University of Alabama in Huntsville

LOUIS

Honors Capstone Projects and Theses Honors College

5-1-2022

Ballistic Modeling of Multi-Grain Solid Rocket Motor Performance

Variability
Justin Alexander Phillips

Follow this and additional works at: https://louis.uah.edu/honors-capstones

Recommended Citation
Phillips, Justin Alexander, "Ballistic Modeling of Multi-Grain Solid Rocket Motor Performance Variability"
(2022). Honors Capstone Projects and Theses. 727.
https://louis.uah.edu/honors-capstones/727

This Thesis is brought to you for free and open access by the Honors College at LOUIS. It has been accepted for
inclusion in Honors Capstone Projects and Theses by an authorized administrator of LOUIS.
 Ballistic Modeling of Multi-Grain Solid Rocket
Motor Performance Variability
by

Justin Alexander Phillips

An Honors Capstone

submitted in partial fulfillment of the requirements

for the Honors Diploma

to

The Honors College

of

The University of Alabama in Huntsville

1 May 2022

Honors Capstone Director: Dr. David Lineberry

Research Engineer, UAH Propulsion Research Center

05/01/2022
_____________________________________________________
Student Date

_____________________________________________________
05/01/2022
Director Date

05/02/2022
_____________________________________________________
Department Chair Date
Digitally signed by William
William Wilkerson Wilkerson
Date: 2022.05.03 10:12:17 -05'00'
_____________________________________________________
Honors College Dean Date
 Honors College
Frank Franz Hall
+1 (256) 824-6450 (voice)
+1 (256) 824-7339 (fax)
[email protected]

Honors Thesis Copyright Permission

This form must be signed by the student and submitted as a bound part of the thesis.

In presenting this thesis in partial fulfillment of the requirements for Honors Diploma or Certificate
from The University of Alabama in Huntsville, I agree that the Library of this University shall make it
freely available for inspection. I further agree that permission for extensive copying for scholarly
purposes may be granted by my advisor or, in his/her absence, by the Chair of the Department,
Director of the Program, or the Dean of the Honors College. It is also understood that due recognition
shall be given to me and to The University of Alabama in Huntsville in any scholarly use which may be
made of any material in this thesis.

Justin Phillips
____________________________

Student Name (printed)

____________________________

Student Signature

05/01/2022
___________

Date
 Ballistic Modeling of Multi-Grain Solid
Rocket Motor Performance Variability
Justin A. Phillips1
The University of Alabama in Huntsville, Huntsville, AL, 35899, United States of America

This project aims to demonstrate the affect the number of grains has on the performance
variability. The hypothesis was that motor grain variability will have less of an effect on the
motor performance the higher the number of grains. Essentially, the 95% interval should
decrease in spread as the higher grain count drives the performance closer to the expected
mean. In order to show this, a ballistics code was developed using the propellant properties
for Aerotech Blue Thunder propellant. Grain bore diameter, grain length, propellant density,
grain outer diameter, nozzle throat diameter, and nozzle erosion rate were varied based on
measured values from four static motor tests using the Aerotech G138 motor. Then, the code
used the Monte Carlo method to vary the grain properties and conduct a ballistic simulation
to calculate total impulse and average thrust. The code was compared against the results from
the Aerotech G138 motor tests. The code then ran several multi-grain motor configurations
using the same propellant with shorter grains. Motors with up to 5 grains were simulated. The
results of the analysis showed that the variation in thrust decreased from 8.5% to 7.2% and
the total impulse decreased from 10.33% to 9.96% between the single grain and the five-grain
motor configuration. Although the reduction in variability was less significant than expected,
the trend confirms that the performance variability decreases as the number of grains in
increased.

