Shock Waves (2008) 18:291–297
DOI 10.1007/s00193-008-0142-1
ORIGINAL ARTICLE
Application of background oriented schlieren for quantitative
measurements of shock waves from explosions
O. K. Sommersel · D. Bjerketvedt ·
S. O. Christensen · O. Krest · K. Vaagsaether
Received: 30 September 2007 / Accepted: 22 April 2008 / Published online: 22 May 2008
© Springer-Verlag 2008
Abstract This paper describes application of a background
oriented schlieren technique in order to obtain quantitative
measurements of shock waves from explosions by processing
high speed digital video recordings. The technique is illus-
trated by an analysis of two explosions, a high explosive test
and a hydrogen gas explosion test. The visualization of the
shock front is utilized to calculate the shock Mach number,
leading to a predicted shock front pressure. For high explo-
sives the method agreed quite well with a standard curve
for side-on shock pressures. In the case of the gas explosion
test we can also show that the shock front is non-spherical.
It should be possible to develop this technique to investigate
external blast waves and external explosions from vented gas
explosions in more details.
Keywords Shock wave · Hydrogen · Gas explosion
PACS 07.05.Pj · 47.40.-x
This paper is based on work that was presented at the 21th Interna-
tional Colloquium on the Dynamics of Explosions and Reactive Sys-
tems, Poitiers, France, July 23–27, 2007.
Communicated by L. Bauwens.
O. K. Sommersel (B) · D. Bjerketvedt · K. Vaagsaether
Department of Technology, Telemark University College,
Kjølnes Ring 56, 3918 Porsgrunn, Norway
e-mail: ole.k.sommersel@hit.no
S. O. Christensen · O. Krest
Norwegian Defence Estates Agency, (NDEA),
Postboks 405 Sentrum, 0103 Oslo, Norway
1 Introduction
This paper describes application of a background oriented
schlieren (BOS) technique in order to obtain quantitative
measurements of shock waves from explosions by process-
ing high speed digital video recordings. To illustrate the tech-
nique we present results from analysis of two explosions, a
high explosive test and a gas explosion test. The experiments
were performed at the Norwegian Defence Estates Agency
test facility at Raufoss, Norway, June 2005, as part of an
IEA-HIA task 19 project on hydrogen safety. The objective
of this paper is to show that the shock wave overpressures
in a field explosion test can be predicted quantitatively by
means of this technique.
2 Image processing
Background oriented schlieren is an optical measurement
technique, and it is used for visualizing density gradients [1].
The principle is the deviation of light rays that pass through a
density gradient, which is well known from the Schlieren and
shadowgraph techniques. The deviation is caused by the vari-
ation of the refractive index of the transparent media (i.e. air)
[2]. The common steps of the method are imaging of a back-
ground through a flow of interest, and then software analysis
that manipulates the data to find the density gradients. In its
simplest form, BOS makes use of simple background pat-
terns of the form of a randomly generated dot-pattern. The
size of the pattern should ideally be optimized according to
the magnification of the set-up. In our case the background
was a forest (i.e. a row of spruce trees). The technique could
probably be improved by using a random dot pattern, but
would be impractical due to scale. The visualized shock sur-
face is not always visible along its entire contour, and present
123
292 O. K. Sommersel et al.

Fig. 1 Principle of the image

processing. Subtraction of an
undistorted image from a
distorted image yields the result
shown in image three [3]

