburg-Marquardt
algorithm (SPSS 2003). A weighting of 1/x has usu-
ally been used to improve the fit in the fines range,
1 1
0.1 1 10 100 with little penalty in the coefficient of determination
Mesh size, mm
value r2. The convergence for the three term func-
tion is usually rapid when xmax, xmax/x50 and b are cho-
Figure 8. Cone crusher product (CSS = 24 mm) of andes-
ite. Function parameters: x50 = 29.5 mm, xmax = 107 mm sen as the fitting parameters and reasonable initial
and b = 2.291. r2 = 0.9987. Data range 0.380 mm. guesses are used.
Ouchterlony (2003). In the case of the three term function the data
range is limited to values of x from 0.252 mm
and upward. Often the P(x) = 100 % value should
around b = 2.5, for the crushing curves around be excluded as the x100-value in this case may be a
b = 2.0. Very few b-values for blasting lie below 1.5. rather inexact estimate of the xmax-value.
High b-values in conjunction with high xmax-values If only linear least squares algorithms are avail-
tend to occur when the sieving curve looks more able, there are alternatives. One is to use the linear
like a Rosin-Rammler curve (Ouchterlony 2003). properties of the Swebrec function over a large part

6
of its range in P vs. x space and to do a linear curve fit In a recent paper (Ouchterlony 2009) it has been
y = A + B x over a suitable part of the sieving data, shown that expressing the fragment size distribu-
which contains x50. The values of x50 and s50 immedi- tion in terms of variables like x/xmax and parame-
ately follow as s50 = B and x50 = (0.5 A)/B. Equation 4 ters like x50/xmax and b, i.e. as P(x/xmax; x50/xmax, b) of
relates b and xmax through b = 4s50x50 ln(xmax/x50). With Equation 1, has advantages. Any explicit depend-
x50 and b0 = 4s50 x50 as known constants Equation 1 ence on material or explosive parameters is con-
may be rearranged to read fined to the equations that define the parameters
and unlike in the Kuz-Ram model they become
P(x) = 1/{1 + [1 + ln(x50/x) b0/b]b} or (6) functions of the same composite charge and bur-
den parameter s.
1/{1 + [1 + ln(x50/x)/ln(xmax/x50)]
b0 ln(xmax/x50)
},

with b and xmax respectively as the only undeter-


3 RELATION TO OTHER DESCRIPTIONS
mined parameter.
It is then easy to search for that xmax-value > x50
3.1 GGS exponent curves
(or b-value), which minimizes r2. xmax should prob-
ably also be chosen larger than the top stone size. The good fit of the five term Swebrec function
The corresponding b-value (xmax-value) and the to the whole range of a set of sieving data allows
Swebrec function P(x; x50, xmax, b) follow. If a good a smoother description of the Gates-Gaudin-
agreement between the function and the data Schuhmann or GGS exponent n used in commi-
is desired in the fines range, the weighting 1/x nution theory. The discrete expression becomes
mentioned above should probably be included. (Ouchterlony & Moser 2006a)
An example of the linear behaviour is given in
Figure 11 the approximate parameter values come ln(Pu Pl ) lnP d( lnP ) P ( x )
out quite close to the best fit ones of Figure 2. n= = = (7)
ln( x u x l ) ln x d( ln x ) P(x) x
The three basic parameter values may then be
used to improve the initial guesses of the param-
here subscripts u and l denote the upper and lower
eters for the five term extended Swebrec function
limits of a fragment size interval. Figure 12 shows
in a non-linear least squares fitting. The parameter
the GGS-exponent curve for the size distribution
a is frequently quite close to the value a = 1 and
in Figure 3, which also shows the corresponding
the convergence tends to be sensitive to the initial
GGS-exponent curve for the basic Swebrec func-
guess. The exponent c usually lies in the range
tion of Figure 2.
0.5 < c < 3. Sometimes a manual setting of c = 2
It is clear from these figures that the lower inflec-
to start with e.g. will give a better feeling for the
tion point in the lnP vs. lnx (or logP vs. logx) curve,
convergence process. The exact value of c has very
which shows up as a minimum in the GGS expo-
little influence on the r2-value. It is conceivable that
nent curve, is only described by the five term func-
c is related to the mineral species in the rock and
tion. The upper inflection point, which Equation 2
their properties, see the discussion about the GGS
minimum below.
GGS exponent or local slope in log-log diagram
Brarp Round 4
Residuals [4]

Discrete values, based on sieving data


2 2 1.5 Continuous, extended Swebrec
1 1
0 0 Continuous, basic Swebrec function
-1 -1
55 55

45 45 1
Passing, %

35 35

25 25

15 15 0.5

5 5
0 100 200 300 400 500
Mesh size, mm
0
0.1 1 10 100 1000
Figure 11. Linear least squares fit to Brarp round 4 Mesh size, mm .

data in range 90500 mm to get estimates of Swebrec


function parameters: x50 = 459.4 mm and s50 = 0.0010224 Figure 12. GGS exponent curves from sieving data and
1/mm. Remaining fit parameters xmax = 1525 mm and Swebrec curve fits, Brarp round 4. Ouchterlony & Moser
b = 2.254. C.f. Fig. 2. (2006a).

7
refers to, is displayed by both the basic and the blasting do not originate in the crushed zone around
extended Swebrec functions. the blast-hole (Svahn 2003, Reichholf 2003).
The GGS exponent curve has the property that
all sieving curves that are parallel in logP vs. logx
3.2 Comminution theory; NBC properties,
space collapse onto the same curve. Figure 13
simple OCS circuit and t10 approach
shows the GGS exponent curves for full-scale
and model-scale blasts in Brarp granitic gneiss. Steiners (1991, 1998) approach to mechanical com-
Ouchterlony and Moser (2006a) show similar sets minution is based on the concept that a material,
of GGS-exponent curves for Hengl amphibolite which is fractured under pure conditions exhibits
and for Vndle granite. a material specific Natural Breakage Characteris-
Figure 13 shows that the model- and full-scale tic (NBC). Some of the tenets are that
curves lie pretty much within their own bands and
Rock broken in the crushing and grinding
hence that a curve in each set is reasonably paral-
sub-circuits of an Optimum Comminution
lel with its own members. In the fines range, from
Sequence (OCS) has the steepest possible cumu-
about 30 mm and down, the sieving curves may be
lative fragmentation curve PNBC(x).
regarded as shifting upward in parallel in a log-log
When the sub-circuit product streams are classi-
diagram when the specific charge increases.
fied, their fragmentation curves are shifted verti-
Secondly, both sets of sieving curves seem
cally upward as the comminution progresses.
to have a minimum at the same fragment size,
When plotted in logP vs. logx space this becomes
x 1 mm. Thus logP vs. logx has an inflection
a parallel shift so that the local slope depends
point there, which depends on the value of c e.g.
basically only on the fragment size.
The same observation holds for Hengl amphibo-
lite, but with x 0.5 mm and for Vndle granite, This concept has been expanded to cover
but with x 1.5 mm. Recent investigations suggest model-scale blasting with different specific charges
(Grasedieck 2006) that this minimum is associated (Moser et al. 2000, 2003a). GGS-exponent curves
with the point where the fragments switch from that collapse onto each other, see above, is a way
containing basically several grains to being mono- to visualize this parallelism. Figure 14 shows that
mineralic, i.e. containing single grains or parts of the OCS comminution and model-scale blasting
single grains. The fast switchover may be explained curves for Hengl amphibolite are quite parallel in
by that fragments consist of two or only a few the range x = 0.110 mm. This also makes it pos-
grains are relatively seen weaker than other con- sible to extend Steiners energy register concept to
figurations (strm 2004). model-scale blasting (Moser et al. 2003b) but an
In addition to this grain size cut-off for small extension to full-scale blasting has not yet been
fragments, the universal fragmentation model achieved. It may never be because the full-scale
(strm & Ouchterlony 2004) describes two other curves are usually flatter than their model scale
cracking mechanisms. The small scale one is a counterparts, see Figure 13.
branching-merging of cracks that explains e.g. the The parallel shift property in logP versus logx
debris torn loose by propagating larger cracks, c.f. space reduces to the statement that the slope
the test results of Miklautsch (2002). This is one n = xP(x)/P(x) remains independent of some
explanation of why much of the fines generated in

100,0
GGS exponent
2.0
Round 1 BA 10-1, 100 mm
1.8 Round 4 BA 10-2, 100 mm
Round 5 BA 1-2, 190 mm
1.6 Round 6 BA 2-2, 190 mm 10,0
Comminution

1.4 Round 7 BA 1-1, 240 mm


Passing [%]

BA 9, 290 mm
1.2 BA 2-1, 290 mm BIT 2B (190/4)
BIT 66 (190/4)
BIT 3 (190/5)
1.0 BIT 79 (190/5)
BIT 2A (240/5)
0.8 1,0
BIT 21 (240/5)
BIT 8 (290/5)
0.6 BIT 78 (290/5)
Blasting BIT 73A (100/5)
BIT 73B (100/5)
0.4
Comminution
Curves
0.2
0,1
0,01 0,1 1 10 100
0.0 Screen size [mm]
0.1 1 10 100
Mesh size, mm
Figure 14. Comparison of fragmentation curves from
Figure 13. Continuous GGS exponent plots for full- and OCS comminution in Figure 3 and model-scale blasts
model-scale Brarp blasts. Ouchterlony & Moser (2006a). with the same amphibolite. Moser et al. (2000).