I. Nomenclature
Ab = total burn surface area
Ae = area of the nozzle exit
At = area of the throat
a = burning rate coefficient
cf = thrust coefficient
c* = characteristic velocity
db = diameter of the bore
dg = diameter of the grain
dt = diameter of the throat
! = area ratio
" = ratio of specific heats
I = total impulse
L = length of the bore
Me = Mach number at nozzle exit
n = burning rate exponent
pa = ambient pressure
pc = chamber pressure
pe = exit pressure
rb = burn rate
re = throat erosion rate
rf = final radius of the bore
ri = initial radius of the bore

1
Honors Student, Mechanical and Aerospace Engineering.

1
#p = propellant density
$ = standard deviation
T = thrust
t = time
tb = burn time
tl = thickness of the liner
∆t = time step
Wp = propellant weight
w = web distance
wf = final web distance

II. Introduction
This paper aims to describe the purpose, method, and results of this project. The objective of this project was to
determine the effect of variability on the performance of multi-grain motors. The hypothesis is that as the number of
grains increase, the effect of variability on the performance of any single grain will even out in a multi-grain case.
Consider a two-grain rocket motor. If one grain underperforms and another grain overperforms, the effect of the
variability on the motor should even out some, most likely to the expected performance of the motor. As the number
of grains increase, a motor should, theoretically, perform more closely to its expected performance. Its percent
uncertainty should be less than it was for one grain.
In order to do this, an Aerotech 29-mm G138 was chosen for comparison. A ballistic simulation and Monte Carlo
analysis was programmed into MATLAB to simulate the burn of this motor. The dimensions of this motor were
measured, and the propellent data recorded; both of which were used in the simulation. The simulation runs five
different cases for one to five grains. The single grain motor was used as a baseline case to compare to actual recorded
data from a single grain G138 motor. Once the ballistic simulation was confirmed via the baseline case, the results
from the multi-grain Monte Carlo analyses were compared to determine the effect that number of grains has on
performance variability.

III. Method of Analysis

The following section helps explain how the ballistic simulation works. The MATLAB code is split into three
separate files: the main file Ballistics.m and the two functions burnArea.m and thrustCalc.m. There is an additional
file named variability_info.m used to calculate the mean values and variability for the variable properties defined in
the main file. The main file contains the Monte Carlo analysis. It calls the function thrustCalc to determine the thrust
curve generated over the burn time of the motor. From this, it can calculate the total impulse of the motor. It then
graphs the results. The thrustCurve function uses the burnArea function to calculate the burn surface area at each time
step. All the code used in this project can be found in the Appendix.
To conduct a Monte Carlo analysis, it is important to vary certain parameters to create the random distribution in
the thrust and total impulse results for each motor. Below are the selected properties, their mean values, and their
variability. Most of the values were obtained by averaging the values recorded and by finding their standard deviation.
Three values were calculated in a different manner. The exit diameter of the nozzle was only recorded once; its value
is the recorded value, and its variability is its smallest decimal place. The throat erosion rate was calculated by taking
the difference in the measurements before and after testing and dividing them by the recorded burn time as seen in
Equation 1. The mean and standard deviation of these results were taken to get the base value and its variability.

'"# − '"$
%! = (1)
)%

The most difficult property to calculate was the density of the propellant. This was due to the fact that the resolution
of the scale used to measure the weight of the motors was ± 1 g, and all of the grains weighed 75 g. Variability was
introduced by setting ± 0.5 g as ± two times the standard deviation. This is because of the empirical rule which states
that “95% of observation falls within two standard deviations of the mean on a normal distribution” [2]. The following
two equations, Equation 2 and Equation 3, show how the density was calculated.

2
 (
'' '% (
*& = + ,- − )' 0 − - 0 1 ∗ 3) (2)
2 2

4&
#& = (3)
*&

Once the propellant density values were obtained, the average was taken. This was the mean value of the property.
The variability was determined by determining the difference in the propellant density of the propellant weight
increased by 0.25 g and decreased by 0.25 g. The mean of these values was considered to be the variability of the
propellant density. This calculation sequence was to estimate these values as they were measured with a scale that did
not have much precision. The mean values and variability for all these values can be seen in Table 1.

