AN INTRINSIC, HETEROGENEOUS MODEL OF COMPOSITE

SOLID PROPELLANT COMBUSTION

by

BRYAN MICHAEL RASMUSSEN

A THESIS

Submitted in partial fulfillment of the requirements

for the degree of Master of Science in Engineering in
The Department of Mechanical and Aerospace Engineering
of
The School of Graduate Studies
of
The University of Alabama in Huntsville

HUNTSVILLE, ALABAMA

1999
 Copyright by
Bryan M. Rasmussen
All Rights Reserved
1999

ii
 THESIS APPROVAL FORM

Submitted by Bryan M. Rasmussen in partial fulfillment of the requirements for the degree of

Master of Science in Engineering with a major in Mechanical Engineering.

Accepted on behalf of the Faculty of the School of Graduate Studies by the thesis committee:

Committee Chair
(Date)

Department Chair

College Dean

Graduate Dean

iii
 ABSTRACT
School of Graduate Studies
The University of Alabama in Huntsville
Degree: Master of Science College/Dept.: Engineering / Mechanical and
in Engineering Aerospace Engineering
Name of Candidate: Bryan Michael Rasmussen
Title: An Intrinsic, Heterogeneous Model of Composite Solid Propellant Combustion

This thesis is a theoretical study of composite solid propellant combustion, built around a

computational model of AP/HTPB propellants. The purpose of the thesis is to investigate the
effect of composite solid structure on nonsteady, nonlinear combustion processes. Of particular
interest is the effect of heterogeneity on the nonlinear, pressure-coupled frequency response of the
system. The model is a system of eight equations, two of which depend on an implicit solution of
temperature profiles in the propellant binder and oxidizer. It is very complicated and
computationally intense, but it can potentially show trends and dependencies that disappear under
the assumptions of other models.
Two issues have become significant in the development of the current model. First, the
model seems to show an over-dependence of flame structure on the burning rates of the
propellant. Second, the concept of “frequency response” appears to be ambiguous for wholly
nonlinear analyses. The thesis contains recommendations on how to address the issues. It also
contains preliminary results, which show how AP mass percentage, mean pressure, pressure

oscillation magnitude, AP particle diameter, and other parameters affect the frequency response.

Abstract Approval: Committee Chair

(Date)

Department Chair

Graduate Dean

Advisor: Dr. Robert A. Frederick, Jr. Assistant Professor

iv
 ACKNOWLEDGEMENTS

This work was sponsored by the California Institute of Technology Multidisciplinary

University Research Initiative under ONR Grant No. N00014-95-1-1338.

I dedicate this thesis to my cat, Harbinger.

v
 TABLE OF CONTENTS

Page
LIST OF FIGURES ................................................................................................................... ix

LIST OF TABLES ..................................................................................................................... xi

LIST OF SYMBOLS ................................................................................................................ xii

Chapter

I. INTRODUCTION.......................................................................................................... 1

A. Solid Combustion...............................................................................................1
B. Experimental Research.......................................................................................5
1. Burning Rate ......................................................................................... 5

2. Temperature Sensitivity ......................................................................... 6

3. Frequency Response .............................................................................. 7

C. Theoretical Research — Steady-state Models .....................................................8

1. Beckstead-Derr-Price Framework .......................................................... 8

2. Separate Surface Temperatures .............................................................. 8

3. Multi-modal Composite Propellants....................................................... 9

D. Theoretical Research — Nonsteady Models .....................................................11

1. Linear Models ..................................................................................... 11

2. Nonlinear Models ................................................................................ 17

3. A New Approach................................................................................. 19

II. STEADY-STATE COMBUSTION MODEL................................................................ 20

A. Theoretical Framework ....................................................................................20

B. Mathematical Development..............................................................................21
1. Mass Flux of AP.................................................................................. 21

2. Mass Flux of Binder ............................................................................ 21

3. Total Mass Flux................................................................................... 22

vi
 Page
4. AP Flame Height ................................................................................. 23

5. Total Flame Height.............................................................................. 23

6. AP Surface Temperature...................................................................... 27

7. Binder Surface Temperature ................................................................ 28

8. Pre-mixed Flame Temperature............................................................. 29

C. Solution ...........................................................................................................30

D. Preliminary Steady-state Results ......................................................................33

1. Burning Rate vs. Pressure .................................................................... 33

2. Initial Temperature Sensitivity............................................................. 37

3. Evolution of System Variables............................................................. 37

III. NONSTEADY-STATE COMBUSTION MODEL ....................................................... 44

A. Nonsteady Foundations ....................................................................................44
B. Solution Method...............................................................................................45
C. Solution Criteria...............................................................................................50
D. Solution Issues .................................................................................................53
1. Issue One: Burning Rate “Dragging” of the Gas Phase......................... 53

2. Issue Two: Response Function Definition ............................................ 60

E. Preliminary Nonsteady Results.........................................................................64

IV. RESULTS AND DISCUSSION ................................................................................... 69
A. Steady-state......................................................................................................69
1. Effect of Particle Diameter................................................................... 69

2. Effect of Turbulence ............................................................................ 71

3. Effect of AP Mass Percentage.............................................................. 73

4. Effect of Initial Temperature................................................................ 75

B. Frequency Response.........................................................................................75
1. Effect of Mean Pressure....................................................................... 75

vii
 Page
2. Effect of AP Mass Percentage.............................................................. 76

3. Effect of AP Particle Diameter............................................................. 81

4. Path of Burning Rate Dependence........................................................ 81

5. Effect of Oscillation Amplitude ........................................................... 88

V. SUMMARY AND CONCLUSIONS ............................................................................ 91

A. Steady-state......................................................................................................91

B. Nonsteady-State ...............................................................................................92
C. Recommendations ............................................................................................92
APPENDIX A STEADY-STATE MODEL.............................................................................. 94
APPENDIX B NONSTEADY-STATE MODEL.................................................................... 103
REFERENCES ...................................................................................................................... 117
BILBIOGRAPHY................................................................................................................... 120

viii
 LIST OF FIGURES

Figure Page
1.1 Sketch of Burning Solid Motor .......................................................................................2

1.2 Example of Pressure-Coupled Frequency Response Function..........................................4

1.3 Crawford Strand Burner..................................................................................................5

1.4 Ultrasonic Burning Rate Measurement............................................................................6

1.5 T-Burner.........................................................................................................................7

1.6 Oxidizer Particle Size Distribution in a Sample Propellant ............................................10

2.1 Conceptual Picture........................................................................................................20
2.2 Total Burning Rate .......................................................................................................22
2.3 Turbulence as a Function of Reaction Flame Height......................................................26
2.4 Burning Rate vs. Pressure .............................................................................................34
2.5 Theoretical vs. Experimental, (90 80/20, 298)...............................................................35
2.6 Theoretical vs. Experimental, (5, 80/20, 298)................................................................36
2.7 Predicted Initial Temperature Sensitivity.......................................................................39
2.8 Mass Fluxes of (90, 80/20, 298) Propellant ...................................................................40
2.9 Linear Burning Rates of (90, 80/20, 298) Propellant......................................................40
2.10 Flame Heights of (90, 80/20, 298) Propellant ................................................................41
2.11 Flame and Surface Temperatures of (90, 80/20, 298) Propellant....................................42
2.12 Surface Fraction of AP as a Function of Pressure ..........................................................43
3.1 Sample Temperature Profile in Propellant (Ideal)..........................................................47
3.2 Sample Temperature Profile in Propellant (Actual) .......................................................47

3.3 Discrete Computational Grid ........................................................................................48

3.4 Response to Step Pressure Input....................................................................................54
3.5 Response to Oscillatory Pressure Input .........................................................................55

3.6 Heat Feedback Model ...................................................................................................56

3.7 Effect of Curve-Fitting on Step Input ............................................................................58
3.8 Effect of Curve-Fitting on Frequency Response ............................................................59

ix
 Page

3.9 Final Oscillation of Mass Flux Under an Oscillatory Pressure Field ..............................62

3.10 RP vs. Frequency Using Several Calculation Methods ...................................................63

3.11 Effect of AP Particle Diameter on RP ............................................................................65

3.12 Effect of Mean Pressure on RP ......................................................................................66

3.13 Effect of Oscillation Magnitude on RP...........................................................................67

3.14 Effect of Oxidizer Mass Percentage on RP .....................................................................68

4.1 Overall Flame Height and Pre-mixed Flame Temperature .............................................70

4.2 Turbulence in the Steady-State Model...........................................................................72

4.3 Flame Temperature and Heat Release as Functions of αap .............................................73
4.4 Effect of αap on Burning Rate .......................................................................................74
4.5 Steady-state Temperature Profiles at Two Mean Pressures ............................................77
4.6 Normalized RP Plot: Effect of Mean Pressure................................................................78
4.7 Steady-state Curves for Different AP Mass Percentages ................................................79
4.8 Normalized RP Plot: Effect of AP Mass Percentage at 10 bar ........................................80
4.9 Normalized RP Plot: Effect of AP Particle Diameter at 10 bar .......................................82
4.10 Effect of Gas-Phase Dragging on Three Propellants ......................................................83
4.11 Steady-state Curves for Different AP Particle Diameters...............................................85
4.12 Effect of AP Particle Diameter on RP at 100 bar............................................................86
4.13 Normalized RP Plot: Effect of AP Particle Diameter at 100 bar .....................................87
4.14 Effect of Oscillation Magnitude on RP Revisited: Method 4 ..........................................89
4.15 Effect of Oscillation Magnitude on RP Revisited: Method 5 ..........................................90

x
 LIST OF TABLES

Table Page
2.1 Dependent Variables in the Steady-State Model ............................................................31

2.2 Physical Constants in the Model ...................................................................................32

3.3 Dependent Variables in the Nonsteady Model...............................................................51

xi
 LIST OF SYMBOLS

Symbol Definition
A pre-exponential factor; constant in response function; area; constant

a factor in burning rate equation

B constant in response function

Cp constant-pressure specific heat

C; c constants

D diameter
d distribution coefficient
D diffusion coefficient
E activation energy
F mass distribution function in PEM
f function
G mass flux
i imaginary number specifier; space vector index
j time vector index
K turbulent diffusion coefficient
k ratio of motor throat area to burning surface area; ZN parameter
l length
M total mass; molecular weight; coefficient in matrix equation
&
m mass flow rate

n pressure exponent; total number of points in a matrix

P pressure
q specific energy

RP pressure-coupled frequency response function

r burning rate; ZN parameter

S surface fraction

s Laplace variable, second

xii
T temperature

t time

V volume
v solution in matrix equation

x distance

α mass fraction; thermal diffusivity

β diffusion exponent

∆ change in
δ x-spacing coefficient
ζ non-dimensional exponent for reaction
θ non-dimensional temperature
λ thermal conductivity
µ non-dimensional mass flux; ZN parameter
ν exponential modifier
ξ non-dimensional distance
πk pressure temperature sensitivity
ρ density
σ mode width parameter in PEM
σp temperature sensitivity
τ non-dimensional time
φ non-dimensional function
Ω non-dimensional frequency

ω frequency

Modifier Description
( )± just above/below
(?) mean or steady-state

xiii
( )´ differential, or nonsteady part

( )ap ammonium perchlorate

( )b binder
( )d diffusion

( )eff effective

( )f flame
( )f-s flame-to-surface

( )g gas

( )i initial
( )lam laminar
( )max/min maximum/minimum
( )ox oxidizer
( )p total propellant
( )r reaction
( )s solid; surface
( )tr transitional
( )tot total
( )turb turbulent
( )v total vaporization
( )vap surface vaporization
bold indicates matrix or vector

xiv
 Chapter I

INTRODUCTION

A. Solid Combustion
Solid Rockets are the simplest types of rockets, and for that reason they have been in use
for centuries longer than any other type of mechanical propulsion device. For over 800 years,
military engineers and pyrotechnic enthusiasts have added various and sundry solid ingredients to
rocket cases in a confusing search for the best propellants. Modern scientific and engineering
techniques have vastly improved on this trial-and-error approach. Research conducted during the
20th century has resulted in better solid motor manufacturing methods and has added to a basic
understanding of solid combustion.
Whereas the primary ingredient in almost all older rockets was gunpowder, modern solid
propellants generally fall into one of two categories:
• Double-base propellants have a heterogeneous solid phase. They consist of solid
nitrocellulose dissolved in nitroglycerin, possibly with a few minor additives. Double-base
propellants are detonable, so they pose a safety risk during manufacturing or storage.
• Composite propellants have a heterogeneous solid phase. They consist of oxidizer particles
suspended in a polymer binder. The most popular oxidizer is ammonium perchlorate

(NH4ClO4), and most binders are polybutadiene variations. Most practical composite
propellants also contain a powdered metallic fuel, such as aluminum. The binder itself will
usually oxidize and burn during combustion, so the term “fuel” often refers to the binder and

not the aluminum.

Regardless of the type of propellant, all solid propellants have some specific properties
that are of interest to motor designers. One of the most important properties is the burning rate.
Solid fuels ideally burn normal to the burning surface, an assumption that leads to a

1
 2
linear burning rate in the units of distance/time. Figure 1.1 contains a simplified sketch of a

burning solid rocket motor to illustrate the concept of a lindear burning rate.

Experiments have shown that macroscopic burning rates typically depend on a power of
pressure through the equation

r = aP n . (1.1)

The relationship in Equation 1.1 is usually accurate over a local pressure range. It

indicates an important parameter in solid motor design: the pressure exponent, n. If the pressure
exponent is low, the propellant will be stable in the particular pressure range. If the exponent is
high (approaching unity), then the propellant could potentially become unstable with a small
change in pressure. Gunpowder, for example, has a high pressure exponent, so it explodes when
pressurized.

Gas
Fuel Grain Velocity
High
Pressure

1-D Burning
Rate

Figure 1.1: Sketch of Burning Solid Motor

 3

Solid propellant burning rate is also a function of initial temperature of the propellant.

