Sunteți pe pagina 1din 8

Proceedings of the 8th International Conference on Structural Dynamics, EURODYN 2011 3345

Leuven, Belgium, 4-6 July 2011


G. De Roeck, G. Degrande, G. Lombaert, G. Müller (eds.)
ISBN 978-90-760-1931-4

Numerical simulations of a pyrotechnic shock test


J.-M. Ndambi, B. Reymen, D. Lecompte, J. Vantomme
Department of Civil Engineering, Royal Military Academy, Avenue de la Renaissance 30, B-1000 Brussels, Belgium
Email: jmndambi@rma.ac.be, Buno.Reymen@rma.ac.be, David.Lecompte@rma.ac.be, John.Vantomme@rma.ac.be

ABSTRACT: Launch vehicles and spacecraft’s use a certain number of pyrotechnic devices during their missions. During
flights, stages and boosters separation of the launcher are performed by means of fast cutting devices using high explosive
charges (electro-explosive devices or EED). Detonation of these cutting devices produces a so-called pyrotechnic shock which
causes the dynamic loading of the launcher structure and its components. In order to verify critical components, such as electro-
pneumatic components, it is necessary to perform pyrotechnic shock qualification tests. Nowadays, engineers work by try and
error in order to reach the pyroshock test specifications required by launch vehicles and spacecraft’s designers. This can lead to
the realisation of many preliminary tests and becomes time consuming and expensive. This paper deals with the numerical
simulation of the pyroshock test using the finite element code ANSYS-AUTODYN®. For this work, the test setup consists of a
steel resonant plate subjected to a free air blast load. The dynamic response of the plate is recorded by means of a set of shock
accelerometers located on the upper surface. This paper focuses on the analysis of parameters that can have any influence on the
shock response spectrum (SRS) such as mesh size, number of blast loading zones, peak pressure time and the arrival time. The
obtained results show that all these four parameters influence the SRS and that the choice of a good set of parameters can lead to
a good correspondence between numerical and experimental results.

KEY WORDS: Pyroshock; Shock response spectrum; Numerical simulation, ANSYS-AUTODYN®.

1 INTRODUCTION shock accelerometers. The experimental tests are performed in


Although pyroshocks rarely damage structural members, order to obtain reference data that are used for the validation
many flight hardware failures in launch vehicles and of the finite element models (FEM’s). In the FEM’s, the real
spacecrafts have been attributed to pyroshocks [1]. This can blast pressure profile is replaced by a triangle profile in
occur in electronic and electro-pneumatic components that are keeping the same main blast characteristics (peak
sensitive to the high-frequency pyroshock energy. In order to overpressure and impulse).
verify spacecraft’s critical components, it is necessary to
perform pyrotechnic shock qualification tests in laboratories.
2 PYROTECHNIC SHOCK TESTS
Most shock specifications in the aerospace industry are
given in terms of SRS’s and the SRS is defined as the The pyroshock tests are prepared and conducted in the
maximum response from a single degree-of-freedom system Laboratory for the Analysis of Explosion Effect (LAEE) of
(SDOF) over the frequency range of interest to the applied the Royal Military Academy. The test room of the LAEE can
shock. The SRS is useful in the verification of the severity of be used for the detonation of non confined charges up to 150
different pyrotechnic shock loads and the response used for gram equivalent TNT and for test components of up to 25 kg.
pyroshocks is the maximum absolute acceleration (positive 2.1 Test configuration
and negative). [2, 3]. In general, manufacturers collect real
shock levels on different components by means of real scale The test setup consists of a square steel plate suspended in the
model tests use them as design specifications. These shock horizontal position by means of 4 elastic cords (figure 1). The
levels have then to be reproduced in laboratory conditions elastic cords simulate free boundary conditions and avoid the
using high explosive charges. Since the number of test transfer of vibrations and shocks to the laboratory walls. The
parameters (type, location and quantity of explosive, standoff plate is made of mild steel and has a thickness of 15mm and a
distance, type and dimensions of the shock plate, …) is surface of 990x1000mm.
important, this can lead to a large number of preliminary tests. The pyroshock tests are performed in free-air blast
To reduce the number of experimental tests, numerical configuration and only spherical explosive charges of C4 are
simulations can be carried out in order to determine the first used. The reference test is realised with a spherical explosive
test parameters (type, location and quantity of explosive, charge of 20gr C4 with a standoff distance of 11cm.
standoff distance, type and dimensions of the shock plate).
This paper deals with the simulation of one type of
experimental pyroshock test setup. This setup consists of a
steel resonant plate subjected to a free air blast load and the
dynamic response of the plate is recorded by means of a set of
Proceedings of the 8th International Conference on Structural Dynamics, EURODYN 2011 3346

Elastic cords Figure 3 presents the steel plate and numbers 1 to 6


represent the locations of the shock accelerometers. The
Steel plate accelerometers are screwed in the plate in order to ensure a
Accelerometers perfect contact between the plate and the transducers.

