NAVAL

POSTGRADUATE
SCHOOL
MONTEREY, CALIFORNIA

THESIS
GENERALIZED WEAPON EFFECTIVENESS MODELING

by

Colin Michael Anderson

June 2004

Thesis Advisor: Morris R. Driels

Approved for public release; distribution is unlimited

THIS PAGE INTENTIONALLY LEFT BLANK
 REPORT DOCUMENTATION PAGE Form Approved OMB No. 0704-
0188
Public reporting burden for this collection of information is estimated to average 1 hour per response,
including the time for reviewing instruction, searching existing data sources, gathering and maintaining
the data needed, and completing and reviewing the collection of information. Send comments regarding
this burden estimate or any other aspect of this collection of information, including suggestions for
reducing this burden, to Washington headquarters Services, Directorate for Information Operations and
Reports, 1215 Jefferson Davis Highway, Suite 1204, Arlington, VA 22202-4302, and to the Office of
Management and Budget, Paperwork Reduction Project (0704-0188) Washington DC 20503.
1. AGENCY USE ONLY (Leave 2. REPORT DATE 3. REPORT TYPE AND DATES COVERED
blank) June 2004 Master’s Thesis

4. TITLE AND SUBTITLE: 5. FUNDING NUMBERS

Generalize Weapon Effectiveness Modeling
6. AUTHOR(S) Colin Michael Anderson
7. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES) 8. PERFORMING
Naval Postgraduate School ORGANIZATION REPORT
Monterey, CA 93943-5000 NUMBER
9. SPONSORING /MONITORING AGENCY NAME(S) AND ADDRESS(ES) 10. SPONSORING/MONITORING
N/A AGENCY REPORT NUMBER

11. SUPPLEMENTARY NOTES The views expressed in this thesis are those of the author and do not reflect the
official policy or position of the Department of Defense or the U.S. Government.
12a. DISTRIBUTION / AVAILABILITY STATEMENT 12b. DISTRIBUTION CODE
Approved for public release; distribution is unlimited

13. ABSTRACT (maximum 200 words)

In this thesis, we compare weapon effectiveness methods to determine if current effectiveness models provide
accurate results. The United States Military currently adheres to a compilation of data and methodologies named
the Joint Munitions Effectiveness Manuals (JMEM) to determine the effectiveness of air delivered weapons against
a variety of ground targets. Since the time these manuals were implemented in the 1960s, progress in technology
has allowed the weapon/target interaction to be more accurately modeled. This thesis investigates the differences
of these high fidelity models for unguided weapons and the JMEM computations in order to determine whether the
older, more simplistic, models need to be upgraded.

14. SUBJECT TERMS 15. NUMBER OF

Weaponeering, Weapon Effectiveness, Monte Carlo Simulations PAGES
129
16. PRICE CODE
17. SECURITY 18. SECURITY 19. SECURITY 20. LIMITATION
CLASSIFICATION OF CLASSIFICATION OF THIS CLASSIFICATION OF OF ABSTRACT
REPORT PAGE ABSTRACT
Unclassified Unclassified Unclassified UL
NSN 7540-01-280-5500 Standard Form 298 (Rev. 2-89)
Prescribed by ANSI Std. 239-18

i
THIS PAGE INTENTIONALLY LEFT BLANK

ii
 Approved for public release; distribution is unlimited

GENERALIZED WEAPON EFFECTIVENESS MODELING

Colin M. Anderson
Ensign, United States Navy
B.S., University of Washington, 2003

Submitted in partial fulfillment of the

requirements for the degree of

MASTER OF SCIENCE IN MECHANICAL ENGINEERING

from the

NAVAL POSTGRADUATE SCHOOL

June 2004

Author: Colin M. Anderson

Approved by: Morris R. Driels

Anthony J. Healey
Chairman, Department of Mechanical and Astronautical
Engineering

iii
THIS PAGE INTENTIONALLY LEFT BLANK

iv
 ABSTRACT

In this thesis, we compare weapon effectiveness methods to determine if

current effectiveness models provide accurate results. The United States Military
currently adheres to a compilation of data and methodologies named the Joint
Munitions Effectiveness Manuals (JMEM) to determine the effectiveness of air
delivered weapons against a variety of ground targets. Since the time these
manuals were implemented in the 1960s, progress in technology has allowed the
weapon/target interaction to be more accurately modeled. This thesis
investigates the differences of these high fidelity models for unguided weapons
and the JMEM computations in order to determine whether the older, more
simplistic, models need to be upgraded.

v
THIS PAGE INTENTIONALLY LEFT BLANK

vi
 TABLE OF CONTENTS

I. INTRODUCTION............................................................................................. 1
II. BACKGROUND.............................................................................................. 3
A. DELIVERY ACCURACY ...................................................................... 3
1. Aiming Errors........................................................................... 3
2. Ballistic Dispersion ................................................................. 7
B. LETHAL AREAS.................................................................................. 9
C. EFFECTIVENESS .............................................................................. 16
III. SINGLE WEAPON VERSUS UNITARY TARGET ....................................... 21
A. JMEM METHOD................................................................................. 21
B. MONTE CARLO SIMULATIONS ....................................................... 22
1. Modeling Weapons with Carleton Damage Function ......... 26
2. Modeling Weapons with Rectangular Cookie Cutter
Approximation ....................................................................... 28
C. RESULTS OF UNITARY TARGET COMPARISONS ........................ 29
1. JMEM versus Carleton Damage Function ........................... 30
2. JMEM versus Rectangular Cookie Cutter Approximation.. 32
IV. SINGLE WEAPON VERSUS AREA TARGET ............................................. 37
A. JMEM METHOD................................................................................. 38
B. MONTE CARLO SIMULATIONS ....................................................... 44
1. Modeling Weapons with Carleton Damage Function ......... 44
2. Modeling Weapons with Rectangular Cookie Cutter
Approximation ....................................................................... 47
C. RESULTS OF AREA TARGET COMPARISONS .............................. 49
1. JMEM versus Carleton Damage Function ........................... 49
2. JMEM versus Rectangular Cookie Cutter Approximation.. 51
V. STICKS OF WEAPONS VERSUS AREA TARGETS................................... 55
A. OLD JMEM METHOD ........................................................................ 57
B. CURRENT JMEM METHOD .............................................................. 65
C. MONTE CARLO SIMULATIONS ....................................................... 67
1. Modeling Weapons with Carleton Damage Function ......... 73
2. Modeling Weapons with Rectangular Cookie Cutter
Approximation ....................................................................... 75
D. RESULTS OF STICK COMPARISONS ............................................. 78
1. Updated JMEM Method versus Old JMEM Method............. 79
2. JMEM versus Carleton Damage Function ........................... 81
3. JMEM versus Rectangular Cookie Cutter Approximation.. 83
VI. SUMMARY OF RESULTS ............................................................................ 87
VII. CONCLUSIONS............................................................................................ 91
VIII. RECOMMENDATIONS AND FUTURE WORK ............................................ 93
vii
APPENDIX A ........................................................................................................... 97
APPENDIX B ......................................................................................................... 103
LIST OF REFERENCES........................................................................................ 111
INITIAL DISTRIBUTION LIST ............................................................................... 113

viii
 LIST OF FIGURES

Figure 2.1: Range and deflection directions with impact points. ............................. 3
Figure 2.2: Gaussian distribution probability density function (From Ref. [1])......... 4
Figure 2.3: Bi-variate Gaussian distribution (From Ref. [1]).................................... 5
Figure 2.4: Definition of REP and DEP in ground plane. ........................................ 6
Figure 2.5: Definition of CEP in ground plane. ....................................................... 7
Figure 2.6: Generalized Weapon Trajectory and Impact Diagram.......................... 8
Figure 2.7: Weapon/target interaction geometry................................................... 11
Figure 2.8: Gaussian approximations of PK contour lines. (From Ref. [1])............ 12
Figure 2.9: Plot of Carleton damage function. ...................................................... 13
Figure 2.10: Elliptical approximation of Carleton damage function (From Ref. [1]). 14
Figure 2.11: Definition of rectangular cookie cutter dimensions (After Ref.
[1]) ...................................................................................................... 16
Figure 2.12: SSPD for a single target against a unitary target. (From Ref.
[1]) ...................................................................................................... 17
Figure 3.1: Two sample iterations of a Monte Carlo simulation. ........................... 23
Figure 3.2: Example of lethal area matrix populated by Carleton damage
function............................................................................................... 25
Figure 3.3: Example iteration of Monte Carlo simulation using Carleton damage
function to fill lethal area matrix. ......................................................... 27
Figure 3.4: Example iteration of Monte Carlo simulation using a rectangular
cookie cutter lethal area matrix........................................................... 29
Figure 4.1: Area target partially covered by weapon lethal area. (After Ref. [1]).... 37
Figure 4.2: Rectangular lethal area against an area target. (From Ref. [1])........... 39
Figure 4.3: Fractional coverage in the range direction. (From Ref. [1]).................. 40
Figure 4.4: Example iteration of Monte Carlo simulation using Carleton damage
function to fill lethal area matrix against an area target. ..................... 46
Figure 4.5: Example iteration of Monte Carlo simulation using a rectangular
cookie cutter lethal area matrix against an area target. ...................... 48
Figure 5.1: Definition of stick and pattern dimensions............................................. 56
Figure 5.2: Modification of weapon AET due to ballistic dispersion. (From Ref.
[1]) ...................................................................................................... 59
Figure 5.3: Weapons overlapping (left) and not overlapping (right) in the
deflection direction. (From Ref. [1]) .................................................... 60
Figure 5.4: Combination of weapons in deflection direction. (From Ref. [1]).......... 61
Figure 5.5: Weapons overlapping (left) and not overlapping (right) in the range
direction. (From Ref. [1])..................................................................... 62
Figure 5.6: Stick pattern dimension and damage function. (From Ref. [1]) ............. 63
Figure 5.7: Example of weapon lethal areas overlapping in stick. (From Ref. [2]) . 66
Figure 5.8: Effects of aiming error on a stick of weapons. .................................... 68
Figure 5.9: Effects of ballistic dispersion error on a stick of weapons................... 69
Figure 5.10: Monte Carlo stick simulations scenario. ............................................. 70

ix
Figure 5.11: Damage contributions of stick weapons. ............................................ 71
Figure 5.12: Example iteration of Monte Carlo stick simulation using rectangular
cookie cutter lethal area matrices against an area target. .................. 77

x
 LIST OF TABLES

Table 2.1: Sample damage matrix. ..................................................................... 11

Table 3.1: Results of JMEM and Monte Carlo for MAEF=5000ft2 ........................ 30
Table 3.2: Results of JMEM and Monte Carlo for MAEF=1000ft2 ........................ 31
Table 3.3: Results of JMEM and Monte Carlo for MAEF=5000ft2 ........................ 32
Table 3.4: Results of JMEM and Monte Carlo for MAEF=1000ft2 ........................ 33
Table 3.5: Results of JMEM (small area) and Monte Carlo for MAEF=5000ft2 .... 34
Table 3.6: Results of JMEM (small area) and Monte Carlo for MAEF=1000ft2 .... 34
Table 4.1: Results of JMEM and Monte Carlo for MAEF=5000ft2 ........................ 50
Table 4.2: Results of JMEM and Monte Carlo for MAEF=1000ft2 ........................ 50
Table 4.3: Results of JMEM and Monte Carlo for MAEF=5000ft2 ........................ 52
Table 4.4: Results of JMEM and Monte Carlo for MAEF=1000ft2 ........................ 52
Table 5.1: Results of new JMEM and old JMEM for weapons of MAEF=5000ft2 . 79
Table 5.2: Results of new JMEM and old JMEM for weapons of MAEF=1000ft2 . 80
Table 5.3: Results of JMEM and Monte Carlo for MAEF=5000ft2 ........................ 81
Table 5.4: Results of JMEM and Monte Carlo for MAEF=1000ft2 ........................ 82
Table 5.5: Results of JMEM and Monte Carlo for MAEF=5000ft2 ........................ 84
Table 5.6: Results of JMEM and Monte Carlo for MAEF=1000ft2 ........................ 84

xi
THIS PAGE INTENTIONALLY LEFT BLANK

xii
 ACKNOWLEDGMENTS

I would like to thank Professor Driels for his patience and guidance
throughout my time at the Naval Postgraduate School. I have learned a great
deal, and am glad I had the chance to work with you.

I would also like to give thanks for the support and love from my parents,
Bill and Janalyn, my RMS family, and the E-Funk Ensigns. I couldn’t have asked
for better friends and family.

xiii
THIS PAGE INTENTIONALLY LEFT BLANK

xiv
 I. INTRODUCTION

The objective of this thesis is to compare different methodologies for

determining unguided weapon effectiveness against targets in order to determine
if current methods used by the United States Military provide reasonable
estimates of damage. The current methods used by the military in determining
weapon damage effectiveness have been used for quite some time and may be
outdated. Over the last 40 years, weapon/target interaction data from weapon
tests have been compiled by the military into a volume of data called the Joint
Munitions Effectiveness Manuals, or JMEM. The JMEM has used this compiled
data to formulate estimates of weapon effectiveness for given weapons and
targets. In order to simplify weapon effectiveness calculations to meet computing
speeds of the time, the JMEM needed to implement approximations of weapon
effectiveness techniques. For this thesis we will compare these JMEM methods
with weapon effectiveness modeling methods that do not implement
approximations and that more closely represent the weapon/target interaction.
We will use Monte Carlo simulations as the basis of our comparison
testing. The first Monte Carlo simulation will use a damage function called the
Carleton damage function, which is based on weapon test data, to represent the
damage inflicted by a weapon of given lethality. In the Monte Carlo simulations,
we represent weapons with a lethal area matrix center on the weapon impact
point. This lethal area matrix will be populated with damage values obtained
from either the Carleton damage function, or an approximation of the Carleton
damage function, depending on the method tested. We assume this Monte Carlo
weapon effectiveness model representing the weapon with a lethal area matrix
populated with the Carleton damage function to be the highest fidelity method in
determining weapon effectiveness. This is because some JMEM methods
generally rely on approximations of the Carleton damage function to calculate
weapon probabilities of damage.

1
 In order to obtain a broad array of data, we will compare the JMEM and
Monte Carlo simulation results for single as well as sticks of weapons against
unitary and area targets. Along with different weapon/target scenarios, we also
test the different methods for various delivery accuracies to try to pinpoint the
reason for discrepancies in the methods’ effectiveness results. By analyzing the
differences between weapon effectiveness results, we hope to determine
whether the current JMEM techniques should be replaced by more complex
modeling methods.

2
 II. BACKGROUND

A. DELIVERY ACCURACY

The probability of damaging or killing a target obviously depends on how

close a weapon is delivered to the target. This delivery accuracy is affected by
many components, especially errors in aiming the weapon and the ballistics of
the weapon used.

1. Aiming Errors

The delivery accuracy of a weapon depends heavily on the ability to aim

that weapon at a desired impact point. Statistical models are used to determine
to what degree these aiming errors affect the probability of impacting a desired
point. First, two directions are defined in the ground plane. The range direction
is the direction of the velocity vector of the aircraft releasing the weapon. The
deflection direction is the direction perpendicular to the range. We next look at a
sample number of weapons released from the same aircraft at the same release
conditions, all aimed at the same point on the ground plane. This will result in a
map of impact points on the ground plane shown in Figure 2.1.

Range

Aim
Point
Impact
s

Deflection

Figure 2.1: Range and deflection directions with impact points.

3
The distribution, or dispersion, of these impact points can be defined statistically.
In order to do this, we assume that the distribution of the impact points in both
range and deflection is Gaussian and independent of each other.
A Gaussian, or normal distribution, is a distribution of data that has
characteristics that match a probability density function (PDF) given in equation
(2.1).