8
parameter that describes the shift. This is not met by that bypasses the crusher to that which passes
the basic Swebrec function itself but by its behaviour the crusher (x) gives the final product stream,
when x 0, xP(x)/P(x) b/[x ln(xmax/x)], which described by the probability density function
is independent of x50 (Ouchterlony 2003, 2005a). Pc(x). This gives the equation for the PDF of the
Though to retain the meaning of xmax as the maxi- simplified OCS circuit
mum allowable and x50 as the median fragment size,
make the substitutions xmax xc and x50 xmax. Pc(x) = F(x; x < a) + PNBC(x, a) a F(t)dt. (11)
Now xc denotes a characteristic size value for the
distribution, which lies outside the acceptable range An integration of Equation 11 would show that
of x-values, 0xmax. Then the following function has the mass flow balance is conserved. It also permits
NBC properties a couple of interesting observations, namely that i)
if F(x) is an NBC function, then so is Pc(x) and ii)
PNBC(x) = [ln(xc/xmax)/ln(xc/x)]b with the new if all feed material passes the crusher then Pc(x) is
x50 = xmax/(xc/xmax)2
1/b1
(8) a NBC function.
Finally, if one inserts the specific form of PNBC(x)
It describes a set of parallel shifted curves when given in Equation 8 into Equation 11, using it both
the value of xmax (formerly x50) is changed but xc for F(x) = PNBC(x, xmax) and PNBC(x, a) then Equa-
and b are kept constant. The function PNBC(x) is tion 11 is trivially satisfied. Thus one can say that
always concave upwards when x < xmax, which is the simplified description of an OCS sub-circuit,
the behaviour of the OCS sub-circuit curves in which led to Equation 11, is indeed compatible
Figure 7 except for very fine material. This makes with the statement that the asymptotic of the basic
it reasonable to call b an NBC exponent and to Swebrec function is a NBC function.
give it a physical meaning. The JKMRC (Napier-Munn et al. 1996) engi-
The same kind of reasoning could be used for neering oriented approach to crushing fragmenta-
the extended Swebrec function with regards to the tion describes comminution circuits and associated
dominating exponential term for super fine mate- individual crushing and grinding breakage func-
rial, 1/{(1 a)[(xmax/x 1)/(xmax/x50 1)]c} in Equa- tions. They use a matrix description of the product
tion 7, which obviously reduces to a set of parallel flow through the system and a one-family descrip-
curves when x 0 and c is kept constant. If the tion of breakage functions based on the t10 concept
substitutions xmax xc and x50 xmax are made like and expressed in terms of splines. t10 is that part of
above and the prefactor 1a removed one obtains a given size fraction of material subjected to a drop
Steiners version of NBC curves. The same kind of weight crushing test, which afterwards is smaller
reasoning doesnt appear to work for the combined than 1/10 of the original feed materials size.
logarithmic and exponential terms though. The proper non-dimensional version of the basic
Another interesting property of PNBC(x) is Swebrec function in Equation 1 with x/xmax = and
related to a simple OCS circuit, in which all mate- x50/xmax = 50 reads
rial x < a bypasses the crusher and the remainder
x a is crushed. See Ouchterlony (2003). Let F(x) P() = 1/{1 + [ln()/ln(50)]b} with max = 1 (12)
be the probability density function (PDF) of the
feed stream of material. The product coming out The coupling to B(x, t) is obtained if we set
of the crusher can be described by the probability = x/t. It has been fitted to AG/SAG mill break-
density function age function data in Table 6:1 (Napier-Munn et al.
t
1996) e.g. The whole t10 family of crusher curves of
(x) = a B(x, t) F(t)dt. (9) may be reduced to the following simple equation

here B(x, t) is the breakage distribution function of tn = 100/{1 + (100/t10 1) [log(n)]b(t10)}. (13)
the crusher, which gives the relative size distribution
after breakage of a particle of size t. One realises here n is the size reduction ratio. When n = 10 e.g.,
(Ouchterlony 2003) that B(x, t) with x t must be a log(n) = 1 and Equation 13 reduces to the straight
PDF function. According to the NBC concept, the line tn = t10 as it should. An approximate expres-
form of B(x, t) is independent of t. Thus sion for b(t10) was obtained from data fitting
t (Ouchterlony 2003, 2005a)
B(x, t) = b(x)/0 b(s)ds = b(x)/N(t) = b(x)/N(a)
= PNBC(x, xmax = a), (10) b = 1.616 + 0.02735 t10, (14)

as the crusher was described as producing no and the results are plotted in Figure 15. Note that
material larger than x a. Adding the stream an arbitrary vertical line connects points on the

9
tn-curves that together make up the P(x)-curve for (Sanchidrin et al. 2007) gives values in the range
the chosen degree of breakage. The full lines were 18 %. A general sieving curve behaviour like in Fig-
obtained using Equation 14, the dashed ones for ure 16 may increase the value further. All Grimshaws
n = 2 and 75 using the value b = 2.174 valid for (1958) sandstone shots have this behaviour but his
t10 = 20 %. It is seen that the variations in the b-value only limestone shot doesnt so more data is needed.
do make a difference. The data are well represented The extended, five parameter Swebrec function
by the isolines, differing by at most 34 %. The t10 cant fit the data in Figure 16 completely. Silvestrov
family of curves has thus received an extremely (2004, Figure 9) uses the Gilvarry (1961) distribu-
compact form compared to the large set of splines. tion to achieve a very good fit with a trimodal
A similar agreement with the other breakage version with seven terms. Each mode contains a
appearance function data in Napier-Munn et al. maximum of three size values that give the rela-
(1996) is expected. tive importance of linear, surface and volume flaws
respectively.
The data in Figure 16 covers an extremely wide
3.3 The ultra-fines range and other size
range, more than 5 orders of magnitude. An addi-
distribution functions
tion of an extra mode to the extended Swebrec
The range of dry sieving usually goes down to 75 distribution might work as well as the Gilvarry
or 63 m. Very few sieving data for rock fragmenta- function but this is beyond the scope of this paper,
tion that include smaller fragments are presented. which focuses on blast engineering.
One exception is Grimshaw (1958). He blasted
18 30 cylinders of Darley Dale sandstone
3.4 Robustness and parameter interdependence
and limestone with decoupled low VOD explosives
to measure the dust generation. The fine mate- The robustness of the parameter fitting of the
rial was measured with a sedimentation method. Swebrec function was described above. Another
The sieving curves go down to 2 m. An example example is the data from Svahn (2003). She blasted
with an extended Swebrec function fit is shown in three 300 mm cylinders of cement mortar with
Figure 16. 40 g/m PETN cord. The cylinders consisted of
The fit to the data is good down to 66 m, includ- 3 concentric layers of differently pigmented mortar;
ing the increasing slope below 1 mm. Below 66 m black, yellow and green. After blasting in a closed
the sieving curve again flattens, indicating a sur- container, the fragments were collected, sieved and
prisingly high amount of ultra-fine material. colour separated. The resulting curves are shown
Besides the respiratory aspect, this dust may rep- in Figure 17.
resent a large surface area that could account for The fragment size distributions for each layer,
some of the energy losses in a blast. The estimated Pb, Py and Pg are reproduced nicely by the Swebrec
fragmentation, i.e fracture surface energy in bench function, r2 < 0.9978 over the range 0.532 mm
blasts has earlier been estimated to be of the order (Ouchterlony 2003, 2005b). The composite fragment
0.1 %1% (Ouchterlony et al. 2003). A later revision size distribution Ptot was equally well reproduced by
the Swebrec function, even if the weighted average

tn, % passing
100 t2
b = 1,616+0,02735.t10
Residuals [4]

2 2
80
b = 2,174 t4
0 0
-2 -2

60
Mass passing, %

10 10
t10
40 1 1

t25
0.1 0.1
20 t50
0.01 0.01
t75 0.001 0.01 0.1 1 10 100 1000
Mesh size, mm
0
0 10 20 30 40 50
Breakage index t10, % .
Figure 16. Swebrec function fit to sieving data for
blasted Darley dale sandstone. Function parameters:
Figure 15. Mass passing isolines for constant degrees of x50 = 84.2 mm, xmax = 601 mm and b = 2.162. r2 = 0.9958.
size reduction, plotted as function of the breakage index Data range 0.002406 mm, 220 m (0.0020.02 mm)
t10. Full lines denote prediction by Equations 13 and 14. excluded in fit.

