Professional Documents
Culture Documents
Abstract
Accurate and bounding predictions of pressure decay rates for gas containing ventilated cavities that are
exposed to time-varying source pressures are a subject of interest for many engineering problems and are often a
significant source of pressure induced loading on local structures. This type of problem is routinely encountered on
launch vehicles and centers around predictions for payload cavity pressure decay during ascent; where the relative
rate of pressure decay between the local free-stream and internal cavity pressure will drive local differential
pressure levels on cavity walls and internal payloads. Due to the relatively large volumes to vent areas involved
the current state of the art used to predict these gas flows for multiple simulation trajectory sets, which can
contain thousands of individual trajectory ascent paths, is to couple flow discharge coefficient models to ideal
solutions of the 1-D steady energy equation at the vent and 1-D unsteady energy equation for the volume as
opposed to to running transient CFD simulations, due to efficiency and timing demands. The overarching goal
of this paper is to bring together various validated compressible flow discharge models in one text as a resource
for the fluid flow modeling community. We will discuss and present various discharge coefficient models that have
successfully predicted the change in venting rate as the flow through the vent transitions to choked flow, as well as
the sharp reduction in venting efficiency as the vent flows Reynolds number drops below 10,000. Both flight and
detailed CFD simulations are utilized to confirm the appropriateness of these predictions methods. In addition
since the fidelity of the external surface pressures near the vent are of equal importance to the overall gas flow
prediction, a discussion on the observed rapid change in surface pressure transonically as the Mach wave passes
over the vent will also be included with supporting information.
1 Nomenclature
Symbols & Definitions
RANS Reynolds Averaged Navier Stokes Equations
OSC Orbital Sciences Corporation
WT Wind Tunnel
CV Control Volume
p Pressure
ρ Density
U, u, or v Velocity
e Internal energy
h Enthalpy
υ Specific volume, υ = 1/ρ
Cp Constant pressure specific heat or Coefficient of pressure
T Temperature
w Work
q heat addition or dynamic pressure
Cv Constant volume specific heat
V Volume
R Gas constant
γ Ratio of specific heats
S Entropy
ṁ Mass flow rate
Mm or MW Molecular mass or weight
mol Moles
µ Viscosity
k Thermal conductivity
A Area
Cd Discharge coefficient
To Total temperature
L Length, for orifices represents thickness of orifice
D Diameter
c Speed of sound
n Normal vector for mass flow rate direction
Cf Skin friction coefficient
τw Wall surface shear stress
L Thickness of orifice,
D Diameter of orifice
Sharp Edge Orifice L/D ≤ 0.05
Thick Orifice L/D > 0.05
Subscripts
i Ideal
c Cross flow
a Actual
mf Mass flux
t Tank
u Universal or Unsteady
s Steady or Static
∞ freestream value
2 Introduction
The problem of defining accurate and bounding pressure decay rates for gas containing ventilated cavities is routinely
encountered on launch vehicles and centers around predictions for payload cavity pressure decay; where the relative
rate of pressure decay between the local free-stream and internal cavity pressure will drive local differential pressure
levels on cavity walls. The local pressure difference will in-turn drive local loads and stress levels that must be
accounted for to ensure a robust vehicle structural design. In addition, the local rate of pressure decay is important
to the payloads being delivered to orbit, since many of these delicate instruments have pressure decay limits derived
from their own expected venting rates and structural limits. It is not uncommon to expect pressure decay requirements
of 0.2psi/sec to 1.0psi/sec for satellite type payloads being transported to orbit. In addition to adhering to maximum
payload cavity depressurization limits, which are typically of most concern during the transonic portions of flight
due to rapid changes in external pressure as the flow becomes sonic, another area of interest has been predicting
the remaining pressure near fairing separation which can affect not only the structural loads on the payload but
the dynamics of fairing separation. The difficulty in the former prediction mainly rests on accurately determining
the external source pressure through the transonic flow regime, where local body geometry and flow conditions
can dramatically affect when the local flow reaches sonic conditions. In the case where we obtain local supersonic
”bubbles” due to super-critical air-flow around the vehicle, not only is the lower pressure in the supersonic ”bubble”
important, but the increasing pressure at the strong shock that terminates this bubble should also be considered.
Now for the latter situation of predicting pressure and depressurization rates near end of mission, the chief issue
stems from the observed dramatic decrease in venting efficiency in the regime of low differential driving pressure
and/or low fluid densities, which squarely places the flow into a low Reynolds number regime. For these types of flow
conditions we can expect viscosity to now become a significant mechanism to impede fluid flow through the vent.