Two variables, σp and πk, describe the temperature sensitivity1. The sensitivity of the burning
rate to initial temperature at constant pressure is σp. Its mathematical definition is

∂ln r ∂ln aP n 1 ∂a
σp = = = . (1.2)
∂Ti p
∂Ti p
a ∂Ti p

The temperature sensitivity of the pressure to initial temperature at a constant

throat/burning surface area (k) is denoted πk. It represents the sensitivity of the mean motor
pressure to the initial temperature of the motor. Note that πk only applies to whole motors, while
σp can be an intrinsic property of the propellant. The definition of πk is

∂ln P 1 ∂P
πk = = . (1.3)
∂T k P ∂T k

A system as complicated as a burning solid propellant will clearly have many other
quantifiable elements that affect the system. Properties such as flame structure, deflagration-
detonation-transition (DDT) environment, and composition of combustion products are all
important to motor designers. A complete list of all parameters that could possibly be significant
would require twenty or more pages.
One final property that is important in this paper, however, is the pressure-coupled
frequency response function. Response functions in combustion are exactly like response
functions in mechanics. They are ratios of the magnitude of an output function to the magnitude
of a driving function. In the case of the pressure-coupled response, the driving function is

pressure, and the output function is burning rate. The pressure-coupled response function, RP, is

m&′m&
Rp = , (1.4)
P ′P

_
where m& is the mass burning rate, (´) denotes a differential change and ( ) denotes a mean
quantity. Pressure-coupled frequency response is an intrinsic property of the propellant that is
very important in motor design. It is a function of frequency of pressure oscillations, so peaks in
 4
RP indicate areas where the propellant could easily become unstable. In Figure 1.2, for example,

the propellant has a peak at around 675 Hz, and motor designers should not use the propellant in a

motor that has a natural acoustic mode anywhere between 400 and 800 Hz.
Most of the rest of this thesis is an attempt to theoretically predict burning rates,

temperature sensitivities, and pressure-coupled frequency response functions of a particular class

of composite propellants. The following two sections introduce experimental and theoretical
means of calculating these three properties.

Figure 1.2: Example of Pressure-Coupled Frequency Response Function2

 5

B. Experimental Research
1. Burning Rate
The most common way to obtain burning rates has traditionally been the Crawford strand

burner3. This device, shown in Figure 1.3, is very simple. A propellant strand with an inhibiting

outer coating has two or more wires threaded through it. As the surface regresses (linearly, by
assumption), it burns through the wires, changing the resistance in the system. Knowing the time

between the resistance change and the initial spacing of the wires allows for a calculation of the

burning rate of the propellant.

There are many other techniques for finding burning rates in a laboratory. One relatively
new technique is to use an ultrasonic transducer to bounce a sound wave off the propellant
surface4,5. By measuring the time it takes to travel from the transducer to the surface and back, it
is possible to calculate the position of the surface in a 50-200µs time interval. The ultrasonic
method has two advantages: 1) it allows for many measurements in one test, and 2) it allows for
measurements under pressure transients. Figure 1.4 is a simplified diagram of the ultrasonic
technique.

Timing Wires

-
+

Figure 1.3: Crawford Strand Burner

 6

Transducer Coupling
Material Propellant

Figure 1.4: Ultrasonic Burning Rate Measurement

Unfortunately for motor designers, solid propellants do not always burn in an operational
motor as they do in a laboratory. In a motor, hot, high-velocity gases are flowing over the surface
of the propellant, thus increasing the burning rate and causing other difficulties. Expensive
measurement methods, such as x-ray pictures6, are usually necessary for measuring burning rates
in motors although heavily instrumented experimental motors often contain ultrasonic
transducers, thermocouples, pressure transducers, and an array of other instruments.
2. Temperature Sensitivity
In theory, finding temperature sensitivity is easy once the burning rate is known. If the
burning rates are r1 and r2 at initial temperatures T1 and T2 respectively, then σp is a simple
relationship. For discrete data, σP is

∂ln r ln r2 − ln r1
σp = ≅ , (1.5)
∂Ti p
T2 − T1

or, alternatively,

∂ln r 1 r2 − r1
σp = ≅ . (1.6)
∂Ti p
r T2 − T1
 7
Similar relationships exist for πk. The problem is that σp and πk are differential

quantities, and discrete data do not generally translate well into smooth differences.

Consequently, experimentally measured values of σp and πk typically have a great deal of

scatter7.

3. Frequency Response

By far the most widely used device for measuring propellant frequency response is the T-
burner, which has a record of over 50,000 tests8. The T-burner is a length of tube with one

propellant grain at each end. Gas flows out of the tube through a nozzle in the center of the tube,

and the two propellant grains force pressure on each other to produce oscillations. The length of
the tube determines the oscillation frequency. Figure 1.5 is a sketch.
T-burners have several disadvantages. Among other problems, they use a large amount
of propellant per test, they do not generally generate reproducible results, and they do not
accurately mimic internal motor conditions. As a consequence, other methods for determining
the pressure-coupled frequency response are now under development. They include, but are not
limited to, microwave burning rate measurements9, exhaust modulation10, magnetohydrodynamic
flow measurement 11, and direct modulated mass-injection (now under development at UAH).
Modern experimental methods still require significant development before they can produce
accurate, cheap, reproducible response predictions.

Propellant

Nozzle

Figure 1.5: T-Burner

 8

C. Theoretical Research — Steady-state Models

Combustion is a very complicated phenomenon, and modeling efforts are rarely able to

quantitatively reproduce experimental data. A few models have been precise enough to stimulate
burning rate “tailoring” for specific motors, but most models predict only qualitative behavior

and trends. The most practical justification for spending money on theoretical combustion

modeling is that theoretical procedures may lead to a better understanding of physical processes.
That is, modeling has scientific value, not necessarily engineering value.

To make matters even more difficult, the heterogeneous structure of composite

propellants complicates the combustion process, making a nearly intractable problem even worse.
This additional complication generally prohibits analytic solutions of burning rate as a function of
pressure and propellant properties. Most composite steady-state models, such as those listed
below, require numerical solutions on computers.
1. Beckstead-Derr-Price Framework
The first successful heterogeneous model for composite propellants was probably the
Beckstead-Derr-Price (BDP) multiple flame model12. This model was the first to recognize that
the flame structure of a composite propellant was not homogeneous. Indeed, diffusion processes
associated with heterogeneity often dominate the combustion process in BDP models.
The BDP concept involves three combustion regions: two kinetics-dominated (reaction)
flames and one diffusion flame. The oxidizer, usually ammonium perchlorate (AP), breaks down
in one reaction flame and sends approximately 30% O2 into the diffusion flame. Binder
decomposition products pre-react in the other reaction flame then rush into the diffusion flame,
where they react further with the oxygen.
Examples of influential parameters in BDP models include the heat of vaporization, the

heat conduction into the solid phase, and the flame standoff distances. In a BDP-type model,
where the combustion occurs is as important as how it occurs.
2. Separate Surface Temperatures

Researchers have added numerous improvements to the original BDP model

since its original publication13. One of the most important improvements is the consideration of
separate surface temperatures for binder and oxidizer. Because the flame structure is different
 9
over the binder and oxidizer, there is no reason why their surface temperatures or heat fluxes

should be the same. Separate surface temperature models are necessarily more elaborate because

they incorporate the solid-phase relationships of two original BDP models — one for the AP and
one for the binder. Nevertheless, they can reproduce observed behavior that other, simpler

models cannot 14.

3. Multi-modal Composite Propellants

Another limitation of the original BDP model was that it could only simulate propellants

with a single oxidizer particle size. Most composite propellants, in contrast, contain a wide

dispersal of oxidizer particle diameters. Such a scattering is actually desirable because

propellants with a single oxidizer particle diameter are limited to slightly more than an 80%
theoretical maximum oxidizer mass fraction. Simply put, small oxidizer particles are necessary
to fit in between the large ones in order to have a high oxidizer percentage.
Most propellants have essentially two or three particle sizes, with a scattering of other
sizes around the mean. Propellants with only one particle size are called mono-modal, those with
two are called bi-modal, and so on. The plot in Figure 1.6, for example, is a bi-modal propellant
with a large scatter around the coarse oxidizer size.
Glick, Condon, and Renie created a new statistical formalism to deal with multi-modal
propellants15,16. The result of their research is called the Petite Ensemble Model (PEM). The
PEM is still a one-dimensional model, but it incorporates a three-dimensional picture of burning
oxidizer particles. The PEM method is to track the evolution of an oxidizer particle from the time
it breaks the surface until its final consumption. Each particle has a separate life cycle, but all
particles for a particular mode can reduce to a representative “petite ensemble” for that mode.
Glick et al. proposed that each oxidizer mode has a mass distribution function, Fox,d, defined as

1  1 ln D − ln D 2 
Fox ,d = exp −  
 , (1.7)
2π σ 
 2  σ  

where σ is a mode width parameter that roughly corresponds to standard deviation, D is the

diameter of a particular particle, and D is the diameter of the oxidizer mode itself.
 10
In Figure 1.6, for instance, the propellant has two distinct modes. The small-diameter

mode is around 15µm with a small σ, and the large diameter is around 150µm with a large σ. A

value of σ that is greater than unity indicates a “polydisperse” propellant.

The total burning rate is therefore the following integral over all the modes of the

propellant:

rd
r = ∫α Fd d (ln D ). (1.8)
D
d

The PEM has been moderately successful in modeling the effects of particle size
distribution on steady-state properties15,16.

Figure 1.6: Oxidizer Particle Size Distribution in a Sample Propellant17

 11

D. Theoretical Research — Nonsteady Models

Nonsteady-simulation of combustion processes is a very tedious endeavor. In addition to
the normal complexities encountered in steady-state modeling, nonsteady models must also

account for thermal lags in the solid phase.

Under most normal pressure transients, the gas phase reacts very quickly. Thermal
capacitance in the gas phase is usually negligible compared to heat accumulation in the solid

phase. Thus, the solid phase thermal relaxation time is the most important factor in determining

the nonsteady response of burning rate to pressure differences. It is what ultimately drives the
pressure-coupled frequency response of the propellant.
This thermal relaxation time, or “characteristic response time”, is a function of the
thermal diffusivity of the propellant, α, and its burning rate. Denoted τ, the thermal relaxation
time is

α
τ= . (1.9)
r2

Under a positive pressure transient, the temperature profile in the propellant is artificially
steep, causing the propellant to burn faster than it would under the exact same pressure at steady-
state. The essence of nonsteady modeling is to determine exactly how the thermal relaxation time
affects the burning rate, given a known driving pressure function.
Because solid combustion processes are so intricate, most models incorporate Quasi-
Steady gas phase, Homogeneous solid phase, One Dimensional (QSHOD) assumptions to
simplify the problem. The acronym is disingenuous, however, because some of the more
advanced models, including this one, do accommodate some properties of Heterogeneous (i.e.,

composite) propellants.
1. Linear Models
Linear pressure-coupled frequency response models have been in existence since the

1940’s. There are essentially two main categories: those that rely on Flame Models (FM) and
those that rely on the Zeldovich-Novozhilov (ZN) method.
 12
Every nonsteady analysis, regardless of its classification, starts with the same basic

relationship. The transient heat conduction equation is a second-order, parabolic partial

differential equation that represents an energy balance at some point x in the solid phase. Its most
general form is

∂T ∂  ∂T  ∂(C p T )
ρC p = 
λ   − ρr , (1.10)
∂t ∂x  ∂x  ∂x

where x = 0 at the burning surface and is positive above the surface. At steady-state, Equation

1.10 becomes a second-order ordinary differential equation. Assuming that the thermal properties
of the propellant are constant, that the temperature is Ts at the surface, and that the temperature is
Ti well below the surface at x = -∞, the steady-state heat conduction equation becomes

∂2 T ∂T
α − r =0. (1.11)
∂x 2
∂x

The ODE has the following solution, which defines the steady-state temperature profile:

 rx 
T ( x ) = Ti + (Ts − Ti )exp
  . (1.12)
α 

Note that Equation 1.12 is not linear, as the steady-state profile would be in a normal heat
conduction equation. This nonlinearity results from the burning rate contribution term in the
transient heat conduction equation. In colloquial terms, the burning rate “bends” the normal
linear profile into an exponential one.
The next step in the development of a linear model is to define a function for the mass

flow rate. A simple Arrhenius expression gives the mass flow rate as a function of surface
temperature only18. The Arrhenius expression is

 E 
G = As ⋅exp−  . (1.13)
 RT 
 s 
 13
The following are some useful non-dimensional terms:

G
µ= , (1.14)
G

T
θ= , (1.15)
Ts

xC p ,s
ξ= , (1.16)
λp

and

Ω = τω. (1.17)

The linearized and non-dimensional form of the Arrhenius expression (Equation 1.13) is
therefore

µ′
= ζ s θ′
s, (1.18)

where the steady-state portion has been subtracted out, and the exponential ζ s, is given by

Es
ζ s = . (1.19)
RTs

If the reader is unfamiliar with the process of linearization, the next section contains an example

using an exponential function. The non-dimensional form of Equation 1.10 is

∂2θ
)∂θ − ∂θ τ = 0 .
− (1 + µ ′ (1.20)
∂ξ 2
∂ξ ∂t

The steady-state solution of the above PDE is Equation 1.12 again. Using the new
notation, Equation 1.12 becomes

θ = θi + (1 − θi )exp(ξ ). (1.21)
 14
Now two assumptions become necessary: 1) pressure input is a sine wave with

frequency; and 2) the surface temperature oscillates at the same frequency as the input, though

not necessarily in the same phase. One can represent this assumption with the familiar complex
exponential,

′ exp (iω t ).
θ = θ + θmax (1.22)

Substituting Equations 1.22 and 1.21 into Equation 1.20, subtracting off the steady-state
portion, assuming that ∂θ ′∂ξ is on the same order as θ´, and linearizing, the new transient heat

conduction equation is

∂2θ ′ ∂θ ′
− (1 − θi )exp(ξ ),
− iΩ θ ′= µ ′ (1.23)
∂ξ 2 ∂ξ

which has the solution,

θ ′= θs′ (1 − θi ) i [exp(ξ )− exp(sξ )],

exp(sξ )+ µ ′ (1.24)
Ω

where s is the familiar Laplace variable, the positive solution to the quadratic equation,

s 2 − s − iΩ = 0 . (1.25)

Another simplifying constant would be useful here. Let A be defined by

A = (1 − θi )ζ s . (1.26)

The next step is to draw an energy balance from x = 0- to x = +∞ as follows:

∂T
GC p Ts + Gqtot = λs + GC p ,g T f . (1.27)
∂x x =0 −

After substituting Equations 1.24, 1.25, and 1.26 into Equation 1.27, and after
considerable manipulation, the nonsteady energy balance becomes

µ ′  C p ,g
(s − 1)+
A
− A = − θ′
f . (1.28)
ζ s  s  C p ,s
 15
Here is where the American and Russian approaches have traditionally diverged.

