Engineering Structures 32 (2010) 886–896

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Impulsive loading events and similarity scaling

I.M. Snyman ∗
Landwards Sciences, Defence, Peace, Safety and Security (DPSS), Council for Scientific and Industrial Research (CSIR), Meiring Naude Road, Brummeria, Pretoria, South Africa

article info abstract

Article history: Geometrically similar scaling is applied to various blast loading experiments and the results are reported.
Received 19 November 2008 The geometry of the experiment was scaled by a single scale factor. The imparted impulse (captured by the
Received in revised form horizontal motion of a pendulum), mid-point deflections of thin plates and an estimate of the strain rates
2 December 2009
are compared relative to the observed and geometrical scale factors. The analysis confirms the similarity of
Accepted 2 December 2009
Available online 13 January 2010
the imparted impulse obtained and indicates the importance of material properties such as yield strength
and strain rate sensitivity on the similarity of the mid-point deflections.
Keywords:
© 2009 Elsevier Ltd. All rights reserved.
Similarity
Scaling
Blast loading
Ballistic pendulum

1. Introduction loading in close proximity to a clamped thin deformable plate

and measuring the imparted impulse and mid-point deflection.
Detonations of large explosive devices are presently a military Singer et al. [9] reviewed the consistency and repeatability of such
threat in certain affected countries. For the military and research experiments by dimensional analysis. Common to this research
environment, the evaluation of the effect of such big explosions and dimensional analyses is the close proximity (13 mm) of
on structures is a costly exercise. However, these large blasts can the charge to the thin plates (producing a localised load). In
be reduced to smaller ones by means of scaling principles outlined addition, the explosive gas-expanding perimeter does not exceed
during the nineteen-fifties. Reviews can be found in [1–3]. The the exposed width of the plate subjected to loading during the
objective of the scaling of a large blast to a smaller one is to obtain a short time duration of the event. Increasing the standoff distance
similar loading on a smaller structure that is geometrically similar (say up to 300 mm) will affect these conditions and Jacob et al. [10]
to the larger structure. brought this into consideration by extending the dimensional
The measuring of the impulse of a blast event is under analysis. In this case, the charges were placed in a tube to facilitate
investigation at a number of institutions. A comparison of three the impulsive loading (now uniform and not localised anymore). In
different methods is given in [4] and other novel methods are addition, the plates, frames and tubes were geometrically similar
discussed in [5]. Some measurement techniques are directed but the charges were not scaled by the geometry factor.
towards finding the impulse imparted by a buried landmine. The Blast similarity scaling is investigated in this paper to find
vertical impact measurement study of saturated sand by Taylor the total imparted impulse by a ballistic pendulum similar to
et al. [6] discusses the use of such a method. Other methods are by field experiments for landmine protection research. The standoff
Held [7] in which the motion of the upward moving steel blocks distance is such that the expanding gas perimeter increases beyond
are captured by Flash X-Ray images and in [8] the impulse is the diameter of the exposed plate on the pendulum during the
obtained by the integration of the force time history provided by loading event (producing a uniform load). In addition, not all
load cells. Operating most of these instruments is expensive and the explosive gases contribute to the impulsive loading on the
time consuming. exposed region of the ballistic pendulum. Replacing the rigid
A ballistic pendulum offers a cheaper alternative, which is a frontal plate by a clamped mild steel quadrangular thin plate with
well-known and established measurement technique for obtaining a circular area exposed to the charge, the mid-point deflections
the imparted impulse of any explosive charge. A number of can also be measured in addition to the imparted impulse. These
researchers have extensively investigated applying the explosive tests were performed in a small-scale laboratory at the Blast
Impact and Survivability Research Unit (BISRU), University of
Cape Town and for the larger blasts at the Detonics, Ballistics
∗ Tel.: +27 0 12 841 4187. and Explosive Laboratory (DBEL) of the Council for Scientific and
E-mail address: [email protected]. Industrial Research (CSIR). In order to assess the methodology,
0141-0296/$ – see front matter © 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engstruct.2009.12.014
 I.M. Snyman / Engineering Structures 32 (2010) 886–896 887

Notations
m material mass
t time
I total impulse
M explosive mass
R stand-off distance
δ deformation
ε̇ Strain rate
l linear dimensions
λ geometrical scale factor Fig. 1. The ballistic pendulum at BISRU, University of Cape Town.

ρ material density

the geometrical scale factor is compared with the observed scale

factor obtained from the ratio of total imparted impulse, mid-
point deflections and an estimate of the strain rate of geometrically
similar scaled prototypes.