Two general models have typically been used based upon multiple experimental data sets. The first model utilizes
a density based correction which is simply a function of internal pressure of the cavity while the second model’s
correction is based upon local Reynolds number through the vent. In addition to Reynolds number based models,
discharge models as a function of pressure ratio and vent aspect ratio as-well as cross flow momentum over the vent
are outlined. Validation cases based upon comparisons to flight data and detailed CFD simulations are also outlined
and reveal the interesting physical mechanisms occurring as the flow passes through the vent. We will first review the
1D venting equations for compressible flow used to define an idealized mass flow and cavity depressurization rates
then we will look at methods to correct for real world affects, an finally we will validate this methodology against
flight data and uRANS CFD simulations with a discussion of the importance of modeling fidelity for external vent
pressures near sonic flight conditions.
p1 u2 p2 u2
+ e1 + 1 = + e2 + 2 (1)
ρ1 2 ρ2 2
υ = 1/ρ (2)
h=p·υ+e (3)
u21 u22
h1 + = h2 + (4)
2 2
We then apply constant specific heats and set one side of the control volume to its total conditions to arrive at
the most important expression in the venting process; one that will allow for the prediction of gas velocity through
the vent or control volume.
u2
Cp · T + 1 = Cp · T0 (5)
s 2
T
u= 2 · Cp · T0 · 1 − (6)
T0
We now utilize the perfect gas relationship for specific heat and relationships for the isentropic change of state
for a perfect gas as determined from the 1st and 2nd laws of thermodynamics for an adiabatic and reversible process.
From the 1st law of thermodynamics where the only work considered is due to displacement of the boundary.
dq − dw = de (7)
dw = p · dV (8)
dq − p · dV = de (9)
Now if we define a process under constant volume we can define a constant volume heat capacity property of the
fluid.
dq = de|V (10)
dq de
= = Cv (11)
dT dT V
Similarly, if we define a process under constant pressure we can define a constant pressure heat capacity property
of the fluid.
dq − p · dV = de (12)
∆q = p · (V2 − V1 ) + (e2 − e1 ) = h2 − h1 = ∆h (13)
dq dh
= = Cp (14)
dT dT P
Now we can relate Cp and Cv through the definition of enthalpy with the assumption of an ideal gas equation of
state.
dh = Cp · dT = de + p · dV (15)
de = Cv · dT (16)
p · dV = R · dT (ideal gas equation) (17)
Cp · dT = (Cv + R) · dT (18)
Cp = (Cv + R) (19)
Now we introduce the definition of the ratio of specific heats, γ = Cp /Cv , and can define Cp and Cv through the
gas constant and ratio of specific heats.
γ·R
Cp = (20)
γ−1
R
Cv = (21)
γ−1
Now if we substitute the 2nd law of thermodynamics into the 1st for a reversible process we arrive at the following
equations. Since we are defining a process that does work as the gas expands by the inclusion of the p · dV term
(i.e. as the volume of the gas increases or density ρ = 1/υ decreases, dV is positive then p · dV is positive and work
is done). It is interesting to note that if the process was an isenthalpic gas expansion, then the elimination of the
p · dV term would result in no predicted gas temperature change.
dq − p · dV = de (22)
dq = T · dS (23)
h = e + p · dV (24)
dh = du + p · dV + V · dp (25)
T · dS = dh − V · dP (26)
dT dp
dS = Cp · −V · (27)
T T
Now we again substitute the ideal gas equation in the above equation.
dT dp
dS = Cp · −R· (28)
T p
(29)
We now can integrate the above equation with dS equals zero (i.e. reversible process) to arrive at the famous
isentropic gas expansion relationships which now allows us to relate changes in temperature to changes in pressure.
γ γ
γ−1
p0 ρ0 T0
= = (30)
p ρ T
(31)
More generally and compactly the above equations can be re-written as follows and are generally referred to as
polytropic process equations.