gaps as shown in the right hand image of Fig. 1. These gaps

are caused by the absence of background, as the shock front
propagates higher than the trees, especially in the right hand
side of the images.
As illustrated in Fig. 1 our image processing technique
consists in principal of subtracting an undistorted image from
a distorted image [3]. The resulting image was then manipu-
lated by a logarithmic intensity transformation. We read the
positions of the shock front from the resulting image man-
ually. If the shock surface was thicker than one pixel, the
position was defined at the outermost pixel. The shock front
gaps mentioned above did not affect the analysis, because of
our manual procedure.
The time vector used in our calculations were the log-
arithm of the product of sound speed in air and the time
extracted from the high-speed movie, made by the basis of
frames per sec in the high-speed movie, as we had one shock
wave reading for each frame. The distance vector was also
logarithmic. The time and distance data set were curve-fitted
using both a 3 and a 4 degree polynomial (Polyfit in MAT-
Fig. 2 Picture of the container used in the hydrogen gas experiment
LAB). Calculations with polynomials of higher degrees had
large oscillations, and could not present the data correctly.
The data were used to calculate the shock front velocity, The hydrogen experiment, test #39, was performed in a
hence the shock Mach number. The predicted shock front standard 20’ ISO container with inner dimensions L = 6.0 m,
pressures were calculated from the Mach number using the W = 2.4 m and H = 2.4 m. The steel walls and roof were
relation
p/ p0 = (2γ /(γ + 1))(M2 − 1), where γ = 1.4. corrugated. The container doors shown on the container left
We have developed numerical algorithms in MATLAB for hand side in Fig. 2 were fully open. The right wall was a
the image processing and pressure calculations. The code solid wall. The hydrogen filling system consisted of a 0.3 m3
also generated videos from the manipulated images. storage tank and a steel tube connecting the tank and the
container. The tank was placed behind the closed end wall, to
the right in Fig. 2. The container was placed approximately
30 m from a shooting range bunker, where the instruments
3 Experimental setup and high speed video cameras were set up. Figures 3 and 4
show a schematic overview of the container with lengths and
The tests reported here were part of a test series consisting pressure monitor placements.
of calibration tests with C-4 high explosives and more than A downwards directed nozzle was mounted at the tube
30 gas explosions tests with inhomogeneous hydrogen air outlet. The nozzle had an inner diameter of 9 mm. The nozzle
clouds in an ISO container. The C-4 test we are referring to in was placed inside the container 1.0 m from the back wall, at
this paper was a test with 100 g C-4 mounted and detonated a height of 1.0 m above the container floor. The storage tank
on a wood pole. The charge was placed on the outside of was filled with hydrogen at 2.4 MPa overpressure. The fuel
the container, 2.5 m from the container door, and 1 m above supply was controlled by a pneumatic valve. Details of the
ground. gas handling units are presented in Figs. 5, 6 and 7.

123
Application of background oriented schlieren
Fig. 3 Top view of the container and filling system
Fig. 4 Side view of the container with measurements
Fig. 5 Gas storage tank and pneumatic valve
In test #39, eight wooden Euro Pallets, with dimension
0.8 m by 1.2 m and a height of 0.12 m, were placed inside the
container to generate turbulence during the explosion. This
turbulence is mainly caused by the interaction of the flow
with the obstacles. The increase in turbulence contributes to
an increase in the overall burning rate, therefore creating a
more violent explosion compared to an empty container. The
distribution of the obstacles is shown in Fig. 8. The container
doors at one end were open. The hydrogen gas was ignited
293
Fig. 6 Nozzle and obstacle
Fig. 7 Nine milimeter nozzle details
15 s after the release of hydrogen inside the container started.
The ignition source was mounted 1.0 m from the solid back
wall, and 50 mm below the roof. Three Kistler 7001 pressure
transducers were monitoring the explosion pressure inside
the container. In addition, two LC-33 pressure transducers
(P4 and P5, respectively), were mounted 1.0 m above ground,
6.5 and 8.7 m from the container door. The pressure signals
were recorded on a transient digital data logger. The hydrogen
storage tank pressure was monitored with a pressure trans-
ducer mounted at the end wall of the tank. Theoretical cal-
culations of the jet release imply that approximately 0.5 kg
hydrogen was released before ignition, as shown in Fig. 9.
A Photron Ultima APX-RS high-speed monochrome
camera with a Nikkor 50 mm f/1.2 lens, was recording the
123
294 O. K. Sommersel et al.