Table 1: Properties Varied for Monte Carlo Analysis

Property Mean Value Variability

#& [367/9:^3 ] 0.0664 2.2140e-04

3) [9:] 3.9110 0.0059

'% [9:] 0.2752 0.0024

') [9:] 0.9900 0.0200

'" [9:] 0.1760 0.0038

)' [9:] 0.0240 0.0022

'! [9:] 0.4170 0.0010

%! [9:/>] 0.0554 0.0052

Random normal distributions with the same number of data points as Monte Carlo iterations were created. These
were shifted by centering their centers to the mean values of each parameter and multiplying by the variability. This
meant that one standard deviation would be equal to that property’s variability.
In order to perform a Monte Carlo analysis on motors with varying number of grains, an initial basic ballistic
simulation had to be developed. To do this, the quasi-steady lumped parameter method was chosen. The quasi-steady
lumped parameter method assumes that the mass flow entering the chamber is equal to the mass flow exiting the
chamber. This fact along with St. Robert’s law helps determine the instantaneous chamber pressure. From this, thrust
and impulse can be calculated [1]. Essentially, this method increments a time step, determines the web distance at this
step, calculates the chamber pressure, and from this determines the thrust produced by the rocket. By integrating this
value over the time of the burn, the total impulse was calculated.
First, the time and web distance were set to 0 s and 0 in. respectively. A time step of 0.001 s was chosen for this
simulation. The final web distance was determined to be the maximum difference between the initial and final bore
radii between all grains, essentially the thickest grain. Then, the equations for the thrust coefficient, cf, and exit
pressure, pe, were defined, Equation 4 and Equation 5.
,
*+, *-, (
(
2" 2 *-, D! * (4)
?# = @ - 0 (1 − - 0 E + (D! − D/ )G!
"−1 "+1 D.

3
 D.
D! = *
" − 1 ( *-, (5)
H1 + 2 I J

Once these equations were defined in anonymous functions. The function went into a loop that continued until the
maximum web distance of the thickest grain had been reached. Each time this loop would calculate the thrust and time
of that thrust and append them to arrays to return once the loop had ended. The first tasks of this loop were to determine
the burning surface area and the area of the throat. From this it could determine the area ratio, !. This important
parameter changes with each iteration due to the erosion of the throat diameter. The burn area changes due to the
change in the web distance and is calculated by a separate function. Its value is added to in multi-grain scenarios by
adding the same equation to the first with different parameters for the new grain.

(
G% = 2+(3% − 2K)(%%$ + K) + 2+L%%# − (%%$ + K)( M (6)

'" (
G" = + - 0 (7)
2

G!
!= (8)
G"

From this information, the Mach number can be numerically calculated. In MATLAB, this is done with the
function fsolve. Assuming choked flow, Mach number equals 1 at the throat, the equation is as shown below in
Equation 9. Mach number is purely based on the area ratio, !, and the ratio of specific heats, ".
*+,
G! 1 2 + (" − 1)I( ((*-,) (9)
= N O
G" I (" + 1)

With this information, the chamber pressure and exit pressure can be calculated. The exit pressure will be
calculated with the already defined anonymous function. This will be used to calculate the thrust coefficient, and
subsequently the thrust produced. Equations 10 and 11 are arguably the center point of this whole simulation.
,
P#& G% ? ∗ ,-3 (10)
D. = - 0
G"

Q = ?# D. G" (11)

This thrust and the current time steps are appended to their respective arrays before the incrementing values are
incremented. Below are the equations needed for the increment step of the loops. The first one of these, Equation 12,
is known as St. Robert’s law. It determines the rate at which the propellant is burning in in/s. This multiplied by the
time step in seconds gives the increment of the web distance as seen in Equation 13. The time is just incremented by
its time step as seen in Equation 14, and Equation 15 shows the calculation for diameter of the throat. The rate of
erosion, re, is constant, so, unlike St. Robert’s law, does not need to be recalculated at each iteration.

%% = PD.3 (12)

4
 K = K + %% ∆) (13)

) = ) + ∆) (14)

'" = '" + %! ∆) (15)

The Monte Carlo analysis takes this code and wraps it in a large for loop that runs for a thousand trials per motor.
Each run changes the random variability input distributions and outputs a thrust curve, time vector and uses this to
determine total impulse and average thrust. The results are then graphed as a thrust curve and two histograms: one for
mean thrust and one for total impulse. The code for this can be seen in Figure 25 and 26 in the Appendix.

IV. Results
This section details the results from the simulations. The first simulation run aimed to model a single grain
Aerotech G138 motor. The results from this could be compared to the actual values of the motor to determine how
well the simulation performed. Below are several figures that show different aspects of the G138’s burn, and a table,
Table 2, including the experimental data from four motor burns.