P = c · ργ (32)
1
ρ=c·T γ−1 (33)
γ−1
T =c·P γ (34)
Now we can utilize the isentropic process equations along with the velocity equation from the 1D energy balance
equation to arrive at a formulation for velocity and mass-flow rate for non-choked flows as a function of upstream
conditions and downstream pressures.
v
u
u 2 · γ · R · T0
" γ−1 #
p γ
u=t · 1− (35)
γ−1 p0
ṁnon-choked = ρ · A · u (36)
v
γ1 u
u 2 · γ · R · T0
" γ−1 #
p p γ
= ρ0 · ·A·t · 1− (37)
p0 γ−1 p0
To u2 u2 · (γ − 1) γ−1
=1+ =1+ =1+ · M2 (38)
T 2 · Cp · T 2·γ·R·T 2
γ γ−1 γ γ
γ−1
p0 ρ0 T0 γ−1 2
= = = 1+ ·M (39)
p ρ T 2
(40)
Therefore at M=1,
1
γ−1 s
2 2
ṁchoked = ρ · A · u = ρ0 · ·A· γ · R · T0 · (41)
γ+1 γ+1
s γ+1
γ−1
2
= ρ0 · A · γ · R · T0 · (42)
γ+1
(43)
Since we would like to determine when choked flow occurs, we can simply compare pressure ratio values for
velocities equal to Mach 1 in the above equation. Therefore if,
γ
− γ−1
p γ+1
< , for example if γ=1.4, then p/p0 = 0.528 (44)
p0 2
Then the flow is choked with a maximum velocity set equal to Mach 1.
Z XZ
∂ ∂ X
ρ · dV + ρ · υ · dA = (ρ · V ) − ṁ = 0 (45)
∂t i=1 A
∂t
V A
i
Z t Z tX
∂
mt = m0 + (ρ · V ) · ∂t = m0 + ṁ = 0 (46)
0 ∂t 0 A
mt · Rt · Tt
pt = (47)
Vt
Z Z Z Z
∂
ρ · e · dV + ρ · u · e · n · dA + p · u · n · dA + p · ẋ · n · dA = Ẇ (48)
∂t |{z}
v A A A energy source
| {z } | {z } | {z } | {z }
change in energy of CV energy entering CV P ·dV work moving boundary work
Z Z Z Z
∂
ρ · e · dV + e · ρ · u · n · dA + p · υ · ρ · u · n · dA + p · ẋ · n · dA = Ẇ (49)
∂t | {z } | {z }
v A ṁ A ṁ A
∂ X X X
(ρ · e · V ) − ṁ · ein − ṁ · p · υin − p · ẋ · A = Ẇ (50)
∂t
A A A
For a non-chemically reacting ideal gas, internal energy and enthalpy are functions of temperature only and the
gas is referred to as thermally perfect. In addition if the specific heats are constant the gas defined as calorically
perfect [1]. Using these concepts with the relationships for enthalpy and specific heats presented earlier we rewrite
the above equation with-out explicit internal energy or enthalpy variables.
Figure 2: CFD Flow Field Simulations for L/D Effects: Pe/Po=0.98 (sharp edge vs. L/D=2)
line to eliminate the downstream pressure fields ability to send acoustic signals to the upstream flow field and alter
its expansion. The data in the figure below from the experimental data of Deckker and Chang’s [3], Calligan and
Bowden’s [2], and this author’s analytical steady RANS CFD simulations show that as the pressure ratio across the
vent is increased past sonic or choke flow conditions, that the length to diameter effects are generally eliminated.
One note of caution, as the length to diameter of the orifice becomes very large (>> 1), significant venting efficiency
losses are realized through out the entire pressure ratio regime due to the momentum losses associated with boundary
layer growth. The next section will provide additional trends for very long tubes.
summarize the methodology with example results presented for various pressure ratios between the two reservoirs,
where the downstream reservoir’s pressure was set to 14.7psi. In the figure below a reasonable engineering level
match to the more rigorous two-dimensional CFD based solutions is clearly evident.
The Fanno equations for pipe flow are based upon a one dimensional or x-directional momentum principle that
includes the effect of the two dimensional boundary layer on the momentum balance through an approximation of
the y-directional velocity derivative as related to the surface shear stress. We can then use our prior knowledge of
the general values of shear stress for a given flow state to relate conditions at one side of the pipe to the other. In
the results presented below the excellent Van Driest II skin friction formulation as defined by White [16] was used
with exit pipe flow conditions (i.e. exit Mach & Reynolds Number) and averaging over the length of the pipe to
approximate the expected surface shear stresses.