1 −( x − µ ) 2
f ( x) = exp[ ] (2.1)
σ 2π 2σ 2

The mean (µ) and the variance (σ2) are calculated from a given data set, or in our
case, the weapon impact data. The mean for a normal distribution is the value of
x that gives the maximum value of f(x). This is illustrated in Figure 2.2.

f(x)

Figure 2.2: Gaussian distribution probability density function (From Ref. [1])

The variance gives an indication of the spread of the data set. A small variance,
for example, would give a PDF that has a sharp peak, whereas a large variance
would produce a flatter PDF. The square root of the variance (σ), or standard
deviation, of the sample set is also an important quantity. In a normal distribution,
approximately 68% of the data will lie in a range +σ. Similarly, 95.5% of data will
lie in the range + 2σ, and 99.7% will lie within + 3σ.

4
 The normal distribution assumption can be directly applied to the two
dimensional impact point data. For the two dimensional case, we look at the two
dimensional, bi-variate probability density function shown in equation (2.2).

( x − µ x )2 ( y − µ y )
2
1
f ( x, y ) = exp− [ + ] (2.2)
2πσ xσ y 2σ x
2
2σ y
2

The subscripts x and y denote the direction of the specified mean and standard
deviation. For the weaponeering problem, the range and deflection directions will
be treated as x and y respectively. A graphical idea of this concept is presented
in Figure 2.3.

f(x,y)
1

DMPI

Deflection
Range

Figure 2.3: Bi-variate Gaussian distribution (From Ref. [1])

Once the impact point distribution has been assumed to be normal, we will
describe the distribution in terms of a range error probable (REP) and deflection

5
error probable (DEP). The REP and DEP are both related to the standard
deviation of the distribution in the range and deflection directions. Where the
standard deviation was defined as the range such that 68% of points will lie
within +σ, the REP is defined as the distance in the range direction such that
50% of the total impact points will lie in the range +REP. The DEP is defined
similarly, but in the deflection direction. These terms are better shown in Figure
2.4.

Deflection

Aim
Point

DEP

DEP
Range

REP REP

Figure 2.4: Definition of REP and DEP in ground plane.

The difference between REP and DEP and the standard deviations is the
percentage of impact points that are contained within those ranges. Using
tabulated values of the PDF for a normal distribution we find equations (2.3) and
(2.4).

REP = 0.6745σ x (2.3)

DEP = 0.6745σ y (2.4)

6
 Another common way of describing the dispersion of impact points
involves using the circular error probable (CEP). The CEP is a radius about the
desired aim point that contains 50% of the impact points, as shown in Figure 2.5.

Deflection

Aim
Point

Range
CEP

Figure 2.5: Definition of CEP in ground plane.

2. Ballistic Dispersion

The delivery accuracy of a weapon is also dependent on the

manufacturing of the munitions used to attack a target. Many details may vary
from one weapon to another. An example of these errors may be seen from a
rifle bolted to the ground firing at a target. Since the rifle is bolted, we can
assume that there are no aiming errors. The rounds fired from the rifle would
create a pattern of impact points about some point on the target. These errors
can be due to anything from variations in bullet shape to different weights of
explosives used inside each bullet. [Ref. 1]
Usually ballistic dispersion errors are defined as a standard deviation in
the normal plane of the weapon, measured in mils (milliradians). The standard
deviation in the normal plane is denoted as σb. Figure 2.6 illustrates a typical

7
weapon release against a target. The normal plane can be seen as the plane
perpendicular to that of the velocity vector of the weapon at its impact point.

Weapon Release Point

Weapon Trajectory

Weapon Velocity
Altitude Slant Range (SR) Vector at Impact

Normal
Impact Plane
Angle ( I )
Ground Plane

Figure 2.6: Generalized Weapon Trajectory and Impact Diagram

When aiming errors and ballistic dispersion errors are both present in a problem,
it is often convenient to write the ballistic dispersion in terms of feet in the ground
plane. Calculating xbr (ballistic dispersion error in the range direction in feet) and
xbd (ballistic dispersion error in deflection direction in feet) are found with in
equations (2.5) and (2.6).

0.6745 × SR × σ b
xbr =
1000sin( I ) (2.5)

0.6745 × SR × σ b
xbd = (2.6)
1000

The slant range (SR) is the straight-line distance from the aircraft at the point of
release to the impact point of the weapon as shown in Figure 2.6. The impact

8
angle I is measured in radians, and is the angle between the weapon velocity
vector at impact and the ground plane.
For individually released weapons, we are able to obtain an equivalent
REP and DEP that combines the ballistic dispersion errors in feet with the aiming
errors in feet. To find the equivalent REP and DEP, the ballistic dispersion error
and aiming errors are root sum squared, and defined as REP’ and DEP’,

REP ' = REP 2 + xbr 2 (2.7)

DEP ' = DEP 2 + xbd 2 (2.8)

Ballistic dispersion is treated differently when we deal with multiple

weapon releases, or sticks of weapons and will be discussed later in this thesis.

B. LETHAL AREAS

There are two major mechanisms in conventional weapons that inflict

damage upon a target. These mechanisms are the blast wave produced by the
high explosive inside of a weapon, and the fragments propelled from the weapon
after it has exploded.
When the high explosive inside a conventional warhead detonates, the
explosive material will almost instantly be converted into a gas at very high
temperature and pressure. These values are typically near 200 atmospheres of
pressure and 5000 degrees Celsius [Ref. 1]. This high pressure will fragment the
case around the weapon and compress the air around the weapon. This
compression forms a blast wave that travels into the surrounding area. Blast
damage is highly effective against most targets that lie near the impact/explosion
point of the weapon. Since the blast wave is a wave of pressurized and
superheated air, it dissipates fairly quickly in the atmosphere, which means its

9
effectiveness greatly decreases with distance from the point of impact. The blast
energy is roughly inversely proportional to the cube of the distance from the
weapon impact.
Once a blast wave has dissipated in the atmosphere, fragmentation
becomes the main source of damage. Primary fragments are the remains of the
weapon after the high explosive charge has shattered the weapon casing. These
fragments are normally fairly small and are propelled up to very high velocities
from the initial explosion of the weapon. These velocities propel the fragments
through, and eventually past the front of the pressure shock wave created by the
high explosive blast. Since the fragments are only slowed by air friction, they
travel further than the blast wave. This is the advantage of fragmenting weapons.
Since fragments can damage targets further from the weapon impact than blast
can, we can accept greater weapon miss distances, and still damage targets.
Depending on the weapon, a hazardous fragment is one having impact energy of
58 ft-lb (79 joules) or greater [Ref. 1].
The damage resulting from fragmentation and blast is dependent on the
size and type of weapon, as well as the kill criteria for the target it is used against.
Undergoing a fairly lengthy analysis described in detail in Reference 1,
weaponeers are able to create a damage matrix that gives the probability of
killing (PK) a target a given distance away from the weapon impact. An example
of a damage matrix is given in Table 2.1. Each table cell shows the probability of
kill given a distance in both range and deflection from the weapon impact point.
Figure 2.7 shows the angles and distances associated with a weapon detonation
near a target. The distance the weapon detonates from the target, which is
denoted x and y in Figure 2.7, are the range and deflection distances given in
Table 2.1. Note Table 2.1 shows only positive deflection distances. This is done
because fragments are symmetrical about the range direction.

10
 Weapon
detonation
point Target location
W
(center of
Deflection vulnerability)

Deflection y SR
angle α
O T
Elevation
Ground angle θ
zero
Range
x
R

Figure 2.7: Weapon/target interaction geometry.

Deflection (ft)
Range▼ 37.9 75.8 113.7 151.6 189.5 227.4 265.3 303.2 341.1 379.0
-114.4 0 0 0 0 0 0 0 0 0 0
-100.1 0 0 0 0 0 0 0 0 0 0
-85.8 0 0 0 0 0 0 0 0 .0001 .0001
-71.5 .0001 0 0 0 0 0 .0001 .0002 .0001 .0001
-57.2 .0011 0 0 0 0 .0003 .0004 .0002 .0001 .0001
-42.9 .0028 0 0 0 .0009 .0008 .0004 .0002 .0001 .0001
-28.6 .0064 .0001 .0006 .0029 .0017 .0009 .0005 .0002 .0001 .0001
-14.3 .1402 .0059 .0099 .0042 .0019 .0009 .0005 .0002 .0001 .0001
0 .5571 .0459 .0127 .0045 .0019 .0009 .0005 .0002 .0001 .0001
14.3 .6794 .0891 .0156 .0045 .0019 .0009 .0005 .0002 .0001 .0001
28.6 .1741 .0927 .0325 .0116 .0041 .0012 .0005 .0002 .0001 .0001
42.9 .0060 .0186 .0258 .0128 .0063 .0034 .0016 .0006 .0002 .0001
57.2 .0007 .0050 .0105 .0118 .0061 .0032 .0017 .0010 .0006 .0003
71.5 0 .0024 .0015 .0072 .0056 .0031 .0017 .0010 .0006 .0004
85.8 0 .0010 .0012 .0011 .0045 .0028 .0017 .0009 .0005 .0003
100.1 0 .0003 .0009 .0005 .0012 .0025 .0015 .0009 .0005 .0003
114.4 0 0 .0006 .0004 .0002 .0011 .0014 .0009 .0005 .0003
128.7 0 0 .0003 .0003 .0002 .0001 .0009 .0007 .0004 .0003
143.0 0 0 .0001 .0003 .0001 .0001 .0001 .0006 .0004 .0003
157.3 0 0 0 .0002 .0001 .0001 0 .0002 .0004 .0002
171.6 0 0 0 .0001 .0001 .0001 0 0 .0002 .0002
185.9 0 0 0 .0001 .0001 .0001 0 0 0 .0001
200.2 0 0 0 0 .0001 0 0 0 0 0
214.5 0 0 0 0 .0001 0 0 0 0 0
228.8 0 0 0 0 0 0 0 0 0 0
243.1 0 0 0 0 0 0 0 0 0 0

Table 2.1: Sample damage matrix. (From Ref. [1])

11
This high fidelity modeling of the weapon target interaction is too difficult to use in
more simplistic weaponeering modeling, so approximations are made. The first
of these approximations is to find constant values of PK in the damage matrix,
and draw lines of contour through them. This results in a plot similar to Figure
2.8. The contours in the range and deflection directions have near Gaussian
distributions, which is also shown in Figure 2.8.

Weapon
Impact Direction of
Point flight

Figure 2.8: Gaussian approximations of PK contour lines. (From Ref. [1])

12
The Gaussian approximation allows us to develop an expression that gives the
PK for a point (x,y), given a certain weapon. This is called the Carleton damage
function which is given in equation (2.9).

 x2 y2 
PK ( x, y ) = exp −  2
+ 2
(2.9)
 WRr WRd 

The Carleton damage function is plotted in Figure 2.9.

0.8

0.6
PK
Pk

0.4

0.2

Deflection Direction Range Direction

Figure 2.9: Plot of Carleton damage function.

The WRr and WRd, or weapon radii in the range and deflection direction are
characteristics of the weapon used and help give information on the overall
lethality of the weapon. This lethality is best described in a term called the lethal
area, or AL. AL is the area under the PK curve for all area in the ground plane.

13
For fragmentation weapons, we call AL the mean area of effectiveness due to
fragments, or MAEF. For fragmentation weapons modeled with the Carleton
damage function, the MAEF is equal to the area under the Carleton damage
function and given in equation (2.10).

MAEF = π × WRr × WRd (2.10)

Even with the numerous assumptions made up to this point, the Carleton
damage function is still too difficult to use in simple modeling. We can
approximate the lethal area of the Carleton damage function as the area of an
ellipse with dimensions WRr and WRd as shown in Figure 2.10.

Range AL

WRr
WRd Deflection

Figure 2.10: Elliptical approximation of Carleton damage function (From

Ref. [1])

The PK inside of this ellipse is equal to unity, and zero outside of the ellipse. This
is done to conserve the lethality during the approximation. Since the PK is unity
inside the ellipse, and zero outside, we refer to this as an elliptical “cookie cutter”.
The cookie cutter nickname pertains to the fact that a target that lies outside of
the ellipse’s boundary will not be damaged, so the ellipse effectively cuts out an
area of certain kill. It is also useful to define the ratio of the weapon radii WRr and
WRd . This ratio (a) is dependent upon the impact angle of the weapon, and
through the use of empirical data has been defined in equation (2.11).

14
 WRr
a= = MAX [1 − 0.8cos( I ), 0.3] (2.11)
WRd

Knowing this ratio of the weapon radii allows us to rewrite the Carleton damage
function in a different form. We first define an effective target length (L’ET) and
width (W’ET) rather than the weapon radii. Equations (2.12) and (2.13) define
these effective lengths

a
L 'ET = 2 × WRr = 1.128 MAEF × (2.12)
π
L 'ET
W 'ET = 2 × WRd = (2.13)
a

We can then express the Carleton damage function in terms of L’ET and W’ET, as
shown in equation (2.14).

 4x2 4 y2 
PK ( x, y ) = exp −  2
+ 2
(2.14)
 L 'ET W 'ET 

The final step of the procedure is approximating the ellipse shown in

Figure 2.10 into a rectangular cookie cutter. Since lethality must be conserved
through this approximation, we must calculate rectangular dimensions that
enclose the same amount of lethal area as the original ellipse. This
approximation is shown graphically in Figure 2.11.

15
 Range

WRr
LET
WRd
Deflection

WET

Figure 2.11: Definition of rectangular cookie cutter dimensions

(After Ref. [1])

We define the rectangle’s dimensions as the effective target length and width, or
LET and WET respectively. These are calculated from our knowledge of the
weapon radii ratio as shown in equations (2.15) and (2.16).

LET = MAEF × a (2.15)

WET = LET / a (2.16)

C. EFFECTIVENESS

Suppose we are dealing with a single aircraft releasing a single bomb

against a target on the ground. In a problem like this, we would be interested in
calculating the single sortie probability of damage, or SSPD. The SSPD is the
probability that the single released weapon will cause damage to the target,
equal to the amount represented by the damage function. A sortie is single pass
by a single aircraft releasing weapons upon a target. For our case, we will deal
with a single released weapon.
16
 We will first only consider the range direction in our analysis. The
accuracy distribution relative to the target is assumed to be a Gaussian normal
distribution, and is called g(x). Also, the damage function of the weapon is
modeled with the Carleton damage function, or c(x). To find the SSPD in the
range direction, the distance the weapon lands from the target needs to be
known. In this example we will call this distance u. Figure 2.12 shows a
graphical representation of these assumptions.

Accuracy function g(x)

Damage function c(x)
SSPD1

target x=u Range x

Figure 2.12: SSPD for a single target against a unitary target.

(From Ref. [1])

The point SSPD1 marks the probability of damaging the target with a single sortie.
This is the value of the damage function c(x) at the target. From our knowledge
of delivery accuracy models, we may assume that the value u has a normal
distribution for a large number of independent trial sorties. If we average the
SSPD value over this large number of trials it will tend toward an expected value.
Using knowledge of statistics and incorporating the damage function and
accuracy distribution, we end up with equation (2.17).

∞
SSPDx = ∫ c( x) g ( x)dx
−∞
(2.17)

17
We can then substitute in the one-dimensional forms of the Carleton damage
function and the accuracy function. These expressions are given in equations
(2.18) and (2.19).

1  x2 
g ( x) = exp  − 2
(2.18)
σx 2π  2σ x 

 4x2 
c( x) = exp  − 2
(2.19)
 L 'ET 

Substituting these equations into equation (2.17) and solving yields an

expression for SSPDx, which is given in equation (2.20).