Table 2: Experimental Data

Thrust Curve [3] Motor 1 Motor 2 Motor 3 Motor 4
Burn Time [s] 1.1 1.356 1.253 1.268 1.28
Max Pressure [psi] 558.93 606.83 753.53 688.1 633.19
Max Thrust [lbf] 42.74 32.66 36.02 37.37 36.9
Mean Thrust [lbf] 31.02 23.69 22.66 25.67 25.75

Figure 1: Simulation Thrust Curve of Single Grain G138 Motor

5
 Figure 1 is a graph of the thrust curve. The thrust curve varies in comparison to the actual G138 motor in two
major ways. First, the thrust is calculated as higher than it actually is, and, second, the burn time is decreased. These
are related, as usually the burn time decreases as thrust increases for grains with the same propellant mass. This is
because more mass is being thrown out of the back of the rocket. This translates to higher thrust, and the propellant
runs out faster. Another reason for the unrealistic burn time is the lack of a transient time. The quasi-steady lumped
parameter method starts with an instantaneous thrust at the first iteration. This is unrealistic as the motor would take
some time to start firing fully, and the thrust would not start instantaneously as it does in the simulation.
Number of Counts

Figure 2: Mean Thrust Distribution of Single Grain G138 Motor

Figure 2 shows the mean thrust distribution for the G138 motor simulation. This is obviously quite off from the
measured value as it is producing almost double the thrust in the simulation as the measured results show. Part of this
probably results from the loss of burn time in the transient response phase as well as other error introduced by using
the quasi-steady lumped parameter method. “If transient effects become important then higher-fidelity models such
as the lumped parameter or ballistic element methods” could be more accurate [1].

6
 Number of Counts

Figure 3: Total Impulse Distribution of Single Grain G138 Motor

While the mean thrust distribution may be off, the total impulse distribution is far closer to the actual value given
on thrust curve, which shows a total impulse of 35.3 lbf-s [3]. While that is slightly on the lower end of the distribution
is still falls within the 95% interval marked by the vertical red lines. Therefore, this method, while not completely
accurate for all parts of modeling a real motor, was deemed acceptable for use in the more theoretical multi-grain
performance comparison section of this project.
After the initial grain was confirmed, the next test was to compare the effects of increasing the grain count.
However, since the original length of the G138 bore was so large, it was decided to decrease the bore length by cutting
it in half for the multiple grain analysis as it would be more realistic. The original mean bore length was 3.9110 in.
with a standard deviation of 0.0059 in. Therefore, the new bore length was 1.9555 in. with a standard deviation of
0.0029 in.

Figure 4: Thrust Curve of Single Grain Motor

As can be seen in Figure 4, the thrust curve changes drastically. Since the dimensions of the grain are the only
thing that was changed, the more parabolic thrust curve is certainly affected by the change in burn surface area. The

7
original bore length contributed to a smoother, more even thrust curve than this. However, it is important to note that
the thrust tapering off towards the end of the burn is amplified due to the nozzle’s throat erosion. As the throat erodes,
the performance decreases.

Figure 5: Thrust Curve of Three Grain Motor

Figure 6: Thrust Curve of Five Grain Motor

It was expected that the higher the number of grains the shorter the burn time would be and the higher the thrust.
This is clearly seen in Figures 4-6. The thrust is highest in the 5-grain motor configuration, and the burn time is
significantly shorter. With more grains comes a higher surface area and, therefore, a higher chamber pressure. This
burns up the propellant very quickly. However, two things must be noted. The thrust curves presented here only
represent the final iteration of their respective Monte Carlo simulations. While this does not present the average thrust
curve, it demonstrates the higher grain count higher thrust phenomenon quite well. Another thing that must be
considered is the throat’s rate of erosion. This property was calculated based on the original data from the G138 motor
tests and may not be so accurate in these multi-grain scenarios. Nonetheless, its effect does not detract from the purpose
of this project and only from the realism of the simulation.

