AIAA 96–3020
1
CFD APPROACH TO FIREARMS SOUND SUPPRESSOR DESIGN
M. Keith Hudson* and Chris Luchini�
Department of Applied Science
University of Arkansas at Little Rock, Little Rock, Arkansas
J. Keith Clutter� and Wei Shyy�
Department of Aerospace Engineering, Mechanics & Engineering Science,
University of Florida, Gainesville, Florida
Abstract
Suppression of muzzle blast is important in both large
and small caliber gun designs. Key goals in the case of
small caliber systems are the reduction in the incidence
of hearing loss due to the acoustic signal and signature
reduction for military applications. Various devices
have been used to reduce the muzzle blast and the design
of these devices have relied heavily on experimental in-
vestigation. The current study evaluates the utility of
computational models in the design of suppressors for
small caliber guns. Experimental measurements are
made for a representative suppressor design and simula-
tions are performed to determine the level of model so-
phistication needed to correctly predict the effects of the
device. The current simulations correctly capture both
the levels and characteristics of the acoustic signal gen-
erated by the bare muzzle and suppressor configura-
tions. These findings support the use of computational
models in the suppressor design process.
Introduction
Devices for the suppression of overpressures from fire-
arms have been known and utilized for some time dating
back to the work of Maxim around the turn of the centu-
ry [1]. Currently, suppressors are used on both large and
small caliber guns for somewhat different purposes. In
the case of large caliber guns, the primary goal of over-
pressure suppression is to reduce the effects of blast on
structures and supporting vehicles. The design process
of the suppression devices has relied heavily on exper-
imental work and the development of empirical data-
bases [2, 3]. Some computational efforts have been un-
* Associate Professor, Member AIAA.
Research Associate, currently at NASA Jet Propulsion Lab
Professor and Chairman, Associate Fellow AIAA.
�
�
�
Doctoral Student, Member AIAA.
Copyright � 1996 by M. Keith Hudson. Published by the
American Institute of Aeronautics and Astronautics, Inc.
with permission.
dertaken [4, 5, 6] but have been limited primarily to
large caliber gun systems.
In the case of small caliber guns, suppressors have been
widely used as clandestine devices in sniper and other
roles in warfare to avoid detection of the shooter. While
this role has been widely accepted for many years other
applications of suppression are being sought, particular-
ly to reduce the acoustic pressure levels from small arms
firing to address hearing loss disability. Interestingly,
while suppression for hearing loss reduction has re-
ceived some study, there has been little reported in the
open literature over the many years that these devices
have seen use. This is most likely due to strict US regu-
lation of these devices in civilian applications.
As in the case of the large caliber suppressors, the design
process for the suppressors has depended heavily on ex-
periments and a cut–and–try procedure. Unlike the large
caliber work, no significant computational effort has
been undertaken. Therefore, the goal of the current
study is to determine the applicability of computational
tools developed for the large caliber suppressors to the
small caliber suppressors. Of primary concern is the
scaling of the blast phenomena and the identification of
the driving physics which dictates the peak overpressur-
es and pressure signals. These two factors are key to the
acoustic signature of the suppressor and need to be cap-
tured by any computational code to be used for suppres-
sor design.
This report summarizes the initial experimental and
computational investigation into suppressors for 22 and
38 caliber / 9 mm guns. The experimental effort tested
a commercial suppressor as well as a cylindrical baffle
design used to evaluate the computational code. The re-
mainder of this document first discusses the experimen-
tal details and highlights some of the predominate
physical occurrences identified. Next the computational
model is reviewed and the simulations for the cylindri-
cal baffle suppressor are presented and discussed. Con-
clusions are then drawn as to the utility of computational
codes in small caliber suppressor design and the driving
physics behind the acoustic signal.
AIAA 96–3020
2
Experimental Investigation
1. Experimental Setup and Description
Firearms suppressor data collection requires that the re-
searcher have a sound insulated laboratory with ade-
quate backdrop for projectile containment, or have the
ability to set up on an outdoor range which has adequate
facilities to support the planned experiment. A suitable
range has been located which offerers a sheltered area
with utilities, but provides an adequate acoustic envi-
ronment to make sound measurements. All testing has
been performed with the instrumentation sheltered from
direct sunlight, but with the firearms muzzle and micro-
phone located just outside of the shelter to avoid direct
sound reflection effects on the collected data.