L 'ET
SSPDx = (2.20)
8σ x 2 + L 'ET 2

We have been only dealing in one dimension so far, but the results in the
deflection direction are similar as shown in equation (2.21).

W 'ET
SSPDy = (2.21)
8σ y 2 + W 'ET 2

Lastly, we find the total SSPD by combing the SSPD in range and in deflection.

SSPD = SSPDx × SSPDy (2.22)

Calculating the single sortie probability of damage is useful for determining

the damage inflicted by a given weapon against a single target. However, this

18
method in determining the effectiveness of a weapon is not used for all weapon/
target interactions. These different effectiveness methods will be discussed later
in this thesis.

19
THIS PAGE INTENTIONALLY LEFT BLANK

20
 III. SINGLE WEAPON VERSUS UNITARY TARGET

A. JMEM METHOD

The Joint Munitions Effectiveness Manual (JMEM) uses a methodology

similar to the case discussed in the Effectiveness section of Chapter II. Option 4
of the MATLAB m-file code provided in Appendix A reproduces the JMEM
methodology for the case of single weapon released against unitary targets. The
JMEM method uses the Carleton damage function to calculate the probability of
damage to a target. This method first root sum squares the aiming errors and
ballistic dispersion errors, to find a REP’ and DEP’ in feet, in the ground plane as
shown in equations (3.1) and (3.2).

REP ' = REP 2 + xbr 2 (3.1)

DEP ' = DEP 2 + xbd 2 (3.2)

From the user inputted values of the weapon MAEF and the weapon
impact angle, we can calculate the weapon radii WRr and WRd with equations
(3.3) and (3.4).
a
WRr = MAEF × (3.3)
π
WRr
WRd = (3.4)
a

Where a is the aspect ratio of the weapon radii calculated from the weapon
impact angle I, and shown in equation (3.5).

WRr
a= = MAX [1 − 0.8cos( I ), 0.3] (3.5)
WRd

21
The JMEM method for a single weapon against a unitary target calculates the
effective target length L’ET and effective target width W’ET of the weapon lethal
area based on the weapon radii. This is shown in equations (3.6) and (3.7).

a
L 'ET = 2 × WRr = 1.128 MAEF × (3.6)
π
L 'ET
W 'ET = 2 × WRd = (3.7)
a

In order to calculate the single sortie probability of damage (SSPD), we

use equation (3.8). This gives the probability of damage for a single weapon with
user inputted lethality and accuracy against a unitary target.

L 'ET × W 'ET
SSPD = (3.8)
(17.6( REP ') + L 'ET 2 )(17.6( DEP ') 2 + W 'ET 2 )
2

B. MONTE CARLO SIMULATIONS

A Monte Carlo simulation is a way of finding the probability of damaging a

target by analyzing the results of a large number of sample weapon versus target
trials. For example, to find the probability of damaging a target using a Monte
Carlo approach, we look at a large number of independent trials of a weapon
aimed at a desired impact point. The difference between iterations is a small
random variation in the delivery accuracy of the weapon. This is shown in Figure
3.1.

22
 ITERATION 1
Deflection
Ballistic
Dispersion
Aim
Point

Impact

Range

Aiming
Error

ITERATION 2
Impact Deflection

Aim
Ballistic Point
Dispersion

Range

Aiming
Error

Figure 3.1: Two sample iterations of a Monte Carlo simulation.

Figure 3.1 shows two sample iterations of a Monte Carlo simulation. By slightly
changing the aiming error and ballistic dispersion error of a weapon, the impact
point of the weapon will change.
The variations in delivery accuracies are obtained by first calculating the
standard deviations of the aiming error and ballistic dispersion in the ground
plane. Given the user inputted REP and DEP, we obtain the standard deviation

23
of the aiming error in the ground plane for both the range and deflection
directions (σaim_range and σaim_deflection respectively) through equations (3.9) and
(3.10)
REP
σ aim_range = (3.9)
0.6745
DEP
σ aim_deflection = (3.10)
0.6745

Given the user inputted slant range (SR), impact angle (I), and ballistic dispersion
measured in mils in the normal plane (σb), we calculate the standard deviation of
the ballistic dispersion in the ground plane for both the range and deflection
directions (xbr and xbd) through equations (3.11) and (3.12).

0.6745 × SR × σ b
xbr =
1000sin( I ) (3.11)

0.6745 × SR × σ b
xbd = (3.12)
1000

All four of the delivery accuracy standard deviations (σaim_range, σaim_deflection, xbr,
and xbd) are then multiplied by a random number, and the standard deviations
are added together in the range and deflection directions. This results in the
range and deflection location of the impact point, and is shown in equations (3.13)
and (3.14).

Impact Point RANGE = (σ aim_range × random #) + ( xbr × random #) (3.13)

Impact Point DEFLECTION = (σ aim_deflection × random #) + ( xbd × random #) (3.14)

The random# term in the previous equations is a random number from a normal
distribution. This normal distribution has a mean value of zero and a standard
deviation of one. Using this normal distribution allows us to assume that the
24
average probability of damage for a large number of weapon trials will equal the
expected value of the damage function at the desired aim point.
This thesis will use two approaches in modeling the weapon lethal area in
a Monte Carlo simulation. In each simulation, the ground plane and the lethal
area are divided into 1ft by 1ft cells. The first weapon model will use a weapon
lethal area matrix with each matrix cell populated with a probability of damage
obtained by the Carleton damage function. An example of a lethal area matrix
populated by the Carleton damage function is given in Figure 3.2.

0.0000 0.0003 0.0013 0.0036 0.0059 0.0059 0.0036 0.0013 0.0003 0.0000
0.0003 0.0022 0.0098 0.0266 0.0439 0.0439 0.0266 0.0098 0.0022 0.0003
0.0013 0.0098 0.0439 0.1194 0.1969 0.1969 0.1194 0.0439 0.0098 0.0013
0.0036 0.0266 0.1194 0.3247 0.5353 0.5353 0.3247 0.1194 0.0266 0.0036
0.0059 0.0439 0.1969 0.5353 0.8825 0.8825 0.5353 0.1969 0.0439 0.0059
0.0059 0.0439 0.1969 0.5353 0.8825 0.8825 0.5353 0.1969 0.0439 0.0059
0.0036 0.0266 0.1194 0.3247 0.5353 0.5353 0.3247 0.1194 0.0266 0.0036
0.0013 0.0098 0.0439 0.1194 0.1969 0.1969 0.1194 0.0439 0.0098 0.0013
0.0003 0.0022 0.0098 0.0266 0.0439 0.0439 0.0266 0.0098 0.0022 0.0003
0.0000 0.0003 0.0013 0.0036 0.0059 0.0059 0.0036 0.0013 0.0003 0.0000

Figure 3.2: Example of lethal area matrix populated by Carleton damage

function.

The dimensions of this lethal area matrix are dictated by the weapon radii. For
the Monte Carlo simulations in this thesis, the size of the lethal area matrix was
set to 4WRr by 4WRd because outside of these dimensions, the probability of
damage from the Carleton damage function is negligible.
The second Monte Carlo simulation will use the rectangular cookie cutter
approximation of the Carleton damage function. This will also use a lethal area
matrix, but the lethal area matrix cells will all be populated by a probability of
damage equal to one. The dimensions of this lethal area are LET and WET as
explained in Chapter II Section B Lethal Areas. These dimensions are calculated
from the user inputs of MAEF and weapon impact angle (which determines the
aspect ratio a as in equation (3.5)) as shown in equations (3.15) and (3.16).
25
 LET = MAEF × a (3.15)

WET = LET / a (3.16)

Again, the probability of damage inside of the lethal area matrix defined by LET
and WET is equal to one, and the probability of damage outside of these
dimensions is zero.
The details of how each Monte Carlo simulation determines a probability
of damage will be explained in the following two sections.

1. Modeling Weapons with Carleton Damage Function

The first Monte Carlo simulation models a single weapon with a lethal area
matrix with 1ft by 1ft cells populated with a probability of damage from the
Carleton damage function. The first step in this simulation is to create the
weapon lethal area matrix from user inputs. The result of which gives a lethal
area matrix similar to the one shown in Figure 3.2. This weapon matrix is held at
a constant throughout the Monte Carlo simulation. The location of the target in
which we are aiming is what varies in each iteration of the simulation. The
process of varying the target location was explained above with equations (3.9)
through (3.14).
Examples of iterations of the Monte Carlo simulation are shown in Figure
3.3. Here we have a lethal area matrix populated by the Carleton damage
function fixed in the ground plane. For this example the location of the target,
which is based on the delivery accuracy of the weapon, is 3 feet past the
aimpoint in the range direction, and 2 feet past in the deflection direction. The
corresponding cell of this target location is shaded in dark grey. This 0.1194
value is the probability of damaging the target for this iteration. The next iteration
finds the target location to be 4 feet short of the aim point in the range direction,
and 3 feet short in the deflection direction. This target location is shaded in light
grey and has a probability of damage equal to 0.0098.
26
 Aim
Deflection Point
0.0000 0.0003 0.0013 0.0036 0.0059 0.0059 0.0036 0.0013 0.0003 0.0000
0.0003 0.0022 0.0098 0.0266 0.0439 0.0439 0.0266 0.0098 0.0022 0.0003
0.0013 0.0098 0.0439 0.1194 0.1969 0.1969 0.1194 0.0439 0.0098 0.0013
0.0036 0.0266 0.1194 0.3247 0.5353 0.5353 0.3247 0.1194 0.0266 0.0036
0.0059 0.0439 0.1969 0.5353 0.8825 0.8825 0.5353 0.1969 0.0439 0.0059
0.0059 0.0439 0.1969 0.5353 0.8825 0.8825 0.5353 0.1969 0.0439 0.0059 Range
0.0036 0.0266 0.1194 0.3247 0.5353 0.5353 0.3247 0.1194 0.0266 0.0036
0.0013 0.0098 0.0439 0.1194 0.1969 0.1969 0.1194 0.0439 0.0098 0.0013
0.0003 0.0022 0.0098 0.0266 0.0439 0.0439 0.0266 0.0098 0.0022 0.0003
0.0000 0.0003 0.0013 0.0036 0.0059 0.0059 0.0036 0.0013 0.0003 0.0000

Figure 3.3: Example iteration of Monte Carlo simulation using Carleton

damage function to fill lethal area matrix.

For the full Monte Carlo simulation, we would repeat this for n iterations, each
individual iteration (i) resulting in its own probability of damage Pd/i. The value n
is the total number of iterations used in the Monte Carlo simulation. The final
probability of damage for the Monte Carlo simulation is the average of all of the
Pd/I shown in equation (3.17).
n

∑P d /i
Pd = i =1
(3.17)
n

The number of iterations (n) used in the Monte Carlo simulation used to
compare the Monte Carlo and JMEM methods was set to ten million. In running
the Monte Carlo simulation multiple times, we ensure that the resulting Pd from
each simulation did not change. This shows that the simulation was converging
to a single value, which we would expect. After testing different values of n, it
was found that ten million provides for this convergence.
A Monte Carlo simulation with a weapon modeled with a lethal area matrix
fill by the Carleton damage function is presented in Appendix A, Option 1.

27
 2. Modeling Weapons with Rectangular Cookie Cutter
Approximation

The second Monte Carlo simulation models a single weapon as a lethal

area matrix rectangular cookie cutter. The dimensions of this lethal area matrix
are determined from user inputs as described earlier in this chapter in equations
(3.5), (3.15), and (3.16). As in the previous Monte Carlo simulation, we place this
lethal area matrix on the ground plane, and for each iteration of the simulation,
the location of the target changes due to variations in the delivery accuracy of the
weapon.
Examples of iterations of the Monte Carlo simulation are shown in Figure
3.4. Here we have a lethal area matrix with dimensions LET and WET fixed in the
ground plane. Since this is a rectangular cookie cutter approximation method,
we have probability of damage equal to one inside of the lethal area matrix, and
zero outside of the lethal area matrix. In this example we assume that the user
inputted values of the weapon effectiveness result in LET = 6ft and WET = 6ft.
This 6ft by 6ft rectangular cookie cutter is shown In Figure 3.4. For this example,
the location of the target, which is based on the delivery accuracy of the weapon,
is 3 feet past the aim point in the range direction, and 2 feet past in the deflection
direction. The corresponding cell of this target location is shaded in dark grey.
Since this target location is inside the weapon lethal area, its probability of
damage is 1.0. Say the next iteration finds the target location to be 4 feet short
of the aim point in the range direction, and 3 feet short in the deflection direction.
This target location is shaded in light grey and has a probability of damage equal
to 0.0.

28
 Aim
Deflection Point

0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
0.0 0.0 1.0 1.0 1.0 1.0 1.0 1.0 0.0 0.0
0.0 0.0 1.0 1.0 1.0 1.0 1.0 1.0 0.0 0.0
0.0 0.0 1.0 1.0 1.0 1.0 1.0 1.0 0.0 0.0

Range
0.0 0.0 1.0 1.0 1.0 1.0 1.0 1.0 0.0 0.0
0.0 0.0 1.0 1.0 1.0 1.0 1.0 1.0 0.0 0.0
0.0 0.0 1.0 1.0 1.0 1.0 1.0 1.0 0.0 0.0
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0

Figure 3.4: Example iteration of Monte Carlo simulation using a

rectangular cookie cutter lethal area matrix.

For a full Monte Carlo simulation, the total Pd is found as it was before with
equation (3.17). For this simulation however, we will only sum ones and zeros
because those are the only values Pd/I can have. Again, the Monte Carlo
simulation using a rectangular cookie cutter lethal area uses ten million iterations
to achieve a convergence of results.

C. RESULTS OF UNITARY TARGET COMPARISONS

The comparisons done this section are to find the differences between
weapon effectiveness methodologies for the case of a single weapon against
unitary targets. In order to get a wide array of results each comparison was done
for different user inputted values of MAEF, REP & DEP and ballistic dispersion in
mils, in the normal plane. The weapon impact angle and release slant range
were kept at a constant 45° and 10,000 ft respectively for each comparison.

29
 1. JMEM versus Carleton Damage Function

The first comparison to the JMEM methodology of calculating the

effectiveness of a single weapon against a unitary target is to compare it to a
Monte Carlo simulation using the Carleton damage function. This is a particularly
useful comparison because the current JMEM method is based on the Carleton
damage function. Since this is the case, we would expect the Monte Carlo
simulation representing the weapon as the Carleton damage function to produce
similar results to the JMEM. Tables 3.1 and 3.2 present the results for the PK
found for both the JMEM method and Monte Carlo simulation. The shaded
values in the Tables are obtained from the JMEM method, while the unshaded
values are those of the Monte Carlo simulation.