2. Scaling law

Geometrically similar scaling yields a law that scales the

dimensions and time by a factor λ, which is referred to in the
following text as the geometrical scale factor. The scaling of the
relevant variables is shown in Table 1 where the superscript M
indicates the model and P the prototype. The mass of the pendulum
was not scaled according to geometric similarity and the mass m
in the table refers to the charge and plate masses that are attached
to the pendulum. Fig. 2. A clamped thin frontal plate.
For geometric similarity it is required that the different
materials have the same material properties. The material Table 1
properties of the clamped mild steel quadrangular thin plates are The scaling of the relevant variables.
very similar, except for some variation in the quasi-static yield Linear dimensions ` are linearly proportional `P = λ`M
stresses as discussed below. The effect of strain rates was not Mass m is proportional to the third order mP = λ3 mM
measured during the tests explicitly and is derived from the total Densities ρ of materials are the same ρP = ρM
Characteristics times t are linearly proportional t P = λt M
impulse and thickness of the clamped steel quadrangular plates.
Deformations δ are linearly proportional δ P = λδ M
The strain rate scales inversely with the geometric factor and this Total impulse I is proportional to the third order I P = λ3 I M
effect would be included in the measured mid-point deflection. Strain rate is inversely proportional ε̇ P = λ−1 ε̇ M

3. Experimental design
Table 2
The model and prototype dimensions.
A series of geometrically similar blast experiments are executed
Model Prototype 1.6 Prototype 2
with the impulse measured by a ballistic pendulum and a mass
damper system. The model charge is 4 g PE4 (a plastic bonded Scale factor – 1.6 2
RDX explosive similar to C4 with a detonation velocity of 8100 Cylindrical charge (diameter to height ratio: 2 to 1)
m/s and density 1.6 g/cm3 ) and the first prototype is scaled 1.6 Mass (g) 4 16.4 32
Radius (mm) 9.3 14.9 18.6
times, yielding a charge mass of 16.4 g PE4. The second prototype Height (mm) 9.3 14.8 18.5
is scaling the model two times, yielding a charge mass of 32 g PE4.
Spherical charge
Of course, if one considers the 16.4 g charge as a new model, one Mass (g) 4 16.4 32
can scale it up by 1.25 times to obtain the second prototype with a Radius (mm) 8.4 13.5 16.8
charge mass of 32 g. The third prototype is obtained by scaling the Dimensions related to ballistic pendulum
model 32 g charge to a 6 kg charge. Stand-off distance (mm) 100 160 200
The first series of experimental tests involves a ballistic Plate exposed width (mm) 200 320 400
pendulum containing a base plate, 16.4 g charge and a rigid Plate thickness (mm) 13 20 25
Deformable plate thickness (mm) 1 1.6 2
frontal plate (Fig. 1). The additional mass used to balance the Inner diameter of frame (mm) 100 160 200
pendulum and the pen recording the horizontal displacement is Frame thickness (mm) 12.5 20 25
clearly visible. In Fig. 2 the rigid frontal plate on the pendulum is
replaced by a frame with a clamped thin quadrangular plate.
The general dimensional set-up is given in Table 2 with the geo- would be typical in a landmine field experiment, a second series of
metrically similar scaled values of the model problem and the pro- tests were done where the charges were positioned on a rigid base
totypes. The standoff distance, dimensions of the frontal plates on plate. In Fig. 3, the positioning is shown for a 4 g cylindrical charge
the pendulum and the charge are scaled by geometric similarity. on a polystyrene column free in air and on a rigid base plate with
Two experimental set-ups were used in this series of tests. a hole through it in order to detonate from the rear (Fig. 4). Special
Firstly, the charges were positioned on a polystyrene column to care was exercised to ensure good contact between the detonator
simulate them being free in air. To assess the effect of reflections as and the explosive. The height of the charge from the floor was kept
888 I.M. Snyman / Engineering Structures 32 (2010) 886–896

Fig. 3. The positioning and detonation of the cylindrical charges (free in air).

Fig. 5. The Scientifically Instrumented Impulse Measuring Apparatus.

the charges. With this in mind, more than one test was done
and care was taken to administer the set-up and detonation of
the charges as best as possible. For the ballistic pendulum, three
spherical charge tests per explosive mass were done while five
tests per explosive mass were done for the cylindrical charges.
Five tests were done on the ballistic pendulum for the scaling
to SIIMA. Four tests were done using SIIMA at DBEL. The total
imparted impulse was obtained for the tests with the rigid frontal
plates exposed to the charges. For the thin plates, the mid-
point deflection was measured in addition to obtaining the total
impulse.