8
 Number of Counts

Figure 7: Mean Thrust Distribution of Single Grain Motor

Number of Counts

Figure 8: Mean Thrust Distribution of Three Grain Motor

9
 Number of Counts

Figure 9: Mean Thrust Distribution of Five Grain Motor

Figures 7-9 confirm what was discovered in the thrust curve plots. The thrust certainly increases as the grain count
is increased for the motor. On top of this is the fact that the uncertainty of the mean thrust does decrease overall as
grain count increases. This corroborates the original hypothesis and can be seen graphically in Figure 10 and
numerically in Table 3. The uncertainty sometimes does increase by a small percentage as motor grain increases, but
the overall trend is to decrease as grain count increases.

Table 3: Uncertainty Comparison

Number of Grains Mean Thrust (%) Total Impulse (%)

1 8.54 10.33

2 7.40 9.70

3 7.44 9.81

4 7.26 9.61

5 7.29 9.96

10
 Figure 10: Mean Thrust Distribution of Five Grain Motor

Table 3 seems to confirm the original hypothesis. The expectation was that by adding grains, the interval on which
the majority of data points lay would decrease. This seems to be the overall trend, even though the percentage
sometimes increases between data points. To really test this hypothesis, one would have to isolate the variability to
only the grain geometry or even the propellant density between grains. The second of these would be difficult to
model, but an interesting next step in this research.
Number of Counts

Figure 11: Total Impulse Distribution of Single Grain Motor

11
 Number of Counts

Figure 12: Total Impulse Distribution of Five Grain Motor

The total impulse distributions seen in Figure 11 and Figure 12 show very little difference in uncertainty. The
number of grains does not really affect this value, but instead the location of the distribution. This makes sense since
total impulse is an integral of the force created by the motor over the time it is burned. In the higher thrust, higher
grain count cases, the area under the curve would increase despite the shorter burn time. Therefore, the total impulse
increases as grain count increases. It is important to note that, like the thrust curves, these integrals are only based on
the last data point of the Monte Carlo simulation. Nonetheless, they demonstrate the effect that grain count has on
total impulse location and distribution.

V. Conclusion
In conclusion, this project demonstrates some of the important aspects of solid rocket motors via a basic Monte
Carlo simulation. However, when compared to actual results from the motor tests, it is obvious that the code requires
much higher fidelity to fully capture and represent the process and results of a solid rocket motor test. Nonetheless, it
demonstrated the effect of grain number on variability in performance. The grain count increased the thrust and
impulse, while it decreased the variation of average thrust and total impulse. A few options for further research
stemming from this project would be interesting to explore. One, a study determining which parameters with which
variability affects the performance the most, and another, a study on optimizing nozzle dimensions for multi-grain
motors. Nonetheless, this project proved useful in determining the overall trend of motor performance as grain count
increases. Indeed, the performance increases as grain count increases and uncertainty decreases.

Appendix
This appendix includes extra figures not displayed in the main body. It also contains the code written for this
project.

12
Figure 13: Thrust Curve of Two Grain Motor

Figure 14: Thrust Curve of Four Grain Motor

13
 Number of Counts

Figure 15: Mean Thrust Distribution of Two Grain Motor

Number of Counts

Figure 16: Mean Thrust Distribution of Four Grain Motor

14
Number of Counts

Figure 17: Total Impulse Distribution of Two Grain Motor

Number of Counts

Figure 18: Total Impulse Distribution of Three Grain Motor

15
Number of Counts

Figure 19: Total Impulse Distribution of Four Grain Motor

16
Figure 20: thrustCalc Function

17
 Figure 21: burnArea Function

Figure 22: variability_info.m Recorded Information for Tested Motors

18
Figure 23: variability_info.m Calculated Mean and Standard Deviation

Figure 24: Ballistic.m Program Constants and Variability Set Up

19
 Figure 25: Ballistic.m Program Monte Carlo Analysis

Figure 26: Ballistic.m Program Plotting

References
Books
[1] Heister, S. D., Anderson, W. E., Pourpoint, T., and Cassady, R. J., Rocket Propulsion, Cambridge University Press, United
Kingdom, 2019.

Electronic Publications
[2] “2.2.7 - The Empirical Rule,” Penn State Eberly College of Science. URL: https://online.stat.psu.edu/stat200/lesson/2/2.2/2.2.7
[retrieved 30 April 2022]

20
[3] “Aero Tech HP-G138T,” thrustcurve.org. URL: https://www.thrustcurve.org/motors/AeroTech/HP-G138T/ [retrieved 1 May
2022]

21