MAEF=5000ft2 Ballistic Dispersion (mils in normal plane)

0 3 5
REP=25ft & DEP=50ft 0.2236 0.1787 0.1363
0.2239 0.1477 0.0978

REP=25ft & DEP=25ft 0.3384 0.2621 0.1905

0.3386 0.2091 0.1284

REP=50ft & DEP=25ft 0.1834 0.1616 0.1335

0.1837 0.1419 0.101

REP=50ft & DEP=50ft 0.1211 0.1103 0.0956

0.1215 0.1002 0.077

Table 3.1: Results of JMEM and Monte Carlo for MAEF=5000ft2

30
 MAEF=1000ft2 Ballistic Dispersion (mils in normal plane)
0 3 5
REP=25ft & DEP=50ft 0.0542 0.042 0.031
0.0546 0.0339 0.0217

REP=25ft & DEP=25ft 0.0999 0.0719 0.0483

0.1003 0.0543 0.0304

REP=50ft & DEP=25ft 0.0509 0.0427 0.0333

0.0511 0.036 0.0237

REP=50ft & DEP=50ft 0.0276 0.0249 0.0213

0.0278 0.0225 0.0169

Table 3.2: Results of JMEM and Monte Carlo for MAEF=1000ft2

It can be seen from Tables 3.1 and 3.2 that the Monte Carlo simulation
does not produce the same results as the JMEM. This is especially true for the
case of a large ballistic dispersion. The lethal area as described by the MAEF
does not seem to alter the error between the JMEM and Monte Carlo result,
which is expected since both methodologies use the MAEF to find the weapon
radii used to describe the lethality of the weapon. This leaves the delivery
accuracy as the prime reason for the differences in results. These differences
are especially seen for the case of large ballistic dispersion. As the ballistic
dispersion is lowered down to zero, we see a decrease in the error between the
two methods. In the cases of no ballistic dispersion for any of the four aiming
error combinations, we obtain results that are within a tenth of a percent of each
other for the two methods. This leads us to believe that the handling of the
ballistic dispersion is the main reason for differences in the results. There are
two viable explanations for this. One possibility is that the Monte Carlo
simulation does not correctly represent the ballistic dispersion by varying it by a
random number in each iteration of the code. Another reason may be from the
fact that the JMEM method root sums squares the aiming error with the ballistic
dispersion (measured in feet in the ground plane). It may be that these ballistic
errors cannot be combined with the aiming error probables as originally thought.

31
 It is also interesting to note that the Monte Carlo result is constantly
greater that the JMEM result for non-zero ballistic dispersion cases, which shows
that the JMEM is a more conservative estimate of damage.

2. JMEM versus Rectangular Cookie Cutter Approximation

The last comparison for single weapons against unitary targets is between
the JMEM method and a Monte Carlo simulation which represents the weapon
as a rectangular cookie cutter. Though interesting, the results of this comparison
have very little practical meaning. JMEM weaponeering methodologies make
use of the rectangular cookie cutter approximation in cases where the Carleton
damage function itself to too complicated to utilize. The case we are looking at
now of a single weapon versus and unitary target does not need to use an
approximation, and the JMEM method uses the Carleton damage function as
was explained in Chapter III Section A. Modeling the weapon as a rectangular
cookie cutter area would only lessen the fidelity of the JMEM method for this
weapon/target case. The results of comparing the Monte Carlo simulation and
the JMEM method are given in Tables 3.3 and 3.4. The shaded values in the
Tables are obtained from the JMEM method, while the unshaded values are
those of the Monte Carlo simulation.
MAEF=5000ft2 Ballistic Dispersion (mils in normal plane)
0 3 5
REP=25ft & DEP=50ft 0.2443 0.1921 0.1438
0.2239 0.1477 0.0978

REP=25ft & DEP=25ft 0.3937 0.2975 0.2108

0.3386 0.2091 0.1284

REP=50ft & DEP=25ft 0.2062 0.1801 0.146

0.1837 0.1419 0.101

REP=50ft & DEP=50ft 0.1218 0.1161 0.0997

0.1215 0.1002 0.077

Table 3.3: Results of JMEM and Monte Carlo for MAEF=5000ft2

32
 MAEF=1000ft2 Ballistic Dispersion (mils in normal plane)
0 3 5
REP=25ft & DEP=50ft 0.0528 0.0408 0.0298
0.0546 0.0339 0.0217

REP=25ft & DEP=25ft 0.1007 0.0713 0.0475

0.1003 0.0543 0.0304

REP=50ft & DEP=25ft 0.0509 0.0422 0.0326

0.0511 0.036 0.0237

REP=50ft & DEP=50ft 0.0266 0.024 0.0206

0.0278 0.0225 0.0169

Table 3.4: Results of JMEM and Monte Carlo for MAEF=1000ft2

As expected, the results from the Monte Carlo simulation and the JMEM
method do not match. This is due mainly to the reasons discussed before
concerning the fact that the rectangular cookie cutter method makes
approximations in the Carleton damage function that the JMEM method does not
make. Along with this reason we suspect that the ballistic dispersion accounts
for some of the error between the methods, as it did for the Monte Carlo
simulation with the Carleton damage function. Also similar to the previous
comparison is that the Monte Carlo simulation consistently gives greater damage
results than the JMEM method, leading to the conclusion that the JMEM is more
conservative than the Monte Carlo for all delivery accuracy inputs.
A better way of comparing the rectangular cookie cutter Monte Carlo
simulation and JMEM method is to look at the JMEM for single weapons versus
area targets. The concept of single weapons versus area target will be
discussed in detail Chapter IV. When the JMEM calculates a probability of
damage for a single weapon against a unitary target, the target dimensions
inputted into OEM 5.0 (Reference 2) are a 0 ft by 0 ft target. When these
dimensions are inputted, the OEM program uses the Carleton damage function
as described earlier in this chapter for a single weapon against a unitary target.
33
When anything other than 0 ft by 0 ft target is inputted, the OEM program uses
the JMEM method for single weapons against area targets. This methodology
for area targets uses an approximation of the Carleton damage function to find
the probability of damage. For the comparison given below in Tables 3.5 and 3.6,
a target dimension of .01 ft by .01 ft was inputted into OEM 5.0 to represent a
unitary target with a small area target. Now we are able to compare results of
more similar methods, rather than comparing a rectangular cookie cutter Monte
Carlo simulation with the Carleton damage function.

MAEF=5000ft2 Ballistic Dispersion (mils in normal plane)

0 3 5
REP=25ft & DEP=50ft 0.2443 0.1921 0.1438
0.2503 0.1601 0.104

REP=25ft & DEP=25ft 0.3937 0.2975 0.2108

0.402 0.2379 0.1407

REP=50ft & DEP=25ft 0.2062 0.1801 0.146

0.2109 0.1595 0.1103

REP=50ft & DEP=50ft 0.1218 0.1161 0.0997

0.1313 0.1073 0.0815

Table 3.5: Results of JMEM (small area) and Monte Carlo for MAEF=5000ft2

MAEF=1000ft2 Ballistic Dispersion (mils in normal plane)

0 3 5
REP=25ft & DEP=50ft 0.0528 0.0408 0.0298
0.0546 0..0347 0.022

REP=25ft & DEP=25ft 0.1007 0.0713 0.0475

0.1071 0.0567 0.0313

REP=50ft & DEP=25ft 0.0509 0.0422 0.0326

0.0541 0.0375 0.0244

REP=50ft & DEP=50ft 0.0266 0.024 0.0206

0.0285 0.023 0.0172

Table 3.6: Results of JMEM (small area) and Monte Carlo for MAEF=1000ft2

34
We can see from Tables 3.3, 3.4, 3.5 and 3.6 that the results for the Monte Carlo
simulation match somewhat better with the JMEM method for a small area target
than they do with the unitary target JMEM model. This was expected due to the
similarities between the rectangular cookie cutter method and the JMEM method
for area targets in approximating the Carleton damage function. Differences in
the results are from the sources previously discussed in this section, as well the
fact that we are dealing with different target areas. The Monte Carlo uses a 1ft
by 1 ft cell to represent the unitary target, while we have to use a .01 ft by .01 ft
area to calculate the JMEM method probability of damage.

35
THIS PAGE INTENTIONALLY LEFT BLANK

36
 IV. SINGLE WEAPON VERSUS AREA TARGET

The area target is best described as an area of targets, or a certain

grouping of target elements. An area target could contain anything from ground
troops to tanks to components of a surface to air missile site. The key to
analyzing area targets is that we must assume that the target elements are
uniformly distributed within the prescribed area.
When determining the amount of damage done to an area target, we do
not calculate the single sortie probability of damage as done with unitary targets.
For area targets we measure a quantity called the expected fractional damage,
or EFD. Since we are dealing with an area of target elements, we are concerned
with what fraction of these target elements are damaged for a given weapon
impact. This is shown in Figure 4.1 below. Figure 4.1 shows an area target
consisting of twelve target elements (denoted by stars), which are uniformly
distributed inside the target area dimensions.

Area Target

Weapon Lethal Area

Figure 4.1: Area target partially covered by weapon lethal area. (After Ref. [1])

In this case, 50% of the target elements are covered by the weapon lethal area
rectangle. If we assume the probability of damage inside of the weapon lethal
area, or PCD, is 1.0, we know the fractional damage done in Figure 4.1. This will
37
remain true for any case in which the weapon lands at a point such that half of
the area target elements are covered by the weapon lethal area. The EFD is
calculated from the expected fractional coverage of the weapon lethal area on
the area target for a large number of trials, shown in equation (4.1)

EFD = E ( FC ) × PCD (4.1)

The same comparisons done for unitary target situations earlier will be
done for area targets. The methods analyzed are Monte Carlo simulations with
two ways of modeling weapon effectiveness, and the JMEM method of
calculating EFD.

A. JMEM METHOD

The first step in the JMEM method for area targets is dealing with the
delivery accuracy of the weapon. The delivery accuracy is handled in the same
fashion as it is for unitary targets by root sum squaring the aiming errors and
ballistic dispersion to find REP’ and DEP’.
We will first look at only at the problem in the range direction. The area
target is assigned length and width dimensions of LA and WA respectively. A
sample weapon/target interaction is presented in Figure 4.2. The figure shows a
target length of LA with a weapon impacting the ground a distance u from the
center of the area target. The weapon is represented by the rectangular
equivalent of the Carleton damage function, where the probability of damage
inside of the rectangular area is equal to unity. As before when calculating SSPD,
the accuracy function of the weapon is represented by g(x), and the rectangular
damage function has length of LET in our one dimensional case.

38
 Accuracy function g(x) Damage

1.0

LET

target x=u Range x

-LA/2 LA/2

Figure 4.2: Rectangular lethal area against an area target. (From Ref. [1])

The example shown in Figure 4.2 would yield no damage to the area
target because the weapon lethal area does not overlap the area target. The
next weapon drop however, may have an impact point that causes the lethal area
to partially cover the area target, which we have defined as fractional coverage.
In order to determine the damage caused by a weapon, we will need to know the
details of the fractional coverage.
With a weapon dropped against an area target, there are an infinite
number of ways in which the weapon lethal area can either partially overlap,
totally cover, or not overlap the area target. This creates a problem when dealing
with weapon lethal areas that have different area target dimensions. Through a
detailed explanation provided in Reference 1, we are able to simplify the problem
of weapon overlap scenarios. This is done first by defining a parameter called
the effective pattern length and width. These are found in equations (4.2) and
(4.3).

LEP = max( LET , LA ) (4.2)

WEP = max(WET ,WA ) (4.3)

39
 The next finding is that there is a way to summarize the fractional
coverage of a weapon lethal area onto an area target as a function of x, or the
distance from the weapon impact point to the center of the area target. This
summary is shown graphically in Figure 4.3.

FR FR = 1

LEP + LA x LEP + LA x
FR = + FR = −
2 LA LA 2 LA LA

-s -t t s

Figure 4.3: Fractional coverage in the range direction. (From Ref. [1])

The values s and t are shorthand notations given in equations (4.4) and (4.5).
LEP + LA
s= (4.4)
2
LEP − LA
t= (4.5)
2
We are then able to find the expected value of the fractional coverage
since we know the fractional coverage given a value of x (from Figure 4.3) and
that the value x itself is a product of the normally distributed aiming error g(x).
We can then define the expected fractional coverage E(FR) in equation (4.6)

∞
E ( FR ) = ∫ F ( x) g ( x)dx
−∞
R (4.6)

Equation (4.6) can be rewritten b\y substituting in the value of FR from Figure 4.3
and using the probability density function of the aiming error g(x) for a normal
distribution with zero mean (given by equation (4.7))
40
 1  − x2 
g ( x) = exp   (4.7)
σx 2π  2σ x 
We can then expand equation (4.6) to equation (4.8)

t  − x2  
 ∫ exp  2 
dx 
−t  2σ x  
+∞  −t   − x  
1  LEP + LA x  2
E ( FR ) = ∫ FR ( x) g ( x)dx =
σ x 2π  −∫s  2 LA
+  +  exp  2 
dx  (4.8)
−∞
LA   2σ x  
 s 
 +  LEP + LA − x  exp  − x dx 
2

∫  2σ x  
 t  2 LA
2
LA 

This may also be written as equation (4.9)

t  − x2   LEP + LA  −t  − x2  
 ∫ exp  dx +   ∫  2σ x 2 
exp  dx 
 2σ x 
2
−t  2 LA  − s 
   −x  −t
 − x  
1  LEP + LA 
s 2 2
x
σ x 2π   2 LA  ∫t ∫− s LA  2σ x 2 dx 
E ( FR ) = +
   exp  2 
dx + exp (4.9)
 2σ x  
 s 
− x exp  − x dx
2

 ∫t LA 2
 x σ 2


Equation (4.9) is easier solved in parts, so a shorthand notation is used to split it

into five smaller integrals. This shorthand is shown in equation (4.10).

E ( FR ) = {I1 + I 2 + I 3 + I 4 − I 5 } (4.10)

The first three integrals in equation (4.10) involve the normal cumulative
distribution given in equation (4.7). In order to compute the normal cumulative
distribution function (CDF) we use the MATLAB function normcdf.m. This
function takes the user inputs of x, µ (mean of the miss distance), and σ

41
(standard deviation of the miss distance) and gives the value of the CDF. In our
case, each integral has the same zero mean and the same standard deviation
calculated from the delivery accuracy of the weapon. Using this function as well
as our previously defined shorthand notation, we solve the first three integrals in
equations (4.11), (4.12), and (4.13).

I1 = [normcdf (t ) − normcdf (−t )] (4.11)

LEP + LA
I2 = [normcdf (−t ) − normcdf (− s )] (4.12)
2 LA
LEP + LA
I3 = [normcdf ( s ) − normcdf (t )] (4.13)
2 LA

By rearranging elements of the fourth and fifth integrals, we can solve the
difference between the integrals analytically as done in Reference 1. The result
of this is given in equation (4.14).

2 2
 s   t 
2σ x −
 σ 2 
−
 σ 2 
I 4 − I5 = [e  x 
−e  x 
] (4.14)
LA 2π

The same procedure is used for the deflection direction. Equations (4.15)
through (4.21) summarize the equations to find E(FD).

WEP + WA
s= (4.15)
2
WEP − WA
t= (4.16)
2

E ( FD ) = {I1 + I 2 + I 3 + I 4 − I 5 } (4.17)

42
 I1 = [normcdf (t ) − normcdf (−t )] (4.18)

WEP + WA
I2 = [normcdf (−t ) − normcdf (− s )] (4.19)
2WA

WEP + WA
I3 = [normcdf ( s ) − normcdf (t )] (4.20)
2WA

2 2
 s   t 
2σ y − 
σy 2 
− 
σy 2 
I 4 − I5 = [e  
−e  
] (4.21)
WA 2π

Once we have the expected fractional coverage in both the range and
deflection direction we combine them in equation (4.22) to find the total
expected fractional coverage.

E ( FC ) = E ( FR ) × E ( FD ) (4.22)

Given the fractional coverage of the weapon, we multiple it by the

reliability of the weapon (R) and the conditional probability of damage (PCD)
inside of the weapon lethal area as shown in equation (4.23). The PCD term is
equal to unity when the weapon lethal area is larger than the target area. When
the target area is larger, the PCD must go down in order to conserve the lethality
of the weapon. This is explained in detail in Reference 1.

 AET 
EFD = E ( FC ) × PCD × R = E ( FC ) ×  × R (4.23)
 LEP × WEP 

43
B. MONTE CARLO SIMULATIONS

For the case of single weapons versus area targets, we run two Monte
Carlo simulations to determine the probability of damaging the target. The first
Monte Carlo represents the weapon as a lethal area matrix consisting of 1 ft by 1
ft cells populated with probability of damage values populated from the Carleton
damage function. The second Monte Carlo simulation represents the weapon
with a lethal area matrix which has dimensions of a rectangular cookie cutter
approximation of the Carleton damage function. The details of these two
simulations are explained in detail in Chapter 3 Section B. This section also
explains the process of varying the delivery accuracy of the weapon in each
iteration of the Monte Carlo Simulation.