4. Thin steel plate properties

All clamped mild steel quadrangular thin plates are of

commercial quality mild steel with a low Carbon content (less
than 0.15 weight percent). The thickness of the plates ranged from
0.99 mm to 1 mm (model), 1.63 mm to 1.65 mm (prototype 1.6)
Fig. 4. The positioning and detonation of the cylindrical charges (on a rigid base
plate).
and 1.92 mm to 2 mm (prototype 2). Five test pieces (50 mm long
and 12 mm wide) cut from each collection of thin plates were
Table 3
subjected to quasi-static tensile tests. The average yield stresses
Model and prototype dimensions for large scale experiment. and elongations of the test pieces are shown in Table 5. The yield
stress of the various mild steel plates used in the spherical charge
Model Prototype 5.72
tests differs more in comparison with those for the case of the
Scale factor 5.72 cylindrical charges.
Explosive mass (g) 32 6000 The tensile tests were done with cross head speeds of 5, 50,
Stand-off distance (mm) 105 600 75 and 100 mm/min. The strain rate is obtained by dividing the
Charge diameter (mm) 37 212
cross head speed by the test specimen length of 50 mm. This
Charge height (mm) 19 106
Plate side (mm) 210 1200 gives strain rates respectively of 0.0017, 0.017, 0.025 and 0.033 per
second.
The stress strain behaviour of steel can be approximated by a
constant, as the height of the ballistic pendulum was not varied power law. In addition, the strain rate can be used to improve the
throughout the experiments. approximation by the introduction of strain rate sensitivity. The
The second series of tests involves scaling a model 32 g power law constitutive equation consists of
charge to a prototype 6 kg charge. The total imparted impulse is
inferred from the ballistic pendulum for the model problem and for σ = H ε n ε̇ m , (1)
the prototype the Scientifically Instrumented Impulse Measuring in which σ is the true stress, H is the hardening constant, ε is the
Apparatus (SIIMA) is used (Fig. 5). SIIMA consists of a mass damper true strain and n an exponent less than one, ε̇ is the strain rate and
system that measures the vertical force–time history response of m the strain rate sensitivity (also less than one). With given true
an 8.5 ton suspended mass to the blast loading by a series of load stress, true strain and strain rate, relation (1) was used to find the
cells. The force time history is then integrated to yield the total strain rate sensitivity of a steel specimen.
impulse (see Snyman and Reinecke [11]). The geometrical scale The strain hardening coefficient H, exponent n and strain rate
factor is 5.72 and the dimensions of the model and prototype set- sensitivity m in (1) were determined by a multiple regression
up for the experiment are shown in Table 3. The diameter to height analysis (see [12]) using true strain data between 10% and 35%. The
ratio is 2 to 1. results are shown in Table 5. The strain rate sensitivity for all the
A brief description of all the tests is given in Table 4. From steel plates is found to be between 0.003 and 0.007. The coefficient
experience, it is almost inevitable that some variance in the test of determination R2 of the various regressions is very close
results will be present because of the set-up and detonation of to 1.
 I.M. Snyman / Engineering Structures 32 (2010) 886–896 889

Table 4
The number of blast tests.
Location Charge shape Position of charge Exposed plate type Number of tests

Sphere Free in air Rigid 9

Sphere Free in air Clamped thin plate in rigid frame 9
Cylinder Free in air Rigid 15
Ballistic pendulum at BISRU, UCT
Cylinder On base plate Rigid 15
Cylinder On base plate Clamped thin plate in rigid frame 15
Cylinder (32 g) On base plate Rigid 5
SIIMA at DBEL, CSIR Cylinder (6 kg) On ground surface Rigid 4

Table 5
The quasi-static yield stress, elongation and strain rate sensitivity of the thin plates.
Charge Thickness Yield stress Elongation Strain hardening Strain hardening Strain rate Coefficient of determination R2
mm (MPa) (%) (MPa) exponent sensitivity

0.99 217 38 633 0.29 0.006 0.9999

Spherical
1.64 301 39 721 0.31 0.007 0.9974
charge
1.91 272 40 694 0.31 0.006 1.0000
0.9 299 38 782 0.30 0.003 1.0000
Cylindrical
1.6 349 34 859 0.33 0.006 0.9994
charge
2.0 331 38 852 0.32 0.005 0.9995