10
Mass passing, % Blasting, lab & full-size
100 Brarp lab tests
Slope s50 at x50, 1/mm
Hengl lab tests
Less Fines lab cylinders CP lab tests
s50 = 0.25/x500.75 NK stroma lab tests
NK krino lab tests
0.100 NK rev lab tests
NK fragment lab tests
10 Brarp bench blasts
Reef blasts (Claude C)
Quarry blasts (Claude C)
Dolomite blasts (Stagg)
0.010
full-size blasts
s50 = 0.12/x500.75
1 All, black+yellow+green
Swebrec function fit all
0.001
Either black, yellow or green
Individual color Swebrec fits
1 10 100 1000
Weighted ave. of indiv. fits x50, mm
0.1
0.1 1 10 100
Mesh size, mm
Figure 18. The relationship between the slope s50 at x50
.

and x50 itself for blasted rock. Ouchterlony (2005b).


Figure 17. Size distributions for blasted mortar cylinder
no. 2. Valid range 0.532 mm. The layer weights used were,
black inner layer 1 = 16.0 %, yellow middle layer 2 = 28.4 %
Table 1. Statistics, meanno. of tests std deviation, for
and outer green layer 3 = 55.6 %. Ouchterlony (2005b).
composite slope product s50 x500.75, including estimate of
Equation 17.

of the three individual functions doesnt add up to Burden q s50 x500.75 0.2 (0.25/B)0.25
the composite Swebrec function mathematically, i.e. Reference m kg/m3 1/mm0.25 1/mm0.25
1 Pb + 2 Py + 3 Pg Ptot with 1 + 2 + 3 = 1.
When the parameters x50, xmax and b of the Swe- Krist. (87) 0.125 0.40.6 0.23108 0.014 0.238
brec function are determined by curve fitting one Krist. (87) 0.250 0.6 0.18012 0.014 0.200
would usually expect them to be independent. An Krist. (90) 0.250 0.6 0.20821 0.021 0.200
Myk. (96) 0.250 0.6 0.18214 0.016 0.200
analysis (Ouchterlony 2005b) of the size distribu-
Krist. (87) 0.375 0.6 0.14004 0.006 0.180
tion data that were used to confirm the Swebrec Quarrying 3.0 0.51.0 0.125 0.107
function gives an interesting result when the slope Open pits 10.0 0.51.0 0.060.10 0.080
value s50 (1/mm with x50 in mm) at x50 is plotted
versus x50 in a log-log diagram. The data collapses
more or less on two parallel lines independent of
the specific charge. The lines are given by with increasing specimen size. Recent tests
(Ouchterlony et al. 2007) in a large open pit mine
s50 = 0.12/x500.75 for full-scale blasting and (15a) with 311 mm holes in a pattern with burden and
s50 = 0.25/x 0.75
for cylindrical models (15b) spacing B S = 7.7 9.7 m in double 15 m benches
50
gave the result that s50 x500.75 0.060.10. See chap-
Figure 18 shows these results for the blasting ter 4 below.
of amphibolite, dolomite, granite, granitic gneiss, In another recent development (Ouchterlony
limestone, syenite and quartzite. The relationship 2009) dimensional analysis is used to derive expres-
s50 1/x50 would e.g. have implied uniformity of sions for the parameters of the size distribution
the fragmentation curves. in terms of one dimensionless parameter, either
Kristiansen (1987) and Mykland (1996) have s q e B0.4 or expressions in which the exponents
blasted cubic blocks of concrete of side lengths 0.25, for e and B vary. s contains the specific charge
0.50 and 0.75 m with many different explosives. q (kg/m3), the explosive energy e (MJ/kg) and
Kristiansen et al. (1990) have blasted 0.25 m side the burden B (m). Equations 11a, 26 and 27 give
length cubic specimens of gabbro, granite, syenite x50/B /s where is a dimensionless geometrical
and quartz rock with two different kinds of dyna- shape factor and the exponent is a constant. Simi-
mite. The charge configuration and the specific larly by Equation 19a (Ouchterlony 2009) we could
charge was pretty much the same in all tests. write s50 B s/. This leads to
Statistics for the slope product s50 x500.75
(1/mm0.25) that would tend to be constant accord- s50 x500.75 (q e B0.4)3/4/( B)0.25 1/( B)0.25. (16)
ing to Equations 15ab are shown in Table 1.
The data in the second last column in Table 1 As a first approximation we have set the expo-
show a definite tendency to decrease significantly nent 3/4 = 0 as the left member appears to be

11
relatively independent of q, see Equations 15ab whether Equation 17 is compatible with the require-
above. Thus in 2D only a dependence on shape ment that harder blasting should give more fines.
through and on size through burden B remains. Equation 18 is important because it means that
Still as a first approximation, assume that the fragment size distribution from regular full-
doesnt vary much, and normalize the composite scale blasting within a very wide range could be
product value for the 0.5 m cubes to 0.2 m. Then described by two parameters only, x50 and xmax e.g.
we may write This is a very strong limitation to the possibilities
of altering the size distribution by ordinary means
s50 x500.75 0.2(Bref /B)0.25 with Bref = 0.25 m. (17) like changing the blast hole diameter etc. This will
be discussed further below.
The last column in Table 2 shows that this
expression captures the main variation in the com-
posite product s50 x500.75. Equation 17 may thus 4 PREDICTING THE SIEVING CURVE
replace 15a and 15b. BASED ON LIMITED DATA
Combining Equations 4 and 17 we obtain a
good starting value for b for normal full-scale A sieving curve for a full-scale blast is a rare occur-
blasting such as quarrying from rence but the wide range of good fit, which the
basic Swebrec distribution has, up to 3 orders of
b 0.4 (Bref/B)0.25 x500.25 ln(xmax/x50). (18) magnitude, opens a number of possibilities to
obtain such a curve. The first and obvious one
here Bref is a new reference burden of about 4 m. is predicting the complete fragment size distri-
It may be fine tuned for a specific full-scale appli- bution from either coarse material or fines data.
cation by a different choice of Bref. When field Examples of this are given in Figures 1920 for the
data are analyzed in more detail it is expected that Brarp bench blasts in granitic gneiss described by
i) the exponent 0.25 may change somewhat, ii) some Moser et al. (2003b). The relatively small rounds,
dependence on q and the size ratios in may emerge about 500 ton, were sieved in their entirety down
in Equations 17 and 18. It also remains to be seen to 90 mm and then quartered twice to obtain the
fines part of the sieving curves from lab sieving.
Figure 19 shows the example of fitting the
Table 2. Sieving data for barrel samples, sum (or basic Swebrec function to the coarse range data,
weighted ave.) and best fit Swebrec function. Param- x 90 mm, and extrapolating into the fines range
eters: x50 = 55.9 mm, xmax = 250 mm and b = 1.387. Range to obtain a prediction. Figure 20 shows how abso-
0.25180 mm. lute sieving data in the range 1 x 22.4 mm is
used for curve fitting and subsequent extrapolating
Barrel 1 2 3 4 14 Swebrec
into the coarse range. This extrapolation becomes
Mesh Pass. Pass. Pass. Pass. Pass. Pass.
mm % % % % % %
Mass passing, %
250 100.0 100.0 100.0 100.0 100.0 100.0 100
180 89.1 91.0 71.3 97.2 86.9 89.2 Swebrec function: x50 = 459 mm
125 79.3 83.2 47.7 89.9 74.7 74.5 xmax = 2260 mm and b = 2.81
90 69.0 69.3 38.4 81.6 64.1 63.0 x50
63 59.0 58.0 30.5 72.0 54.4 52.9 10
45 50.9 50.0 25.3 63.8 47.1 45.3
31.5 43.6 41.5 20.2 55.3 39.8 38.9
22.4 36.9 34.6 16.7 47.2 33.5 34.0 1
16 31.7 29.1 14.3 40.8 28.7 30.1
11.2 29.5 25.8 12.3 36.9 25.8 26.6 Rosin-Rammler function
8 27.0 22.9 11.0 32.9 23.2 24.0 x50 = 468 mm and n = 1.063
5.6 24.4 20.0 9.9 29.4 20.7 21.5 0.1
4 22.3 17.9 9.1 25.9 18.6 19.6
2 19.7 15.7 8.1 21.9 16.2 16.5
1 17.7 14.3 7.4 19.2 14.5 14.1
0.5 16.2 13.0 6.7 17.1 13.1 12.2 0.01
4
0.25 13.3 9.7 5.4 13.7 10.4 10.7 0.1 1 10 100 1000
0.125 7.2 4.4 2.7 7.3 5.3 9.5 Mesh size, mm .