1. Modeling Weapons with Carleton Damage Function

The first Monte Carlo simulation models the weapon as a lethal area
matrix populated with the Carleton damage function. This process begins with
the user inputting values of the MAEF and impact angle (I) of the weapon. This
information provides us the means to calculate the aspect ratio (a) of the weapon
radii which gives us the weapon radii (WRr and WRd), as shown in equations
(4.24), (4.25), and (4.26).

WRr
a= = MAX [1 − 0.8cos( I ), 0.3] (4.24)
WRd

a
WRr = MAEF × (4.25)
π
WRr
WRd = (4.26)
a

Now that the weapon radii are calculated, the lethal area matrix can be filled
with values from the Carleton damage function in equation (4.27). The values x
44
and y are the distances in the range and deflection direction respectively from
the weapon impact point. For our Monte Carlo simulation, the weapon lethal
area matrix is fixed in the ground plane and we will look at the distance from the
target to the center of the weapon impact point in order to find the probability of
damaging the target.

 x2 y2 
P ( x, y ) = exp −  2
+ 2
(4.27)
 WRr WRd 

An example iteration of a Monte Carlo simulation modeling the weapon

as a lethal area matrix filled with the Carleton damage function are shown in
Figure 4.4. The target we are dealing with is 3 ft by 3 ft for this example. We
first look at a target location where the center of the target is placed at 3 ft past
the aim point in the range direction, and 1 ft short of the aim point in the
deflection direction. Again, this target placement based on the delivery
accuracy of the weapon. By knowing the location of the center of the target, we
can then superimpose the 3 ft by 3ft target onto the lethal area matrix on the
ground plane. We superimpose the target area onto the lethal area by dividing
the target area into 1ft by 1ft cells. This way, each division of the target area will
overlap a specific value of the lethal area matrix. This is shown by the dark grey
3 ft by 3ft target area in Figure 4.4.

45
 Deflection Aim
Point
0.0000 0.0003 0.0013 0.0036 0.0059 0.0059 0.0036 0.0013 0.0003 0.0000
0.0003 0.0022 0.0098 0.0266 0.0439 0.0439 0.0266 0.0098 0.0022 0.0003
0.0013 0.0098 0.0439 0.1194 0.1969 0.1969 0.1194 0.0439 0.0098 0.0013
0.0036 0.0266 0.1194 0.3247 0.5353 0.5353 0.3247 0.1194 0.0266 0.0036
0.0059 0.0439 0.1969 0.5353 0.8825 0.8825 0.5353 0.1969 0.0439 0.0059
0.0059 0.0439 0.1969 0.5353 0.8825 0.8825 0.5353 0.1969 0.0439 0.0059 Range
0.0036 0.0266 0.1194 0.3247 0.5353 0.5353 0.3247 0.1194 0.0266 0.0036
0.0013 0.0098 0.0439 0.1194 0.1969 0.1969 0.1194 0.0439 0.0098 0.0013
0.0003 0.0022 0.0098 0.0266 0.0439 0.0439 0.0266 0.0098 0.0022 0.0003
0.0000 0.0003 0.0013 0.0036 0.0059 0.0059 0.0036 0.0013 0.0003 0.0000

Figure 4.4: Example iteration of Monte Carlo simulation using Carleton

damage function to fill lethal area matrix against an area
target.

Finding the probability of damaging the area target for each iteration, we need
to account for the probability of damage in each 1ft by 1 ft division of the target
area. The total probability of damage is obtained by averaging the individual
target cells. This is shown in equation (4.28).

∑P
j =1
d / cell ( j )

Pd / i = (4.28)
j

Equation (4.28) sums the probability of damage Pd/cell(j) from each of the j total
number of 1 ft by 1 ft cells in the area target. It then divides this sum by the total
number of cells j to find the average Pd/i, which is the probability of damaging
the area target for a given iteration i. For our example this would give results
shown in equation (4.29).

0.5353 + 0.1969 + 0.0439 + 0.5353 + 0.1969 + 0.0439 + 0.3247 + 0.1194 + 0.0266 (4.29)
Pd / i = = .2247
9

46
 For the full Monte Carlo simulation, we repeat this for n iterations, each
individual iteration (i) resulting in its own probability of damage Pd/i. The value n
is the total number of iterations used in the Monte Carlo simulation. The final
probability of damage for the Monte Carlo simulation is the average of all of the
Pd/I shown in equation (4.30)

∑P d /i
Pd = i =1
(4.30)
n

A Monte Carlo simulation with a weapon modeled with a lethal area

matrix fill by the Carleton damage function is presented in Appendix A, Option 1.
This is the same option for the single weapon versus unitary target case, so the
user must input an area target size.

2. Modeling Weapons with Rectangular Cookie Cutter

Approximation

The second Monte Carlo simulation to find the probability of damaging an

area target models a single weapon as a lethal area matrix rectangular cookie
cutter. The dimensions of this lethal area matrix are determined from user inputs
as described earlier in Chapter III, equations (3.5), (3.15), and (3.16). As in the
previous Monte Carlo simulation, we place this lethal area matrix on the ground
plane, and for each iteration of the simulation, the location of the target changes
due to variations in the delivery accuracy of the weapon.
An example iteration of the Monte Carlo simulation representing the
weapon as a rectangular cookie cutter approximation of the Carleton damage
function is given in Figure 4.5. For this sample case, we assume that the user
inputted values of weapon lethality produce a LET = 6ft and a WET = 6ft. These
are the dimensions of the rectangular cookie cutter lethal area. This lethal area
matrix is then fixed to the ground plane. For this example, the center of the 3 ft

47
by 3 ft area target is 3 ft short of the aim point in the range direction, and 2 ft
short of aim point in the deflection direction. This location is based on the
delivery accuracy of the weapon. This is shown by the dark grey 3 ft by 3ft
target area in Figure 4.5.

Aim
Deflection Point
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
0.0 0.0 1.0 1.0 1.0 1.0 1.0 1.0 0.0 0.0
0.0 0.0 1.0 1.0 1.0 1.0 1.0 1.0 0.0 0.0
0.0 0.0 1.0 1.0 1.0 1.0 1.0 1.0 0.0 0.0

Range
0.0 0.0 1.0 1.0 1.0 1.0 1.0 1.0 0.0 0.0
0.0 0.0 1.0 1.0 1.0 1.0 1.0 1.0 0.0 0.0
0.0 0.0 1.0 1.0 1.0 1.0 1.0 1.0 0.0 0.0
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0

Figure 4.5: Example iteration of Monte Carlo simulation using a rectangular

cookie cutter lethal area matrix against an area target.

Again, finding the probability of damaging the area target for each iteration, we
need to account for the probability of damage in each 1ft by 1 ft division of the
target area. The total probability of damage is obtained by averaging the
individual target cells. This is shown above in equation (4.28). For our example
iteration we would compute Pd/I = .6667 as shown in equation (4.31).

0.0 + 1.0 + 1.0 + 0.0 + 1.0 + 1.0 + 0.0 + 1.0 + 1.0

Pd / i = = .6667 (4.31)
9

As before, for the full Monte Carlo simulation, we repeat this for n
iterations, each individual iteration (i) resulting in its own probability of damage
Pd/i. The final probability of damage for the Monte Carlo simulation is the
average of all of the Pd/I shown in equation (4.30).

48
 A Monte Carlo simulation with a weapon modeled by a lethal area matrix
filled by a rectangular cookie cutter approximation of the Carleton damage
function is presented in Appendix A, Option 2. This is the same option for the
single weapon versus unitary target case, so the user must input an area target
size.

C. RESULTS OF AREA TARGET COMPARISONS

The comparisons done this section are to find the differences between
weapon effectiveness methodologies for the case of a single weapon against
area targets. In order to get a wide array of results each comparison was done
for different user inputted values of MAEF, REP & DEP and ballistic dispersion in
mils, in the normal plane. The weapon impact angle and release slant range
were kept at a constant 45° and 10,000 ft respectively for each comparison. The
size of the area target used for all comparisons was a 50 ft by 50 ft area.

1. JMEM versus Carleton Damage Function

The first comparison done was the JMEM method results for computing
damage probabilities of single weapons versus area targets against a Monte
Carlo simulation modeling a weapon with the a lethal area matrix populated with
the Carleton damage function. We expect these methods to give different results
since the JMEM approximates the fractional coverage and the effectiveness of
the weapon against the target using equations (4.8) through (4.22) while the
Monte Carlo simulation uses a lethal area matrix from the Carleton damage
function. Tables 4.1 and 4.2 present the results for the PK found for both the
JMEM method and Monte Carlo simulation. The shaded values in the Tables are
obtained from the JMEM method, while the unshaded values are those of the
Monte Carlo simulation.

49
 MAEF=5000ft2 Ballistic Dispersion (mils in normal plane)
0 3 5
REP=25ft & DEP=50ft 0.2079 0.1694 0.131
0.2285 0.1521 0.1007

REP=25ft & DEP=25ft 0.3089 0.2445 0.181

0.3596 0.2226 0.135

REP=50ft & DEP=25ft 0.1743 0.1551 0.1288

0.1987 0.152 0.1066

REP=50ft & DEP=50ft 0.1168 0.1073 0.0929

0.1263 0.1039 0.0794

Table 4.1: Results of JMEM and Monte Carlo for MAEF=5000ft2

MAEF=1000ft2 Ballistic Dispersion (mils in normal plane)

0 3 5
REP=25ft & DEP=50ft 0.0499 0.0394 0.0297
0.0487 0.0321 0.021

REP=25ft & DEP=25ft 0.0877 0.0655 0.0454

0.0881 0.0511 0.0295

REP=50ft & DEP=25ft 0.0471 0.04 0.0317

0.0487 0.0349 0.0232

REP=50ft & DEP=50ft 0.0265 0.0242 0.0207

0.0269 0.0219 0.0166

Table 4.2: Results of JMEM and Monte Carlo for MAEF=1000ft2

It can be seen in Tables 4.1 and 4.2 that the probability of damage results
for the two methodologies does not match. The lethal area as described by the
MAEF does not seem to alter the error between the JMEM and Monte Carlo
result, which is expected since both methodologies use the MAEF to find the
weapon radii used to describe the lethality of the weapon. We can imply then
50
that the delivery accuracy of the weapon is the root of the error. This is
especially true for cases involving weapon ballistic dispersion. This was also
case for the single weapon versus unitary target cases. From this, we came
make one of two assumptions. One possibility is that the Monte Carlo simulation
does not correctly represent the ballistic dispersion by varying it by a random
number in each iteration of the code. Another reason may be that the root sum
squaring of the weapon ballistic dispersion with the aiming error as done in the
JMEM method may not be a realistic approximation of the weapon target
interaction.
Unlike the comparison between the JMEM and Monte Carlo simulation
methods performed for single weapons against unitary targets, from Table 4.1
and 4.2 we see an appreciable difference in the results for cases of zero ballistic
dispersion. This error could be due to the fact that the JMEM method
approximates the fractional coverage of the weapon on the target in terms of
Figure 4.3. This differs from the Monte Carlo simulation, in that the Monte Carlo
simulation computes the fractional coverage directly from the ground plane,
where no approximations are used.
It is also interesting to note that the Monte Carlo result is constantly
greater that the JMEM result for non-zero ballistic dispersion cases, which shows
that the JMEM is a more conservative estimate of damage. However, the results
for zero ballistic dispersion cases are generally more conservative in the Monte
Carlo simulation.

2. JMEM versus Rectangular Cookie Cutter Approximation

The last comparison for single weapons against area targets is between
the JMEM method and a Monte Carlo simulation which represents the weapon
as a rectangular cookie cutter. This is a particularly interesting comparison
because both methods represent the weapon lethality in terms of a rectangular
lethal area. The Monte Carlo simulation always represents the weapon lethal
area as a rectangular cookie cutter approximation of the Carleton damage
51
function, which is based off of the user inputted values of weapon lethality. Again,
this rectangular cookie cutter has dimensions of LET and WET. The JMEM method,
however, uses a rectangular cookie cutter dimensions based on the sizes of the
target and LET and WET as shown in equations (4.2) and (4.3). Table 4.3 and 4.4
present the results for the PK found for both the JMEM method and Monte Carlo
simulation. The shaded values in the Tables are obtained from the JMEM
method, while the unshaded values are those of the Monte Carlo simulation.

MAEF=5000ft2 Ballistic Dispersion (mils in normal plane)

0 3 5
REP=25ft & DEP=50ft 0.2253 0.181 0.1381
0.2285 0.1521 0.1007

REP=25ft & DEP=25ft 0.3564 0.2756 0.1994

0.3596 0.2226 0.135

REP=50ft & DEP=25ft 0.1958 0.1718 0.1405

0.1987 0.152 0.1066

REP=50ft & DEP=50ft 0.1237 0.1127 0.0969

0.1263 0.1039 0.0794

Table 4.3: Results of JMEM and Monte Carlo for MAEF=5000ft2

MAEF=1000ft2 Ballistic Dispersion (mils in normal plane)
0 3 5
REP=25ft & DEP=50ft 0.0486 0.0381 0.0286
0.0487 0.0321 0.021

REP=25ft & DEP=25ft 0.0879 0.0647 0.0445

0.0881 0.0511 0.0295

REP=50ft & DEP=25ft

0.0466 0.0394 0.0309
0.0487 0.0349 0.0232

REP=50ft & DEP=50ft

0.0258 0.0233 0.02
0.0269 0.0219 0.0166

Table 4.4: Results of JMEM and Monte Carlo for MAEF=1000ft2

52
 The trends shown in Tables 4.3 and 4.4 are very similar to those
presented for the other Monte Carlo versus JMEM comparisons. The most
recognizable trend is the fact that error between the JMEM and Monte Carlo
simulation is greatest for cases involving high weapon ballistic dispersion. For
this comparison less error is present for cases with zero ballistic dispersion than
there was for the comparison using the Monte Carlo simulation representing the
weapon with the Carleton damage function. A viable explanation for this is that
the JMEM and Monte Carlo using the rectangular cookie cutter model the
weapon lethal area more in a similar fashion. This is not the case that was
presented earlier in this chapter, because the Monte Carlo simulation
representing the weapon with the Carleton damage function does not use an
approximation of the Carleton damage function in calculating the weapon lethal
area matrix.

Again, a prevalent similarity between all the comparisons presented is that

the Monte Carlo simulation consistently gives greater damage results than the
JMEM method, leading to the conclusion that the JMEM is more conservative
than the Monte Carlo for all delivery accuracy inputs.

53
THIS PAGE INTENTIONALLY LEFT BLANK

54
 V. STICKS OF WEAPONS VERSUS AREA TARGETS

There are times when an aircraft will drop several weapons against an
area target, rather than just a single weapon. By dropping several weapons
during one aircraft sortie we increase the damaged area in the ground plane. At
times this can increase the probability of damaging the target area depending on
the size and delivery accuracy of the stick.
When a stick of weapons is released against an area target, the pattern of
weapon impacts on the ground determines the effectiveness of the stick. Figure
5.1 shows an example of the pattern of impact points for a stick of four weapons.
Each impact point is modeled with a small explosive picture. The dotted
rectangles represent the lethal area of each weapon. By knowing the location of
each impact point and the size of the lethal area for each weapon, we define
dimensions of the stick, and impact point pattern. We define the stick length and
width, LS and WS, as the dimensions of the smallest rectangle in the ground
plane that encloses each impact point of the weapons in the stick. These impact
points are also the centers of each respective weapon’s lethal area. The pattern
length and width, LP and WP, are defined as the dimensions of the smallest
rectangle that encloses all of the weapons’ lethal areas.