Figure 1 shows the general layout of the equipment and
tested firearm for all experimental trials. The equip-
ment used includes a Competitor Corp. 38 Special cali-
ber action for all nominally 38 caliber / 9 mm testing and
a AMT Lightning rifle for all 22 LR testing. Both ac-
tions have been modified to allow fitting of a commer-
cial suppressor shell, utilizing a GEMTECH Model Vor-
tec 9 for 38 / 9 mm and a Vortec 2 for all 22 LR testing.
Barrel length on the Competitor action is 10 inches
while the 22 rifle has a length of 20 inches. The cylindri-
cal baffle suppressor dimensions are given in figure 2.
Handloaded ammunition has been used in the 38 / 9 mm
unit consisting of a 160 grain Speer jacketed bullet in a
38 Special casing, over 8.6 grains of Alliance Blue Dot
Powder. The 22 LR has used commercially available
CCI Blazer brand ammunition. During firing, the 38 /
9 mm unit is held on a sandbag, while the 22 rifle is
shoulder fired in the normal manner. Care is taken to en-
sure the same relative alignment of the pressure gages
for each firing.
Acoustic data is collected using a Bruel and Kjaer 4135
condenser microphone powered by a 2801 power sup-
ply. Calibration data indicated that this unit is accurate
to 100 KHz and provides an output of 3.39 mV/Pa. The
microphone is positioned upright (pointed up) on a tri-
pod and positioned between 3 and 20 inches from the
muzzle. The firearm is then positioned to a point paral-
lel to the microphone, and then pulled back up to 10 in-
ches from the microphone to establish a grid of measure-
ments (Table 1). The microphone is read by a LeCroy
Model 9400A, 175 MHz 8–bit digital storage scope.
Computer readouts of the sound tracings during firings
are not available so peak data is recorded by hand. If
there appeared to be two major sound peaks, each peak
is recorded. Measurements from three firings are made
at each gage position. Firings are made with the bare
muzzle in all positions, followed by a similar set of fir-
ings with the suppressor attached. For all experimental
firings, the suppressors consisted of a right circular cyl-
inder body with one copper baffle held in place one third
of the distance down the suppressor body by aluminum
spacers (Figure 3). Limiting firing has been carried out
using the commercial suppressor on the 22 to show the
cylindrical suppressor to be used in the computational
code evaluation produce similar pressure reductions.
2. Experimental Results and Discussion
All the experimental measurements are presented in
Table 2 where ‘‘Sup” denotes the cylindrical baffled
suppressor and ‘‘Com” the commercial suppressor.
Scope traces from the unsuppressed firearms show a
single high–intensity peak with only minimal ringing
type peaks seen over the rest of the measurement period.
This of course correlates with the sharp, high–intensity
crack heard by the ear upon firearms discharge. For the
positions further from the muzzle the sound is seen to di-
minish with distance from the microphone, as would be
expected, and the tracing pattern remains essentially the
same except for the overall intensity changes.
Scope traces for the firearm firings using the cylindrical
baffle suppressor show a characteristic intensity spread-
ing. The large single peak seen with the bare muzzle is
gone, replaced typically by a set of peaks of similar in-
tensity, often by two peaks of almost the same amplitude
especially in the 38 / 9 mm data. The values of the two
peaks are given in Table 2 and are denoted with the 1 and
2 following the suppressor designation. Also for the sup-
pressor configuration, the smaller peaks which appear
as ringing type peaks in the bare muzzle tests are rela-
tively larger when compared to the peak signals. This
is in agreement with the suppressor acting to ”spread”
the discharge sound out over a larger time scale, mini-
mizing the peak value, but giving a longer duration to
the overall sound. Audibly, this is heard by the authors
as a change in the characteristics of the sounds to less of
a crack and more of a loud hissing noise. Also, audibly,
the sound is suppressed to a level where it is not objec-
tionable to the un–protected ear. The control firings
made using the full commercial set of baffles is noted to
be very quiet, although still sounding like a firearm in
general. Another distinct acoustic signal noted during
testing is the sonic crack generated by the supersonic
bullet. This is especially true in the 22 LR trials.
Computational Model
1. Governing Equations
The computational model used for the current study is
a finite volume based computational fluid dynamics
(CFD) code developed to aid in the design of gun
muzzle devices. The governing equations for the gun
blast problem are the full Navier–Stokes equations for
a multi–species chemically reacting flow. The current
study focuses on the inviscid and real gas aspects of the
AIAA 96–3020
3
problem to determine their relative role in the genera-
tion of the acoustic signature. Therefore, the equations
to be solved are the Euler equations for a multi–species
flow with variable specific heats. When discretized, the
equations take the form
�J Q
�t �
�J F
��
��J G�� � J H � 0 (1)
where the dependent variable and flux vectors are
Q �
������
�
�
�u
�v
�E
��1
�
��NS�1
������
�
�
, F �
�������
�
�U
�uU � �xP
�vU � �yP
U �E � P
��1U
�
��NS�1U
�������
�
�
,
G �
�������
�
�V
�uV � �xP
�vV � �yP
V �E � P
��1V
�
��NS�1V
�������
�
�
, H � 1y
������
�
�v
�uv
�v2
v �E � P
��1v
�
��NS�1v
������
�
�
.