Fig. 8 Distribution of obstacles in the container during the hydrogen

gas experiment. Sketch made from the FLACS package

Cumulative mass flow

0.7

0.6

0.5
Mass H2 (kg)

0.4

Fig. 10 Manipulated image from the C-4 test [3]

0.3

0.2 C4 Kingery & Bulmash (in F.P. Lees)

2
10 C4 BOS, n = 3
C4 BOS, n = 4
0.1
Pressure side−on [kPa]

0
0 5 10 15 20 25 30
Time (s)

Fig. 9 Cumulative mass of hydrogen released in the experiment show

approximately 0.5 kg at time of ignition (15 s) 1
10

explosion events at a rate of 3,000 fps. The distance between

the experimental set-up and the camera was approximately
30 m. 0
10
1
10
Distance [m]

Fig. 11 Predicted shock front pressures from the C-4 test. Third and
4 Results and discussion fourth degree polynomial

A manipulated image of the C-4 test is shown in Fig. 10.

Fig. 11 shows the predicted shock front pressures from the
C-4 test compared with a pressure–distance curve based on
the Kingery and Bulmash free air burst TNT curve [4]. This
curve is based on TNT data, so in order for us to compare
with our C-4 experiment, the curve needed to be corrected to a
C-4 equivalent. According to [5], the factor for a C-4 to TNT
equivalence conversion is 1.2 at this pressure, hence 100 g
C-4 equals 120 g TNT, shown as the solid line in Fig. 11.
The length scale is based on the number of pixels in the
BOS movie, corrected by the known length of the container,
yielding a factor of 24.43 pixels/m. This factor is an impor-
tant parameter, as to get the length scales defined correctly.
Figure 11 shows a good agreement between the corrected
C-4 curve and the experimental data. The end points of the
experimental curves are deviating from the theoretical curve.
This is caused by the polynomial curve fitting. Calculations
with polynomials of higher degrees had large oscillations,
and could not present the data correctly. Even though the
predicted shock front pressures deviate in the end points of
our data set, the average value is following the Kingery and
Bulmash curve quite well. The result shows that it is possi-
ble to estimate shock front pressures from a high explosive
detonation by using a high speed digital video recording of
the event.
The hydrogen experiment was violent and the pressure
sensors located inside of the container failed at an early stage.

123
Application of background oriented schlieren 295

25
P4
20
Pressure, kPa

15
10
5
0
−5
0.81 0.82 0.83 0.84 0.85 0.86
Time, s
15
P5
Pressure, kPa

10

−5
0.81 0.82 0.83 0.84 0.85 0.86
Time, s

Fig. 12 Pressure records from hydrogen gas explosion test. Pressure

sensor P4 and P5 were mounted outside of the container, 6.5 and 8.7 m
from the container door

Fig. 13 Manipulated image from the hydrogen gas explosion test with
The first peak from monitor P1 was recorded though, and defined axes
shows a maximum pressure of 83 kPa. The two pressure sen-
sors located outside of the container gave pressure records as
shown in Fig 12. In order to determine more thoroughly the 0.018

maximum overpressure generated in this explosion, we ana- 0.016

lyzed the high-speed films. The explosion in the experiments
generated a shock wave that was visible on the high-speed 0.014
films. Based on the theory of the BOS technique, we were
0.012
able to extract pictures of this shock wave propagating from
Time [s]

the container opening. 0.01

Figure 13 shows one of the manipulated images from the
0.008
hydrogen gas explosion test. From the manipulated images
we have extracted two sets of shock trajectories along the 0.006 H2 − Vertical
vertical and horizontal axes. We defined upper corner of the H − Horizontal
0.004 2
container as a reference point. Figure 14 shows the trajec- H − Vertical second shock
2
tories in a time-distance diagram for the shock front propa- 0.002
gation along the vertical and horizontal axes. Figure 14 also 1 2 3 4 5 6 7
Distance [m]
show the time–distance results for the second shock wave,
read along the vertical direction. The data from the hydrogen Fig. 14 Shock propagation in the hydrogen gas explosion test along
gas explosion test were treated in the same manner as the the axes defined in Fig. 13
C-4 experiment. The estimated shock front pressure versus
time can be found in Fig. 15. It is clear that the shock in the
horizontal direction propagates at a higher velocity than in It is interesting to note that on the manipulated image video
the vertical direction. We have made some preliminary CFD several shock waves follow the first shock wave. Figure 16
simulations with the FLACS code [6] and our own CFD code. shows an image of the manipulated video where the first
Both these codes indicated that the flow velocity ahead of the shock wave is followed by a second shock. The second shock
shock cannot explain the differences in horizontal and verti- wave is harder to recognise, and appears only in the late half
cal shock velocities. The asymmetrical shock is expected to of the movie.
be a result of the reflection off ground and the momentum Figure 17 shows the estimated shock front pressure ver-
in horizontal direction of the jet from the explosion inside sus distance. Preliminary FLACS simulations and measured
the container and possibly also the external explosion. Such pressures are also shown in the figure. In the FLACS sim-
phenomena have been investigated among others by Forcier ulations we assumed a homogeneous 20% hydrogen-air gas
and Zalosh [7], Chiu et al. [8] and Harrison and Eyre [9]. cloud in the upper half of the container volume. The estimated