55
 Direction LP
LS of Flight

WS

WP

Figure 5.1: Definition of stick and pattern dimensions.

The location of the impact points on the ground plane is highly dependent
on each of the weapons’ release conditions. In the range direction, the impact
pattern is determined by the intervalometer setting of the aircraft. The
intervalometer is a timer in the aircraft that sends electrical pulses to the
mechanism that releases the weapons from the aircraft. For example, an
intervalometer setting of .5 seconds, which means, one or more weapons will be
released every .5 seconds. For a given stick delivery we define the number of
intervalometer pulses sent, or nr, and the number of weapons released for each
pulse, or np. In the deflection direction, the stick pattern is also affected by the
release conditions. This dependence stems from the location of the weapon on
the aircraft. We would expect a weapon released from under the aircraft
fuselage to land in a different position than one released from the tip of the
aircraft’s wing. In this thesis, we will not study the effects of different weapon

56
release positions on the aircraft. We instead assume that half of the weapons fall
from the same position one wing, and the other half on the same position on the
opposite wing.
Another determining factor of the impact points is the delivery accuracy of
the stick. We define the delivery accuracy differently for a stick of weapons than
we do a single weapon.
In this thesis we will compare the probability of damaging an area target of
dimensions 50 ft by 50 ft using different methodologies. As before, we will
compare the JMEM method of computing the probability of damaging an area
target to two Monte Carlo simulations; the first determining damage using the
Carleton damage function, and the other using a rectangular cookie cutter
approximation of the Carleton damage function to compute damage.

A. OLD JMEM METHOD

Recently the JMEM changed its methodologies to determine the

effectiveness of sticks of weapons against area targets. The old JMEM method
for sticks takes heavily into account the possible overlap situations between the
weapon lethal areas of the stick weapons. The following explains this method in
detail.
The old JMEM method first calculates the effectiveness of the individual
weapons in the stick. For the calculations in this thesis, we assume that each of
the weapons in the stick is identical. From the user inputted value of MAEF,
impact angle I, and the slant range SR, we find the effective target length and
width (LET and WET). This is shown in equations (5.1), (5.2), and (5.3).

WRr
a= = MAX [1 − 0.8cos( I ), 0.3]
WRd (5.1)

LET = MAEF × a (5.2)

57
 WET = LET / a (5.3)

The next step is to determine the effect of ballistic dispersion on the stick
effectiveness. We apply the ballistic dispersion to each weapon by enlarging the
area of effectiveness. This is done because we cannot combine the ballistic
dispersion error with the aiming error when dealing with sticks. Ballistic
dispersion in sticks affects each weapon in the stick, rather than the stick as a
whole. A detailed explanation of the delivery accuracy of sticks is presented in
Section C of this chapter. The new dimensions of the effective area are called LB
and WB. To calculate LB and WB, we first need to find the ballistic dispersion as a
standard deviation in feet in the range and deflection directions in the ground
plane, or xbr and xbd. These values are calculated using equations (5.4) and (5.5).

SR × σ b
xbr =
1000sin( I ) (5.4)

SR × σ b
xbd = (5.5)
1000

One thing to note about these values is that they are similar to the xbr and xbd
calculated for the single weapon cases, but they are without the factor of 0.6745.
For sticks, we do not use the ballistic dispersion as an error probable because
we do not root sum square the values with the aiming error probables. Therefore,
we keep the ballistic dispersion in terms of standard deviations. Once we have
xbr and xbd we can find LB and WB through equations (5.6) and (5.7).

LB = LET 2 + 8 xbr 2 (5.6)

WB = WET 2 + 8 xbd 2 (5.7)

58
With LB and WB, along with the user inputted values of stick length (LS) and stick
width (WS), we can define the pattern length (LP) and width (WP). This is shown
in equations (5.8) and (5.9).

LP = LS + LB (5.8)

WP = WS + WB (5.9)

Since we have increased the effective area of the weapon with the ballistic
dispersion, we need to change the conditional probability of damage inside the
rectangle LB by WB in order to preserve the lethality of the weapon. The new
conditional probability of damage, PCD1, is a fraction of the original probability of
damage, PCD. This is found through equation (5.10).

LET WET
PCD1 = PCD (5.10)
LBWB

A graphical representation of how the old JMEM method enlarges the area of
effectiveness of each weapon is shown in Figure 5.2.
WB
WET

A=AB
A=AET LET PCD1 LB
PCD

Before ballistic
dispersion After ballistic
dispersion

Figure 5.2: Modification of weapon AET due to ballistic dispersion.

(From Ref. [1])
59
 The next step in the old JMEM method for sticks is to account for different
overlap scenarios of the stick weapons’ effective areas. For the overlap in the
deflection direction, we define a degree of overlap in the deflection direction, or
nod, which is calculated from the stick width (WS), pattern width (WP), and the
number of weapons released per intervalometer pulse (np). This is shown in
equation (5.11).

n pWB
nod = (5.11)
WP

In Figure 5.3 we show how weapons may overlap (left) or may not overlap (right)
in the deflection direction.

WB WB

WP WP

Figure 5.3: Weapons overlapping (left) and not overlapping (right) in the
deflection direction. (From Ref. [1])

This nod term can also be defined as what percentage of the np released
weapons can fit into the stick pattern width. We can tell a great deal from the
degree of overlap. If nod >1, we know there is overlap between the stick
weapons in the deflection direction. If this is the case, we find the conditional
probability of damage in the deflection direction, or PCD/d, by using the survival
rule and powering up the single weapon conditional damage PCD1 by the degree
of overlap. This is shown in equation (5.12).

60
 PCD / d = 1 − (1 − PCD1 ) nod (5.12)

However, if nod <1, we have no overlap in the deflection direction. In this case
PCD/d is found by equation (5.13).

WB
PCD / d = n p PCD1 (5.13)
WP

This analysis in the deflection direction results in a rectangle that combines

weapons in the deflection direction that has a damage function equal to PCD/d.
This rectangle is shown in Figure 5.4.

PCD/d LB

WP

Figure 5.4: Combination of weapons in deflection direction. (From Ref. [1])

A similar analysis is performed for overlap in the range direction. Rather

than using np to determine the conditional probability, we use the number of
intervalometer pulses, or nr. We define the degree of overlap in the range
direction in equation (5.14).

nr LB
nor = (5.14)
LP

61
In Figure 5.5 we show how weapons may overlap (left) or may not overlap (right)
in the deflection direction.

LB
LP
LP LB

Figure 5.5: Weapons overlapping (left) and not overlapping (right) in the
range direction. (From Ref. [1])

If nor >1, we know there is overlap between the stick weapons in the range
direction. For this is the case, we find the conditional probability of damage for
the whole pattern, or PCDS, by powering up PCD/d by the degree of overlap in the
range direction. This is shown in equation (5.15).

PCDS = 1 − (1 − PCD / d ) nor (5.15)

However, if nor <1, we have no overlap in the range direction, and PCDS is found
by equation (5.16).

LB
PCDS = nr PCD / d (5.16)
LP

Now we have taken the stick of weapons and represented the stick as a single
effectiveness area rectangle with dimensions LP and WP, with a conditional
probability of damage of PCDS as shown in Figure 5.6.

62
 LP PCDS

WP

Figure 5.6: Stick pattern dimension and damage function. (From Ref. [1])

The pattern shown in Figure 5.6 is now considered a single effectiveness

area. The old JMEM treats this case similarly to the single weapon versus area
targets discussed in Chapter IV. The fractional coverage analysis is identical, but
the effective pattern length and width (LEP and WEP) is found by taking the
maximum of the stick pattern dimensions and the target size. The target size has
dimensions of La in the range direction, and Wa in the deflection direction. The
effective pattern dimensions are calculated with equations (5.17) and (5.18).

LEP = MAX ( LP , La ) (5.17)

WEP = MAX (WP , Wa ) (5.18)

The following is a summary of the equations used to find the expected

fractional coverage of a single effectiveness area on a target area. For a detailed
explanation of the equations used to find fractional coverage and fractional
damage, please refer to Chapter IV Section A.

63
Range Direction

LEP + LA
s= (5.19)
2
LEP − LA
t= (5.20)
2
E ( FR ) = {I1 + I 2 + I 3 + I 4 − I 5 } (5.21)

I1 = [normcdf (t ) − normcdf (−t )] (5.22)

LEP + LA
I2 = [normcdf (−t ) − normcdf (− s )] (5.23)
2 LA
LEP + LA
I3 = [normcdf ( s ) − normcdf (t )] (5.24)
2 LA
2 2
 s   t 
2σ x −  
σx 2 
−  
σx 2 
I 4 − I5 = [e −e ] (5.25)
LA 2π

Deflection Direction

WEP + WA
s= (5.26)
2
WEP − WA
t= (5.27)
2
E ( FD ) = {I1 + I 2 + I 3 + I 4 − I 5 } (5.28)

I1 = [normcdf (t ) − normcdf (−t )] (5.29)

WEP + WA
I2 = [normcdf (−t ) − normcdf (− s )] (5.30)
2WA
WEP + WA
I3 = [normcdf ( s ) − normcdf (t )] (5.31)
2WA
2 2
 s   t 
2σ y − 
σy 2 
− 
σy 2 
I 4 − I5 = [e  
−e  
] (5.32)
WA 2π
64
Combined Range and Deflection

E ( FC ) = E ( FR ) × E ( FD ) (5.33)

After finding the expected fractional coverage, or E(FC), we can combined

what we’ve calculated, as well as our knowledge of the weapon reliability (R), to
find the expected fractional damage (EFD) of the stick against an area target.
This is shown in equation (5.34).

 LW 
EFD = E ( FC ) ×  P P  × R × PCDS (5.34)
 LEPWEP 

B. CURRENT JMEM METHOD

The current JMEM method for computing the effectiveness of sticks of

weapons is similar to that of the JMEM method it replaced. The difference
between the methods is the handling of the possible overlap conditions of the
lethal areas of each stick weapon.
Computing the lethal area of each stick weapon in the JMEM method
begins with the handling of ballistic dispersion effects on the individual weapons
of the stick. This is done by enlarging the effective lethal area of each weapon to
account for ballistic dispersion. The enlarged dimensions are defined as LB and
WB, and are calculated with same formulas that were used for the old JMEM
method. The formulas are restated in equations (5.35) and (5.36).

LB = LET 2 + 8 xbr 2 (5.35)

WB = WET 2 + 8 xbd 2 (5.36)

65
The updated JMEM method then calculates the possible overlap conditions for
the lethal enlarged effective areas of the weapons in the stick upon each other.
An example of the possible overlap conditions of the lethal weapon areas is
shown in Figure 5.7.

Figure 5.7: Example of weapon lethal areas overlapping in stick. (From Ref. [2])

The JMEM then calculates the conditional probability of damage in each

area shown in Figure 5.7. To find the conditional probability of damage for each
of the overlapping areas we implement the survival rule, which is given in
equation (5.37). This equation gives the PCD of the overlap area, based on the
conditional probability of damage for each weapon lethal area (PCD/weapon) and the
number of weapons that overlap the specific area.

PCD = 1 − (1 − PCD / weapon ) n (5.37)

66
Once these different damage probabilities for each overlap area have been
calculated, the JMEM performs a fractional coverage analysis of the stick against
the area target.
The process of calculating the conditional probability of damage for each
overlap area, as well as the detailed fractional coverage analysis is explain in
great detail in Reference 2.

C. MONTE CARLO SIMULATIONS

For the case of a stick of weapons versus area targets, we run two Monte
Carlo simulations to determine the probability of damaging the target. The first
Monte Carlo simulation uses the Carleton damage function to determine the
damage caused to the area target. The second Monte Carlo simulation uses a
rectangular cookie cutter approximation of the Carleton damage function to find
the damage to an area target.
As before with single weapons, delivery accuracy plays a large role in
determining the probability of damaging a target. In the case of sticks, we cannot
directly combine the aiming error and ballistic dispersion error. For sticks, the
aiming error affects the accuracy of the stick of weapons as a whole. The
ballistic dispersion, on the other hand, affects the impact point of each individual
weapon in the stick. Figure 5.8 illustrates two situations where a stick of 4
weapons is aimed at the aim point in the center. The aiming accuracy of the each
stick is different, so they do not land in the same location. As can be seen in
Figure 5.8, the stick and pattern dimensions no not change, only the location of
the stick.

67
 Stick 2 LP
LS

WS

Stick 1
LS WP
LP

WS Aim Point

WP

Figure 5.8: Effects of aiming error on a stick of weapons.

Where the aiming error affects the location of the entire stick, ballistic dispersion
errors change the pattern length (LP) and width (WP) of the stick. As described
before, these dimensions create the smallest rectangle that encloses all of the
lethal areas of each weapon. The ballistic dispersion changes the pattern
dimensions by affecting the individual impact point of each weapon. This is
shown in Figure 5.9.

68
 LP
LP

WP

WP

Stick Before Stick with added

Ballistic Dispersion Ballistic Dispersion

Figure 5.9: Effects of ballistic dispersion error on a stick of weapons.

We can see in Figure 5.9 how the ballistic dispersion affects the impact point of
each weapon, thus changing the pattern dimensions of the stick.
The Monte Carlo simulations run in this thesis are for a stick of four
weapons that has a stick length of 100 ft and a stick width of 100 ft. Figure 5.10
gives a visual representation of this by placing a Cartesian coordinate system on
the ground, and centering the area target about the origin (shown as a dashed
rectangle). With no aiming error or ballistic dispersion present in the problem,
the stick would land as shown in Figure 5.10. We can see that the weapons are
evenly spaced at a distance of 100 ft in the stick.

69
 Range

(50,50)
(-50,50)

(0,0) Deflection

(-50,-50) (50,-50)

Figure 5.10: Monte Carlo stick simulations scenario.

The first step in the Monte Carlo simulations is to find the standard
deviation of the aiming error and ballistic dispersion in the ground plane. This is
calculated from the user inputted values of the range error probable (REP),
deflection (DEP), slant range (SR), and impact angle (I). This is shown in
equations (5.38), (5.39), (5.40), and (5.41).

REP
σ aim _ range = (5.38)
0.6745
DEP
σ aim _ deflection = (5.39)
0.6745
SR × σ b
xbr =
1000sin( I ) (5.40)

SR × σ b
xbd = (5.41)
1000

70
All four of the delivery accuracy standard deviations (σaim_range, σaim_deflection, xbr,
and xbd) are then multiplied by a random number from a normal distribution. This
normal distribution has a mean value of zero and a standard deviation of one.
Using this normal distribution allows us to assume that the average probability of
damage for a large number of stick trials will equal the expected value of the
damage function at the desired aim point.
With the ballistic dispersion error affecting each weapon individually, and
the aiming error affecting the stick as a whole, we are able to find the location of
each individual weapon impact point. At each impact point, we represent the
weapon either by a lethal area matrix filled by the Carleton damage function or a
rectangular cookie cutter, depending on the simulation.
The final step of the Monte Carlo simulations is to determine the
probability of damage to the target from each of the four stick weapons. This is
accomplished by dividing the 50 ft by 50 ft target into 1ft by 1ft cells, and using
the survival rule to find the probability of damage in each target cell. Figure 5.11
shows an example of how we combine the probabilities of damage from each of
the four weapons in a stick on a specific cell in the target area. The probability of
damage contributions from the four weapons are defined as Pd1, Pd2, Pd3, and
Pd4.

Impact Point 1 Impact Point 3

Pd1

Pd3

Pd2 Target Area Cells

Pd4
Impact Point 2 Impact Point 4

Figure 5.11: Damage contributions of stick weapons.