Equation 1.28 contains two variables: θf and µ´. Another equation is necessary in order to obtain

a solution. One approach is to simply assume an equation based on an exponential, pressure-

related model of the flame. This is the so-called flame model approach.

For example, Denison and Baum19 used a simple expression to link the gas phase to the

burning rate. They proposed

 Ef 
G = cP n exp− , (1.29)
 2 RT 
 f 

or, linearized,

P′ θ′
µ ′= n + ζ
f
f , (1.30)
P θf

where ζ f is analogous to ζ s. The definition of ζ f is

Ef
ζ f = (1.31)
2 RT f

Yet another constant will simplify the notation even more. “B” has the definition

C p ,g ζ
B= s
θf , (1.32)
C p ,s ζ f

Solving Equations 1.28 and 1.30 together leads to the following definition for the

pressure-coupled response function:

nB
Rp = . (1.33)
(s − 1)+ A
− A+ B
s

(Equation 1.33 often takes different forms in the literature, with slightly different
definitions for A and B.)
 16
The Zeldovich-Novozhilov method 20 is an alternative approach for creating linear

models. It is a phenomenological description that depends on steady-state properties of the

propellant to replace Equation 1.29.

One form of the ZN response function looks very similar to the flame model result21. It is

n + (nr − µk )(s − 1)
Rp = , (1.34)
1 + r (s − 1)− k (s − 1) s

where k, r, and µ are combinations of rate properties as follows:

k = (Ts − Ti )σ p , (1.35)

∂Ts
r= , (1.36)
∂Ti P

and

1  ∂Ts 
µ=   . (1.37)
 
Ts − Ti ∂ln P T i

Ideally, the ZN approach would yield a better response function because it does not rely
on an a priori flame model. In practice, however, ZN models often suffer from lack of accurate
data because the parameters r, k, and µ are very difficult to measure precisely in the laboratory.
Brewster and Son thoroughly analyzed both ZN and flame models in 1994 and concluded

that the simple Arrhenius expression with no pressure dependence was inadequate for nonsteady
analysis22. They proposed the use of a different expression, called zeroth order decomposition,

which had been in use since the early 1970’s23.

The zeroth order decomposition relationship is

 2E s 
As ρ s Ts2 C p ,sα s exp− 
 RT 
 s 
r2 = . (1.38)
[
2 E s C p (Ts − Ti )− Qs 2 − ƒs q / rρ s]
 17
Equation 1.38, however, really applies only at steady-state. Because of the consideration of

condensed-phase energy storage, a proper unsteady version must include an integral term to

account for solid phase capacitance, shown here as

 2E s 
As ρ s Ts2 C p ,sα s exp−  (1.39)
 RT 
 
r2 =
s
.
 0
∂T 
2 E s C p (Ts − Ti )− ρs ∫C p dx − Qs 2 − ƒs q / rρ s 
 −∞ ∂t 

Brewster is continuing work in this area by modifying the initial temperature in the zeroth order

decomposition equation 24.

Another question for linear models is how to extend them to heterogeneous propellants.
One tactic is to perturb and linearize a BDP-type steady-state model. This approach usually leads
to a functional form similar to Equation 1.33, but with significantly more complicated constants25.
Glick et al. tried a different approach by extending the PEM to unsteady situations26.
Their work implies that a multi-modal propellant will behave like a series of homogeneous
propellants, so it is possible to combine the end results of several homogeneous response
functions. Indeed, they did report multi-peaked response functions that were consistent with
several QSHOD models in series.
2. Nonlinear Models
As mentioned above, all nonsteady solid combustion models employ the transient heat
conduction equation and share some of the same basic assumptions. Linear models, however,
generally have an analytic solution, due to the simplification process.
The standard way to linearize a function is to take the Taylor series expansion and
disregard everything higher than order two. For example, consider the following function:

f (x ) = A exp (Bx ). (1.40)

 18
At any given x, it is possible to represent the independent variable using a mean and a

perturbation around the mean. Let x = x + x ′

. Substituting and expanding the original function

in a Taylor series, one obtains

f = A + AB( x + x ′
)+ AB 2 ( x + x ′
)+ AB 3 ( x + x ′
)+K.
1 2 1 3
(1.41)
2 6

The perturbations around the mean are, by assumption, relatively small. Surely, then, the

square of the perturbations would be even smaller, in fact negligible. Neglecting everything of
second-order or higher, the linearized form of f is

f = A + AB( x + x ′
). (1.42)

Setting f = f + f ′and subtracting off the steady portion, one obtains

f ′= ABx ′
. (1.43)

Equation 1.43 is linear and thus is much simpler than the transcendental Equation 1.40.
If a model contains a system of linear equations, it will have a simple analytic solution, whereas a
system of transcendental equations may not have an analytic solution and will require numerical
solution techniques.
Nonlinear models do not consider the perturbations to be necessarily small, and thus they
include second-order or higher terms. These additional terms do complicate models significantly,
but nonlinear models can account for effects that linear models cannot.
Nonlinear models can, for instance, account for large pressure spikes and possibly predict
extinguishment and deflagration-to-detonation thresholds. They can predict the evolution of a
system over time, and they can give response functions in terms of both amplitude and frequency.

To create a nonlinear model, it is necessary to preserve at least some effects of order two
or greater. One approach is to take the Taylor series expansions of functions, as in the linear
case, but leave in terms of progressively higher order27. Another approach is to perform no

reduction whatsoever. These are the most complicated attempts, and they require relatively

sophisticated computers and programming techniques. Researchers have been able to attack the
nonlinear problem since the 1970’s, but modern computing power has sped up the process28.
 19
Because of the added complexity of the mathematical circumstances, nonlinear models

typically incorporate many simplifying assumptions about the physical circumstances. No

nonlinear models in the open literature have been able to account for the complex gas phase and
heterogeneous solid phase of a BDP-type analysis. Some nonlinear models, however, have been

able to account for changing thermo-physical properties in the solid phase28,29,30. Variable-

property models have shown a reduction of amplitude in the frequency response function, as well
as a shift to higher frequencies. The observed effects are probably due to temperature profiles

that are steeper in variable-property simulations than in fixed-property simulations. As discussed

in Chapter IV, steeper temperature profiles typically diminish response amplitude and shift the
peak to higher frequencies.
3. A New Approach
The purpose of this thesis is to combine some of the best aspects of the different types of
models into one comprehensive model with an intrinsically heterogeneous view of composite
solid combustion. The following chapters describe a model of solid propellant combustion that is
very similar to the BDP steady-state description, but with time-dependent terms to account for
thermal lags in the solid phase.
The model is a completely nonlinear analysis that contains no Taylor series expansions.
It is a description of mono-modal propellants only although a PEM or other surface description
might not be too difficult to merge into the model at some later date. The model does not account
for changes in constant-pressure specific heat and thermal conductivity in the solid phase.
Essentially, it is an attempt to marry the mathematical and numerical complexity of a nonlinear
model to the more physically accurate view of a BDP steady-state model.
 Chapter II

STEADY-STATE COMBUSTION MODEL

A. Theoretical Framework
The theoretical model presented here is a modification of the original BDP multiple
flame model for composite propellants. As in the BDP, the present model contains three types of
flames, as shown in Figure 2.1.
The pre-mixed flame is a kinetics flame that emerges due to the exothermic
decomposition of AP. The most reactive product of this flame is the approximately 30% O2 that
results from AP decomposition. The reaction flame is also a kinetics flame, but it receives its
chemical energy through a reaction between perchloric acid from the AP flame and gaseous
decomposition products from the polymer binder. Finally, the flame occurs above the kinetics
region where the products of the previous two flames diffuse into each other and form the final
decomposition products.

diffusion mass flux pre-mixed flame

flame
reaction flame

AP

BINDER

Figure 2.1: Conceptual Picture

20
 21
B. Mathematical Development

The model comprises eight interrelated, dependent variables, which are elucidated below.

In all the following equations, the coordinate system is one-dimensional, with x = 0 at the surface,
x = -∞ far below the surface, and x = +∞ far above the surface. This Lagrangean coordinate

reference frame moves relative to a “laboratory” reference frame. Rather than picturing a

propellant surface that regresses to a base, the simplest way to view the system is to picture a
“river” of propellant that flows to the surface, where it vaporizes.

1. Mass Flux of AP

Because the model is one-dimensional, a mass flux (mass flow rate per unit area) can
represent the mass flow. The mass flux is the same, by assumption, at any point on the surface of
an AP particle.
The burning rates of many materials seem to be related almost solely to surface
temperature. In this model, the Arrhenius surface pyrolysis relationship18 is

 E s ,ap 
G ap = As ,ap exp− . (2.1)
 RT 
 s ,ap 

Although this expression does not account for sub-surface effects as in Equations 1.38 and 1.39,
other expressions in the model do account for them. Thus, the Arrhenius expression is probably
adequate in this case.
2. Mass Flux of Binder
The mass flux of the binder is essentially the same expression as Equation 2.1 with
different thermo-physical constants. The binder Arrhenius expression is

 E 
Gb = As ,b exp− s ,b . (2.2)
 RT 
 s ,b 

There has been some recent discussion about the activation energy for HTPB under
combustion heating conditions 31. The binder seems to have a lower activation energy under
higher heating rates. This is probably due to physical processes, not changes in polymer
 22
chemistry. The activation energy used here is 8.8 kcal/mole, a value that corresponds to a

relatively low heating rate32.

3. Total Mass Flux

Because the model is one-dimensional, another equation is necessary to combine the

mass fluxes of binder and AP. Modelers have used several different relationships in the past,

some more elaborate than others. The PEM, for example, is probably the most accurate, but, of
course, it is also one of the most tedious.

The purpose of this model is not to create a perfect steady-state description. Hence,

simplicity wins out in this situation, and the simplest relationship that conserves physical
principles is the best.
The mass flux of the propellant must be an algebraic combination of the mass fluxes of
the binder and oxidizer. Assuming that the propellant burns linearly, the average combined
burning rate should be the total amount of time that a propellant burns, divided by the total
length. Figure 2.2 shows the concept graphically.

t = t0

l
r=
l ∆t

t = t1

Figure 2.2: Total Burning Rate

 23
Simply picking a characteristic length of l, the total mass flux over that length becomes

ρpl 1
Gp = = . (2.3)
M ap Mb α ap 1 − α ap
+ +
G ap A Gb A G ap Gb

4. AP Flame Height
The reaction flames of the model are, by assumption, second-order flames. In common

terms, this assumption means that the reactions result from two particles colliding, not from

commingled reactions involving three or more particles. Mathematically, the assumption means
that the total flame height is inversely proportional to the square of the pressure.
The pre-mixed flame height is also a function of the gas velocity moving through it. It is
therefore directly proportional to the mass flux of AP. It also depends on the activation energy of
the reaction in an Arrhenius-type expression. The expression for xf,ap in the current model is

G ap
x f ,ap = , (2.4)
 E g ,ap 
P 2 Ag ,ap exp− 
 RT 
 f ,ap 

which is similar to kinetics-dominated flame heights in many other models12,13,14.

5. Total Flame Height

The expression for total flame height is somewhat more complicated. It too has a
kinetics-related component that is proportional to the inverse square of the pressure, but it has a
diffusion component as well. The reaction flame height is very similar to Equation 2.4, and its
equation is

Gp
xr = . (2.5)
 E 
P 2 Ar exp− r 
 RT 
 f 

Diffusion flames, as their names imply, result from mixing processes where one material
diffuses into another. Bunsen burners and cigarette lighters are examples of diffusion flames.
 24
Burke and Schumann were probably the first researchers to thoroughly analyze diffusion flames

in macroscopic environments, and they accomplished this in 1928 33. Their analysis is still in

common use today.

Under relatively fast burning rates, the Burke-Schumann analysis reduces to
2
G p D*ap
xd = , (2.6)
Adiff (ρ g Dg )eff

where Dap * is the characteristic diameter of an oxidizer particle. It represents the average

diameter of an oxidizer particle while the propellant is burning. It is related to the mass fluxes of
AP and binder, as well as the surface geometry, through the following:

2 Dap
D*ap = . (2.7)
ρ p G ap
6
ρ ap G p

Two effects contribute to the value of the diffusion coefficient. The dominant
contribution at low pressures is ordinary laminar diffusion mixing, which is itself related to a
reference diffusion coefficient and temperature 34. It has the form

(ρ D )
T f ,ap
g g lam = ρ g D0 . (2.8)
P

Substituting the ideal gas law to write the laminar diffusion coefficient as a function of
temperature alone, the laminar coefficient is

(ρ D )
g g lam = D0 T f ,ap
β− 1 M
. (2.9)
R
 25
The second effect on the diffusion coefficient is turbulent mixing. This is a relatively

high-pressure phenomenon that is near zero below a threshold. The following equation represents

turbulent mixing in the model:

  1 
(ρ D )
g g turb = KG p D*ap tan − 1 (C1 C 2 )+ tan − 1 C 2  − C1 .
  
(2.10)

  xr 


In the above equation, K is a constant that is on the order of one, C1 and C2 are constants that

control the onset of turbulence, and the arctangent function is a convenient way to model the

growing effects of turbulent mixing on the system.