(2)
The dependent variable �i is the mass fraction of ith spe-
cies with the fluid being defined by NS total species.
Note that the mass fraction of the NSth species is not ex-
plicitly modeled since the total density is included and
the relationship � ��
NS
i�1
��i holds.
The suppressor design to be simulated are axisymmetric
and therefore the axisymmetric form of the equations is
used and the effects of the third dimension are included
by incorporating the source term H.
The grid Jacobian J and the contravariant velocities are
defined as
J � x�y�� x�y�
U � �xu � �yv
V � �xu � �yv
. (3)
The effects of the projectile are included in the simula-
tion by making a constant velocity assumption and de-
termining at each time interval the appropriate location
of the projectile. The cells which contain the projectile
are identified and an additional source term is added to
denote the appropriate volumetric change and imperme-
able surface boundary conditions are applied to model
the projectile’s surface.
2. Gas Properties
The equation of state is derived by assuming the ideal
gas equation is valid for each species and has the form
[7]
P � �RuT�
NS
i�1
�i
Mi
(4)
The temperature during the calculations must be ex-
tracted from the conserved quantity of internal energy
using the relationship
e ��
NS
i�1
�ihi �
P
�
hi � hofi � �
T
TR
CpidT
(5)
where TR is the reference temperature for the gas prop-
erties. The specific heat, Cpi, of each species is a known
function of temperature. The representation of specific
heats can vary from assuming they remain constant to
a quadratic dependence on T. If a high order function is
used for Cpi then an iterative procedure must be used to
extract the temperature in each cell at each time level.
Here, a compromise between efficiency and sophistica-
tion is made by representing Cp as a linear function of
T over the temperature range to be encountered during
the simulations. By using the linear relationship, the
temperatures at each point in the field can be extracted
by solving a simple equation while introducing the ef-
fects of varying specific heats.
3. Fluid Dynamics Operator
The fluid dynamics aspects of the problem are modeled
using an explicit schemes. To maintain second–order
accuracy, the fluid dynamics operator must be second
order and here a prediction–correction scheme is used
of the form [8]
Q* � Qn � �t2 ���F
(1) ���G(1) � Hi , j�
n
Qn�1 � Qn � �t���F(2) ���G(2) � Hi , j�
* (6)
with
��F � Fi�12 , j
� Fi�12 , j
��G � Gi , j�12
� G
i , j�12
(7)
AIAA 96–3020
4
and where the superscripts * and n denote the time level
at which the fluxes are computed and the superscripts
(1) and (2) denote the spatial order of the numerical
fluxes. Note the fluxes in � and � are computed at the
cell faces and the axisymmetric source term is com-
puted based on the cell average.
The scheme used to define the inviscid numerical fluxes
is the Steger–Warming flux vector splitting algorithm
which has been extended to model multi–species flows
[9]. The flux vector splitting algorithm decomposes the
inviscid fluxes into non–negative (K+) and non–positive
(K–) components based on the eigenvalues of the Jaco-
bian A � �F�Q and likewise for G. The split fluxes take
the form
K� � ��1K1 � ��2K2 � ��3K3 (8)
where the eigenvalues are
��k �
1
2
��k � |�k|�
�1 � �k
�2 � �k � c|�k|
�3 � �k � c|�k|
(9)
with
k � k
~
xu � k
~
yv
k
~
x �
kx
|�k| k
~
y �
ky
|�k|
|�k| k2x � k2y�
(10)
The split flux components are
K1 �
�� 1
�
�������
�
�
u
v
�ht � c2�� 1�
�1
�
�NS�1
�������
�
�
K2,3 �
1
2�
��������
�
�
�u � k~ xc�
�v � k~ yc�
�ht � kc�
�1
�
�NS�1
��������
�
�
(11)
The above formulation gives K=F when k=� and K=G
when k=�. For the multi–species chemically reacting
flow, c is the frozen speed of sound where
c2 � � �P�
� and � is the effective specific heat ra-
tion.