0.063 3.7 2.2 1.4 3.7 2.7 8.5


0 0.0 0.0 0.0 0.0 0.0 0.0 Figure 19. Prediction of fine fractions for Brarp round
4 from sieving data in range 90500 mm. Comparison of
Mass, kg 350.3 367.9 358.5 322.9 1399.6 Swebrec and Rosin-Rammler extrapolations. Ouchterlony
(2005a).

12
Mass passing, % xmax = B Mass passing, %
100 100
Hengl full-scale blast
50 Data points in curve fit
Swebrec function: x50 = 460 mm Data points not used
x50 - points 1-50 and 1200 mm
xmax = 1800 mm and b = 2,41
- points 0.063-50mm
10 10 - points 100 and 1200 mm

1 1

0.1 0.1
0.1 1 10 100 1000 0.1 1 10 100 1000
Mesh size, mm .
Mesh size, mm .
.

Figure 20. Using Brarp round 4 sieving data in range Figure 21. Constructing a complete sieving curve for
122.4 mm plus estimate of xmax to make Swebrec predic- Hengl blast 420-4 through i) straight line interpolation,
tion of x50 and coarse fractions. Ouchterlony (2005a). ii) interpolation with basic Swebrec function and
iii) extrapolation with extended Swebrec function.

an interpolation if a reasonable estimate of xmax can


be used and then the prediction gets better. Both The curve fit based on the extended Swebrec
figures show that the extrapolation/interpolation function in Figure 21 is based on sieving data in
gives prediction values that lie quite close to the the range 0.063 x 50 mm. This produces pre-
measured ones. diction of xmax of 424 mm which is unrealistically
The statistical description of the fragmentation low but the curve fit in the fines range is well suited
is generally expected to become poorer as fragment for a calculation of the GGS-exponent curve.
size increases. Firstly, the number of fragments The curve fit based on the basic Swebrec func-
in each size class decreases rapidly. Secondly, the tion uses only the concave part of the sieving data,
Swebrec function approaches the 100 % line at a 1 x 50 mm. There are two alternatives, to use
tangent, which makes the curve fit more sensi- the xmax estimate of 1200 mm (interpolation) or not
tive. Thirdly, because of the finite size of a blast- to use it (extrapolation). It was found that using
ing round, the largest boulder represents a finite this xmax estimate gave the most reasonable value for
percentage of it. Thus the 100 % line will never be the slope product, s50 x500.75 = 0.131, which is quite
reached in practice and the measured x100 value is a close to the expected value 0.12 in Equation 15a.
rather inexact upper limit. The ensuing prediction of x50 became 316 mm.
This means that a prediction of the coarse range This method could also be used for a crusher with
fragmentation based on fines data will probably a known open side setting where the fines fractions
be more exact than the reverse procedure. This part of the product is sieved off and weighed using
could also serve as a justification for using the 1/x a belt scale. Figure 22 shows the example of a rotary
weighting in the curve fitting procedure. jaw crusher at the Vndle granite quarry. Fitting the
One such case is the blasting round at Hengl basic Swebrec function to the concave part of the
analysed by Ouchterlony and Moser (2006a). For fines data, 1 x 22.4 mm, plus a open side setting
a 2000 ton blast in amphibolite, the material for of xmax 300 mm produces the curve in the figure.
which x 100 mm was sieved and found to consti- For the muck piles of larger blasts even less data
tute 10.9 % of the round. From this material sieve would be obtained. Ouchterlony et al. (2006) con-
samples were taken. The xmax-value was estimated struct the complete sieving curves of four 1215000
to be about 1200 mm. Figure 21 shows three fitting ton muck piles in the Vndle granite quarry in the
alternatives, a linear interpolation from the fines following way. From each pile every 25th bucket
data to xmax in log-log space, the basic Swebrec from the backhoe was loaded on a dumper and used
function and the extended Swebrec function. to build a test pile of about 400 ton. From each pile
The linear extrapolation is a bad alterna- 100 ton were taken and run over grizzlies (//250 and
tive because it gives a GGS-exponent curve with 150 mm) and then sieved twice (#100 and 40 mm).
strangely constant value in the interpolation range From both sieving fractions, 0100 mm and
and the predicted x50-value of 448 mm is unrealisti- 040 mm, oil drum samples were taken for lab
cally large. sieving down to 0.063 mm. The sieving samples

13
Mass passing, %
Residuals

0.2 0.2
100
uncertainty of
-0.1 -0.1 90 boulder limit
correction for 400-500 mm
80
flakiness = 1.1
70
Mass passing, %

10 10 60
50 Test pile Round 1-L
sieved mtrl 0-100 mm
40 PowerGrid data
1 1
30 interpolated data
grizzly data
20 PowerGrid data - flakiness corrected
0.1 0.1 fixed points Swebrec distribution
0.05 0.5 5 50 500 10
on linear part Kuz-Ram distribution
Mesh size, mm

0 100 200 300 400 500 600


Figure 22. Using sieving data for granite for rotary jaw Mesh size, mm.

crusher product in range 0.522.4 mm plus open-side set-


ting estimate of xmax 300 mm to predict missing frac- Figure 23a. Constructed sieving curve for test pile from
tions. Swebrec function parameters: x50 = 77 mm and round 1L. Note linear scales for P and x. Ouchterlony
b = 2.33. et al. (2006).

showed little scatter and were more or less overlap- Mass passing, %
ping in the range x 22.4 mm. 100
The construction of the sieving curves was done 90 Test pile Round 1-L
as follows, see Figures 23ab. sieved mtrl 0-100 mm
80 PowerGrid data
interpolated data
1. The basic Swebrec function is relatively linear 70
grizzly data
within the range of 1060 % passing and the 60 - flakiness corrected
cumulative sieving data for 40 and 100 mm Swebrec distribution
50
mesh sizes lie well within this range. They are Rosin-Rammler
the fixed points of the construction. Data 40
Rosin-Ramler distribution
points for 45, 63 and 90 mm mesh sizes were 30 only good around x50 = 167 mm
taken from a straight line interpolation and the errors for fines and coarse mtrl
20
125 mm point taken as an extrapolated value on
the same line. These extra points give the central 10
data range a stronger weighting and they could
be left out. 0.01 0.1 1 10 100
Mesh size, mm
2. The lab samples from the 0100 mm fractions .

are assumed to be relatively uninfluenced Figure 23b. Constructed sieving curve for test pile
by the sampling procedure for the material from round 1-L. Note log scale for x. Ouchterlony et al.
x 40 mm. As 40 mm is missing in the lab (2006).
screen series, the 40 mm data point was inter-
polated from the 31.5 and 45 mm data points.
The lab sample data for all smaller screen sizes run and a new flakiness value chosen until the
were thus scaled by the factor P(40 mm). This minimum was found. This value turns out to be
makes the lab sample data a continuation of 1.10 or 1.15 and we chose the value 1.10.
the interpolation line into the fines region. 4. The amount of oversize that was put aside
The continuation is expected to be quite, if not during digging was estimated at 12 %, which
exactly smooth. means that the sieving curve of the test piles
3. A grizzly with a given gap lets through larger wouldnt differ very much from those of the cor-
fragments than a square mesh with the same responding muck piles. The representativity of
dimensions. To account for this the grizzly gap the test piles might contain larger errors so the
values 250 and 150 mm have been multiplied final adjustment to the oversize wasnt made.
with a flakiness factor. A lower limit is given by Furthermore a lower cut off size in the boul-
the interpolation line of step 1). We let the flaki- der piles is not sharp, it started at about 0.25 m.
ness factor be determined by the curve fit that A visual judgement gave 0.40.5 m as a reason-
gives the highest r2 value. i.e. a flakiness factor able largest fragment size in the test piles. For
was chosen, the curve fit program TableCurve2D the construction of the fragmentation curves we