A postulate of the model is that turbulence commences in the diffusion flame as reaction
height falls. When the reaction height is large, turbulence is negligible, but when it shortens at
higher pressures, the arctangent function in Equation 2.10 becomes significant. This is a purely
utilitarian assumption, as it is necessary to define some sort of criterion for when turbulence
should appear and a full turbulence analysis would be too complicated for the current project. In
Equation 2.10, C1 controls the point at which turbulence begins, and C2 determines the length of
the transitional region. Figure 2.3 is a graphical representation.
Setting (ρgDg)eff = (ρgDg)lam + (ρgDg)turb and xf = xr + xd (that is, by combining Equations
2.5, 2.6, 2.7, 2.9, and 2.10 ), the total flame height equation becomes

2
G p D*ap
xf = (2.11)
   1 
+ KG p D*ap tan − 1 (C1 C 2 )+ tan − 1 C 2  − C1 
β− 1 M
Adiff D0 T f ,ap
    

R    xr 
Gp
+
 E 
P 2 Ar exp− r 
.
 RT 
 f 
 26

3
Fully-Developed Turbulence

2.5

2
tan-1 ((( 1/x )-C )*C ) + tan-1 (C *C )
2

Transitional Region
1

1.5
2
1

1
r

0.5 No Turbulence

0
10 1 0.1 0.01 0.001
x (µm)
r

Figure 2.3: Turbulence as a Function of Reaction Flame Height

 27
6. AP Surface Temperature

The oxidizer and binder have separate surface temperatures in the model, and equations

for both come from energy balances. Consider an energy balance from deep in an oxidizer grain
(x = -∞ ) to just above the surface (x = 0+). Assuming that the only energy going into the surface

is conduction from the gas phase, the energy balance is

∂T
G ap C p ,s ,ap Ti + λg ,ap (2.12)
∂x x=0 +

∂T
0

+ G ap q v ,ap = G ap C p ,s ,ap Ts ,ap + ∫ρ

−∞
s ,ap C p ,s ,ap
∂t
∂x .

Ammonium perchlorate has three distinct crystal phases. The first phase is only present
at very low temperatures and is not of interest in practical rocket applications. The second phase
is an orthorhombic phase, which is the natural state from cold temperatures up to approximately
513K. The third phase is a cubic phase that persists until sublimation/melting. For the purpose of
this model, however, use an average specific heat at reference temperature, Tref=500K.
From a steady-state viewpoint, the phase transition does not matter, except that it draws
energy out of the system. The qv,ap is positive (exothermic), and it is the sum of three energies:

qv,ap = qtr + qvap,ap + qr , (2.13)

where qtr is the specific energy required to force the phase transition, qvap,ap is the energy required
to vaporize the AP at the surface assuming 70% sublimation and 30% degradation, and qr is the
energy of exothermic reactions in the thin melt layer12.
Now the only remaining unknown term is the derivative that defines conduction into the
surface. To obtain it, one can reasonably postulate an exponential temperature profile in three

regions in the flame.

One can postulate a profile above the oxidizer and below xf,ap,

 
T = T f ,ap − (T − Ts ,ap )exp− ν x 
 , (2.14)
 x f , ap 
f ,ap

 
 28
above the binder,

 x 
T = Tf − (Tf − Ts, b )exp− ν  , (2.15)
 xf 
 

and between xf,ap and xf,

 x − x f ,ap 
T = T f − (T f − T f ,ap )exp− ν . (2.16)
 x f − x f ,ap 
 

In the three profiles above, ν is a constant that modifies the steepness of the profile. Use
ν=2.5 because that will get the temperature to within 1/e2.5 of the maximum temperature
difference at the characteristic height. For example, at x = xf,ap over the oxidizer, the temperature
is

(T − Ts ,ap )
T = T f ,ap − = T f ,ap − 0.082(T f ,ap − Ts ,ap ).
f ,ap
(2.17)
exp( 2.5 )

Now it is possible to calculate the derivative term from the assumed profile as follows:

∂T ν
= (T f ,ap − Ts ,ap ). (2.18)
∂x x =0+ x f ,ap

Assuming that the integral term in Equation 2.12 is zero (steady-state), the solution for
oxidizer surface temperature is

ν
G ap C p ,s ,ap Ti + λg ,ap T f ,ap + G ap q v ,ap (2.19)
x f ,ap
Ts ,ap = .
ν
G ap C p ,s ,ap + λg ,ap
x f ,ap

7. Binder Surface Temperature

The expression for the surface temperature of the binder is almost exactly the same as the
one for the oxidizer. Start again with the binder energy balance. Polymer binders are obviously
not crystalline, and there is no phase transition until the surface, so the balance reduces to
 29
∂T
Gb C p ,s ,b Ti + λg ,b (2.20)
∂x x =0 +

∂T
0

+ Gb q v ,b = Gb C p ,s ,b Ts ,b + ∫ρ
−∞
s ,b C p ,s ,b
∂t
∂x .

Differentiate Equation 2.15 to get the derivative term and solve, again assuming that the

integral term in Equation 2.20 is zero. The solution for Ts,b is

ν
Gb C p ,c ,b Ti + λg ,b T f + Gb q v ,b (2.21)
xf
Ts ,b = .
ν
Gb C p ,s ,b + λg ,ap
xf

8. Pre-mixed Flame Temperature

One way to calculate the pre-mixed flame temperature is simply to define it as the
adiabatic flame temperature of self-deflagrating ammonium perchlorate. Such an assumption
might not be realistic, however, because of the effects of the diffusion flame.
If the AP were burning without a separate fuel, then Tf,ap would clearly be the adiabatic
flame temperature. The addition of a fuel to the mix adds extra chemical energy, some of which
must flow back into the pre-mixed flame. Thus, an energy balance in the diffusion part of the gas
phase is necessary to calculate the pre-mixed flame temperature. The energy balance is, in fact,
very similar to Equations 2.12 and 2.20:

∂T
G p C p ,g , p T f + λg + G p q f ,d = G p C p ,g , p T f . (2.22)
∂x x = x f ,ap

The above equation does not contain an unsteady integral term. The gas phase is assumed to bea

quasi-steady, a reasonable assumption except under very high pressure transients.

 30
Again, using the assumed temperature profile in the diffusion flame to get the derivative

term and solve for Tf,ap, the final expression is

 νλg , p 
T f G p C p ,g , p + − Gp q f (2.23)
 x − x 
=  
f f ,ap
T f ,ap .
νλg , p
G p C p ,g , p +
x f − x f ,ap

Equation 2.23 leads to an important point — the gas phase does not move immediately to

steady-state, even though it is “quasi-steady”. It moves instead to a state that would be steady-
state for a particular value of Gp. In other words, the total propellant mass flux is not quasi-
steady, so it “drags” the gas phase with it. This point will become significant in the unsteady
portion of the model.

C. Solution
The steady-state model developed in the previous section is a system of eight equations
that must be solved simultaneously. Equations 2.1, 2.2, 2.3, 2.4, 2.11, 2.19, 2.21, and 2.23
represent the system. Table 2.1 contains a list of the dependent variables and their relationships
to one another.
The model contains six “floating” parameters: K, C1, C2, Adiff, Ar, and Ag,ap. The first three
parameters define the turbulent onset (the shape of the curve in Figure 2.3), so they effectively

constitute one floating parameter. That is, the model really has four floating parameters- Adiff, Ar,
Ag,ap, and the shape of the turbulent mixing transition.
These floating parameters help to “calibrate” the model. Because it is generally

impossible to find accurate values of the parameters from experiments, they are completely
adjustable. Increasing or decreasing the parameters can move the final result of the model to a
reasonable approximation of the experimental data. See Table 2.2 for a list of all

physical/mechanical/chemical properties in the model, including the deduced floating parameters.

Table 2.2 contains some properties, such as solid-phase thermal conductivity, that
 31

Table 2.1: Dependent Variables in the Steady-State Model

Function of Variable
Variable Eq. #
Gp Gap Gb xf,ap xf Ts,ap Ts,b Tf,ap

Gp 2.3 P P

Gap 2.1 P

Gb 2.2 P

xf,ap 2.4 P P

xf 2.11 P P P

Ts,ap 2.19 P P P

Ts,b 2.21 P P

Tf,ap 2.23 P P P
 32

Table 2.2: Physical Constants in the Model

Constant Value Constant Value

Adiff† 11 Es,b 3.43·104 J·mole-1

Ag,ap† 5.0 · 10-4 m K† 2

Ar† 2.2·10-4 m M 0.0262 J·mole-1·K-1

As,ap 9.6·105 kg·m-2·s-1 qf* 2.07·106-3.63·106 J·kg-1

As,b 1.225·103 kg·m-2·s-1 qr 0 J·kg-1

C1† 30 m-1 qtr -8.78·104 J·kg-1

C2† 0.075 m qv,b -2·105 J·kg-1

Cp,cu,ap T·(1.717)+669.9 J·kg-1·K-1 qvap,ap 5.1·105 J·kg-1

Cp,g,ap 1254 J·kg-1·K-1 Tf* 1587-2993 K

Cp,g,b 2100 J·kg-1·K-1 λg,ap T·(7.2·10-5)+6·10-3 W·m-1·K-1

Cp,g,p* 1787-2870 J·kg-1·K-1 λg,b T·(4.33·10-4)-0.15 W·m-1·K-1

Cp,or,ap T·(1.717)+586.2 J·kg-1·K-1 λg,p T·(1.08·10-4)+0.0133 W·m-1·K-1

Cp,s,b T·(3.559)+1047 J·kg-1·K-1 λs,ap T·(-3.854·10-4)+0.628 W·m-1·K-1

Do 7.585·10-5 m 2·s-1 λs.b T·(5.43·10-5)+0.184 W·m-1·K-1

Eg,ap 6.28·10
4
J·mole
-1
ρap 1950 kg·m
-3

Er 1.26·105 J·mole-1 ρb 920 kg·m-3

Es,ap 9.6·104 J·mole-1

*
Linearly interpolated from thermo-chemical-equilibrium calculations at various oxidizer mass
percentages.
†
Floating parameter.
 33
are not necessary in the steady-state model but do contribute to the nonsteady model. Also, some

of the properties in Table 2.2 are averages of various results reported in the literature.

There are of course, many different numerical techniques for solving the system of
equations. It is not a simple problem, due to the non-linearity of the equations and the wide range

of conditions under which they must be solved.

The next section contains some preliminary results from the steady-state model. Results
come from Mathcad 7.0.3 calculations over a wide range of pressures and initial temperatures.

Mathcad’s numerical solution algorithm is a variation of the MINPACK public domain algorithm

published by the Argonne National Laboratory. The MINPACK algorithm is itself a version of
the Levenberg-Marquardt method35.
Appendix A contains the steady-state solution sheet.

D. Preliminary Steady-state Results

This section contains some general results of the steady-state model. For more explicit
results showing the effect of changes in specific variables, along with discussion of the physical
meaning of those results, see Chapter IV.
1. Burning Rate vs. Pressure
Figure 2.4 is a plot of burning rate vs. pressure. All propellants are 80% AP / 20% HTPB
at an initial temperature of 298K. Experimental data were available for 5µm and 90µm oxidizer
particle diameters, so the model is calibrated to those data. The experimental data come from

ultrasonic tests conducted at the Office Nationale d’Etudes et de Recherches Aérospatiales

(ONERA) in 1996 in Palaiseau, France. The extra trends at 50µm and 200µm are purely

theoretical predictions. Figures 2.5 and 2.6 show the data from Figure 2.4 in a different way.

They are plots that show the percentage difference between theoretical predictions and
experimental data for (90, 80/20, 298)‡ and (5, 80/20, 298) propellants respectively. The

theoretical predictions are a baseline for the difference calculations.

‡
If a propellant is (a, b/c, d), then the oxidizer particle diameter is “a” µm, the oxidizer/binder mass

fraction is “b/c”, and the initial temperature is “d”K.

 34

1000

Experimental (5µm)
Model (5 µm)
Model (50 µm)
Experimental (90µm)
Model (90 µm)
Model (200µm)

100
Burning Rate (mm/s)

10

1 10 Pressure (Bar) 100 1000

Figure 2.4: Burning Rate vs. Pressure

 35

100

Model (90µm)
+10%
-10%
Experimental (90µm)
Burning Rate (mm/s)

10

1
1 10 100
Predicted Burning Rate (mm/s)

Figure 2.5: Theoretical vs. Experimental, (90 80/20, 298)

 36

100

Model (5µm)
+10%
-10%
Burning Rate (mm/s)

Experimental (5µm)

10

1
1 10 100
Predicted Burning Rate (mm/s)

Figure 2.6: Theoretical vs. Experimental, (5, 80/20, 298)

 37
2. Initial Temperature Sensitivity

Figure 2.7 is a plot of predicted initial temperature sensitivity as a function of pressure.

The three traces are for 80%AP / 20%HTPB at three different particle diameters. Values of σp
come from applying Equation 1.5 to two separate simulations at 219K and 333K initial

temperature.

Theoretical initial temperature sensitivity predictions do agree generally with

experimental data. Practical AP/HTPB propellants occasionally exhibit higher sensitivities, but

mono-modal propellants are probably less sensitive than others. The general trend in composite

propellants is for propellants with wide oxidizer particle diameter distributions to have higher
sensitivities7. Obviously, mono-modal propellants have the tightest distribution possible, so they
should have lower values of σp.
3. Evolution of System Variables
The rest of the charts in this section represent system variables as functions of pressure
for (90, 80/20, 298) propellant. All data are theoretical, and in fact some of the following
variables would be nearly impossible to find experimentally.
Figure 2.8 is a plot of the relative mass fluxes in the system. The mass flux for AP grows
much higher than that of the binder at high pressures, even though total mass is conserved.
Dividing by the densities, one obtains Figure 2.9, a plot of the linear burning rates.
Figure 2.10 is a plot of the flame heights in the system. The total flame height, xf, is the
sum of the reaction flame height, xr, and the diffusion flame height, xd. This plot will become
important in the following chapter, as quasi-steady flame heights and temperatures are necessary
for calculating reasonable response amplitudes.
Figure 2.11 is a plot of Tf,ap, Ts,ap, and Ts,b as functions of pressure. The surface

temperatures approach the flame temperatures as the pressure builds and the flame heights fall,
bringing more conductive energy into the propellant surface. The surface temperature of the AP
is a function of xf,ap, and surface temperature of the binder is a function of xf. The adiabatic flame

temperature is a constant 2309K for the 80/20 formulation.