5. Experimental results From the 95% confidence interval (t-distribution) it is evident that
the impulse values are reproducible, that is, if a subsequent test
The horizontal displacement of the ballistic pendulum was is performed, the total imparted impulse of say the 4 g charge
captured from which the initial velocity of the oscillating mass will have a 0.95 probability of lying in the interval 3.5 Ns ±7%.
is derived. Multiplying the initial velocity with the pendulum This is the case for both the exposed rigid and deformable frontal
mass yields the change in momentum, which is considered as plate on the ballistic pendulum. The mid-point deflections of thin
the total impulse imparted by the charge on the pendulum. The deformable plates in a frame with a circular hole show a spread
total impulse contains the loading of the positive and negative of data as is also evident from the standard deviation and the
phases of the reflected pressure on the exposed plate of the confidence interval. The estimates of the strain rates also show a
ballistic pendulum. Due to the presence of the laboratory walls spread of data that is confirmed by the strain rate sensitivity in
and floor, small reflections will also be included. The mass of the Table 5. Examples of the deformation of the thin plates are shown
ballistic pendulum and SIIMA are such that the explosive loading is in Fig. 6 from which it is evident that the loading is uniform. The
completed before the onset of displacement. The rigid frontal plate number of holes in the frame (11 mm diameter) that holds the
exposed to the blast loading and the frames, which holds the thin thin plate increases with the size of the plate, namely eight for the
plates have not deformed during any of the experiments. model, ten for prototype 1.6 and 12 for prototype 2. The four holes
The rate at which the mid-point deflection occurred was not (13 mm in diameter) in the corners are for the attachment to the
measured but the maximum rate can be estimated. It is not pendulum.
expected that this rate is constant throughout the loading event. To The results of the 45 tests for the cylindrical charges are shown
find the maximum, assume that the force producing the measured in Table 7. The total impulse imparted onto a ballistic pendulum
impulse acts only on the exposed circular plate for the duration by the 4 g model charge and the scaled prototypes 16.4 g and 32 g
of the momentum transfer. In reality, the blast pressure wave is are clustered about the average as the standard deviation confirms.
acting on the clamped plate initially and later on the frame (and From the 95% confidence interval, it is evident that the impulse
also with varying intensity). The fraction acting on the frame alone values are reproducible. This is the case for both the exposed
is not known and including it in the calculation over predicts the rigid and deformable frontal plate on the ballistic pendulum and
strain rate for the clamped plate. However, of interest is the ratio shows the same trend as for the spherical charges. The mid-point
between the respective strain rates (which yields the observed deflections of thin deformable plates in a frame with a circular
scaling factor). This is not affected by considering only the fraction hole also show a spread of data as is evident from the standard
(same for all the events) that interacts with the thin plate. The deviation and the confidence interval (but less so than for the
estimates of the strain rates are therefore only used to identify the spherical charges). The estimates of the strain rates also show a
scaling trend with regard to the mid-point deflections. spread of data but are somewhat less in comparison with the data
Dividing the impulse by the mass of the circular exposed plate for the spherical charges.
yields the velocity by which rate it will deform. If we divide this The impulse increases as the conditions under which the
velocity with the deformation we arrive at an estimate of the strain explosive charge is detonated changes from the free in air to the
rate. Due to the over-prediction, this strain rate is the maximum rigid base plate and changing the frontal rigid plate to a thin
that the plate will experience. The shape of the plate exposed to clamped plate inside a frame. Some of the changes are hardly
the blast is approximated by a spherical dome and the surface area noticeable for the small charges but more significant for the larger
is calculated from the mid-point deflection and the original radius. charges as is evident from Fig. 7. Replacing the rigid frontal plate
Subtract the surface area of the original cylindrical exposed plate with a frame and a thin deformable plate introduces a cylindrical
from the surface area of the dome shaped plate after the blast. This cavity in the frontal plate of the pendulum. This is a gas trap that
yields an area with a radius equal to the mid-point deflection which increases the total imparted impulse. From the results, it is evident
now is used to estimate the deformation of the plate. that the gas trap effect scales with the geometry. A significant
The results of the 18 tests for spherical charges are shown in increase in the total imparted impulse is found by positioning the
Table 6. The total impulse imparted onto a ballistic pendulum by charge on a base plate especially for the larger charges. It appears
the 4 g model charge and the scaled prototypes 16.4 g and 32 g that the effect of the reflection of the pressure of the charge from
are clustered about the average as the standard deviation confirms. the base plate on the pendulum also scales with the geometry.
890 I.M. Snyman / Engineering Structures 32 (2010) 886–896

Fig. 6. The mid-point deflection of the model and scaled prototypes for spherical charges.