2.2 Measurement system and test procedure


During the pyroshock tests, the response signals are recorded
by means of 6 shock accelerometers PCB, type 3054A,
characterized by a sensitivity of 0.477mV/g. The signals from
accelerometers are first conditioned by a Brüel & Kjær
amplifier, type Nexus 2693, DeltaTron conditioning
amplifiers units, before they are sent to the computer. The
amplified signals are simultaneously sampled with two 14-bit
Figure 1: test facility of the LAEE
data acquisition cards type SPECTRUM MI.4022, installed in
an industrial computer. The sample rate of 1MHz is adopted
The explosive charge is suspended from the plate by means
with a length of 32768 data points to avoid the problem of
of the detonator power cables (figure 2). For this work, the
aliasing.
spherical shape of the explosive charge is chosen for the
facility of modelling in the numerical simulations. In this 5 pyroshock tests are performed using the same explosive
paper, the case of contact explosions is not considered charge of 20gr C4. For each measurement point and each test,
a SRS is calculated. The average of the 5 SRS’s is used for the
validation of the numerical models while the envelope curves
Explosive charge of maxima and minima of these SRS’s are used as the
confidence interval.

2.3 Test results


The shock response spectra are calculated from measured
acceleration signals by considering a damping ratio of 5 %
(Q=10) using a script written in MATLAB® [4]. The SRS’s
are calculated, from 80Hz to 10 kHz, with a resolution of 48
points per octave. This frequency range corresponds to the
range generally encountered in the pyroshock specifications
[4]. Figure 4 presents the measured acceleration signal at the
Figure 2: shape and position of the explosive measurement point 4, while figure 5 presents the
superposition of the 5 SRS’s obtained from the 5 pyroshock
6 measurement points are defined on the half upper surface tests (point 4).
of the steel plate. Due to the symmetry of the plate, the
measured signals on the half surface can be reported on the
other half part of the plate (figure 3).
1000

1 4
680
2 5
990

500

320 3 6

Figure 4: Acceleration signal at point 4


140

320

Figure 3: measurement points


Proceedings of the 8th International Conference on Structural Dynamics, EURODYN 2011 3347

3 NUMERICAL SIMULATIONS
The main objective is to develop a numerical model allowing
to predict the dynamic response of a steel plate subjected to
pyroshocks. The SRS will be used as the validation tool and
the numerical simulations are performed using the explicit
finite element code ANSYS-AUTODYN®, version 12[6].
3.1 Finite element model (FEM)
3.1.1 Steel resonant plate
The finite element model of the steel plate is made using the
3D ANSYS-AUTODYN shell solver. As the frequency range
considered for the analysis runs from 80 Hz to 10 kHz, the
number of cells to consider should allow a correct description
of the dynamic response of the plate in this frequency range.
Figure 5: 5 SRS’s calculated from acceleration signals Wattiaux has shown that for this kind of plate, and for this
measured at point 4 frequency range, a finite element model with a grid of at least
56x56 elements is sufficient [7].
Figure 6 presents the calculated SRS (average) and the The mechanical properties used in the numerical models have
minimum and maximum envelope curves. The peak been validated by comparing eigenfrequencies, from
difference between the max and min SRS does not exceed experimental and numerical modal analyses, of the first 15
6dB (see figure 7), which is specified as the test control eigenmodes (0-650Hz). This analysis is limited to the first 15
tolerance in the NASA STANDARD PYROSHOCK TEST eigenmodes because of difficulties in the identification of the
CRITERIA [2]. This shows the repeatability of the higher experimental mode shapes. The result of this analysis
experimental pyroshock tests. leads to a steel plate with a Young Modulus of 200GPa and a
density of 7850k/m³. A linear equation of state and the
Johnson Cook strength model are used in order to complete
the description of this material in ANYS-AUTODYN.
3.1.2 Numerical approach: explosion phenomenon
In general, Euler and Lagrange solvers are used for the
simulation of the interaction between blast waves and
structures in explicit finite element codes. In that case, the
Euler solver is used for the description of the detonation
phenomenon and the blast wave propagation and the Lagrange
solver is used to represent structures. The two solvers are then
combined by defining a solver interaction [6]. For pyroshock
tests, Mauleon and Ben Ali have shown that this way of
working is very expensive in computation time and does not
Figure 6: Min Max SRS envelope curves allow to reach the acceleration levels observed in experiments
[8, 9]. In this paper, the explosion phenomenon is replaced by
its action on the resonant plate, i.e. distributions of the blast
overpressure and impulse.
3.1.3 Determination of the blast loading
The blast load on the resonant plate is determined by using the
module “Load on Structure” of the CONWEP software [10].
This module allows the calculation of the blast overpressure
and impulse distributions on the surface impinged by the blast
wave. Figure 8 and 9 presents the blast overpressure
distribution for an explosion of 20gr of composition C4 placed
in the centre at a standoff distance of 11cm. The pressure and
impulse distributions are circular and make it easier the
modelling of the blast loading in ANSYS-AUTODYN. The
number of zone on these figures can be adapted by the user in
order to optimize the simulations. In the following, it will be
Figure 7: Difference Min-Max SRS shown that the number of zones influences the SRS’s.
Proceedings of the 8th International Conference on Structural Dynamics, EURODYN 2011 3348