So first we take the x-directional Navier Stokes equation with Stokes hypothesis (i.e. mean pressure or average
normal stresses on a fluid element are equal to the thermodynamic pressure), negligible body forces, constant viscosity,
2 2
and ∂∂xu2 << ∂∂yu2 , which is expected with-in the boundary layer where the viscous forces are dominant. We arrive
∂p 2 2
at the following equation, ρ · u · ∂u ∂ u ∂ u
∂x = − ∂x + µ · ∂y 2 . We now rearrange, ρ · u · ∂u + ∂p = µ · ∂y 2 · ∂x and realize
∂u
that we can utilize the definition of shear stress at the wall, τw = µ · ∂y , to approximate the double derivative term,
∂2u −τw /µ−0
assuming that the velocity gradient at the center of the pipe is zero, yielding: ∂y 2 = ∂
∂y · ∂u
∂y = D/2 + 0−τD/2
w /µ−0
=
2 4·Cf
4·τw
D·µ = − ρ·u
2 · D·µ . Now combining this with the above equation we arrive at the common form of the Fanno flow
f −2·ρ·u2 ·C
differential relationship: ρ · u · ∂u + ∂p = D · ∂x . Now following the steps and variable substitutions by
Anderson [1] this differential equation can integrated into the following relationship.
This equation relates the Mach numbers at the sections in tube to the integrated effect of friction between the
sections. Now we can utilize adiabatic (i.e. constant total temperature), continuity, ρ1 ·u1 = ρ2 ·u2 , and the definition
of speed of sound, a2 = γ · p/ρ, relationships to determine temperatures and pressures at the sections as a function
of the change in Mach number between the sections.
T2 2 + (γ − 1) · M12
= (59)
T1 2 + (γ − 1) · M22
1/2
2 + (γ − 1) · M12
p2 M1
= · (60)
p1 M2 2 + (γ − 1) · M22
We now simply calculate a set of mass flow rates for various pipe exit total pressure’s to the downstream reservoir
pressure using one dimensional venting equations, where the predicted venting exit velocity and temperatures are
used as the exit static conditions for use with the Fanno flow relationships. We then calculate the corresponding pipe
entrance conditions with the Fanno flow equations above, and then calculate the mass-flow rate from the upstream
reservoir to the pipe entrance pressure. We then look for the unique condition when the mass-flow rates are equal and
compare this to the expected ideal mass flow rate for flow from the upstream reservoir pressure to the downstream
reservoir pressure with an orifice area equal to the pipe area. Figure 4 illustrates the resulting analytically calculated
discharge coefficients.
An alternative to a explicit Reynolds number model has been previously widely used and therefore is included
here. Investigators [10] [11] [13] [14] have shown through experimental data that tank blow down and inflation can
be well predicted by using 1-D inviscid, compressible, and isentropic flow relationships for determination of mass
flow rate with a simple correction factor applied to the discharge coefficient. The form of this correction factor is:
Cde
|Pi − Pnon2 |
Cd1 = (61)
Pnon1
Where Pi is the current internal tank pressure and Pnon2 is set equal to the driving pressure for an inflation
situation or set to zero for a deflation situation. Pnon1 is also set to the driving pressure for an inflation situation
or set to initial pressure for a deflation situation. Cdc and Cde are experimentally derived coefficients and typical
correlated bounding values for minimum mass flow rates are 0.6 and 0.2 and for maximum mass flow rates are 0.85
and 0.125, respectively.
After investigation of the above equation, a simple interpretation would indicate that the equation reduces the
discharge coefficient as the internal pressure and the forcing pressure approach each other, in other words, as the
differential pressure across the vents approaches zero. This would seem to be in-line with the conclusion made by
Livesay [9], that as the differential pressure approaches zero, flow Reynolds numbers decreases and viscous effects
become important, and the 1-D inviscid-compressible-isentropic flow relationships start to increasingly over-predict
the mass flow rate. His argument rests on the fact that the limit of the inviscidly predicted mass flow rate divided
by the differential pressure as the differential pressure approaches zero goes to infinity.
Now a question the reader must certainly be asking is, which model is the best? Currently this author has shown
very good matches to various experiments with the explicit Reynolds number model and its clear dependence on the
most important parameter regarding viscous and inertia forces make it the clear first choice for modeling purpose’s.