Table 6
The total impulse and mid-point deflection results for spherical charges.
Set-up Variable Charge #1 #2 #3 Average Standard Confidence
mass (g) deviation interval (%)

4 3.4 3.6 3.7 3.5 0.14 ±7

Free in air, rigid plate Impulse (Ns) 16.4 15.1 15.1 14.5 14.9 0.35 ±4
32 31.4 30.5 28.7 30.2 1.35 ±8
4 3.5 3.4 3.8 3.6 0.20 ±10
Impulse (Ns) 16.4 13.9 15.2 16.6 15.2 1.36 ±16
32 33.6 34.1 33.6 33.8 0.30 ±2
4 3.6 3.0 2.7 3.1 0.47 ±28
Free in air, thin deformable plate Mid-point deflection (mm) 16.4 3.2 3.8 5.7 4.2 1.30 ±57
32 6.6 7.9 6.9 7.1 0.70 ±18
4 15 914 18 803 23 532 19 417 3846 ±37
Estimated Strain rate (1/s) 16.4 17 218 15 683 11 514 14 805 2952 ±37
32 11 062 9 306 10 497 10 288 897 ±16

Table 7
The total impulse and mid-point deflection results for cylindrical charges.
Set-up Variable Charge #1 #2 #3 #4 #5 Average Standard Confidence
mass (g) deviation interval (%)

4 4.9 4.1 4.4 3.8 4.2 4.3 0.43 ±11

Free in air, rigid plate Total impulse (Ns) 16.4 19.9 17.1 17.3 17.9 17.5 17.9 1.13 ±7
32 32.1 33.9 34.6 36.7 36.1 34.7 1.82 ±6
4 5.4 6.2 6.9 5.2 5.7 5.9 0.68 ±13
Base plate, rigid plate Total impulse (Ns) 16.4 22.4 23.5 22.5 23.6 22.7 22.9 0.57 ±3
32 33.1 36.0 37.4 38.8 40.9 37.2 2.92 ±9
4 6.6 5.5 6.0 7.2 7.0 6.4 0.70 ±13
Total impulse (Ns) 16.4 24.9 26.9 30.3 25.8 26.8 26.9 2.04 ±9
32 51.7 54.6 52.8 48.2 55.5 52.6 2.86 ±6
4 7.1 5.8 6.8 6.6 7.9 6.8 0.78 ±13
Mid-point deflection
Base plate, thin deformable plate 16.4 10.0 10.3 10.5 6.7 11.3 9.8 1.75 ±25
(mm)
32 12.3 14.3 15.3 12.6 13.0 13.5 1.25 ±11
4 16 717 17 337 15 912 19 554 16 141 17 132 1462 ±10
Estimated Strain rate (1/s) 16.4 11 032 11 563 12 788 16 984 10 561 12 585 2595 ±24
32 9 519 8 651 7 830 8 660 9 662 8 864 745 ±10

Fig. 7. The total imparted impulse and charge mass.