Figure 10: real blast overpressure profile


Figure 8: impulse distribution on the steel plate
P(t)

Pm
i

t
tA tP tA+td(tE)

Figure 11: triangular blast overpressure profile

Figure 9: overpressure distribution on the steel plate In ANSYS-AUTODYN, the blast loads are applied as a series
of stress boundary conditions (SBC). This means that, for a
3.1.4 Blast load in ANSYS-AUTODYN number of zones considered, the same number of stress
boundary conditions have to be defined. Each boundary
The FEM of the steel plate is divided in a number of zones condition is characterized by four parameters: peak
corresponding to those defined in the CONWEP code; each overpressure (Pmi); start time (tAi), peak time (tPi) and end time
zone being loaded with a blast load corresponding to those (tEi) (see figure 11). For each zone, the value of the peak
calculated by CONWEP. In ANSYS-AUTODYN, the real overpressure considered is the linear average of its border
blast loading profile (figure 10) is simplified in a triangle values; the start time corresponds to the arrival time tA, the
(figure 11) by keeping the same blast loading main peak time varies between the stat time and the end time and is
characteristics: peak blast overpressure (pm) and blast impulse calculated in order to obtain the value of impulse calculated
(i). In the triangular profile, the value of the peak with CONWEP. Figure 12 describes the parameters used in
overpressure, pm, is those obtained from CONWEP, but the the definition of the ANSYS-AUTODYN stress boundary
positive phase duration, td, is calculated in such a way that the conditions.
impulse is kept the same. Equation (1) presents the formulae
used for the calculation of the new positive phase duration. It 3.2 Numerical results
should also be noted that the negative impulse is neglected The numerical accelerations are captured by means of 6
here. numerical gauges placed on the FEM of the steel plate at the
same locations as in experiments. The obtained numerical
accelerations are used for the calculation of the numerical
2i SRS’s that are at the end compared to the experimental
td = (1) results. 5 sensitivity analyses on the SRS are performed: FEM
pm mesh size, number of blast loading zones, arrival and peak
time of the blast load.
Proceedings of the 8th International Conference on Structural Dynamics, EURODYN 2011 3349

P ƒ compared to experimental results, the FEM’s


underestimates the SRS in the lower and higher
frequency ranges.
Pm1 This analysis shows that from the mesh size of 300x300 and
above, the deviation in SRS’s is limited at 4.3dB over all the
frequency range. In the rest of this paper, the mesh size of
Pm2 300x300 will be used, because more than 48 hours of
computation time is necessary for larger mesh sizes.
Pm3
Pm4
Pm5
Pm6
Pm7
tA1 tA2 tA3 tA4 tA5 tA6 tA7 t
tE1 tE2 tE3 tE4 tE5tE6 tE7