 38
Finally, Figure 2.12 is a plot of oxidizer surface fraction as a function of pressure for a

(90, 80/20, 298) propellant, where the surface fraction comes from a mass balance,

G AP S AP + Gb ( 1 − S AP ) = G p . (2.24)

Rearranging, the mass balance becomes

G p − Gb
S AP = . (2.25)
GAP − Gb
 39

0.003

0.0025

0.002
σ (K )
-1

0.0015
P

0.001
Model (5 µm)
Model (90 µm)
Model (200 µm)

0.0005

0
1 10 P (Bar) 100 1000

Figure 2.7: Predicted Initial Temperature Sensitivity

 40
100

Gp
Mass Flux (kg·s -1·m-2 )

Gap
Gb
10

1
1 10 100 1000
Pressure (Bar)

Figure 2.8: Mass Fluxes of (90, 80/20, 298) Propellant

100

Propellant
AP
Binder
Burning Rate (mm/s)

10

1
1 10 100 1000
Pressure (Bar)

Figure 2.9: Linear Burning Rates of (90, 80/20, 298) Propellant

 41

300

250

200

xf xr
Distance (µm)

x x
f,ap d
150

100

50

0
1 10 Pressure (Bar) 100 1000

Figure 2.10: Flame Heights of (90, 80/20, 298) Propellant

 42

1500

T
s,b
T
s,ap
T
f,ap
Temperature (K)

1000

500
1 10 100 1000
Pressure (Bar)

Figure 2.11: Flame and Surface Temperatures of (90, 80/20, 298) Propellant
 43

0.9

AP Surf. Fraction
AP Vol. Fraction
0.8
Fraction

0.7

0.6

0.5
1 10 100 1000
Pressure (Bar)

Figure 2.12: Surface Fraction of AP as a Function of Pressure

 Chapter III

NONSTEADY-STATE COMBUSTION MODEL

A. Nonsteady Foundations
Equations 2.12 and 2.20 contain two integral terms that are zero under steady-state
conditions. Leaving them in the energy balances leads to two different expressions for the
surface temperatures of the oxidizer and binder:

ν ∂T
0

G ap C p ,s, ap Ti + λg ,ap Tf ,ap

x f , ap
+ G ap q v ,ap − ∫ρ
−∞
s ,ap C p , s,ap
∂t
dx (3.1)
Ts ,ap = ,
ν
G ap C p ,s ,ap + λg , ap
x f ,ap

and

ν ∂T
0

G b C p ,c ,b Ti + λg , b Tf
xf
+ G b q v ,b − ∫ρ
−∞
s,b C p,s,b
∂t
dx (3.2)
Ts , b = .
ν
G b C p ,s , b + λg , b
xf

The integral terms in the numerators represent the only difference between these two
equations and the previous expressions for surface temperature. In fact, the integral terms are the
only two nonsteady contributions in the model.

Essentially, the two integral terms represent a “capacitance” in the solid phase of the
propellant. Solid materials store energy through their temperatures and specific heats, and, just as
in an electrical capacitor, it takes time to discharge this stored energy. The heat discharge time is

44
 45
related to the thermal conductivity of the system, just as the discharge time of a capacitor in an

RC circuit is related to the total impedance.

The best way to solve the integrals is to go back to the transient heat conduction equation.
The following explanation considers either the binder or the oxidizer as a homogeneous

propellant for now, though the temperature profiles in both must be solved simultaneously,

according to the eight variables in the model. At any given point x in the solid phase of a
homogeneous propellant with constant thermal properties, the heat conduction equation reduces

to

∂T ∂2 T ∂T
=α − r . (3.3)
∂t ∂x 2
∂x

Hence, if the temperature profile at any given time is known, it is then possible to calculate
∂T ∂t across the whole propellant.
In the model presented in this paper, however, the propellant is certainly not
homogeneous. There are in fact two temperature profiles in the system — one in the binder and
one in the oxidizer Hence, the model must incorporate two different versions of Equation 3.3 in
order to come to a solution. The equations are structurally identical, but they have different
thermo-physical constants and different burning rates.
To reiterate, the unsteady model is almost exactly the same as the steady-state model,
except that Equations 2.12 and 2.20 have been replaced by Equations 3.1 and 3.2 respectively.
The trick here is to calculate the integral terms that make the unsteady equations unique, using a

different version of Equation 3.3 for both the binder and the oxidizer.

B. Solution Method
There is no analytic solution for the two integral terms in Equations 3.1 and 3.2. To solve
the system of eight equations, one must numerically calculate the binder and oxidizer temperature
profiles at each time step.
From a conceptual standpoint, the easiest way to calculate temperature profiles is to use

the temperature profile from the previous time step in Equation 3.3 to get the ∂T/∂t at each x. The
 46
next temperature profile is the old profile, plus ∂T/∂t multiplied by the length of the time step.

Explicit methods such as these have a very serious limitation in heat transfer problems. The time

step must be very small in order for the equations to converge36. Specifically,

∆t ≤
(∆x )2 . (3.4)
2α

For example, consider a propellant with a thermal diffusivity of 1·10-7 m2·s-1. Say the

particle diameter is 100µm, so that near the surface, the ∆x should be at least as small as 0.05µm.

The maximum time step size is

∆t ≤
(∆x )2 =
(5 ⋅10 )
−8 2
= 1.25 ⋅10 − 8 seconds (!). (3.5)
2α 2 ⋅10 −7

Considering that the response time of the system, as given by Equation 1.9, is on the
order of a few milliseconds, the ∆t calculated above would require thousands of explicit solutions
for even a very short simulation. Computation time for an explicit method is therefore
prohibitively large, especially because the temperature profile must be calculated many times at
each time step in the course of finding a simultaneous solution to the eight nonlinear equations of
the model. Clearly, a better method is necessary.
One common numerical technique for calculating transient temperature profiles is the
Crank-Nicolson method 37. It has many variations, but the underlying idea is very simple. To

calculate a temperature profile at time tj+1, use an average of the temperature profile at tj and tj+1 in
all of the ∂/∂x terms. The temperature profile at tj+1 is unknown, so one must solve for the whole
profile at once. This type of solution is known as an implicit solution, and it is stable even for

large time steps. Of course, smaller time steps do lead to better numerical accuracy.
The best way to elucidate the idea is to show a sample case. Consider, for example, the
AP and binder temperature profiles known at n points in the solid phase of the propellant at time

tj. The task is to calculate the new temperature profiles, given the old profiles and new surface
temperatures at time tj+1. Figures 3.1 and 3.2 are representative steady-state temperature profiles
in the binder and AP for a (90, 80/20, 298) propellant.
 47
xn-1 … xi+1 xi xI-1 … x3 x2 x1 x0

Tap

Tb

Figure 3.1: Sample Temperature Profile in Propellant (Ideal)

1000

800 T
ap
T
b
T
i

600

T (K)

400

200

0
-100 -80 -60 -40 -20 0
X (µm)

Figure 3.2: Sample Temperature Profile in Propellant (Actual)

 48
First, note that the x-spacing in the calculated profile is not constant. (The circles and
lines represent calculation points.) The reason for the non-constant spacing is that the
temperature, as shown in Equation 1.12, grows exponentially near the surface. An x-spacing that
becomes exponentially smaller near the surface is therefore preferable because it has a fine
resolution near the surface where it is needed and a coarse resolution deep inside the propellant
where the temperature changes slowly with distance. For the present model, the equation
describing the x-spacing is

x i = x depth ⋅exp[
(i + 1 − n )⋅ν X ]− x depth ⋅exp[(1 − n )⋅ν X ]. (3.6)

The following derivation considers just one generic profile for now, though the model
contains two solid-phase calculations. Denoting, Ti,j+1 as the new temperature and Ti,j as the
previous temperature at some xi, the discrete mathematical environment looks like Figure 3.3.

i+1 , j+1 i , j+1 i-1 , j+1

Time

Distance

i+1 , j i,j i-1 , j

Figure 3.3: Discrete Computational Grid

 49
Equation 3.3 now becomes a discrete algebraic problem. The derivatives come from an

average between time tj+1 and time tj. In finite-difference form, the derivatives are

∂T
≈
1
(T
i,j+ 1 − Ti , j ), (3.7)
∂t x = xi ∆t

∂T 1 Ti − 1, j − Ti + 1, j  1 Ti − 1, j + 1 − Ti + 1, j + 1 
≈  +  , (3.8)
∂x x = xi 2 xi− 1 − xi+ 1
 2 x − x
  i− 1 i+ 1



and

 Ti − 1, j − Ti , j Ti , j − Ti + 1, j   Ti − 1, j + 1 − Ti , j + 1 Ti , j + 1 − Ti + 1, j + 1 
 −   −  (3.9)
∂T2
1  xi− 1 − xi xi − x i + 1  1  x i − 1 − xi x i − xi + 1 
≈  +  .
∂x 2 x = xi 2 xi− 1 − xi+ 1  2 xi− 1 − xi+ 1 
 2   2 
   

Let δ
xi = xi-1 – xi+1. Substituting Equations 3.7 - 3.9 into Equation 3.3 and placing all the j+1
(unknown) terms on one side, the final form is

 − α r j+ 1 
Ti − 1, j + 1  + + (3.10)
δx ⋅(x − x ) 2 ⋅δx 
 i i− 1 i i 
1 α α 
Ti , j + 1  + + +
 ∆t x − x xi − xi+ 1 
 i− 1 i 
 r j+ 1 α  Ti , j r j Ti − 1, j − Ti + 1, j 
Ti + 1, j + 1 − −  = −  +
 2 ⋅δx −  ∆t 2 
 i x i x i+ 1   x i − 1 − xi + 1 

α Ti − 1, j − Ti , j Ti , j − Ti + 1, j .
 − 
xi− 1 − xi+ 1  x − x xi − xi+ 1 
 i− 1 i 

Replacing the coefficients with constants, Equation 3.10 reduces to

Ti − 1, j + 1 Ai + Ti , j + 1 Bi + Ti + 1, j + 1 C i = Vi . (3.11)
 50
Boundary conditions dictate that V0=Ts, A0=An-1=1, and Vn-1=Ti. To simplify the equation

further, define a new matrix, M, by

B0 0 0 0 L 0 
A B1 C1 0 L 0 
(3.12)
 1 
0 A2 B2 C2 L 0 
M= .
0 0 A3 B3 L 0 
M M M M O 0 
 

0 0 0 0 0 Bn − 1 


The new temperature profile, as a function of the previous temperature profile and new
surface temperature, is therefore the solution of the matrix equation,

MT = V. (3.13)

Because M is a tri-diagonal matrix, a large number of very efficient and quick algorithms for
solving Equation 3.13 are available38.

C. Solution Criteria
Now, finally, there exists a nonsteady, nonlinear model. It consists of Equations 2.1, 2.2,
2.3, 2.4, 2.11, 2.23, 3.1, and 3.2, all solved simultaneously. Moreover Equations 3.1, and 3.2
depend on simultaneous temperature profile solutions, determined from different versions of
Equation 3.13. Table 3.3 contains a list of the dependent variables, with their equation numbers
and dependencies.

The only remaining questions from a computational standpoint are how to apply the
model to a given input and how to maintain stability in a time-dependent solution. Here, the
characteristic response time is a critical parameter. Its definition, from Chapter I, is

α
τ= . (1.9)
r2
 51

Table 3.3: Dependent Variables in Nonsteady Model

Function of Variable
Variable Eq. #
Gp Gap Gb xf,ap xf Ts,ap Ts,b Tf,ap

Gp 2.3 P P

Gap 2.1 P

Gb 2.2 P

xf,ap 2.4 P P

xf 2.11 P P P

Ts,ap 3.1 P P P P

Ts,b 3.2 P P P

Tf,ap 2.23 P P P
 52
The system obviously has two characteristic response times: one in the AP and one in the

binder. Denote τmax as the larger of the two and τmin as the smaller. The smallest possible time

step, therefore, should be the smallest characteristic response time divided into sufficiently small
increments. “Sufficiently small” in this case might mean at least 59 increments per response

time, i.e.,

τ min
∆t ≤ . (3.14)
59

In practice, however, numerical stability requirements are more restrictive, so the above

criterion rarely dominates. Characteristic response times are on the order of a few milliseconds,
yet the simulation usually requires a ∆t of around 10-5 seconds to maintain stability.
Different types of simulations, too, require different step sizes. In a simulation of a step-
or exponentially increasing pressure function, the lower of 10-5 or τmin/59 would certainly suffice.
Harmonic pressure oscillations at high frequency, however, might require a smaller step size in
order to obtain the appropriate number of increments per pressure oscillation.
For a simulation of a harmonic pressure input of frequency ω, the time step should be the
minimum of 10-5 s, 1/(ω·59), and τmin/59. In other words, the criterion is

τ min 1 
∆t = min
 , 10 − 5 , 
. (3.15)
 59 ω ⋅59 

Just as τmin governs the step size, τmax governs the length of the simulation. Under a step-
or exponentially-increasing pressure, the transient behavior is the region of interest. Thus, the

simulation should run until the transient behavior dies down, usually at three or four times τmax.
In oscillatory burning, however, the region of interest is after the response has developed
a condition of dynamic equilibrium. Again considering a harmonic pressure input of frequency

ω, run the simulation to either 10*τmax or 10 pressure oscillations, whichever is longer.

Figure 3.4 is an example of a step pressure input from 10 bar to 20 bar. Figure 3.5 is an
example of an oscillatory pressure at 2000 Hz, 10 bar mean and 20% oscillation amplitude. The
time scale for both figures is t/τmax with (90, 80/20, 298) propellant. Both figures incorporate the

“curve-fit” method of flame feedback, discussed in the following section.

 53
The nonsteady model resides as a Mathcad 7.0.3 file, with some of the more

computationally intensive routines written in Microsoft Visual C++. The nonsteady Mathcad

solution sheet is in Appendix B.