14
chose 400 mm as a representative grizzly gap, with Split Desktop. The two systems were used to
which was then multiplied by the flakiness fac- obtain estimates of x50 and x80.
tor of step 3). This data point was also used in The Split On-line system had been calibrated
the curve fitting. using belt cuts from the primary, gyratory crusher.
These belt cut samples were also well described by
The predicted fragment size distribution for one the basic Swebrec function. See Figure 10. A closer
of the test piles is shown in Figures 23ab, in a lin- inspection of the images taken as the trucks dumped
ear diagram and in a semi-log diagram. revealed a number of shortcomings in the data col-
The test piles from which the 100 ton sieving lection (empty or partially filled trays e.g.), errors
samples were also photographed and digital image in the edge definition (splitting of large pieces and
analysis (Split Desktop) was used to obtain the cor- merging of small ones e.g.) and an appreciable
responding fragment size distributions, which were sensitivity to light conditions (time of day), which
quite different than the ones constructed using the undermined the credibility of the On-line data in
Swebrec function. A relatively thorough discussion absolute terms (Ouchterlony et al. 2007).
about the two methods, sieving and image analy- The corresponding Desktop images contained
sis, was made (Ouchterlony et al. 2006) and it was the same type of edge definition errors but were
concluded that the sieving data were more reliable. manually edited to correct for them. The correla-
One point was that the mathematical form of the tion between the On-line and Desktop data is not
fines description in the image analysis is erroneous, very high, see Figure 25 and the predicted x50 and
another that the digital edge definitions had to be x80 values differ quite much, see Table 3.
manually corrected, primarily for the splitting of It is practically impossible to calibrate either
large blocks. of the systems for the blast fragmentation. Split
A situation with even less information was at hand On-line has the theoretical possibility to shift ma-
in an investigation made in the Aitik copper mine. terial from one bin to another to emulate sieving
The blast in this case consisted of about 190 kton results but the sieving of a whole 180 ton truckload
(Ouchterlony et al. 2007) of gneiss and schist. has yet to be done. Split Desktop is more restricted
Oil barrels with fine material were taken from the but has an experience based fines correction factor
muck pile. The material was sieved in the size range ff, which quantifies the amount of fine material that
0.063 x 250 mm, see Table 2 and Figure 24. the system doesnt see. The factor ff will of course
The samples were quite well described by the vary with the position (of the photo) in a muck pile
basic Swebrec function. Now of course what por- and a sieving of the corresponding material is hard
tion of the muck pile these samples represent is not
known. The other accessible fragmentation data
was that from image analysis; an installation of x50,online, mm
Split On-line where the trucks dumped their loads dig inside round
into the primary crusher and images taken of the 1000 dig outside round
surface of the truck trays after loading for analysis mean x80, online
best fit slope
2best fit slope
Mass passing, % 750
100 best fit slope
90 Sieve samples round 4141-2
Barrel 1 x50,online =
80 Barrel 2 0.389x50,desktop
Barrel 3 500
70
Barrel 4 x50,online = 203 mm
60 Ave. barrels 1-4
50
250
40 0,5-125 mm
30
20
0
10 0 250 500 750 1000
0 x50,desktop, mm
0.1 1 10 100
Mesh size, mm .
Figure 25. Correlation between Split On-line and Desk-
top data for x50 values for contents in same truck tray. The
Figure 24. Lab sieving data for a total of 1400 kg of trend in the best fit line is not significant (Ouchterlony
barrel material. Ouchterlony et al. (2007). et al. 2007).

15
Table 3. Split data and Swebrec function parameters for Mass passing, %
100
muck pile curves in Fig. 18. DT-0 means Desktop program
x50+x80, Desktop ff = 0 %
with fines factor ff = 0%. OL means On-line program. 90
x50+x80, Desktop ff = 20 %
80
Mtrl Swebrec param. Primary crusher product
70 x50+x80, On-line data
Anal. x50 x80 125 x50 xmax s50 x500.75
meth. mm mm % mm mm b r2 1/mm0.25 60

DT-0 585 990 15.0 587 1536 1.772 0.9999 0.094 50


DT-20 425 850 22.4 426 1565 1.834 0.9998 0.078 40
OL 170 600 43.7 175 2922 2.462 0.9987 0.060
30
20
10
Residuals [6]

0.6 0.6
0.3 0.3 0
0 0 1 10 100 1000
-0.3 -0.3
Mesh size, mm .
Mass passing, %

Figure 27. Alternative constructed sieving curves for


10 10 muck pile rock (lower three curves) and estimated siev-
ing curve for primary crusher product. Ouchterlony et al.
(2007).
1 1

0.1 0.1
0.02 0.2 2 20
Mesh size, mm
200 2000 measurements on the gyratory crusher could prob-
ably be used because the upper dashed, On-line
Figure 26. Constructed sieving curve from ave. barrel curve predicts that 2040 % of the feed is crushed
data in Table 3 (0.5125 mm) and Split Desktop data and the lower dashed, Desktop curve predicts that
DT-20 for x50 and x80. Even weights. Data for Swebrec fit 3565 % or twice the amount is crushed.
given in Table 3. Table 3 also contains values for the slope factor
s50 x500.75 that lie in the range 0.060.10. See also
section 3.4 above. Equation 17 suggests that 0.08
to do. This makes the x50 and x80 values obtained would be reasonable value so from this point of
from image analysis relatively unreliable in abso- view none of the three alternatives are excluded.
lute terms. More Aitik rounds have been monitored. In a
All three sets of x50 and x80 values in Table 3 MSc report (Demenegas 2008) blast 5162 of about
plus the averaged lab sample data in the range 400 kton was studied. It had been divided in two
0.5 x 125 mm with an unknown fines multiplier halves. One domain had the normal specific charge
were used as a basis for the fitting procedure. For of 1.08 kg/m3 (B S = 7.5 9.5 m) and the other
each set of x50 and x80 values the multiplier value 1.35 kg/m3 (B S = 6.5 8.1 m). The round was
that maximized r2 was chosen. Figure 26 gives part of a test series in which the goal was to improve
one example. The associated parameter values are the throughput in the primary autogenous mill.
given in Table 3, the curves are shown in Figure 27 Drilling, charging and fragmentation were
together with an estimate of the sieving curve of closely monitored like in the previous investiga-
the crusher product. tion. The On-line system at the primary crusher
Table 3 and Figure 27 show that the amount didnt work during the hauling so the investigation
of fine material x 125 mm predicted by the dif- took a different direction.
ferent curve fits is quite wide, from 15 to 44 % of More than 250 photos of truck loads were
the material fed to the crusher. In order to narrow taken and the 60 best ones chosen for evaluation
the range further, the estimated sieving curve of with Split Desktop, 30 for each domain. Two truck
the crusher product is also shown in Figure 27. The samples were taken from specified positions in
crushers feed curve can of course never cross the each domain, run over a 100 mm grizzly and quar-
product curve but how close can it lie? tered until two 350 kg samples remained. These
The report by Ouchterlony et al. (2007) discusses were sieved.
this matter, looks at the top size data etc. and As before, Split Desktop data for x50 and x80
concludes that the real fragment size distribution with ff = 20% (DT-20) were obtained from the
probably lies somewhere between the two dashed photos and combined with the lab sieving data
curves, with a fine material fraction of 22 to 44 %. in the range 190 mm. The data in Table 4 were
To choose between the alternatives, power draw obtained. Note that the value x50 = 233 mm places

16
Table 4. Fragmentation data (DT-20) for Aitik round Mass passing, %
100
5162, given as mean std deviation. N = normal specific
charge, H = high. 90 Crusher model data
Blasted, Split Desktop
80 Crusher model output
q x20 x50 x80 x100
70 Measured Split On-line
Round kg/m3 mm mm mm mm
60
5162-N 1.08 70 19 233 47 463 108 838 190
50
5162-H 1.35 61 14 200 54 419 126 802 214
40
30
20
the fragmentation curve midway between the two
dashed lines in Figure 27. 10
A statistical analysis showed that of these sets of 0
data only the x50-values are significantly different. 1 10 100 1000
Mesh size, mm
Even if On-line data from the truck dumping into .

the primary crusher wasnt available, On-line data Figure 28. Crusher model product () for measured
from the product belt were for the times when the blast fragmentation () at normal specific charge com-
blasted rock was dumped. It was decided to apply pared with measured product (x). Demenegas (2008).
the crusher model developed by Bergman (2005)
to different input or feed distributions; the one just
constructed, the Kuz-Ram model with the Swebrec Mass passing, %
function and the CZM model (Kanchibotla 1999). 100
The results for the measured feed are shown in 90 Crusher model data
Figure 28 and those for the calculated CZM feed in Blasted, CZM input
80 Crusher model output
Figure 29. The Kuz-Ram cum Swebrec function feed
70 Measured Split On-line
gave a curve nearly identical to the one in Figure 28.
Figures 2829 permit a couple of observations. 60
Firstly, the CZM model feed, whose coarse part 50
looks quite like the measured fragmentation has
40
vastly more fine material than the one measured.
It has more fines even than the measured crusher 30
output. This is of course impossible. It may be that 20
a better calibration of the CZM model parameters 10
would remove this contradiction but the rather
abrupt change in slope, in Figure 29 at P = 70 %, is 0
1 10 100 1000
also unrealistic and it is built-in into the model. Mesh size, mm .