 I.M. Snyman / Engineering Structures 32 (2010) 886–896 891

There is also an increase in impulse changing the shape of the The scaling of 32 g to 6000 g yields 15 scaling factors, as
charge from a sphere to a cylinder. there were five BISRU pendulum tests and only three SIIMA tests
The total impulse imparted onto a ballistic pendulum by the (Table 11). In Fig. 10, the observed scaling factors obtained from
32 g model cylindrical charge positioned on a rigid base plate the imparted impulse are shown relative to the geometrical scaling
and the scaled prototype 5.72 measured by SIIMA are shown in factors of 5.72. The vertical bars represent an 8% error related to the
Table 8. The standard deviation indicates that the impulse values geometrical scale factor.
are clustered about the average and from the confidence interval, The observed average scaling factor between the imparted
it is evident that the impulse values are reproducible. The fourth impulse of 32 g and 6000 g is 5.28, which is 8% smaller than the
test with SIIMA did not record the force time history correctly. geometrical scaling factor of 5.72. It appears that the difference
of the reflecting surface (rigid plate for the BISRU pendulum and
soil for SIIMA) may have contributed somewhat to the size of the
6. Observed scale factors
observed scaling factor. The standard deviation is small, indicating
that the observed scaling factors are clustered about the average.
Analysing the observed scale factors gives an indication of
The 95% confidence interval indicates good reproducibility, for
the success of geometrically similar scaling with regard to
example, the next test in the scaling event will yield (with 0.95
capturing the information. The observed scale factors obtained
probability) an observed scale factor in the interval 5.28% ± 0.8%.
from the comparison of the total imparted impulse and mid-point
deflection are analysed in this section with regard to the relevant
6.2. Scale factors obtained from comparing the mid-point deflections
geometrically scaled factors 1.25, 1.6 and 2.
The second set of observed scale factors are obtained by
6.1. Observed scale factors obtained from comparing the total comparing the mid-point deflection of the clamped thin plate in
imparted impulse on the pendulum the various scaled tests with one another. From Table 1 it follows
that the observed scale factor associated with the mid-point
The first set of observed scale factors are obtained by comparing deflections is the ratio of the two-scaled mid-point deflections.
the total imparted impulse of the various scaled tests with one For the spherical charges, three tests per explosive mass yield
another. From Table 1 it is evident that the observed scale factor nine scale factors for each model to prototype scaling. These scale
associated with the imparted impulse is the cube root of the ratio of factors are shown in Fig. 11. The vertical bars represent a 5% error
the scaled impulse values. For the spherical charges three tests per related to the geometrical scale factor.
explosive mass yield nine scale factors for each model to prototype The observed average scale factors are shown in Table 12 and
scaling. These observed scale factors for the imparted impulse to vary considerably in relation to the geometrical factors (up to
a rigid plate and a thin plate attached to a rigid frame are shown 60%). The standard deviation is small indicating that the observed
in Fig. 8 against the geometrical scale factors. The vertical bars scale factors are clustered around the average. The 95% confidence
represent a 5% error related to the geometrical scale factor. The interval supports the spread of results with regard to mid-point
observed scale factors are slightly larger than the geometrical scale deflection.
factors but fall within the 5% error margin. There are 25 scale factors for the cylindrical tests associated
The average observed scale factor associated with the total with the mid-point deflections (as there were five tests per scaling
imparted impulse for each scaling event is shown in Table 9. event). In Fig. 12, the observed scaling factors are shown relative to
Comparing the geometrical scale factor with the observed average the geometrical scaling factors of 1.25, 1.6 and 2. The vertical bars
scale factor shows a difference of 5% or less. It is of interest that represent a 5% error related to the geometrical scale factor. From
the scaling factor varies the most for the case with clamped thin the graph, the spread of values is evident, although not so large as
frontal plate on the pendulum. The standard deviation (taken over in the case for the spherical charges.
the nine values) is small and the 95% confidence interval indicates The observed average scale factors are shown in Table 13 and
are fairly close to the geometrical factors (varies up to 8%). The
good reproducibility, for example the next test in the scaling event
standard deviation is small indicating that the observed scale
of 1.25 times will have a 0.95 probability of yielding an observed
factors are clustered around the average. The 95% confidence
scale factor in the interval 1.27% ± 1.1%.
interval supports the spread of results with regard to mid-point
There are 25 scale factors for the cylindrical tests (as there were
deflection.
five tests per scaling event). In Fig. 9, the observed scale factors for
the imparted impulse are shown against the geometry scale factors 6.3. Scale factors obtained from comparing the strain rates
of 1.25, 1.6 and 2. The vertical bars represent a 5% error related to
the geometrical scale factor. The observed scale factors for a rigid The last set of observed scale factors are obtained by considering
plate (charge free in air) and for a thin plate attached to a rigid the scaling of the strain rates between the clamped thin plates.
frame are close to geometrical scale factors. The case where the From Table 1 it follows that the observed scale factor is the ratio
explosive was positioned on a base plate and the imparted impulse of the two scaled strain rates. For the spherical charges, three
was measured by a rigid plate lies slightly outside the error bars for tests per explosive mass yield nine scale factors for each model to
the scale factors 1.25 and 2. prototype scaling. The observed average scale factors are shown
The average scale factor for each scaling event is shown in in Table 14 and vary between −5% and 16% in relation to the
Table 10. The observed scale factors for the rigid frontal plate and geometrical factors. However, the standard deviation is fairly large
free in air cylindrical charges are very close (within 2%) to the and the observed scale factors are therefore not clustered around
geometrical scale factors. The same can be said for the clamped thin the average. The 95% confidence interval supports the spread of
frontal plate with the charge positioned on a base plate. However, results.
the 1.25 and 2 times scaling events with the rigid frontal plate There are 25 scale factors for the cylindrical tests associated
with the charge positioned on a base plate yield observed scale with the mid-point deflections and strain rates. The observed
factors that are 6% to 7% smaller than the geometrical scale factors. average scale factors are shown in Table 15 and differ somewhat
The standard deviation is small, implying that all the scale factors from the geometrical factors (up to 12%). The standard deviation is
are clustered around the average value (even when it is 6% to 7% large (except for the scaling factor 2) indicating that the observed
smaller than the geometrical factor). The 95% confidence interval scale factors are not clustered around the average. The 95%
indicates good reproducibility. confidence interval supports the spread of results.
892 I.M. Snyman / Engineering Structures 32 (2010) 886–896

Table 8
Total impulse of scaled 32 g and 6 kg cylindrical charges.
Charge mass (g) #1 #2 #3 #4 #5 Average Standard deviation Confidence interval (%)

32 37 37 38 37 38 37 0.62 ±2
6000 5209 5612 5716 No data X 5512 268 ±9

Fig. 8. The observed scale factor obtained from the imparted impulse for spherical charges.