D. Solution Issues
Completely nonlinear nonsteady models present a set of challenges to the programmer
that test the limits of both computational accuracy and theoretical validity. Several notable

difficulties have appeared during the course of the current study.

1. Issue One: Burning Rate “Dragging” of the Gas Phase

Perhaps the most troubling issue is the apparent over-influence of burning rate on the gas
phase. In Equations 2.4, 2.11, and 2.23, the flame heights and flame temperatures are linearly
related to the mass fluxes and quadratically related to the pressure.
Even a linear dependence on burning rate, however, seems to override much of the
nonsteady behavior of the propellant system. Figure 3.6 is a simplified sketch of the feedback
from the flame into the propellant surface. Both the heat flux into the solid phase and the heat
flux out of the gas phase are functions of burning rates in the current model. Each functional
dependency acts like a form of damping, which tends to drag the model, shifting the response into
shorter times with smaller amplitudes.
Some modelers have completely disregarded the dependency of qf-s,ap and qf-s,b on burning
rate29,30 while others have included it in a simplified form28. Many researchers have investigated
the effect of flame modeling on transient response, but to the author’s knowledge, no one has yet
investigated multiple-flame, BDP-type flame structures in nonlinear, nonsteady regimes. The

over-dependence observed here is not encouraging.

 54

11 50
Gp

Gp,bar (20 bar)

45
10
Gp,bar (10 bar)
40
9

35
Mass Flux (kg·m )
-2

Pressure (bar)
8
30

7
25

6 20

15
5 Pressure (bar)

10
4
-1 0 1 t/τ 2 3 4
max

Figure 3.4: Response to Step Pressure Input

 55

G 20
p

G (10 bar)
7.6 p,bar

18
7.4

7.2 16
Mass Flux (kg·m ·s )
-1
-2

Pressure (bar)
7
14

6.8

12
6.6

6.4 10

6.2
8
Pressure Function
6
0 0.2 0.4 0.6 0.8 1 1.2 1.4
t/τ
max

Figure 3.5: Response to Oscillatory Pressure Input

 56

qs,b = q(rb , qf-s,b) qf-s,b = q(r’s , P)

Binder

AP
qs,ap = q(rap , qf-s,ap) qf-s,ap = q(r’s , P)

Figure 3.6: Heat Feedback Model

 57
All is not lost, however, as it is still possible to obtain response results that compare

favorably to those of other models. The method is simple; it involves eliminating some of the

dependencies shown in Figure 3.6 to minimize the damping in the system. Specifically, the
following algorithm will eliminate the gas phase dragging:

§ Run a steady-state simulation.

§ Fit the output of xf, xr, xf,ap, and Tf,ap as functions of pressure.
§ Substitute these curve-fits into the nonsteady model in place of Equations 2.4, 2.11, and 2.23.

The above procedure effectively decouples the gas phase from the burning rate.

Although this may not be an ideal simplification, it will produce results until a better method
comes along. Figure 3.7 is a plot of various methods of calculating response of a (90, 80/20, 298)
propellant to a step pressure input from 10 bar to 20 bar. “Full calculation” means that the
quantity depends on burning rate through the proper equation , and “curve-fit” means that the
quantity comes from a curve-fit of the steady-state data as a function of pressure only.
Figure 3.8 shows the same effect as Figure 3.7 in a frequency-response plot. The
propellant is (90, 80/20, 298) subjected to harmonic pressure oscillations at a mean of 10 bar,
with 20% oscillation magnitude. All curves come from a “peak-average” calculation method, as
discussed in the following section.
The figures indicate that the equations for Tf,ap and xf,ap are the most significant
contributions to damping in the system. This result matches expectations because Tf,ap and xf,ap
define the heat feedback to the AP, which is approximately 80% of the system. The large
difference between the full gas phase calculation and the curve-fit gas phase calculation in Figure
3.7 and Figure 3.8 is interesting. Clearly, the burning rate dependence adds significant damping
to the system.

All simulations for the remainder of the paper will use a curve-fit gas phase, unless stated
otherwise. To reiterate, this is not a perfect assumption, but it gets results.
 58

11
curve-fit gas phase c.f. : curve fit
f.c. : full calculation
10.5 x f.c. All other c.f.
d

10 x f.c. All other c.f.

f

x f.c. All other c.f.

Burning Rate (mm/s)

f,AP
9.5

8.5
T f.c. all others c.f.
f,AP

8
full gas phase quasi-steady

7.5

7
0 0.005 0.01 0.015 0.02
Time (s)
Figure 3.7: Effect of Curve-Fitting on Step Input
 59

1.2
curve-fit gas phase f.c. : Full Calculation
c.f. : Curve-Fit

1 x f.c. All other c.f.

d

0.8
T f.c. All other c.f.
f,ap
R (real part)

0.6
P

0.4

full calculation x f.c. All other c.f.

f,ap
0.2

0
0 500 1000 1500 2000
ω (Hz)
Figure 3.8: Effect of Curve-Fitting on Frequency Response
 60
2. Issue Two: Response Function Definition

The pressure-coupled response function, as previously mentioned, is the ratio of a

percentage change in burning rate to the percentage change in pressure. From Chapter I, the
pressure-coupled frequency response is

m&′m& r ′r
Rp = = . (1.4)
P ′P P ′P

_
The difficulty comes in the definition of r and r´. Linear models do not have a problem.
The output of a linear model, based on a harmonic input, is harmonic itself around the mean
_
steady-state solution. Thus, r and r´ are well-defined and the definition in Equation 1.4 is
unambiguous.
In nonlinear models, the output due to a harmonic forcing function is not itself harmonic,
the “mean” is not necessarily the steady-state solution at the mean of the forcing function, and the
peaks of the output are not symmetric about the arithmetic mean. Fortunately, in most cases the
output is periodic with a frequency equal to that of the driving function. Such periodicity does
not eliminate the ambiguous nature of Equation 1.4, but it does allow for an answer.
For example, Figure 3.9 is a plot of the final oscillation of Gp in response to a 125 Hz
harmonic driving pressure with a mean of 10 bar and 20% oscillation magnitude. The propellant
is (90, 80/20, 298), and the simulation has run to ten times the maximum characteristic response
time.
_
The “steady-state” line is simply the solution at P = P (10 bar). The “arithmetic mean” is

the average of all the points in the curve, and the “peak-average” is the sum of the top and bottom
peaks divided by two. All three of these methods are candidates for calculating the r̄ in Equation
1.4.
 61
Although difficult to see in Figure 3.9, the arithmetic mean does not exactly equal the

peak average. Thus, there are at least five different ways to calculate RP from the Figure 3.9:

1) r = arithmetic mean; r´ = rmax - r . 4) r = steady-state at P ; r´ = rmax - r .

2) r = arithmetic mean; r´ = r - rmin. 5) r = steady-state at P ; r´ = r - rmin.
3) r = (rmin + rmax)/2; r´ = either .

Method 3 is the recommended method because it usually returns an answer that is

between the extremes of the four other methods and because it returns only one value of RP for a

given r . Figure 3.10 is a plot of RP vs. frequency for all five of the calculation methods. The
two lines plotted against the secondary abscissa show the percentage difference between r and
the mean burning rate, using the mean burning rate as a baseline. The methods differ the most at
low frequency, and the percentage difference between the means is largest below 100Hz. Figure
3.10 represents a (90, 80/20, 298) propellant excited by a pressure oscillation of 20% about a
mean of 10 bar.
 62

8
Mass Flux (kg·m ·s )
-1
-2

6 Calculated
Steady State
Arithmetic Mean
Peak-Average

5
0 0.002 0.004 0.006 0.008
Time (s)

Figure 3.9: Final Oscillation of Mass Flux Under an Oscillatory Pressure Field
 63

1.2 8
Method 1)

1 Method 2) 6
Method 3)

% Difference Between "Means"

Method 4)
0.8 4
Method 5)
P

0.6 2
R

0.4 0

0.2 -2
Arithmetic Mean Difference
Peak-Average Difference

0 -4
0 200 400 600 800 1000
Frequency (Hz)

Figure 3.10: RP vs. Frequency Using Several Calculation Methods

 64

E. Preliminary Nonsteady Results

The following pages contain RP vs. frequency plots that illustrate the effect of various
parameters on the system.

Figure 3.11 is a frequency response plot for 80% AP, 20% HTPB propellants at 298K

initial temperature. The driving pressure in all cases is a harmonic function with a mean at 10 bar
and 20% pressure oscillations.

Figure 3.12 is a frequency response plot for a (50, 80/20, 298) propellant. The driving

pressure in all cases is a harmonic function with 20% oscillation magnitude, but the mean
pressure ranges from 10 bar to 100 bar.
Figure 3.13 is a frequency response plot for a (50, 80/20, 298) propellant. The driving
pressure in all cases is a harmonic function with a 10 bar mean pressure, but the oscillation
magnitude ranges from 5% to 30%.
Figure 3.14 is a frequency response plot for a 50µm AP particle diameter propellant at
298K initial temperature, while the AP mass percentage ranges from 73% to 87%. The driving
pressure in all cases is a harmonic function with a mean at 10 bar and 20% oscillation magnitude.
 65

2.5

5 µm
50 µm
2 90 µm
200 µm

1.5
R (real part)
P

0.5

0
0 500 1000 1500 2000
ω (Hz)
Figure 3.11: Effect of AP Particle Diameter on RP
 66

1.6

10 Bar
1.4
50 Bar

100 Bar
1.2
R (real part)

0.8
P

0.6

0.4

0.2
0 500 1000 1500 2000
ω (Hz)
Figure 3.12: Effect of Mean Pressure on RP
 67

1.6

1.4

1.2

1
R (real part)

0.8
P

0.6

5%
0.4 10%
20%
30%
0.2

0
100 1000
ω (Hz)
Figure 3.13: Effect of Oscillation Magnitude on RP
 68

α = 73%
α = 77%
α = 80%
α = 85%
1.5 α = 87%
R (real part)

1
P

0.5

0
0 500 1000 1500 2000
ω (Hz)
Figure 3.14: Effect of Oxidizer Mass Percentage on RP
 Chapter IV

RESULTS AND DISCUSSION

A. Steady-state
To a motor designer, the most important conditions of the steady-state model are the
effect of AP particle diameter, the effect of AP mass percentage, and the initial temperature
sensitivity. In addition, the theoretical turbulence modeling developed in this thesis has added
some complexity which deserves consideration. The following sections are a discussion of some
of the trends and relationships in these four areas.
1. Effect of Particle Diameter
The AP particle diameter size affects the model through the size of the diffusion flame.
The surface of a burning solid propellant is analogous to an array of Bunsen burners, where the
AP particle diameter controls the size of the burner nozzle. The diffusion flame then causes two
effects, both of which slow down the overall propellant burning rate.
First, large diffusion flames pull the total flame height high above the surface. Thus, xf is
larger in Equation 2.15, and the heat flux into the binder surface is smaller.
Second, the diffusion flame indirectly lowers the heat flux into the AP by changing the
pre-mixed flame temperature. When xf is large, the AP flame temperature drops according to

Equation 2.23. The heat flux into the AP surface falls because even though xf,ap stays relatively
small, the temperature at that point drops significantly.
Figure 4.1 is a plot of diffusion flame height and AP flame temperature for (5, 80/20,

298), (50, 80/20, 298), (90, 80/20, 298), and (200, 80/20, 298) propellants across a wide pressure
range. Notice how the smallest-diameter (5µm) propellant has almost no diffusion flame.
Consequently, it has the highest AP flame temperature. (See also Figure 2.4.)

69
 70

500 1500

400

Pre-Mixed Flame Temperature (K)

5 µm
Diffusion Flame Height (µm)

1000
5 µm
300 50 µm 50 µm

90 µm 90 µm
200 µm
200 µm
200
500

100

0 0
1 10 Pressure (Bar) 100 1000

Figure 4.1: Overall Flame Height and Pre-mixed Flame Temperature

 71

2. Effect of Turbulence

The diffusion flame height depends heavily on turbulent mixing at high pressures. The
“humps” in Figure 4.1 would have a constant, upward slope if not for the turbulent mixing that

kicks in with a low reaction flame height.

This should sound alarm bells in the alert reader’s ear. It is possible that the model is
“faking out” nature by employing unrealistically low flame heights to make up for deficiencies in

other areas. For example, radiation might play an important role in the burning rate, especially at

high pressure as the flame heights get lower.

One should remember, though, that the flame heights are really characteristic flame
heights. They do not represent the actual bright zone of gas that is visible over a strand of
burning propellant. They represent the point at which the temperature is the following (over
either binder or AP):

(T − Ts )
T( x f ) = T f −
f
. (4.1)
eν

Therefore, no one can say exactly how small the “flame heights” should be in an actual
burning propellant. Moreover, the purpose of the model is to study nonsteady heterogeneous
effects, so the gas phase is not as important as the solid phase, where most of the thermal lag
resides. As long as the gas model provides a reasonable heat-feedback relation, it is doing its job
splendidly.
Figure 4.2 contains two plots. The bottom portion is a plot of burning rate vs. pressure
for (90, 80/20, 298) and (5, 80/20, 298) propellants, compared to experimental data. The dotted

lines represent the theoretical calculations without turbulence. The top portion is a plot of total
flame height for the same cases. The diffusion flame in the plot continues to get higher with
pressure when turbulence is neglected. This causes an underestimation of burning rate at high

pressures.
 72

4
10

100
Burning Rate (mm/s)

1000
5 µm
5µm - No Turb. 10
90 µm
90µm - No Turb.

Flame Height (µm)

Experimental (5µm)
100 Model (5 µm)
Model (5µm - No Turb.)
1
Experimental (90µm)
Model (90 µm)
Model (90µm - No Turb.)