Secondly, the crushing doesnt seem to create


any progeny smaller than about 3050 mm. This is Figure 29. Crusher model product () for calculated
highly suspect when one compares with a breakage CZM feed () at normal specific charge compared with
function description like in Equation 12 and Fig- measured product (x). Demenegas (2008).
ure 15. Thus it is quite probable that the crusher
model used under-predicts the amount of fine
material created during the crushing process. large however that it would definitely be advisable
The gap between measured and predicted to use crusher effect measurements to support the
amounts of fine material is a factor of two at analysis.
75 mm and it increases with decreasing fragment A final note, even if criticism has been voiced
size. This too supports the conjecture above about about the image based fragmentation systems used
fines under-prediction. above there is no reason to believe that they give
We have shown how one can make predictions any worse or any better fragmentation data than
of the sieving curves of blasted and crushed mate- any other commercial systems in widespread use.
rial, using the basic Swebrec function but based on
less and less information as the size of the muck
pile increases. The less input information that is 5 DESIGN CURVES FOR BLAST
used naturally, the lower the expected accuracy FRAGMENTATION
of the prediction. We have also shown how these
data could be applied to studies of crushing in pro- The final step in fragmentation prediction is to be
duction conditions. The uncertainties are now so able to predict how the parameters x50, xmax and

17
possibly b depend on blast design factors like spe- Table 5. Back calculated rock mass constant A from
cific charge q (kg/m3), explosive weight strength s constructed sieving curves. L = low specific charge,
and the rock properties. This is usually called a frag- H = high.
mentation model and several exist (Cunningham
Round 1-L 1-H 2-H 2-L
1983, 1987; Kanchibotla et al. 1999 & Djordjevic
1999 e.g.). The main crux of the models lie in quan- x50 (mm) 167 141 190 180
tifying the influence of the rock properties and q (kg/m3) 0.52 0.63 0.68 0.49
many authors have modified the complicated rock sANFO (%) 85 85 85 85
mass factor A of the Kuz-Ram model (Cunningham Q (kg/hole) 86 78 93 77
1987) to suit their own needs. A 3.89 3.89 5.41 4.07
In a more limited situation, in a single quarry or
mine where the rock conditions are relatively con-
stant or well known domains exist, there is still the and 1-H, that value was used in the construction of
need to predict how fragmentation changes when the design curves, c.f. Equation 15a (Ouchterlony
specific charge is changed, either by expanding or et al. 2006). The idea that xmax would be unaffected
shrinking the blast hole pattern, by altering the by the specific charge was further rejected because
blast hole diameter or some other choice. it led to the unrealistic result that blasting with a
Ouchterlony et al. (2006) developed such design lower specific charge would for sufficiently small
curves for the Vndle granite quarry. The purpose fragments yield a higher portion of fines than blast-
of the work was to provide the quarry owner with a ing with a higher specific charge. Instead it was pro-
tool that would tell him how to change the blasting posed that the rock in the muck piles for sufficiently
so that the desired feed into the primary crusher small fragments obeys the NBC characteristic of
was obtained. that material, just as in model-scale blasting and
The fragmentation in four blasts was monitored. OCS comminution (Moser et al. 2000 & 2003a).
The specific charge was changed in the range 0.49 This amounts to setting b constant. Working
to 0.68 kg/m3. This was accomplished by expanding backwards from a b-expression like Equation 18
the original pattern B S = 3 4 m to 3.2 4.25 m with Bref constant, the result became an equation
and by shrinking it to 2.8 3.75 m in production similar to the Kuz-Ram one in Equation 19. It
blasts with four rows of 90 mm holes charged reads xmax 500 (0.55/q)0.6 and describes how xmax
with bulk emulsion and a shallow V-initiation with changes in the range 0.3 < q 0.8 kg/m3.
the Nonel Unidet system. The resulting design curves are shown in Fig-
The Kuz-Ram prediction equation for the ure 30 and the corresponding Swebrec function
median fragment size reads (Cunningham 1987) parameters are given in Table 6. The two middle
curves, for which q = 0.5 to 0.6 kg/m3, are basically
x50 = A Q1/6 (115/sANFO)19/30/q0.8, (19) what was measured in the quarry. The other curves
are extrapolations. Together they make it possible
where x50 is given in cm, Q is the charge weight per to predict the effect of changing the blast hole pat-
hole (kg) and sANFO is the weight strength of the tern on all fractions in the muck pile.
explosive relative to ANFO (%). The rock mass As an example take the road base material at
factor A in the model is determined by a fairly Vndle, the fraction x 32 mm, that is produced
complicated equation based on rock and jointing and sieved off directly after the primary crusher.
properties. It has a range from about 1 to 20 and is The quarry sells 2030 % of its production as road
easily the most influential factor in Equation 19. base material and the crusher settings and screen
An independent assessment of the Vndle con- size are chosen accordingly. It was found that the
ditions gave the value A = 4.76. The measured crusher and the blasting contribute roughly equal
fragmentation is more compatible with a value of amounts. Figure 30 put the amount of such mate-
A 3.9, see Table 5 and the following discussions. rial in the muck pile as 1314 %. The figure also
As Cunningham (2005) accepts a correction factor says that by changing the specific charge in the
C(A) = A/A in the range 0.5 C(A) 2 the value range 0.3 < q 0.8 kg/m3 we would expect the road
A/A 3.9/4.76 = 0.82 comes quite close to the base material in the muck pile to lie between 10
independent assessment. This gave confidence to and 20 %.
the use of Equation 19 to describe the median frag- This reasoning could be repeated for any muck
mentation at Vndle but with factor 0.05 instead pile fraction, like x30 or x80, and the design curves thus
of 0.06 in A. provide a help for the quarry blast management.
To obtain the design curves, equations or values The design curves in Figure 30 is repeated
for xmax and b are still needed. As the composite with linear axes in Figure 31. For each the tan-
parameter s50 x500.75 0.125 for the two most reli- gent line at x50, Pt(x) = 100 [0.5 + s50 (x x50)] has
able of the constructed Vndle sieving curves, 1-L been plotted as a solid straight line. The lines are

18
Mass passing, % Mass passing, %
100 100

90 Blasting at Vndle 90 0,8 kg/m3


fragmentation when q varies
0,8 kg/m3 0,7 kg/m3
80 xmax = varies, b constant 80
95% slope
0,7 kg/m3
70 70 q = 0,6 kg/m3
q = 0,6 kg/m3 60
60
50 50
0,5 kg/m3 0,5 kg/m3
40 40
0,4 kg/m3 0,4 kg/m3
30 30
0,3 kg/m3 0,3 kg/m3
20 20
10 10

1 10 32 mm 100 1000 0 100 200 300 400 500 600


Mesh size, mm . Mesh size, mm .

Figure 31. Blast design curves of Figure 30 in diagram


Mass passing, % with linear axes. Tangent lines at x50 added to illustrate
100
range of linear behavior of Swebrec function.
0,8 kg/m3
50 0,7 kg/m3
q = 0,6 kg/m3 The Kuz-Ram prediction equation for x50
0,5 kg/m3 worked in the Vndle quarry. The q-exponent
0,4 kg/m3 = 0.8 was roughly right, the A-value too. Table 4
0,3 kg/m3 says that raising the specific charge q from 1.08
10
to 1.35 kg/m3 at Aitik lowered x50 from 233 mm
to 200 mm. This corresponds to = ln(200/233)/
Blasting at Vndle ln(1.08/1.35) 0.68, which is not too far from
fragmentation when q varies 0.8 considering that only two domains were
xmax = varies, b constant
monitored.
The weight strength factor wasnt tested at
1 Vndle, nor was the size factor Q1/6 as the charge size
1 10 32 mm 100 1000
Mesh size, mm .
was constant. In a later paper (Ouchterlony & Moser
2006a) it was concluded both that the size factor
Figure 30. Blast design curves; the effect on fragmenta- Q1/6 couldnt fully explain the size dependence of
tion of changing the specific charge at Vndle by expanding blast fragmentation and that there also appeared
or contracting the drilling pattern when the hole diameter to be some effect of blast size on A and even on the
is kept constant. Upper graph: linear scale on y axis. Lower q-exponent.
graph: log scale on y axis, showing parallel NBC character
of curves in fines region. Ouchterlony et al. (2006).
Recent developments (Ouchterlony 2009) shed
new light on this part. One observation is that
the quantities xmax/B and b to a first approxima-
Table 6. Calculated values of Swebrec parameters for tion both depend on exactly the same material and
Vndle design curves. s50 x500.75 = 0.125 (1/mm)0.25 and explosives parameters as x50/B through the com-
b = 2.03. bination s q e B0.4, see Ouchterlony (2009)
Equations 11b and 18. This is e.g. respected by
q (kg/m3) 0.3 0.4 0.5 0.55 0.6 0.7 0.8 Equation 18 above. Had this been known during
the Vndle tests then one or both of the choices for
x50 (mm) 259 205 172 159 148 131 118
xmax and b made above would probably have been
xmax (mm) 710 600 525 500 475 435 405
a bit different.
Another observation is that the equation x50/
B /s is an extension of Equation 19 in which
excellent approximations of the sieving curves the Q-dependence has been replaced by a combina-
within the range 20% P(x) 60% for all curves tion of explosive energy e (MJ/kg) and burden B,
but one, q = 0.3 kg/m3. C.f. Figure 4 and step 1 of see Ouchterlony (2009) Equation 30a, and in which
the construction procedure in chapter 4. It is also bench height H and hole spacing S also enters. The
apparent that by lowering the slope of the tangent result is
for the latter curve the same range of validity could
be achieved. x50/B = A [(H/B) (S/B)/k]1/3/(q e B0.4). (20)