Figure 12: definition of the 7 stress boundary conditions for


the 7 blast loading zones
3.2.1 Mesh size sensitivity
The importance of the mesh size on the accuracy of numerical
simulation is well known. Within the framework of this study,
a mesh size sensitivity analysis is performed in order to find Figure 13: mesh size sensitivity
the optimal mesh size that allows a correct description of the
dynamic response of the steel resonant plate. For this work, 6 3.2.2 Number of loading zones
mesh sizes are analyzed: 100x100; 200x200; 300x300; The blast load on the steel resonant plate is not uniform; it is
400x400; 500x500 and 750x750. For each case, the SRS’s for very high in the centre (in front of the explosive charge) but
all the 6 points are calculated. Table 1 resumes the input quickly decreases with distance. This evolution is simulated
parameters used in the definition of the SBC’s. It should be by defining a number of loading zones on the steel plate (see
noted that the arrival time is the same for all SBC’s and that figure 12). Theoretically, this number is infinite, but in
there is no delay between the arrival time and the peak time. practice it is limited by the number of CONWEP colours, i.e.
Table 1: input parameters: mesh sensitivity 24. In this paper 6 cases are investigated: 1, 3, 5, 7, and 10
loading zones. The input parameters are those presented in
Loading Peak Arrival Peak End table 1, but the mesh size is kept constant at 300x300.
Zones pressure time time time Figure 14 presents the evolution of the SRS at the
[MPa] TA [ms] TP [ms] TE [ms] measurement point 4 in function of the number of loading
1 50 1E-5 1E-5 3.4E-5 zones considered. This figure shows that the FEM using:
2 30 1E-5 1E-5 3.3E-5
• 1 loading zone overestimates the SRS in all the
3 15 1E-5 1E-5 3.1E-5 frequency range (maximum deviation: +15dB);
4 4 1E-5 1E-5 6.0E-5
• 3 loading zones leads to SRS’s close to the
5 1.25 1E-5 1E-5 1.14E-4 experimental results but overestimates the SRS’s in
6 0.5 1E-5 1E-5 1.90E-4 the intermediate frequency range (max deviation:
7 0.2 1E-5 1E-5 2.10E-4 +11dB; -7dB );
• 5 loading zones is the configuration that gives the
Figure 13 presents the calculated SRS’s for the 6 mesh sizes closest SRS’s to the experimental one (max
at the measurement point 4 (for the location of the deviation: +3.8dB; -10dB);
measurement points, see figure 3). This figure shows that:
ƒ in the lower and intermediate frequency ranges (80- • 7 loading zones underestimates the SRS in the higher
500Hz; 500Hz-3kHz) and for the mesh sizes considered, frequency range (deviation: +1.2dB; -13dB).
the influence of the mesh size on the SRS is small and • 10 loading zones underestimates the SRS’s in the
limited at about 3.3dB; higher frequency range (deviation: +5dB, -19dB).
ƒ in the higher frequency range (3kHz-10kHz), the
influence of the mesh size on the SRS reaches a value of This analysis shows that the best results are obtained when
5.3dB; using 5 loading zones. In this configuration, the difference
between experimental and numerical SRS’s is limited at
Proceedings of the 8th International Conference on Structural Dynamics, EURODYN 2011 3350

+3.8dB and -10dB. In the following analyses, only 5 loading


zones are considered.

Figure 15: influence of the arrival time on the SRS


3.2.4 Peak time
Figure 14: influence of the number of loading zone on the The objective of this section is the evaluation of the influence
SRS of the peak time on the SRS. 7 cases are considered as shown
in figure 16. For this analysis, the arrival time tA, the end time
3.2.3 Arrival time
tE, the peak pressure Pm and the impulse i are kept constant,
The blast wave does not reach all the surface of the steel only the peak time varies. 5 loading zones are considered and
resonant plate at the same time. As the FEM of the steel plate the mesh size of 300x300 is used.
is divided in blast loading zones, each zone is characterized by
its arrival time. In this section, the effect of the delay in the Figure 17 presents the calculated SRS’s. This figure shows
arrival time on the SRS is analyzed. Table 2 presents the input that the influence of the peak time in the lower frequency is
parameters for this study. 5 loading zones are considered and limited at 2dB in the lower and intermediate frequency ranges
the mesh size of 300x300 is used. (80Hz-3kHz) and reaches a value of 4dB in the higher
frequency range.
Table 2: input parameters: evolution of the SRS vs. arrival
time
P
Loading Peak Arrival Peak End
Zones pressure time time time
[MPa] TA [ms] TP TE [ms] 1 2 3 4 5 6 7
1 50 3E-5 3E-5 5.4E-5 Pm
2 25 4E-5 4E-5 6.6E-5
3 6 9E-5 9E-5 1.3E-4
4 1.25 1.9E-4 1.9E-4 2.94E-4
5 0.3 3.2E-4 3.2E-4 4.8E-4

t
Figure 15 shows that when considering the same arrival
time for all 5 loading zones, the SRS is underestimated in the tA tPi tE
lower and higher frequency range. The best results are
obtained when considering different arrival time for each
loading zone. The maximum deviation is in this last case Figure 16: peak time: cases considered
limited at ± 5dB.
Proceedings of the 8th International Conference on Structural Dynamics, EURODYN 2011 3351