4.5 Cross-Flow
The phenomena of external air momentum interfering with the venting efficiency of simple flush orifice are well
pictorialized by figure 6(a) [6]. The underlining mechanisms are due to the alteration of the external pressure
distribution by the interacting flow streams of the external free stream flow and the exhausting venting flow. Due
to the complexity of the resulting mixing flow field which can be expected to be a strong function of local geometry,
boundary layer state, up-stream Mach number, and exhausting gas properties, models have generally been created
with experimental data sets obtained from wind tunnel test program [2] [4] [15]. More recent investigations [7],
have successfully shown that modern CFD tools can sufficiently predict many facets of this interaction. Figure 6(b)
provides an example of the venting discharge efficiency changes with relative external flow momentum changes for
a sharp edge orifice on a flat plate. This figure is a reprint from the experimental work of Walters, Glasgow, and
Baker [15] overlaid with steady RANS CFD simulations conducted at OSC and clearly shows that as the mass flow
ratio, defined as the mass flow of the exhausting gas through the vent to the mass flow of an external stream tube
of an area equal to the vent, approaches unity, cross-flow effect are virtually eliminated.
Now to illustrate the above concept and to provide a validation point to the 1D venting equations and discharge
models presented earlier, Figure 8 overlays flight and predictions for OSC’s Taurus vehicle. The predictions utilize
the concepts presented earlier with both Reynolds number and pressure ratio dependent discharge models employed.
The models appear to do a great job of not only predicting the magnitudes of the internal fairing or shroud pressure,
but also the depressurization levels. For this situation the transonic Mach wave does indeed define the maximum
depressurization level for the payload as noted as a potential earlier.
As another validation point that also provides insight into the spatial gradients expected in the shroud due to the
venting process, a unsteady CFD simulation (uRANS) was conducted. Figure 9 and 10 compare predicted shroud
pressures and temperatures between uRANS and 1D venting equation results for a typical medium sized launch
vehicle ascent profile with adiabatic shroud walls and no internal energy sources. As can be seen a very good match
to with-in 3% is observed at times near the max depressurization point. The uRANS CFD simulations allow us to
define the zone of influence around the vent where the 1D cavity energy balance for zero momentum air is strictly
valid. For the test case analyzed, the data clearly shows that at location of one diameter or greater away from the
vent inside the cavity, for the case of air leaving the cavity, negligible gradients in pressure are present, which would
validate the use of 1D methods for this problem
dP
(a) Spatial Variation Pressure (b) Spatial Variation dt
dP
(c) dt
6 Conclusion
One dimensional compressible venting equations provide a powerful and very efficient method to predict cavity venting
rates. When these equations are coupled with robust discharge models that depend on vent geometry, pressure ratio
across the vent, and Reynolds number through the vent, a high fidelity simulation of cavity venting can be expected.
While these models and examples provided stem from work done on small to medium sized launch vehicles, it is
expected that these concepts are amenable to any field where compressible gas discharge modeling is required.
References
[1] J. Anderson. Modern Compressible Flow. McGraw Hill, 1990.
[2] Bowden and Callaghan. Investigation of Flow Coefficients of Circular, Square, and Elliptical Orifices at High
Pressure Ratios. NASA TN 1947, 1949.
[3] B. Deckker and Y. Chang. An Investigation of Steady Compressible Flow Through Thick Orifices. Proc Instn
Mech Engrs, Vol 180 Pt 37, 1966.
[4] Dewey and Vick. An Investigation of the Discharge and Drag Characteristics of Auxiliary-Air Outlets Discharging
into a Transonic Stream. NASA TN 3466, 1955.
[5] Fox and McDonald. Introduction of Fluid Mechanics. Wiley, 1992.
[6] J. Francisco. Compartment Venting. NASA-SP-8060, 1970.
[7] Z. Gillcoatt. OSP Venting with Cross-Flow Effects. OSC TAD2075rB, 2009.
[8] I. Idelchik. Handbook of Hydraulic Resistance 3rd Edition. Begal House, 1996.
[9] R. Livesay. Method for calculation of gas flow in the whole pressure regime through ducts of any length. J. Vac.
Sci. Tech, 2000.
[10] Mironer and Regan. Venting of Space Shuttle Payloads. NASA 83-2600, 1983.
[11] Mironer, Urquhart, and Krebs. Venting of Sounding Rocket Payloads. NASA 82-1729, 1982.
[12] Moraes and Pereira. Verification of the Pressure Equalisation Inside the Satellite Compartment of the Brazilian
Satellite Launch Vehicle. J. of the Braz. Soc. of Mech. Sci. & Eng., 2005.
[13] M. Mousseux. Leak Down Test Analysis and Venting Program Calibration. OSC TAD246, 1997.
[14] A. Murri. Payload Venting in Worst-case Shuttle Environment. NASA 87-1598, 1987.
[15] Walters, Glasgow, and Baker. Experimental Determination of Generalized Venting Characteristics. NASA
CR-61241, 1968.