19
The choice = 19/30 0.633 emulates the 100
Ave. fragment size x50, mm
Kuz-Ram exponent for the explosive strength aggr.& plaster debris
in Equation 19. If the blasting pattern (BSH) x50 = 52.4/q0.76
is kept constant and q varies because of chang- 50
ing bore-hole diameter would also express the free cylinder
q-dependence, otherwise not. x50 = 18.5/q0.84
At Vndle and Aitik though, H and S/B were
kept constant and the blasting pattern S B was aggr.debris
x50 = 38.1/q0.40
changed for a given bore-hole diameter to change
Free
as the specific charge. Using Q/q = BSH to elimi- power fit
nate all explicit dependence on B retrieves a factor Aggregate debris
10
Q1/3 0.2, in which the exponent becomes 0.21 and power fit
a q-factor in the denominator q0.8 + 1/3 in which the Aggr. & plaster debris
power fit
exponent becomes 0.84.
Thus the more general Equation 20 has the capac- 0.1 1 5
ity to explain in better detail than Equation 19 what Specific charge, kg/m3
happens to fragmentation when various parameters
in the blasting pattern are changed. The work by Figure 32. Effect of debris confinement on the
Ouchterlony (2009) doesnt reconcile the difference fragmentation of mortar cylinders blasted at specific
between how q is defined in bench blasting and in charges in the range 0.22.6 kg/m3. Johansson et al.
(2007).
the small scale cylinder tests though.
So far the results presented have been obtained
for blasting towards a free surface at which the
incident compressive wave is reflected as a tensile in the confined interior rows might superficially
wave, which helps the breakage. Johansson (2008) be considered 3D blasts crater like rather than 2D
and Johansson et al. (2007, 2008) present model ones. This would require some changes in Equa-
blasting data for mortar cylinders in which the frag- tion 20 (Ouchterlony 2009). The bore-holes in the
mentation under confinement by aggregate debris interior of such a blast have in reality not only a
was studied. The cylinders were 140 280 mm free top surface but also side surfaces confined by
and had central, through-going holes in which blasted rock. It might thus be worthwhile to test
decoupled PETN cord of different strengths were the approach above in which the effect of the con-
detonated. The coupling ratio was kept reasonably finement is characterized by an effective specific
constant. charge.
Four debris confinements were tested and 160
free and confined shots made. The sieving curves
followed the Swebrec distribution quite well also 6 CONCLUSION
for the confined shots, except close to the critical
specific charge (burden). This paper summarizes the last decade of research
Figure 32 shows some results. The debris con- at Swebrec and its predecessor SveBeFo (Swedish
finement makes the fragmentation considerably Rock Engineering Research), work in which the
coarser. Denote the elastic properties of the mortar Less Fines project had an important part. This
by m (density) and cm (P-wave velocity) and of the has led to a new fragment size distribution with
aggregate a and ca respectively. Next denote the an unexpected ability to reproduce sieving curves
impedances Zm = m cm and Za = a ca. Then the of all kinds of rock and concrete/mortar that has
amplitude of the reflected wave when a 1D rectan- been blasted or crushed in almost every conceiv-
gular wave arrives at a mortar-aggregate interface able way.
is given by multiplying the incident wave by the The Swebrec function is described in some detail
reflection coefficient Re = (Za Zm)/(Za + Zm). The and examples of curve fits are given as is some
energy content of the reflected wave is obtained by advice about the fitting procedure. Its relations
multiplying by Re2. with other comminution concepts such as t10 from
In the final regression analysis (Johansson 2008), JKMRC are explored. The result that the param-
the effect of the debris confinement was quite well eters of the Swebrec function are not independent
explained by multiplying the specific charge q or is of special importance and a more general ver-
its energy q e by Re2 in an equation like 19. This in sion of this dependence is presented.
a sense makes q Re2 an effective specific charge. How to use the Swebrec function for full-scale
Full-scale blasts, like large open pit rounds, blasts is described and especially for determining
where the material movement changes from hori- the fragment size distribution for large blasts where
zontal in the first few rows to near vertical swelling the available information is very limited.

20
The final section treats design curves, in short 3rd EFEE World Conf. on Explosives and Blasting,
determining developed versions of the Kuz-Ram Brighton, UK, 1316 September, pp. 201210. Reading,
model which can predict the changes in fragmenta- UK: European Federation of Explosives Engineers.
tion that are caused by changes in specific charge Demenegas, V. 2008. Fragmentation analysis of opti-
mized blasting rounds in the Aitik mine. MSc thesis
and other geometrical parameters of a bench blast. 2008: 071. Lule: Lule Univ. Techn.
Some recent results finally suggest that the effect of Djordjevic, N. 1999. Two-component model of blast
confining debris at the free face could be accounted fragmentation. In Symp. series S21, Proc. 6th Int.
for by multiplying the specific charge by the wave Symp. on Rock Fragmentation by Blasting, Johannes-
reflection factor of 1D wave dynamics. burg, South Africa, 812 August, pp. 213219. Johan-
nesburg: SAIMM.
Franklin, J. & Katsabanis, T. (eds). 1996. Measurement of
ACKNOWLEDGEMENT Blast Fragmentation. Rotterdam: Balkema.
Gilvarry, J.J. 1961. Fracture of brittle solids. I. Distrib-
The author wants to give special thanks to profes- tion function for fragment size in single fracture (theo-
sor Peter Moser of MU-Leoben with whom he retical). J. Appl. Physics 32(3): 391399.
Grasedieck, A. 2006. The natural breakage characteris-
has long collaborated in matters of blast fragmen- tics (NBC) of rocks in blasting. PhD thesis, 217 pp.
tation. This paper is based on a manuscript pre- Leoben, Austria: Montanuniv., Dep. of Mining Eng. &
ceding a paper in German (Ouchterlony & Moser Mineral Economics.
2006b). An updated version was presented as a ple- Johansson, D. 2008. Fragmentation and waste rock com-
nary lecture at the ISEM 2008 in Tokyo, April 24, paction in small-scale confined blasting. Licentiate
2008. This paper is a second update. thesis 2008: 30. Lule: Lule Univ. Techn.
MU-Leoben financed the lab scale tests with Johansson, D., Ouchterlony, F. & Nyberg, U. 2007. Blast-
the Brarp and Vndle specimens with internal ing against aggregate confinement, fragmentation and
resources. The EU financed the Hengl part of this swelling in model scale. In P. Moser et al. (eds.), Proc.
4th EFEE World Conf. on Explosives and Blasting,
work through the Less Fines project, EU project pp. 1327. Reading: EFEE.
GRD-2000-25224 and Hengl Bitustein AG sup- Johansson, D., Ouchterlony, F., Edin, J., Martinsson, L. &
plied the test site. SveBeFo and supporting Swed- Nyberg, U. 2008. Blasting against confinement,
ish industries financed the work in the Brarp fragmentation and compaction in model scale. In
quarry. Emmaboda Granit AB supplied the test H. Schunnesson & E. Nordlund (eds.), MassMin
site. The MinBaS Development Programme, the 2008, Proc. 5th Int. Conf. & Exhib. on Mass Mining,
Swedish Geological Survey, the Swedish Mineral pp. 681690. Rotterdam: Balkema.
Processing Research Association (MinFo) and the Kanchibotla, S.S., Valery, W. & Morell, S. 1999. Mod-
Swedish Energy Administration financed the work elling fines in blast fragmentation and its impact on
crushing and grinding. In C. Workman-Davies (ed.),
in the Vndle quarry. Swerock AB supplied the Proc. Explo 1999 Conf., pp. 137144. Carlton, VIC:
test site. The site owners have made many in-kind AusIMM.
contributions. King, R.P. 2001. Modeling & simulation of mineral
processing systems. Boston: Butterworth-Heineman.
Kristiansen, J. 1987. Utvikling av en ny parameter for
REFERENCES karakterisere sprengstoffers evne til fragmentere
fjellbelyst ved sprengningsforsk i betong. Cand Sci
strm, J.A. 2004. Personal communication. Thesis, 87 + 73 pp. Univ. Oslo: Inst. Geol. In Norwegian.
strm, J.A. & Ouchterlony, F. 2004. Universal fragmen- Kristiansen, J., Kure, K., Vestre, J. & Bergqvist, I. 1990.
tation in D dimensions. Physical Rev Letters 52(24): An investigation of heave and fragmentation related
245506-1 to 4. to explosive properties. Proc. 3rd Int. Symp. on Rock
Bergman, P. 2005. Optimisation of fragmentation and Fragmentation by Blasting, pp. 8390. Parkville, VIC:
comminution at Boliden Mineral, Aitik operation. AusIMM.
Licentiate thesis 2005: 90. Lule: Lule Univ. Techn. Kuznetsov, V.M. 1973. The mean diameter of the frag-
Blair, D.P. 2004. Curve-fitting schemes for fragmentation ments formed by blasting rock. Soviet Mining Sciences
data. FragblastInt. J. Blast. Fragment. 8(3): 137150. 9(2): 144148.
Cunningham, C.V.B. 1983. In R. Holmberg & A. Rustan Maripuu, R. 1968. Investigation of sieving analysis and
(eds.), Proc. 1st Int. Symp. on Rock Fragmentation by fragment shape of SLC rock from LKAB, Kiruna. Dipl.
Blasting, Lule, Sweden, 2226 August, pp. 439453. Thesis. Stockholm: Royal Inst Techn. In Swedish.
Lule: Lule Univ. Techn. Miklautsch, A. 2002. Experimental investigation of the
Cunningham, C.V.B. 1987. Fragmentation estima- blast fragmentation behaviour of rock and concrete.
tions and the Kuz-Ram modelfour years on. In Dipl. work: 161 pp. Leoben, Austria: Montanuniv.,
W.L. Fourney & R.D. Dick (eds.), Proc. 2nd Int. Symp. Dep. of Mining Engng & Mineral Economics.
on Rock Fragmentation by Blasting, 2326 August, Moser, P. 2003. Less fines production in aggregate and
pp. 475487. Bethel, CT: SEM. industrial minerals industry. In R. Holmberg (ed.),
Cunningham, C.V.B. 2005. The Kuz-Ram fragmenta- Proc. EFEE 2nd World Conf. on Explosives & Blasting
tion model20 years on. In R. Holmberg (ed.), Proc. Techn, pp. 335343. Rotterdam: Balkema.