Figure 17: evolution of the SRS in function of the peak time Figure 19: deviation between numerical and experimental
SRS
4 COMPARISON OF NUMERICAL AND
EXPERIMENTAL RESULTS 5 CONCLUSION
Comparing numerical and experimental results, these analyses The numerical simulation of one pyroshock test setup is
have shown that the best numerical results are obtained with a studied in this paper with the finite element code ANSYS-
FEM characterized by: AUTODYN. The test setup consists of a steel plate subjected
to a free air blast load. The explosive charge of 20gr C4 is
ƒ a mesh size of 300x300;
placed in the centre and at a standoff distance of 11cm from
ƒ 5 blast loading zones on the steel resonant plate; the plate. In the simulations, the blast load is replaced by a set
of 5 stress boundary conditions and the steel plate is modelled
ƒ different arrival time for each blast loading zone;
with the 3D ANSYS-AUTODYN Shell solver. 4 sensitivity
ƒ and a correct peak time for each loading zone. analyses on the SRS are performed: FEM mesh size, number
of blast loading zones, arrival and peak time of the blast load.
Figure 18 presents the experimental SRS superposed on the Next to numerical simulations, some experimental tests are
numerical SRS calculated with the above-mentioned performed and their results serve as reference data for the
parameters while figure 19 presents the deviation between the validation of the finite element models. From experimental
two SRS’s. Figure 19 shows that, except the frequency range and numerical acceleration signals, the SRS are calculated and
from 80 to 90Hz where the deviation reaches a value of - compared.
7.1dB, in general the deviation remains smaller than the
experimental confidence interval of 5.3 dB. From numerical and experimental results, it has been shown
that a good choice of the set of the FEM parameters can lead
to a good correspondence between numerical and
experimental results. The observed difference between
experimental and numerical SRS’s does not exceed, on the
one hand 5.3dB which is the confidence interval for the
experimental tests and on the other hand 6dB which is known
as the test control tolerance in the NASA STANDARD
PYROSHOCK TEST CRITERIA. These results show that the
approach followed in this paper by replacing the blast loading
by it equivalent set of stress boundary conditions allow the
good estimation of the dynamic response of resonant plate
expressed in terms of SRS.

REFERENCES
[1] Himelblau, H., Kern, D. L., Piersol, A. G; The Proposed NASA
Pyroshock Test Criteria Standard - Part I, California, USA, 18-Nov-
1996.
[2] NASA TECHNICAL STANDARD, PYROSHOCK TEST CRITERIA,
May, 18, 1999.
[3] Harris C.M., Piersol A.G., HARRIS SHOCK AND VIBRATION
HANDBOOK, fifth Edition, McGraw-Hill, 2002,
Figure 18 comparison of experimental and numerical SRS:
[4] MATLAB, Rev. 2009a, The Mathworks Inc, 2009,
point 4 [5] Reymen B.; Ndambi J.-M.; Vantomme J.; Bachmann J.b; Dreer T.b,
Critical Evaluation of Actual Problems in Pyrotechnical Shock Testing
Proceedings of the 8th International Conference on Structural Dynamics, EURODYN 2011 3352

of Space Equipment, CDROM Proceeding on On-Board Energetic


Equipment 2004.
[6] ANSYS-AUTODYN ver. 12, ANSYS® AUTODYN® Explicit Software
for Nonlinear Dynamics, 2009.
[7] Wattiaux D, Verlinden O, Conti C, De Fruytier C., Prediction of the
vibration levels generated by pyrotechnic shocks using an approach by
equivalent mechanical shock, Journal of sound and acoustics, 2008.
[8] Mauleon S., Numerical and experimental study of a pyroshock test set
up for small spacecraft components, Master Thesis, Royal Military
Academy, 2010.
[9] Ben Ali H., Etude Paramétrique D’un Dispositif De Test
Pyrotechnique, Master Thesis, Royal Military Academy, 2010.
[10] CONWEP: Conventional Weapons Effect; D. W. Hyde; USAEWES / SS-
R;. Collection of conventional weapons effects calculations from the
equations and curves of TM 5-855-1, "Fundamentals of Protective
Design for Conventional Weapons", 20 Aug 1992.

S-ar putea să vă placă și