Table 9
Summary of the observed scale factor obtained from the impulse of spherical charges.
Set-up Scale Scale factor average Difference with theory (%) Standard deviation Confidence interval (%)

Scale 1.25 1.27 1 0.019 ±1

Rigid frontal plate Scale 1.6 1.61 1 0.021 ±1
Scale 2 2.04 2 0.035 ±1
Scale 1.25 1.31 4 0.034 ±2
Clamped thin plate in rigid frame Scale 1.6 1.62 1 0.049 ±2
Scale 2 2.11 5 0.034 ±1

Fig. 9. The observed scale factor obtained from the imparted impulse for cylindrical charges.

7. Discussion ratio deviates by about 7%. A similar trend is evident when the
scaling from 16.4 g to 32 g is considered (scale factor 1.25). Scaling
The ratio of the observed scale factor and the geometrical from 32 g to 6 kg yields a scaling ratio of 0.92.
scale factor should be as close to one as possible to confirm a The scale factor ratios associated with the mid-point deflection
geometrically similar scaling. The ratios resulting from the impulse of the clamped quadrangular plates are shown in Fig. 14. The
associated with the various test set-ups are shown in Fig. 13. spherical charges yield a scaling ratio different from one by a large
Scaling the model 4 g charge to the prototype 1.6 yields ratios very margin (up to 35%). The cylindrical charges that are positioned on
close to one (within 2%). The same applies when considering the a base plate yield ratios fairly close to one (within 4%).
ratios for the prototype 2 for the spherical and cylindrical charges The quasi-static yield stresses of the various clamped quad-
free in air, but when the charge is positioned on a base plate, the rangular plates (1 mm, 1.6 mm and 2 mm thick) represent a
 I.M. Snyman / Engineering Structures 32 (2010) 886–896 893

Table 10
Summary of the observed scale factor obtained from the impulse of cylindrical charges.
Set-up Charge position Scale Scale factor average Difference with Standard deviation Confidence
theory (%) interval (%)

Scale 1.25 1.25 0 0.030 ±1

Rigid frontal plate Free in air Scale 1.6 1.61 1 0.057 ±2
Scale 2 2.01 1 0.068 ±1
Scale 1.25 1.18 −6 0.030 ±1
Rigid frontal plate Base plate Scale 1.6 1.58 −1 0.056 ±2
Scale 2 1.86 −8 0.078 ±2
Scale 1.25 1.25 0 0.035 ±1
Clamped thin plate in rigid frame Base plate Scale 1.6 1.61 1 0.066 ±2
Scale 2 2.02 1 0.076 ±2

Fig. 10. The observed scale factor obtained from the imparted impulse for cylindrical charges (BISRU pendulum and SIIMA).

Table 11
Summary of the observed scale factor obtained from the impulse of cylindrical charges (BISRU pendulum and SIIMA).
Set-up Charge position Scale Scale factor Average Difference Standard deviation Confidence interval

Rigid frontal plate Base plate or ground surface Scale 5.72 5.28 −8% 0.078 ±1%

Fig. 11. The observed scale factor according to mid-point deflection of spherical charges.

Table 12
Summary of the observed scale factors obtained from mid-point deflection for spherical charges.
Set-up Scale Scale factor average Difference with theory (%) Standard deviation Confidence interval (%)

Scale 1.25 1.79 30 0.461 ±19

Clamped thin plate in rigid frame Scale 1.6 1.38 14 0.411 ±22
Scale 2 2.33 14 0.361 ±12
894 I.M. Snyman / Engineering Structures 32 (2010) 886–896

Fig. 12. The observed scale factor according to mid-point deflection of cylindrical charges.