21
Moser, P. 2005. Less Fines in aggregate and industrial Ouchterlony, F. & Moser, P. 2006b. Die Swebrec Funktion:
minerals productionResults of an European research Eine neue Korngrenverteilungsfunktion und ihr
project. In R. Holmberg (ed.), Proc. 3rd EFEE World praktischer Einsatz in der Sprengtechnik. BHM, Berg-
Conf. on Explosives & Blasting, Brighton, UK, 1316 und Httenmnnische Monatshefte 151(10): 389404.
September, pp. 567574 England: EFEE. Ouchterlony, F., Nyberg, U., Bergman, P. & Esen, S.
Moser, P., Cheimanoff, N., Ortiz, R. & Hochholdinger, R. 2007. Monitoring the Blast Fragmentation of Boli-
2000. Breakage characteristics in rock blasting. In den Minerals Aitik Mine. In P. Moser et al. (eds.),
R. Holmberg (ed.), Proc. 1st EFEE Conf. on Explo- Proc. 4th EFEE World Conf. on Explosives & Blasting,
sives & Blasting, pp. 165170. Rotterdam: Balkema. pp. 4762. EFEE.
Moser, P., Grasedieck, A., Arsic, V. & Reichholf, G. Ouchterlony, F., Nyberg, U., Olsson, M., Bergqvist, I.,
2003a. Charakteristik der Korngrssenverteilung von Granlund, L. & Grind, H. 2003. On the energy balance
Spreng-hauwerk im Feinbereich. Berg- und Htten- of production blasts at Nordkalks Klinthagen quarry.
mnnische Monatshefte 148: 205216. In R. Holmberg (ed.), Proc. EFEE 2nd World Conf.
Moser, P., Olsson, M., Ouchterlony, F. & Grasedieck, A. on Expl. & Blasting Techn., pp. 193203. Rotterdam:
2003b. Comparison of the blast fragmentation from Balkema.
lab-scale and full-scale tests at Brarp. In R. Holm- Ouchterlony, F., Olsson, M., Nyberg, U., Andersson, P. &
berg (ed.) Proc. EFEE 2nd World Conf. on Explosives & Gustavsson, L. 2006. Constructing the fragment size
Blasting Techn., pp. 449458. Rotterdam: Balkema. distribution of a bench blasting round, using the new
Mykland, J. 1996. En teoretisk og eksperimentell under- Swebrec function. Proc. 8th Int. Symp. on Rock Frag-
skelse av sprengstoffenes egenskaper relatert till mentation by BlastingFragblast 8, Santiago, Chile,
fragmentering och kast av bergarter. Cand Sci Thesis, 711 May, pp. 332344. Santiago: Editec S.A.
159 + 83 pp. Univ. Oslo: Inst. Geol. In Norwegian. Reichholf, G. 2003. Experimental investigation into the
Napier-Munn, T.J., Morrell, S., Morrison, R.D. & characteristic of particle size distributions of blasted
Kojovic, T. 1996. Mineral comminution circuitsTheir material. PhD thesis, 223 pp. Leoben, Austria: Monta-
operation and optimisation. JKMRC Mon. Series Min- nuniv., Dep. of Mining Engng & Mineral Economics.
ing and Mineral Proc. Brisbane, QLD: JKMRC. Sanchidrin, J.A., Segarra, P. & Lpez, L.M. 2007. On
Ouchterlony, F. 2003. Bend it like Beckham or a wide- specific surface and fragmentation energy of blasted
range yet simple fragment size distribution for blasted rock. In P. Moser et al. (eds.), Proc. 4th EFEE World
and crushed rock. EU project GRD-2000-25224. Less Conf. on Explosives & Blasting, pp. 6373. EFEE.
Fines project int. techn. rpt no. 78. Leoben, Austria: Sanchidrin, J.A., Segarra, P., Ouchterlony, F. & Lpez, L.M.
Montanuniversitt. 2009. On the accuracy of fragment size measurement by
Ouchterlony, F. 2004. Influence of blasting on the size image analysis in combination with some distribution
distribution and properties of muckpile fragments, a functions. Rock Mech Rock Engng 42(1): 95116.
state-of-the art review. Rpt project P2000-10: Energy Silvestrov, V.V. 2004. Application of the Gilvarry distri-
optimisation during comminution. Stockholm: Swed- bution to the statistical description of fragmentation
ish Min. Res. Org. MinFo. of solids under dynamic loading. Comb. Expl. & Shock
Ouchterlony, F. 2005a. The Swebrec function: link- Waves 40(2): 225237.
ing fragmentation by blasting and crushing. Mining SPSS. 2003. TableCurve2D, Automated curve fitting &
Techn. (Trans. of the Inst. of Mining & Met. A) 114: equation discovery, ver. 4. Users Manual. Chicago:
A29A44. SPSS, Science Marketing Dep.
Ouchterlony, F. 2005b. What does the fragment size distri- Steiner, H.J. 1991. The significance of the Rittinger
bution from blasting look like? In R. Holmberg (ed.), equation in present-day comminution technology, In
Proc. 3rd EFEE World Conf. on Explosives & Blasting, Proc. XVII Int. Minerals Processing Cong., Dresden,
pp. 189199. England: EFEE. I: 177188.
Ouchterlony, F. 2009. A common form for fragment Steiner, H.J. 1998. Zerkleinerungstechnische Eigenschaf-
size distributions from blasting and a derivation ten von Gesteinen. Felsbau 16: 320325.
of a generalized Kuznetsovs x50-equation. In J.A. Svahn, V. 2003. Generation of fines in bench blasting.
Sanchidrin (ed.), Proc. 9th Int. Symp. on Rock Frag- Licentiate thesis A104. Gothenburg: Chalmers Univ.
mentation by BlastingFragblast 9, Granada, Spain, Tech.
1317 September. Rotterdam: Balkema. Wimmer, M., Ouchterlony, F. & Moser, P. 2008. The
Ouchterlony, F. & Moser, P. 2006a. Likenesses and fragment size distribution of Kiruna magnetite, from
differences in the fragmentation of full-scale and model-scale to run of the mine. In H. Schunnesson &
model-scale blasts. Proc. 8th Int Symp. on Rock Frag- E. Nordlund (eds.), MassMin 2008, Proc. 5th Int. Conf. &
mentation by BlastingFragblast 8, Santiago, Chile, Exhib. on Mass Mining, pp. 691703. Rotterdam:
711 May, pp. 207220. Santiago: Editec. Balkema.

22

Das könnte Ihnen auch gefallen