10

0.1

1 1 10 100 1000
Pressure (Bar)

Figure 4.2: Turbulence in the Steady-State Model

 73
3. Effect of AP Mass Percentage

Whereas most of the other parameters change the physics of the system, the AP mass

percentage affects the chemistry of the system. The parameters qf, Tf and Cp,g all come from
linear, single-variable interpolation of thermo-chemical-equilibrium calculations done at

ONERA, where the single variable is oxidizer mass percentage. In addition, the AP mass

percentage is a critical component of the total mass flux combination (Equation 2.3) and the total
solid propellant density. Figure 4.3 is a plot of the flame temperature and diffusion flame heat

release for a propellant at 298K initial temperature.

Raising the AP mass percentage will obviously raise the burning rate, which is why
motor designers often try to get αap as high as possible by using multi-modal propellants. Figure
4.4 shows the effect of AP mass percentage on the burning rate of a 50µm propellant at 298K
initial temperature. Higher values of αap seem to wash out the slowing effect of the diffusion
flame, which is why the curves in Figure 4.4 with higher values of αap seem to have shallower
“dips”.

6
3000 4 10

6
3.5 10

2500
T (K)

T (K) q (J/kg) 3 106

f

f f
q (J/kg)
f

2000
6
2.5 10

6
1500 2 10
0.72 0.74 0.76 0.78 0.8 0.82 0.84 0.86 0.88
α ap

Figure 4.3: Flame Temperature and Heat Release as Functions of αap

 74

1000

α = 73%
ap

α = 77%
ap

α = 80%
ap
100
α = 85%
Burning Rate (mm/s)

ap

α = 87%
ap

10

1
1 10 100 1000
Pressure (bar)

Figure 4.4: Effect of αap on Burning Rate

 75
4. Effect of Initial Temperature

Initial temperature takes two paths to affect the model. The first, most obvious, path is

the Ti term in Equation 2.19 and again in Equation 2.21. These two terms come from the heat
capacity times the difference between Ti and Ts. Intuitively, when something is cold, it takes

more energy to heat it than it does when the object is already warm.

The other path that Ti takes to affect the model is through the flame temperature. Simply
doing an energy balance at steady-state leads to

(T − Ti ,ref ).
C p ,s , p
T f = T f ,ref + i (4.2)
C p ,g , p

Because Tf is so far from the surface of both the binder and the AP, the effect of Equation
4.2 on the model is mild. It is a significant contribution, however, and the model must account
for it. To see how significant it is, note the correlation between Figure 4.1 and Figure 2.7. As the
diffusion flame gets higher, the contribution of Tf to the burning process becomes smaller, and the
temperature sensitivity, which depends on Tf though Equation 4.2, goes down. When xf falls
again at higher pressure, the temperature sensitivity goes back up.

B. Frequency Response
The following sections contain a discussion of the nonsteady model and the related,
dependent nature of its variables.
1. Effect of Mean Pressure

A rise in the mean pressure changes the nonsteady response in a very direct manner.
Higher pressures induce faster burning rates, which in turn shift the frequency response of a
propellant in a predictable pattern.

Equation 1.9 shows that the characteristic response time of a propellant is inversely
proportional to the square of the burning rate. A propellant with a higher burning rate has a much
shorter characteristic response time, and thus the pressure-coupled frequency response peak is
shifted toward higher frequencies.
 76
Moreover, the temperature profiles in the binder and AP are much shorter in faster

burning propellants. The integral terms in Equations 3.1 and 3.2 are less significant, and the

response amplitude should correspondingly be smaller. Figure 4.5 is a plot of steady-state binder
and AP temperature profiles for a (50, 80/20, 298) propellant at 10 bar and 100 bar pressure. The

profiles at 100 bar are clearly much steeper.

The mean pressure also contributes to the frequency response by changing the zero-
crossing of the response curve. This point represents the pressure exponent, which changes with

pressure. On a log-log plot of burning rate vs. pressure, the pressure exponent is the slope of the

burning rate curve (as in Figure 2.4).

In sum, an increasing pressure should shift the peak of the pressure coupled frequency
response curve to higher frequencies while simultaneously diminishing the amplitude of the curve
although the whole curve can shift up or down depending on the value of n at the particular
pressure. Figure 3.12 shows this predicted trend. The curve at 100 bar is higher than the curve at
50 bar because the pressure exponent is higher for a (50, 80/20, 298) propellant at 100 bar than at
50 bar. Figure 4.6 is a plot of the same data, where each curve is normalized by subtracting the
pressure exponent from RP.
2. Effect of AP Mass Percentage
The AP mass percentage also affects the frequency response through the burning rate and
pressure exponent. Higher burning rates lead to response peaks at higher frequency with lower
peak amplitude, while the pressure exponent shifts the entire curve either up or down. Figure 4.7
is a plot of the steady-state curves of a 50µm AP particle diameter propellant at around 10 bars
and 298K initial temperature. It is essentially a blown-up version of Figure 4.4. Figure 4.8 is an
n-normalized plot of the data from Figure 3.14. The expected trends are evident.
 77

1000 T (10 bar)

binder

T (10 bar)
ap

T (100 bar)
binder

T (100 bar)
800 ap

T
i
Temperature (K)

600

400

200
-5 -5 -5 -5 -5
-5 10 -4 10 -3 10 -2 10 -1 10 0
Position (µm)
Figure 4.5: Steady-state Temperature Profiles at Two Mean Pressures
 78

10 bar
0.8
50 bar

100 bar
0.6

0.4
R -n
P

0.2

-0.2

-0.4
0 500 1000 1500 2000
ω (Hz)
Figure 4.6: Normalized RP Plot: Effect of Mean Pressure
 79

10
n = 0.33
n = 0.35

n = 0.53
Burning Rate (mm/s)

n = 0.71

n = 0.90
α ap = 73%
α = 77%
ap
α = 80%
ap
α = 85%
1 ap
α ap = 87%
2 3 4 5 6 7 8 9 10
Pressure (bar)

Figure 4.7: Steady-state Curves for Different AP Mass Percentages

 80

1.5

α = 73%
ap
α = 77%
ap
1 α = 80%
ap
α = 85%
ap
α = 87%
ap

0.5
R -n
P

-0.5

-1
0 500 1000 1500 2000
ω (Hz)
Figure 4.8: Normalized RP Plot: Effect of AP Mass Percentage at 10 bar
 81
3. Effect of AP Particle Diameter

In the nonsteady-state, just as in the steady-state, the AP particle diameter changes the

system mostly through the diffusion flame. Larger AP particle diameters induce lower AP flame
temperatures and higher overall flame heights.

Composite propellants with small AP particles typically burn faster than their

counterparts with larger AP particles. As in the previous two sections, Equation 1.9 indicates that
the characteristic response time is lower. The response peak should therefore occur at higher

frequencies with a smaller amplitude.

Oddly enough, Figure 3.11 does not show such a trend. The frequencies are certainly
higher for smaller AP particle diameters, but the amplitudes actually increase! The pressure
exponent for the 5µm propellant is high compared to the others, but this does not explain the very
large peak. Figure 4.9 is an n-normalized version of Figure 3.11.
4. Path of Burning Rate Dependence
The previous three sections lead to some interesting hypotheses about the nature of gas-
phase “dragging” done by the burning rate. Faster burning propellants have lower response
amplitudes than slower burning propellants, as long as the AP particle diameters are equal. In
contrast, a propellant with a smaller AP particle diameter will almost certainly have a higher
response amplitude, even though it burns faster. All the evidence points toward one culprit for
the gas-phase dragging: the diffusion flame.
As the AP particle diameter gets smaller, the combustion model tends to resemble a
homogeneous propellant. The diffusion flame, specifically, becomes very small, and the pre-
mixed flame temperature becomes larger, almost matching the adiabatic flame temperature.
Figure 4.1 shows the effect quite clearly.

Because the response amplitude is higher in smaller AP particle diameter propellants, one
would expect the dragging effects to be less significant. Unfortunately, this does not seem to be
the case. Figure 4.10 shows that the dragging is just as significant in a (5, 80/20, 298) propellant

as it is in a (90, 80/20, 298) or (50, 80/20, 298) propellant. The path through which the gas phase
affects the frequency response is therefore not a simple, linear relationship.
 82

1.5

5 µm
50 µm
90 µm
1 200 µm
R -n

0.5
P

-0.5
0 500 1000 1500 2000
ω (Hz)
Figure 4.9: Normalized RP Plot: Effect of AP Particle Diameter at 10 bar
 83

2.5

(5, 80/20, 298) c.f.

(5, 80/20, 298) f.c.
(50, 80/20, 298) c.f.
2 (50, 80/20, 298) f.c.
(90, 80/20, 298) c.f.
(90, 80/20, 298) f.c.

1.5
R (real part)
P

0.5

0
0 500 1000 1500 2000
ω (Hz)
Figure 4.10: Effect of Gas-Phase Dragging on Three Propellants
 84
Another possible explanation for the trend of higher response amplitude with lower AP

particle diameter is that propellants with large exponents will show high response peaks, even if

the shift is factored out. That is, the pressure exponent amplifies the peak in addition to shifting
it.

Consider, for example, a pressure region where the pressure exponent is not necessarily

larger with decreasing AP particle diameter. Figure 4.11 is a plot of steady-state burning rate
curves for various AP particle diameters of an 80%AP / 20%HTPB propellant at 298K initial

temperature around 100 bar pressure. The 200µm propellant has the highest pressure exponent,

followed by the 5µm, 90µm, and 50µm propellants in descending order.

Figure 4.12 shows the response calculated with a mean pressure of 100 bar and an
oscillation magnitude of 20%. The 5µm curve still has the highest peak, followed by the 200 µm,
90µm, and 50µm propellants. Figure 4.13 is an n-normalized version.
In Figure 4.13, the peak amplitude is definitely higher for the 5µm propellant, even
though its pressure exponent is lower than that of the 200µm propellant. The general trend is
lower response with smaller AP particle diameter, except in the case of the 5µm propellant. The
pressure exponent may cause a slight amplification in addition to the shift, but it is not enough to
cause the 200µm propellant to mimic the very high response amplitude of the 5µm propellant.
Thus, the diffusion flame remains as a likely culprit in diminishing response amplitude, while the
pressure exponent mostly shifts the RP curve up or down.
There is, as yet, no direct evidence of exactly how the diffusion flame or any other aspect
of the gas phase influences response amplitude. Neither the pre-mixed flame temperature nor the
total flame height alone can account for the problem, as Figures 3.7 and 3.8 illustrate.
One conclusion remains clear — the over-simplified notion of a pressure-dependent-only

gas phase is physically unrealistic. Future research in this area must either “fix” the gas phase
model or develop a phenomenological explanation of why multiple-flame models resembling the
one in this thesis are fundamentally incapable of predicting adequate response amplitudes.
 85

10
n = 0.579

n = 0.448
Burning Rate (mm/s)

n = 0.544

n = 0.0.671

200 µm
90 µm
50 µm
5 µm
1
2 3 4 5 6 7 8 9 10
Pressure (bar)

Figure 4.11: Steady-state Curves for Different AP Particle Diameters

 86

1
5 µm
50 µm
0.9 90 µm
200 µm

0.8
R (real part)

0.7

0.6
P

0.5

0.4

0.3
4
100 1000 10
ω (Hz)
Figure 4.12: Effect of AP Particle Diameter on RP at 100 bar
 87

0.4
5 µm
50 µm
0.3 90 µm
200 µm

0.2
R -n

0.1
P

-0.1

-0.2
4
100 1000 10
ω (Hz)
Figure 4.13: Normalized RP Plot: Effect of AP Particle Diameter at 100 bar
 88
5. Effect of Oscillation Amplitude

Nonlinear effects should show up most profoundly through the oscillation magnitude. In

a linear system, the response to a harmonic input is itself harmonic, so the oscillation magnitude
is completely irrelevant. In nonlinear systems, however, the response becomes “less harmonic”

as the input magnitude goes up. Thus, one should see a definite trend of some sort as the driving

pressure magnitude goes from 5% to 30%.

Figure 3.13 does not show much of a trend at all. There may be a simple explanation for

its absence; the peak-average method of calculating RP could be minimizing the nonlinear effects.

Again, the basic problem is that the definition of RP is inherently linear, so any RP taken
from a nonlinear simulation will be somewhat contrived and arbitrary. The effect of oscillation
amplitude could differ greatly, depending on the calculation method. For example, Figures 4.14
and 4.15 are plots of RP vs. frequency for various oscillation magnitudes. The curves in Figure
4.14 come from method 4, where r̄ is the steady-state burning rate and r´ is the maximum positive
change in burning rate over the course of one oscillation. The curves in Figure 4.15 come from
method 5, where r̄ is also the steady-state burning rate, but r´ is the maximum negative change in
burning rate. They both show a relatively high dependence on oscillation magnitude compared to
Figure 3.13. Oddly enough, the trends in Figures 4.14 and 4.15 are reversed. In Figure 4.14,
increasing oscillation magnitude diminishes the response peak. In Figure 4.15, it is the opposite.
 89

1.6

1.4

1.2

1
R (real part)

0.8
P

0.6

5%
0.4 10%
20%
30%
0.2

0
100 1000
ω (Hz)
Figure 4.14: Effect of Oscillation Magnitude on RP Revisited: Method 4
 90

1.6

1.4

1.2
R (real part)

0.8
P

0.6 5%
10%
20%
0.4 30%

0.2
100 1000
ω (Hz)
Figure 4.15: Effect of Oscillation Magnitude on RP Revisited: Method 5
 Chapter V

SUMMARY AND CONCLUSIONS

A. Steady-state
In general, the steady-state predictions seem to match the available experimental data
very well. Figures 2.5 and 2.6 show that the theoretical predictions are usually within 10% of
experimental measurements done at ONERA on a class of mono-modal propellants. Temperature
sensitivity data are sparse and not very accurate anyway, so it would be difficult to compare
Figure 2.7 to any real propellants. Work is ongoing in this area.
There is a problem, however, in comparing theoretical predictions to experimental data.
The chemical properties of HTPB can vary widely, depending on the manufacturer, curative, cure
cycle, etc. The molecular weight, for example, can vary from under 1200 gm/mol to over 5500
gm/mol 39. There is no guarantee that the properties of HTPB used here, which are based on
French HTPB, will correspond to HTPB manufactured in the U.S. or anywhere else. The model
might require completely different values in Table 2.2, given a different type of binder.
Regardless of these issues, the steady-state part of the model certainly does seem to
return excellent results. Although the model may still be mimicking nature with unrealistic

parameters rather than solid theoretical reasoning, it seems at least to pass all the obvious checks.
As the model makes more steady-state predictions that seem reasonable, it gains credibility. Only
further experimentation and validation will tell.