123
296 O. K. Sommersel et al.

200
H2 − Vertical, n = 3
180
H − Vertical, n = 4
2
160
H2 − Horizontal, n = 3

Pressure (side−on) [kPa]

140
2
H2 − Horizontal, n = 4
Pressure [kPa]

10
120

100

80

60

40

20
0 0.005 0.01 0.015
1
Time [s] 10
0 1
10 10
Distance [m]
Fig. 15 Estimated shock front pressure versus time for the hydrogen
explosion test H2 − Vertical, n = 3
H2 − Vertical, n = 4
H2 − Horizontal, n = 3
H2 − Horizontal, n = 4
FLACS − horizontal
FLACS − vertical
Pressure transducer (P4 and P5)

Fig. 17 Estimated shock front pressure versus distance for the hydro-
gen explosion test

direction. We observed also that the first shock wave was

followed by several other shock waves.
It should be possible to use this technique to investigate
external blast waves and external explosions from vented gas
explosions.

Acknowledgments Financial support from the Norwegian Research

Council Programmes, Strategiske høgskoleprosjekter and RENERGI,
is gratefully acknowledged.

Fig. 16 A second shock wave following the first shock. Manipulated

image from the hydrogen gas explosion test
References

shock front pressures seem reasonable compared with the
FLACS simulation and pressure measurements.
5 Conclusions
We have shown that this BOS technique can be used for
estimating explosion overpressures from high explosives and
gas explosions. For high explosives the method agreed quite
well with a standard curve for side-on shock pressures.
For the gas explosion we found that the shock wave prop-
agated faster in the horizontal direction than in the vertical
1. Klinge, K., Kirmse, T., Kompenhans, J.: Application of quantitative
background oriented schlieren (BOS): investigation of a wing tip
vortex in a transonic wind tunnel. In: Proceedings of PSFVIP-4
2003 (2003)
2. Venkatakrishnan, L., Meier, E.A: Density measurements using the
background oriented schlieren technique. Exp. Fluids 37, 237–247
(2004)
3. Sommersel, O.K., Bjerketvedt, D., Christensen, S.O., Krest, O.,
Vaagsaether, K.: Application of background oriented schlieren
(BOS) for quantitative measurements of shock waves from explo-
sions. Presented at the 21st ICDERS 2007 (2007)
4. Lees, FP.: Loss Prevention in the Process Industries, Hazard Iden-
tification, Assessment and Control, vol. 2, 2nd edn. Reed Edu-
cational and Professional Publishing Ltd (ISBN 0-7506-1547-8),
pp. 17/129–132 (1996)

123
Application of background oriented schlieren 297

5. Design and Analysis of Hardened Structures to Convential 8. Chiu, K.W., Lee, J.H., Knystautas, R.: The blast waves from asym-
Weapons Effects (1997). Joint Departments of the Army, Air Force, metrical explosions. J. Fluid Mech. 82(2), 193–208 (1977)
and Navy and the Defense Special Weapons Agency, TM 5-855- 9. Harrison, A.J., Eyre, J.A.: External explosions as a result of explo-
1/AFPAM 32-1147(I)/NAVFAC P- 1080/DAHSCWEMAN-97 sion venting. Combust. Sci. Technol. 52, 91–106 (1987)
6. FLACS User’s Guide 98, GexCon AS, Norway
7. Forcier, T., Zalosh, R.: External pressures generated by vented gas
and dust explosions. J. Loss Prev. Proc. Ind. 13, 411–417 (2000)

123