71
To calculate the probability of damage due to the contributions from each of the
four weapons we use the survival rule. The survival rule is a formula that
calculates the probability of damaging a target given multiple damage
contributions from the four weapons. This is shown in equation (5.42).

Pd cell = 1- (1- Pd1 ) *(1- Pd 2 ) *(1- Pd3 ) *(1- Pd 4 ) (5.42)

We repeat this survival rule on each of the j number of cells in the target matrix.
To find the overall probability of damage for a given Monte Carlo iteration i, we
sum all of the Pd/cell, and divide it by the total number of cells j. This is shown in
equation (5.43).

∑P
j =1
d / cell ( j )

Pd / i = (5.43)
j

For the full Monte Carlo simulation, we repeat this for n iterations, each
individual iteration (i) resulting in its own probability of damage Pd/i. The value n
is the total number of iterations used in the Monte Carlo simulation. The final
probability of damage for the Monte Carlo simulation is the average of all of the
Pd/I shown in equation (5.44).

∑P d /i
Pd = i =1
(5.44)
n

72
 1. Modeling Weapons with Carleton Damage Function

The first Monte Carlo simulation utilizes the Carleton damage function to
determine the amount of damage done to a target element given the distance
from the element to the weapon impact points in a stick of weapons. In this
simulation, we will assume that each of the weapons in the stick is identical in
their effectiveness. We determine this weapon effectiveness in terms of the
weapon radii WRr and WRd. The weapon radii are found from the user inputted
values of MAEF, impact angle I, and the slant range SR. This is shown in
equations (5.45), (5.46), and (5.47).

WRr
a= = MAX [1 − 0.8cos( I ), 0.3] (5.45)
WRd

a
WRr = MAEF × (5.46)
π
WRr
WRd = (5.47)
a

These values of weapon radii stay constant through each iteration of the Monte
Carlo simulation. The values that change with each iteration will be the individual
weapon impact points, as discussed before.
Unlike the single weapon cases against unitary and area targets, there is
not a lethal area matrix used in the Monte Carlo simulation for sticks using the
Carleton damage function. Instead, we determine the location of each weapon
impact point in the stick and determine the distance from the impact points to the
target area. To do this, we divide the target area into 1ft by 1ft cells, and find the
distance from each these cells to the impact points. The Carleton damage
function is given in equation (5.48).

 x2 y2 
Pd ( x, y ) = exp −  2
+ 2
(5.48)
 WRr WRd 
73
We have already calculated the weapon radii, and we define the x and y terms in
equation (5.48) as the distance in range and deflection respectively from a
weapon impact point to a given cell in the area target. For a given cell in the
area target, we will get a Pd from the Carleton damage function, for each of the
four weapons in the stick. For example, let us say have a scenario where the
weapon radii are both 10 ft (WRr =10 and WRd = 10). We will also say we are
looking at a target cell 3 ft from the impact point of the first weapon (in a stick of
four) in the range direction, and 5 ft in the deflection direction. We can find the
probability of damage due to this weapon on that target cell as shown in equation
(5.49).

 32 52 
Pd1 = exp −  2 + 2  = 0.7118 (5.49)
10 10 

This value of 0.7118 will be called Pd1, since it is the probability of damage
caused by the first weapon in the stick. We will assume the other three weapons
in the stick cause probabilities of damage equal to 0.5312, 0.3041, and 0.7590
respectively. Now that we have Pd1, Pd2, Pd3, and Pd4, we use the survival rule
explain in equatrion (5.42)to find the overall probability of damage caused by the
example stick. The overall Pd of the cell for this example is shown in equation
(5.50).

Pd cell = 1- (1- 0.7118) *(1- 0.5312) *(1- 0.3041) *(1- 0.7590) = 0.9773 (5.50)

The overall Pd of the target cell is 0.9773. This process is then repeated for all of
the cells in the target area. To finish the iteration, the Pd’s of each of the area
target cells are averaged to find a total probability of damaging the area target.

74
 For the full Monte Carlo simulation, we would repeat this process for n
iterations, each individual iteration (i) resulting in its own probability of damage
Pd/i. The value n is the total number of iterations used in the Monte Carlo
simulation. The final probability of damage for the Monte Carlo simulation is the
average of all of the Pd/I shown in equation (5.51).

∑P d /i
Pd = i =1
(5.51)
n

The number of iterations (n) used in the Monte Carlo simulation used to
compare the Monte Carlo and JMEM methods for sticks was set to one hundred
thousand. In running the Monte Carlo simulation multiple times, we ensure that
the resulting Pd from each simulation did not change. This shows that the
simulation was converging to a single value, which we would expect. After
testing different values of n, it was found that one hundred thousand provides for
this convergence.
A Monte Carlo simulation modeling a stick of weapons using with the
Carleton damage function is presented in Appendix B, Option 1.

2. Modeling Weapons with Rectangular Cookie Cutter

Approximation

The second Monte Carlo simulation utilizes a rectangular cookie cutter

approximation of the Carleton damage function to determine the amount of
damage done to a target element given the distance from the element to the
weapon impact points in a stick of weapons. We again assume that each
weapon in the stick has identical lethality. We define this lethality as we have
before with equations (5.52), (5.53), and (5.54).

WRr
a= = MAX [1 − 0.8cos( I ), 0.3] (5.52)
WRd
75
 LET = MAEF × a (5.53)

WET = LET / a (5.54)

The calculation of the probability of damage against the area target is very
similar to the Monte Carlo simulation using the Carleton damage function. We
first determine the location of each weapon impact point in the stick and
determine the distance from the impact points to the target area. To do this, we
divide the target area into 1ft by 1ft cells, and find the distance from each these
cells to the impact points. At each of the weapon impact points, a rectangular
cookie cutter with dimensions LET and WET is placed to represent the weapon.
An example iteration of a Monte Carlo simulation using rectangular cookie
cutter representations of the weapons is shown in Figure 5.12. In this example
we have four weapons, each represented by rectangular cookie cutters of
dimensions LET and WET= 3ft. The stick of weapons dropped against a 3ft by 3ft
area target which is shaded in light grey. Due to the assumed delivery accuracy
errors, the stick has not fallen centered upon the area target.

76
 Aim
Deflection Point
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
0.0 1.0 1.0 1.0 0.0 0.0 1.0 1.0 1.0 0.0
0.0 1.0 1.0 1.0 0.0 0.0 1.0 1.0 1.0 0.0
0.0 1.0 1.0 1.0 0.0 0.0 1.0 1.0 1.0 0.0
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0

Range
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
0.0 1.0 1.0 1.0 0.0 0.0 1.0 1.0 1.0 0.0
0.0 1.0 1.0 1.0 0.0 0.0 1.0 1.0 1.0 0.0
0.0 1.0 1.0 1.0 0.0 0.0 1.0 1.0 1.0 0.0
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0

Figure 5.12: Example iteration of Monte Carlo stick simulation using

rectangular cookie cutter lethal area matrices against an
area target.

Unlike the Monte Carlo simulation using the Carleton damage function, for the
rectangular cookie cutter Monte Carlo simulation we do not need to use the
survival rule to find the probability of damage caused by the stick. Since each
weapon is represented by a rectangular cookie cutter, the Pd inside each weapon
effective area is equal to one. Therefore, each 1 ft by 1 ft cell of the area target
is either completely damaged by the stick (Pdcell= 1) or not at all (Pdcell=0). For
the example shown in Figure 5.12 we see that the top left hand cell, which is
shaded in dark grey, is completely damaged by the top left hand weapon of the
stick. This Pdcell would equal 1 whether or not another weapon of the stick
contributes damage to the cell. The probability of damaging the whole area
target for this iteration, or Pd/i, is equal to the average of all the Pd/cell of each cell
in the target area. For the example iteration above we find that the Pd/i is shown
in equation (5.55)

1.0 + 0.0 + 0.0 + 1.0 + 0.0 + 0.0 + 0.0 + 0.0 + 0.0

Pd / i = = 0.2222 (5.55)
9

77
 For the full Monte Carlo simulation, we would repeat this process for n
iterations, each individual iteration (i) resulting in its own probability of damage
Pd/i. The value n is the total number of iterations used in the Monte Carlo
simulation. The final probability of damage for the Monte Carlo simulation is the
average of all of the Pd/I shown in equation (5.56).

∑P d /i
Pd = i =1
(5.56)
n

The number of iterations (n) used in the Monte Carlo simulation used to
compare the Monte Carlo and JMEM methods for sticks was set to one hundred
thousand. In running the Monte Carlo simulation multiple times, we ensure that
the resulting Pd from each simulation did not change. This shows that the
simulation was converging to a single value, which we would expect. After
testing different values of n, it was found that one hundred thousand provides for
this convergence.
A Monte Carlo simulation with a stick modeled as rectangular cookie
cutter approximations of the Carleton damage function is presented in Appendix
B, Option 2.

D. RESULTS OF STICK COMPARISONS

The comparisons done this section are to find the differences between
weapon effectiveness methodologies for the case of a stick of weapons against
area targets. In order to get a wide array of results each comparison was done
for different user inputted values of MAEF, REP & DEP and ballistic dispersion in
mils, in the normal plane. The weapon impact angle and release slant range
were kept at a constant 45° and 10,000 ft respectively for each comparison. The

78
size of the area target used for all comparisons was a 50 ft by 50 ft area. Each
test case dealt with a stick of four weapons with a stick length and width equal to
100 ft.

1. Updated JMEM Method versus Old JMEM Method

The first comparison completed was the updated JMEM method results for
computing damage probabilities of sticks of weapons versus area targets against
the old JMEM method. We would expect these methods to produce different
results due to the way each method calculates the possible overlap scenarios for
weapons in the stick. Tables 5.1 and 5.2 present the results for the PK found for
both the JMEM methods. The shaded values in the Tables are obtained from the
new JMEM method, while the unshaded values are those of the old JMEM
method.

MAEF=5000ft2 Ballistic Dispersion (mils in normal plane)

0 3 5
REP=25ft & DEP=50ft 0.4933 0.3099 0.2093
0.3636 0.3422 0.2877

REP=25ft & DEP=25ft 0.589 0.35 0.2241

0.4334 0.4159 0.3522

REP=50ft & DEP=25ft 0.421 0.3057 0.2144

0.3888 0.3428 0.2928

REP=50ft & DEP=50ft 0.3526 0.2707 0.2003

0.3261 0.2813 0.2385

Table 5.1: Results of new JMEM and old JMEM for weapons of
MAEF=5000ft2

79
 MAEF=1000ft2 Ballistic Dispersion (mils in normal plane)
0 3 5
REP=25ft & DEP=50ft 0.1311 0.0732 0.0475
0.0773 0.08 0.0674

REP=25ft & DEP=25ft 0.1826 0.0893 0.0527

0.0863 0.0966 0.0845

REP=50ft & DEP=25ft

0.1208 0.077 0.0502
0.0797 0.0781 0.0687

REP=50ft & DEP=50ft

0.0867 0.0631 0.0453
0.0714 0.0647 0.0547

Table 5.2: Results of new JMEM and old JMEM for weapons of
MAEF=1000ft2

It can be seen from Tables 5.1 and 5.2 that the damage results do not
match, as was expected from the different methods. We see that the new JMEM
method is generally more conservative in its damage estimates for cases of zero
ballistic dispersion than the JMEM method it replaced. For the cases involving
ballistic dispersion, the new JMEM gives larger damage estimates. We would
assume that the ballistic dispersion is the cause of the differences in damage
estimates, but both JMEM methods treat the ballistic dispersion in the same way;
by enlarging the effective target area of the weapon. This leaves the difference
in modeling the overlap conditions of the weapons in the stick. The older JMEM
method defines degrees of overlap of stick weapons in the deflection and range
directions. The new JMEM method calculates conditional probabilities of
damage in each of the areas where the weapon effectiveness rectangles overlap,
as discussed in detail in Reference 2.

80
 2. JMEM versus Carleton Damage Function

The next comparison done was between the JMEM method results for
computing damage probabilities of sticks of weapons versus area targets and a
Monte Carlo using the Carleton damage function to calculate damage
probabilities. The JMEM and Monte Carlo simulation are expected to yield
different damage probabilities. This is due to a difference in incorporating
weapon ballistic dispersion and the calculation of damage itself. Tables 5.3 and
5.4 present the results for the PK found for both the JMEM method and Monte
Carlo simulation. The shaded values in the Tables are obtained from the JMEM
method, while the unshaded values are those of the Monte Carlo simulation.

MAEF=5000ft2 Ballistic Dispersion (mils in normal plane)

0 3 5
REP=25ft & DEP=50ft 0.3443 0.2992 0.241
0.3636 0.3422 0.2877

REP=25ft & DEP=25ft 0.4097 0.3569 0.2838

0.4334 0.4159 0.3522

REP=50ft & DEP=25ft 0.3373 0.2966 0.2448

0.3888 0.3428 0.2928

REP=50ft & DEP=50ft 0.2827 0.249 0.2085

0.3261 0.2813 0.2385

Table 5.3: Results of JMEM and Monte Carlo for MAEF=5000ft2

81
 MAEF=1000ft2 Ballistic Dispersion (mils in normal plane)
0 3 5
REP=25ft & DEP=50ft 0.0803 0.0711 0.0563
0.0773 0.08 0.0674

REP=25ft & DEP=25ft 0.0971 0.0861 0.0673

0.0863 0.0966 0.0845

REP=50ft & DEP=25ft

0.0804 0.0704 0.0572
0.0797 0.0781 0.0687

REP=50ft & DEP=50ft

0.0666 0.0581 0.0479
0.0714 0.0647 0.0547

Table 5.4: Results of JMEM and Monte Carlo for MAEF=1000ft2

It can be seen in Tables 5.3 and 5.4 that the probability of damage results
for the two methodologies does not match. This was expected due to the
significant difference between the JMEM and Monte Carlo simulation methods.
When looking at the results, a majority of the comparisons show that the JMEM
method is more conservative than the Monte Carlo simulation. The few times
where the JMEM results are greater than the Monte Carlo simulation are cases
with a MAEF = 1000 ft2. This leads us to believe that the differences in the
dealing with the lethality of the stick weapons in the two methods leads to
differences in the damage caused by those weapons. This is intuitive because
the JMEM method increases each weapons lethal area due to the ballistic
dispersion, while the Monte Carlo simulation obtains the damage directly from
the Carleton damage function. The Monte Carlo simulation deals with ballistic
dispersion by moving the individual weapon impact points of the stick.
The calculation of the damage to each 1 ft by 1 ft cell of the area target is
also handled differently in the JMEM method and Monte Carlo simulation. As
discussed earlier, the JMEM method takes each impact point of the stick
82
weapons, and places an effective lethal area on each impact point to represent
the weapon lethality. The method then calculates the conditional probability of
each lethal area, and each overlap area where the lethal areas of two or more
stick weapons have overlapped. The JMEM then uses fractional coverage
analysis to find the damage done to the whole area target. The Monte Carlo
simulation represents each weapon with a Carleton damage function. The fact
that the Monte Carlo simulation uses the Carleton damage function directly leads
to assumption that the Monte Carlo is more accurate than the JMEM method,
which uses approximations of the Carleton damage function.

3. JMEM versus Rectangular Cookie Cutter Approximation

The last comparison for sticks against area targets is between the JMEM
method and a Monte Carlo simulation which represents the stick weapons
rectangular cookie cutters. Again, we do not expect the same results for both of
these models. We would expect, however, that the Monte Carlo simulation using
rectangular cookie cutters to be closer to the JMEM results than the Monte Carlo
simulation which used the Carleton damage function directly. This is because
both the JMEM and the Monte Carlo simulation using the rectangular cookie
cutter make approximations of the Carleton damage function in order to find the
probability of damaging the area target. Tables 5.5 and 5.6 present the results
for the PK found for both the JMEM method and Monte Carlo simulation. The
shaded values in the Tables are obtained from the JMEM method, while the
unshaded values are those of the Monte Carlo simulation.