In summary, some of the most important conclusions from the steady-state model are as
follows:
• Propellants with smaller AP particle diameters typically burn faster. This is probably due to

the lessening effect of the diffusion flame.

91
 92
• Higher AP mass percentages induce faster burning rates as well.

• Predicted temperature sensitivity, σP, is on the order of 0.002 K-1, a reasonable value for

AP/HTPB, mono-modal composites.

• The model predicts that propellants with larger AP particle diameters will have higher values

of σP at lower pressure, but the effect is not pronounced.

B. Nonsteady-State
The nonsteady regime, in contrast to the steady state, is less encouraging. By far, the

most troubling aspect of the nonsteady model is the apparent over-dependence of flame heat
feedback on the burning rates. Although the assumption of a solely pressure-dependent gas phase
will produce results, this is an unsatisfactory approximation. Models are supposed to grow more
accurate with fewer assumptions, not less accurate, and it does seem odd that a more realistic
model leads to less realistic results. Some of the most important conclusions from the nonsteady
model results are as follows:
• Faster burning propellants typically have a pressure-coupled frequency-response peak at
higher frequencies of pressure oscillations. Also, their frequency-response peaks are of lower
amplitude.
• The pressure exponent of the propellant tends to shift the RP curve up or down.
• One exception to the above rules is that propellants with extremely fine AP particle diameters
tend to demonstrate a high frequency response. The diffusion flame, or lack thereof, may be
responsible for this observed effect.
• Pressure oscillation amplitude has very little effect on frequency response. RP has an

ambiguous definition in nonlinear models, however, so this result may not be too meaningful.

C. Recommendations
The next few paragraphs are recommendations for future work and study in the area of
composite propellant combustion modeling.
First, it should be possible to either “fix” the nonsteady model so that it returns higher

response amplitudes with full calculation or develop a phenomenological explanation of why

 93
BDP-type models such as these are fundamentally unable to predict reasonable response

amplitudes.

Second, future models should include the effect of additives, such as the ubiquitous
aluminum. Many researchers have been working on the “aluminum problem” for some time, so

this is not going to be an easy feat.

Third, the model should eventually include the effect of multi-modal propellants. Mono-
modal propellants are rare in practical motors because of the limited AP mass percentage, so if

theoretical models such as the one in this thesis are to exhibit any practical use, they must

eventually describe propellants with multiple oxidizer particle diameters.

Fourth, as a philosophical recommendation, researchers should continue to do theoretical
modeling because it is a worthy scientific endeavor. Experimental research is an excellent way to
catalogue observed physical phenomena and indeed is absolutely essential for scientific progress.
Theory, however, contributes to the understanding of physical phenomena, as opposed to the
classification of it. This alone makes theory valuable, even if it has no immediate application.
APPENDICES
 APPENDIX A

Steady-State Model

The following pages are a direct copy from a Mathcad 7.0.3 sheet. This Appendix

contains a version of the steady-state model, shown immediately after a sample calculation.

94
95
96
97
98
99
100
101
102
 APPENDIX B

Nonsteady-State Model

The following pages are a direct copy from a Mathcad 7.0.3 sheet. This Appendix

contains a version of the nonsteady-state model, shown immediately after a sample calculation.

103
104
105
106
107
108
109
110
111
112
113
114
115
116
 REFERENCES

1
Kubota, N. 1984. “Survey of Rocket Propellants and Their Combustion Characteristics,”
Fundamentals of Solid Propellant Combustion, Kuo, K., and M. Summerfield (Eds.). AIAA
Progress in Aeronautics and Astronautics, Vol. 90, pp. 1-52.
2
Strand, L., and R. Brown. 1992. “Laboratory Test Methods for Combustion-Stability
Properties of Solid Propellants,” Nonsteady Burning and Combustion Stability of Solid
Propellants, De Luca, L., E. Price and M. Summerfield, (Eds.). AIAA Progress in Aeronautics
and Astronautics, Vol. 143, pp. 689-718.
3
Crawford, B. L., Jr., C. Huggett, F. Daniels and R. E. Wilfong. 1947. “Direct Determination
of Burning Rates of Propellant Powders,” Analytical Chemistry, Vol. 19, No. 9, pp. 630-633.
4
Cauty, F. and J. Demarais. 1990. “Ultrasonic Measurement of The Uncured Solid Propellant
Burning Rate,” Proceedings of the 21st ICT International Congress, Karlsruhe, Germany.
ONERA TP 1990-90.
5
McQuade, W. 1998. Ultrasound Technique Resolution for Solid Propellant Burning Rate
Measurement, M.S. Thesis, University of Alabama in Huntsville, Huntsville, AL.
6
Razdan, M., and K. Kuo. 1984. “Erosive Burning of Solid Propellants,” Fundamentals of
Solid Propellant Combustion, Kuo, K., and M. Summerfield (Eds.). AIAA Progress in
Aeronautics and Astronautics, Vol. 90, pp. 515-598.
7
Cohen, N., and D. Flanigan. 1983. A Literature Review of Solid Propellant Burn Rate
Temperature Sensitivity Morton Thiokol Corporation, Huntsville, AL. AFRPL-TR-83-042.
8
Price, E. 1992. “Solid Rocket Combustion Instability- An American Historical Account,”
Nonsteady Burning and Combustion Stability of Solid Propellants, De Luca, L., E. Price and M.
Summerfield, (Eds.). AIAA Progress in Aeronautics and Astronautics, Vol. 143, pp. 1-16.
9
Strand, L., A. Schultz and G. Reedy. 1974. “Microwave Doppler Shift Technique for
Determining Solid Propellant Regression Rates,” Journal of Spacecraft and Rockets, Vol. 11, No.
2, pp.75-80.
10
Traineau, J., M. Prévost, and P. Tarrin. 1994. “Experimental Low and Medium Frequency
Determination of Solid Propellants Pressure-Coupled Response Function,” AIAA 94-3043.
11
Cauty, F., P. Comas, F. Vuillot and M. Micci. 1996. “Magnetic Flow Meter Measurement
of Solid Propellant Pressure-Coupled Responses Using an Acoustic Analysis,” Journal of
Propulsion and Power, Vol. 12, No. 2, pp.436-438.
12
Beckstead, M., R. Derr and C. Price. 1970. “A Model of Solid-Propellant Combustion
Based on Multiple Flames,” AIAA Journal, Vol. 8, No. 12, pp.2200-2207.
13
Cohen, N. 1980. “Review of Composite Burn Rate Modeling,” AIAA Journal, Vol. 18, No.
3, pp. 277-293.

117
 118

14
Cohen, N., and L. Strand. 1982. “An Improved Model for the Combustion of AP
Composite Propellants,” AIAA Journal, Vol. 20, No. 12, pp.1739-1746.
15
Condon, J., J. Osborn and R. Glick. 1976. “Statistical Analysis of Polydisperse,
Heterogeneous Propellant Combustion: Steady-state,” 13th JANNAF Combustion Meeting, CPIA
Publication 281, Vol. 2, pp.313-345.
16
Condon, J., J. Renie and J. Osborn. 1979. “Oxidizer Size Distribution Effects on Propellant
Combustion,” AIAA Journal, Vol. 17, No. 8, pp.878-883.
17
Sutton, G. P. 1992. Rocket Propulsion Elements. Wiley, New York.
18
Ramohalli, K. 1984. “Steady-State Burning of Composite Propellants under Zero Cross-
Flow Situation,” Fundamentals of Solid Propellant Combustion, Kuo, K., and M. Summerfield
(Eds.). AIAA Progress in Aeronautics and Astronautics, Vol. 90, pp. 409-477.
19
Denison, M., and E. Baum. 1961.“A Simplified Model of Unstable Burning in Solid
Propellants,” Journal of the American Rocket Society, Vol. 31, 1961, p.1112.
20
Novozhilov, B. 1992. “Theory of Nonsteady Burning and Combustion Stability of Solid
Propellants Using the Zeldovich-Novozhilov Method,” Nonsteady Burning and Combustion
Stability of Solid Propellants, De Luca, L., E. Price and M. Summerfield, (Eds.). AIAA Progress
in Aeronautics and Astronautics, Vol. 143, pp. 601-639.
21
Barrère, M. 1992. “Introduction to Nonsteady Burning and Combustion Instability,”
Nonsteady Burning and Combustion Stability of Solid Propellants, De Luca, L., E. Price and M.
Summerfield, (Eds.). AIAA Progress in Aeronautics and Astronautics, Vol. 143, pp. 17-58.
22
Brewster, Q, and S. Son. 1995. “Quasi-Steady Combustion Modeling of Homogeneous
Solid Propellants,” Combustion and Flame, Vol. 103, pp. 11-26.
23
Lengellé, G. 1970. “Thermal Degradation Kinetics and Surface Pyrolysis of Polymers,”
AIAA Journal, Vol. 8 No. 11, pp. 1989-1998.
24
Group discussions at a meeting for the Multidisciplinary University Research Initiative
(MURI), Cleveland, OH; July 1998.
25
Cohen, N. 1981. "Response Function Theories That Account for Size Distribution
Effects— a Review," AIAA Journal, Vol. 19, No. 7, pp. 907-912.
26
Condon, J., J. Osborn and R. Glick. 1976. “Statistical Analysis of Polydisperse,
Heterogeneous Propellant Combustion: Nonsteady-state,” 13th JANNAF Combustion Meeting,
CPIA Publication 281, Vol. 2, pp.209-223.
27
Brown, R., and R. Muzzy. 1970. “Pressure-Coupled Combustion Instability,” AIAA
Journal, Vol. 8, No. 8, pp. 1492-1500.
28
Galfetti, L., G. Riva and C. Bruno. 1992. “Numerical Computations of Solid-Propellant
Nonsteady Burning in Open or Confined Volumes,” Nonsteady Burning and Combustion
 119

Stability of Solid Propellants, De Luca, L., E. Price and M. Summerfield, (Eds.). AIAA Progress
in Aeronautics and Astronautics, Vol. 143, pp. 643-687.
29
Louwers, J., and G. Gaidot. 1997. “Model for the Nonlinear Transient Burning of
Hydrazinium Nitroformate,” International Workshop: Combustion Instability of Solid Propellants
and Rocket Motors, Milano, Italy.
30
Louwers, J. and G. Gaidot. 1996. “Nonlinear Transient Burning of Composite Propellants:
The Effect of Solid Phase Reactions,” Proceedings of the 4th International Symposium on Special
Topics in Chemical Propulsion: Challenges in Propellants & Combustion 100 Years after Nobel,
Stockholm, Sweden.
31
Chiaverini, M., G. Harting, L. Yeu-Cherng, K. Kuo, A. Peretz, S. Jones, B. Wygle and J.
Arves. 1997. “Pyrolysis Behavior of Hybrid Rocket Solid Fuels Under Rapid Heating
Conditions,” AIAA 97-3078.
32
Chen, J., and T. Brill. 1991. “Chemistry and Kinetics of Hydroxyl-terminated
Polybutadiene (HTPB) and Diisocyanate-HTPB Polymers during Slow Decomposition and
Combustion-Like Conditions,” Combustion and Flame, Vol. 87, pp. 217-232.
33
Burke, S., and T. Schumann. 1928. “Diffusion Flames,” Industrial and Engineering
Chemistry, Vol. 20, p. 998.
34
Penner, S. S. 1957. Chemistry Problems in Jet Propulsion. Wiley, Los Angeles.
35
Mathcad User’s Guide. (Version 6.0). 1995. MathSoft, Inc. Cambridge, MA.
36
Holman, J. 1927. Heat Transfer. McGraw Hill, New York..
37
Kreyszig, E. 1993. Advanced Engineering Mathematics. John Wiley & Sons, New York.
38
Press, W., B. Flannery, S. Teukolsky and W. Vetterling. 1987. Numerical Recipes.
Cambridge University Press, New York, p. 40.
39
Jeppson, M., M. Beckstead and Q. Jing. 1998. “A Kinetic Model for the Premixed
Combustion of a Fine AP/HTPB Propellant,” AIAA 98-0447.
 BILBIOGRAPHY

1. Beckstead, M. 1981. “A Model for Solid Propellant Combustion,” 18th Symposium

(International) on Combustion, The Combustion Institute, pp. 175-185.

2. Bellec, R., J. Duterque and G. Lengellé. 1996. Modélisation de la Combustion des

Propergols Solides Aluminisés, ONERA Rapport Technique 37/7128 EN.

3. Condon, J. 1978. The Effect of Oxidizer Particle Size on the Steady-state Combustion,
Erosive Burning and Nonsteady Combustion of Polydisperse Composite Solid
Propellants, Ph.D. Dissertation, Purdue University, May.

4. King, M. 1996. “Examination of Chemical Approaches to Stabilizing Composite-

Propellant Combustion” Journal of Propulsion and Power, Vol. 12, No. 3, pp. 554-563.

5. Renie, J. 1982. Combustion Modeling of Composite Solid Propellants, Ph.D.

Dissertation, Purdue University.

6. Rasmussen, B., and R. A. Frederick, Jr. 1998. “Issues in Nonlinear, Nonsteady Solid
Combustion Modeling” 1998 JANNAF CS/PSHS/APS Joint Meeting; Tucson, AZ.

7. Rasmussen, B., R. A. Frederick, Jr., M. D. Moser and G. Lengellé. "A Theoretical

Pressure-Driven Response Function for Composite Solid Propellants," 1998 JANNAF
Propulsion Meeting; Cleveland, OH.

8. Rasmussen, B., R. A. Frederick, Jr. and G. Lengellé. 1997. "Pressure-Coupled Frequency

Response Models of Solid Propellants," 1997 JANNAF CS/PSHS/APS Joint Meeting;
W. Palm Beach, FL.

120