As indicated in equation 7, the fluxes are evaluated at
the cell faces and are either 1st or 2nd order representa-
tions. The flux at the face is a function of the states in
the neighboring cells and can be symbolically repre-
sented by
F
i�12 , j
� F��QL
i�12 , j
�� F��QR
i�12 , j
� (12)
If a 1st order spatial representation is used, then
QL
i�12 , j
� Qi , j , QRi�12 , j � Qi�1 , j . To achieve
2nd order accuracy, a MUSCL approached is used in
which cell–center values are extrapolated to the inter-
faces [10]. Also, to guard against the interpolation
introducing any nonphysical extremes into the field in
the region of large gradients, a limiter must be used. The
formula for the neighboring states takes the form
QL
i�12 , j
� Qi , j ���i�12 , j
QR
i�12 , j
� Qi�1 , j ���i�12 , j
(13)
where the limiting function is
��
i�12 , j
�
li�1 , j
2 mmod
�Q�i�1 , j,�Q�i�1 , j
��i�12 , j
�
li , j
2 mmod
�Q�i , j,�Q�i , j
(14)
with
�Q�i , j �
2�Qi�1 , j � Qi , j�
li�1 , j � li , j
�Q�i , j �
2�Qi , j � Qi�1 , j�
li , j � li�1 , j
(15)
Here the popular minmod limiter is used where
mmod [X, Y] �
sign(X) max[0., min(|X|, Ysign(X))] . (16)
Note l i,j , the cell–length, is used to provide weighting
for nonuniform grid spacing. The same extrapolation
procedure is carried out for the fluxes in � and can be
performed on either the dependent or primitive vari-
ables. Previous investigations have shown that using
primitive variables gives better performance for flows
AIAA 96–3020
5
with strong shocks and this is the method used here [10].
4. Boundary Conditions
The present predictive code has been designed to model
the launch phase of the ballistics problem and therefore,
it is assumed that boundary conditions near the muzzle
exit are known. This alleviates the need to recompute
the interior ballistics phase for each computations
which reduces the computational time when conducting
design studies for muzzle devices. Typical boundary
condition information needed includes temperature,
pressure, and velocity time histories near the muzzle
exit as well as the gun propellant used. This information
can be obtained either from an interior ballistic code or
from experimental measurements. For the current study,
the simulations were carried out in parallel to the experi-
ments so some assumptions had to be made as to the
boundary conditions. The exact boundary conditions
achieved during the experiments may vary somewhat
from those assumed, however, the relative effects of the
muzzle device should be evident in the simulations.
The particular boundary conditions used for the simula-
tions of the 38 / 9 mm are a peak pressure of 6,000 psi,
peak velocity of 1000 fps, and a peak temperature of
2400 F. It is assumed that all quantities decayed to atmo-
spheric conditions over a time period of approximately
4 ms. For the 22, the peak pressure is lowered to 2,000
psi but the remaining variables were kept the same.
The simulations presented here model the flow field as
a combination of three species, these being the O2 and
N2 found in the ambient air and the gun propellant gas.
The properties for oxygen and nitrogen are available in
various sources [11]. The gun propellant is known to be
composed primarily of the active agents CO and H2 as
well as the inert N2 and to a smaller extent the combus-
tion products H2O and CO2 resulting from the interior
ballistic process. Therefore, the properties used for the
gun propellant (F) are formulated to represent a mixture
of CO and H2 and the boundary conditions imposed near
the muzzle exit specify the mass fraction to be
�F� .64 and �N2 � .36. These assumptions which
simplify the gun gas composition are done to reduce the
number of governing equations. Similar processes have
been used previously with good results even when fur-
ther combustion is included in the modeling [6].
5. Results and Discussions
The only experimental data available for code evalua-
tion is the peak pressures measured in the experiments.
Therefore, the only judgement as to the utility of the
computational code that can be made is whether the
code correctly simulates the general effect of the sup-
pressors in reducing the pressure levels and in turn the
acoustic signal. This data can also be used to determine
if the inviscid and real gas effects being modeled are
dominate players in the determination of the peak pres-
sures and the acoustic signals. The data from the experi-
ments and simulations are presented with respect to the
gage location. The locations of the gages are given in
table 1. The distances are measured from the exit of the
muzzle in the cases with no suppressor and from the exit
of the suppressor when it is used.
A comparison between the simulated and measured
pressures for the bare muzzle 38 / 9 mm is presented in
figure 4 as well as data for the 38 / 9 mm with the sup-
pressor present. The curves denoting the experimental
measurements are fit to the average of the three firings
made for each configuration and gage location. During
the firing with the suppressor, two distinct peaks were
measured by the gages