Table 13
Summary of the observed scale factors obtained from mid-point deflection fo cylindrical charges.
Set-up Scale Scale factor average Difference with theory (%) Standard deviation Confidence interval (%)

Scale 1.25 1.43 13 0.323 ±9

Clamped thin plate in rigid frame Scale 1.6 1.44 10 0.283 ±8
Scale 2 1.99 0 0.272 ±5

Table 14
Summary of the observed scale factors obtained from estimated strain rate for spherical charges.
Set-up Scale Scale factor average Difference with theory (%) Standard deviation Confidence interval (%)

Scale 1.25 1.45 14 0.275 ±20

Clamped thin plate in rigid frame Scale 1.6 1.35 16 0.348 ±27
Scale 2 1.90 −5 0.358 ±20

Table 15
Summary of the observed scale factors obtained from estimated strain rate for cylindrical charges.
Set-up Scale Scale factor average Difference with theory (%) Standard deviation Confidence interval (%)

Scale 1.25 1.43 12 0.292 ±11

Clamped thin plate in rigid frame Scale 1.6 1.40 12 0.248 ±9
Scale 2 1.94 −3 0.215 ±6

Fig. 13. The observed scale factor ratio of the various tests based on the total imparted impulse.

material property that should not vary from model to proto- similar scaling. In Fig. 15, the yield stress ratios are shown for
type. Thus, the ratio of the yield stresses of the scaled plates the spherical and cylindrical charges with ±10% error bars for
(obtained from Table 5) should be one to support geometrically the geometrical scale factor. For the plates used in the tests
 I.M. Snyman / Engineering Structures 32 (2010) 886–896 895

Fig. 14. The observed scale factor ratio of the various tests based on the mid-point deflection of the clamped quadrangular plate.

Fig. 15. Yield stress ratios of the scaled clamped quadrangular plates.

Fig. 16. The observed scale factor ratio associated with the estimated strain rates.

for the spherical charges, the quasi-static yield stress of the In Fig. 16 the ratio of the scale factors associated with the
mild steel thin plates varied considerably as is evident from estimated strain rates are plotted against the geometry scale
the ratios larger than one. The yield stress ratios for the plates factors. Scaling from 4 g to 32 g charges (cylindrical and spherical)
used in the test set-ups for the cylindrical charges are closer to yield a scaling ratio of almost 1, while for the other scaling events
one. the ratio is different from 1 (between 12% and 16%).
896 I.M. Snyman / Engineering Structures 32 (2010) 886–896

8. Conclusion References

The geometrically similar scaling between the various explosive [1] Baker WE. Explosions in air. Austin: University of Texas Press; 1973.
[2] Baker WE, Westine PS, Dodge FT. Similarity methods in engineering dynamics.
charges is successful with regard to the total imparted impulse Amsterdam: Elsevier; 1991.
as the observed and geometrical scale factors are close to one [3] Jones N. Structural impact. Cambridge: Cambridge University Press; 1989.
another. Mixed results were obtained for scaling of the mid-point Chapter 11.
[4] Smith PF, Mostert FJ, Snyman IM. Comparison of methods to measure the blast
deflections of the clamped quadrangular plates. The observed
impulse loading of an explosive charge. In: 24th International symposium of
scaling factor between the 4 g and 32 g cylindrical charges is very ballistics, 2008.
close to the geometrical scaling factor, but the other scaled charges [5] Ames R, Murphy M. Diagnostic techniques for multiphase blast fields. In: 24th
indicated a margin of difference. This difference is also apparent International symposium of ballistics, 2008.
[6] Taylor LC, Skaggs RR, Gault W. Vertical impulse measurements of mines buried
by the scaling trend of the estimated strain rates. The material in saturated sand. Fragblast 2005;9:19–28.
properties such as the strain rate sensitivity of the plates (and to a [7] Held M. Momentum distribution of anti-tank mines. In: 20th International
lesser degree, the quasi-static yield stress) affected the mid-point symposium of ballistics, 2002.
deflections and therefore the scaling results in a number of cases. [8] Reinecke JD, Snyman IM, Ahmed R. Blast characterization through impulse
measurements. In: Science Real and Relevant Science: 2nd CSIR biennial
conference, 2008.
Acknowledgements [9] Singer J, Arbocz J, Weller T. Buckling experiments: Experimental methods in
buckling of thin walled structures. New York: John Wiley; 1998. 251–4.
[10] Jacob N, Nurick GN, Langdon GS. The effect of stand-off distance on the failure
The author would like to express his gratitude to Prof GN Nurick, of fully clamped circular mild steel plates subjected to blast loads. Eng Struct
Dr S Chung Kim Yuen from BISRU for the preparation and execution 2007;29:2723–36.
of the tests. Mr DJ Reinecke and Me R Ahmed from CSIR and Mr FJ [11] Snyman IM, Reinecke JD. Measuring the impulse from an explosive charge. In:
Ballistics symposium South Africa, Denel-OTB, Bredasdorp, 2006.
Beetge from ARMSCOR are thanked for their support during the [12] Shi MF, Meuleman DJ. Strain rate sensitivity of automotive steels. In:
course of the project. International congress and exposition, 1992.