83
 MAEF=5000ft2 Ballistic Dispersion (mils in normal plane)
0 3 5
REP=25ft & DEP=50ft 0.3754 0.3129 0.2491
0.3636 0.3422 0.2877

REP=25ft & DEP=25ft 0.4527 0.3754 0.2944

0.4334 0.4159 0.3522

REP=50ft & DEP=25ft 0.372 0.3148 0.2547

0.3888 0.3428 0.2928

REP=50ft & DEP=50ft 0.3077 0.2599 0.2152

0.3261 0.2813 0.2385

Table 5.5: Results of JMEM and Monte Carlo for MAEF=5000ft2

MAEF=1000ft2 Ballistic Dispersion (mils in normal plane)

0 3 5
REP=25ft & DEP=50ft 0.0797 0.0706 0.0566
0.0773 0.08 0.0674

REP=25ft & DEP=25ft 0.0962 0.0863 0.0676

0.0863 0.0966 0.0845

REP=50ft & DEP=25ft 0.0804 0.0713 0.0576

0.0797 0.0781 0.0687

REP=50ft & DEP=50ft 0.066 0.0582 0.0477

0.0714 0.0647 0.0547

Table 5.6: Results of JMEM and Monte Carlo for MAEF=1000ft2

The trends shown in Tables 5.5 and 5.6 are very similar to those
presented for the other Monte Carlo versus JMEM comparison for stick weapons.
But, as expected, the differences in this comparison are smaller than they were
for the previous comparison between the JMEM method and the Monte Carlo
simulation using the Carleton damage function. From the results, it seems that
since the compared methods shown in Tables 5.5 and 5.6 both use
84
approximations of the Carleton damage function to represent the stick weapons’
lethality, they would give similar results. Since they results are not the same, we
can assume that the differences in the results stem from the treatment of ballistic
dispersion and aiming errors. As discussed earlier, the JMEM uses the ballistic
dispersion to enlarge the effective lethal area of each weapon in the stick. The
JMEM method then does a fractional coverage analysis on the weapon lethal
areas, as well as any areas where the weapon lethal areas overlap. The Monte
Carlo simulations use the delivery accuracy errors to place the actual impact
point of each weapon in the stick. The ballistic dispersion error affects each
weapon individually, while the aiming error moves the location of the entire stick.
We assume that the difference between the damage results of the JMEM and
Monte Carlo methods are due to these large differences in methodologies.

85
THIS PAGE INTENTIONALLY LEFT BLANK

86
 VI. SUMMARY OF RESULTS

This thesis successfully compared and contrasted different methodologies

in determining weapon effectiveness for three different attack scenarios. These
scenarios were a single weapon dropped against a unitary target, a single
weapon against an area target, and a stick of weapons against an area target.
For each weapon/target interaction, we tested an array of ballistic dispersion and
aiming errors. A summary of the method comparisons is detailed below.

• The first set of comparisons completed was for the case of a single
weapon dropped against a unitary target. The three methodologies
compared for this weapon/ target interaction were two Monte Carlo
simulations against the results obtained from the method utilized by the
Joint Munitions Effectiveness Manuals (JMEM). The first Monte Carlo
simulation represented the single weapon as a lethal area matrix with
probabilities of damage obtained from the Carleton damage function.
When comparing the damage results from this Monte Carlo simulation and
the JMEM method results, we found that a majority of the comparisons
showed that the JMEM method gave a conservative estimate of damage.
For cases of zero ballistic dispersion, however, we saw that the Monte
Carlo and JMEM produced near the same result for damage against a
unitary target. The second Monte Carlo simulation represented the single
weapon as a rectangular cookie cutter approximation of the Carleton
damage function. The result of the comparison between the rectangular
cookie cutter Monte Carlo simulation and the JMEM is similar to that of the
previous Monte Carlo simulation in that the JMEM damage result was
consistently more conservative than the Monte Carlo method. However,
for this comparison, the damage results were not similar for cases of zero
ballistic dispersion.

87
• The next set of comparisons completed was for the case of a single
weapon dropped against an area target. As in the previous case for a
single weapon against a unitary target, the JMEM method is compared to
two Monte Carlo simulations. The Monte Carlo simulation that
represented the weapon as a lethal area matrix populated with values
from the Carleton damage function, when compared to the JMEM method,
resulted in higher damage values for all non-zero values of ballistic
dispersion. The zero ballistic dispersion error cases showed that the
Monte Carlo simulation results were generally more conservative than
those of the JMEM. Unlike the previous comparison for unitary targets,
the JMEM and Monte Carlo simulation did not produce similar results.
The same observations were made for the comparison between the
rectangular cookie cutter Monte Carlo simulation and the JMEM method.
The JMEM produced a more conservative damage estimate for all non-
zero ballistic dispersion cases, but a less conservative estimate for the
zero ballistic dispersion cases.

• The last set of comparisons was for the case of a stick of four weapons
released against an area target. There were four weapon effectiveness
methodologies compared in this section of the thesis. These methods
were the old JMEM method, the new JMEM method, a Monte Carlo
simulation using the Carleton damage function, and a Monte Carlo
simulation using the rectangular cookie cutter damage function. Both
Monte Carlo simulations, when compared to the new JMEM method,
consistently gave smaller damage estimates of the stick on the area target.
The only times the Monte Carlo simulations gave higher damage
estimates were for some zero ballistic dispersion cases. When the two
JMEM methods were compared, the older JMEM gave damage results
that were significantly lower than those produced by the new JMEM
method for cases on non-zero ballistic dispersion. For cases of zero
88
 ballistic dispersion we saw that the damage results from the old JMEM
method were appreciably greater than the new JMEM results.

• The number of iterations needed for the Monte Carlo simulations, as well
as the computational time it took to run the simulations, was documented.
The computer platform used to run all Monte Carlo simulations was the
Matlab program running on an Intel Pentium 4 processor at 2.4 GHz. For
the single weapon versus unitary target scenario, both Monte Carlo
simulations required 10,000,000 iterations, and averaged a computing
time of 10 minutes. The single weapon versus area target simulations
also required 10,000,000 iterations, but averaged a computing time of 15
minutes. Lastly, the stick versus area target Monte Carlo simulations
needed 100,000 iterations and averaged a computing time of 20 minutes.

• The Matlab program was initially used because it was convenient at the
time. It is understood that other program languages, such as C or
FORTRAN, are better suited for high speed calculations, and may have
reduced computing time.

As summarized above, each comparison showed that the weapon

effectiveness models for different weapon/target interactions differed in their
damage estimates. These differences were expected for all of the comparisons
due to the different approximations and assumptions made with in each of the
methods’ damage calculations.

89
THIS PAGE INTENTIONALLY LEFT BLANK

90
 VII. CONCLUSIONS

The results of the comparisons run in this thesis have lead to the
formulation of key conclusions. These conclusions help to give an overview of
the findings of the comparisons between the JMEM methods and Monte Carlo
simulations.

• The Monte Carlo simulations used to find the probability of damage for
the single weapon against unitary and area targets needed ten million
iterations to converge to a single answer. For sticks, these simulations
required one hundred thousand iterations. These are a significant
number of iterations, and may not be feasible to utilize in the
operational arena due to slow computing speeds.

• The damage probability of the single weapon versus unitary and area
target cases, calculated using the Monte Carlo simulations, were
consistently greater than the JMEM method results for cases of non-
zero ballistic dispersion. As the ballistic dispersion increased, the
difference between the Monte Carlo and JMEM result increased, with
the Monte Carlo result always being larger.

• For the single weapon versus a unitary target scenario with zero
ballistic dispersion, the results from the Monte Carlo simulation
representing the weapon as a lethal area matrix filled with values from
the Carleton damage function matched the JMEM results within 1%.
This was expected since the JMEM uses the Carleton damage function
to determine the probability of damage, as does the Monte Carlo
simulation.

91
• For the single weapon versus an area target scenario with zero ballistic
dispersion, the results from the Monte Carlo simulation representing
the weapon as a rectangular cookie cutter approximation of the
Carleton damage function matched the JMEM results within 5%. This
closeness was expected because the JMEM method for a single
weapon versus an area target utilizes a rectangular cookie cutter
approach to solving for weapon damage.

• The damage probability of the stick cases calculated by the Monte

Carlo simulations was consistently smaller than the JMEM method
results for cases of non-zero ballistic dispersion. It is interesting to
note that the behavior of the results is opposite that of the single
weapon cases. As the ballistic dispersion increased for the stick
weapons, the resulting difference between the Monte Carlo and JMEM
methods stayed generally constant, with the Monte Carlo result always
being smaller than the JMEM result.

• The comparisons done between the new and old JMEM methods for
sticks show that the new JMEM is a great deal more conservative than
the old JMEM for zero ballistic dispersion cases. This may be the
reason for the old JMEM being replaced.

By analyzing the data obtained in this thesis we have been able to

make conclusions about the JMEM methods and Monte Carlo simulations.
The goal of the Monte Carlo simulations was to give a more realistic way
of modeling the weapon/target interaction. By running these Monte Carlo
simulations, we have insight into the degree of the approximations used in
the JMEM methods.

92
 VIII. RECOMMENDATIONS AND FUTURE WORK

The goal of this thesis was to compare current weapon effectiveness

methods with Monte Carlo simulations to determine whether the current JMEM
methods should be replaced. This analysis included comparing damage results
produced from the methods, as well as computational time needed to run the
methods. By analyzing the data produced by the different methods, it is
concluded that the Monte Carlo simulations better represent what actually occurs
in the weapon target interaction than the current JMEM techniques. This is
probably due to the amount of approximations that the JMEM incorporates into
its methods. It is believed that by running the Monte Carlo simulation without
these approximations of the Carleton damage function, we obtain damage
estimates not tainted by approximating errors in the method. Due to this, the
Monte Carlo simulation should be eventually integrated by the military in weapon
effectiveness calculations. Though the Monte Carlo simulations are a higher
fidelity modeling technique, it is not practical for the military to implement these
methods as this time. This is due to the large number of iterations needed to
obtain a correct result from the Monte Carlo simulations. These required
iterations cannot be completed fast enough with current computing speeds to
provide a quick reliable prediction of damage in the battlefield. However, while
computing speeds keep increasing in the future, Monte Carlo simulations will
become a practical alternative to the current JMEM methods.
The data collected for the JMEM and Monte Carlo simulations in this
thesis were limited in time. Listed below are suggestions for follow on work in the
topic of weapon effectiveness techniques.

• As of now, the methods all require user inputs of delivery accuracy and
weapon lethality. It would be practical to include weapon release
conditions from the aircraft as user inputs as well. The user could choose

93
 a weapon release mechanisms which would include weapon trajectory
analysis to determine the delivery accuracy of the weapon or stick.

• To test the trends of the data obtained through the comparisons of the
JMEM methods and Monte Carlo simulations, a wider array of weapon
lethalities should be tested (different user inputted MAEF values) for a
greater number of delivery accuracy combinations.

• The Matlab codes provided in Appendices A and B were not fully

debugged and written in the most effective of manners due to time
constraints on this thesis. By rewriting the Monte Carlo simulation codes
for better performance, the speed by which a damage result is calculated
may decrease.

• In order to have all methods coded in Matlab, the new JMEM method for
sticks of weapons versus area targets should be coded along the lines of
code given in Appendix B. This would enable all of the stick methods
detailed in this thesis to be accessed by one program.

• The stick Matlab code provided in Appendix B should be written to enable

the user to input the number of stick weapon elements are being dropped,
This could also implement weapon release conditions so that the Matlab
stick program could cover all stick configurations, rather than being hard
coded for a four weapon stick with defined weapon impact points.

• As mentioned before, Matlab was chosen as the programming language

for the JMEM and Monte Carlo simulation methods because the program
was convenient at the time. It was originally thought that we could exploit
other faculties of Matlab to simplify the coding of the Monte Carlo
simulations. Coding the methods discussed in this thesis in programming

94
languages like C or FORTRAN could say valuable computing time, for it is
known that C and FORTRAN are generally more efficient coding
languages.

95
THIS PAGE INTENTIONALLY LEFT BLANK

96
 APPENDIX A

%INPUTS SAME AS JMEM/AS

%Computes and compares the EFD based on four options.

%1) Lethal area matrix filled using the Carleton Damage function,
%2) Rectangular cookie cutter with Monte Carlo simulation
%3) JMEM single weapon vs. area targets with rectangualr cookie cutter-Ch8
%4) JMEM single weapon vs. unitary target with rectangular cookie-Ch7

function [x]=efdjmem(model)

%Define number of total trials in Monte Carlo Simulation

maxiter=1000000

%-------------------------------------------------------------------------
%USER INPUTS
REP=25; %range error probable
DEP=25; %deflection error probable
SR= 10000; %slant range
I=45; %impact angle degrees
sigma_b=0; %Ballistic dispersion in mrads, standard deviation
MAEF=1000;
I_radians=I*pi/180;

a=max(1-.8*cos(I_radians),.3);

% Calculate the ballistic dispersion errors in ft in the ground plane

x_br=0.6745*SR*sigma_b/(1000*sin(I_radians));
x_bd=0.6745*SR*sigma_b/1000;

%Calculate weapon radii and blast radius

WRr=sqrt(MAEF*a/pi);
WRd=WRr/a;
RBl=0;
%-----------------------------------------------------------------------

%Create Ground plane grid

matrixsize=1001;
dmatrix=zeros(matrixsize);

%Define size of weapon matrix

weapon_size_x=round(4*WRr)+1;
weapon_size_y=round(4*WRd)+1;
weaponmatrix=zeros(weapon_size_x,weapon_size_y);

%Initialize target matrix

size_t_range=50;
size_t_deflection=50;
t=zeros(size_t_range, size_t_deflection);

%--------------------------------------------------------------------------

%Offset the center of the weapon matrix

offset=ceil(matrixsize/2);

%Define and place the weapon matrix in the ground plane using the Carleton damage function

h=1;
for i=0:weapon_size_x-1 %------------looping through the columns
for n=0:weapon_size_y-1 %------------looping through the rowa

97
 g=exp(-(((i-floor(weapon_size_x/2))*(i-floor(weapon_size_x/2))/(WRr*WRr))+(((n-floor(weapon_size_y/2))*(n-
floor(weapon_size_y/2)))/(WRd*WRd))));
r=sqrt(((i-floor(weapon_size_x/2))^2)+((n-floor(weapon_size_y/2))^2));

if (r<=RBl)
g=1;
end
dmatrix(i+offset-(floor(weapon_size_x/2)),n+offset-(floor(weapon_size_y/2)))=g;

%Fills in weapon
weaponmatrix(i+1,n+1)=g;

h=h+1;
n=n+1;
end %end of n loop
i=i+1;
end %end of i loop

%Define mean values and standard deviations of the range and deflection
%for...

% Aiming Error
meanrange_aim=offset;
meandeflection_aim=offset;
sigmarange_aim=REP/0.6745;
sigmadeflection_aim=DEP/0.6745;

%Ballistic Dispersion
meanrange_bd=0;
meandeflection_bd=0;
sigmarange_bd=x_br;
sigmadeflection_bd=x_bd;

%Initialize final sum of target cells

tfinalaverage=0;

%**************************************************************************
%**************************************************************************

%Lethal Area Matrix based on Carleton damage function

if model==1

%Trial loop
for z=1:maxiter

rrange_aim=(randn*sigmarange_aim);
rdefl_aim=(randn*sigmadeflection_aim);

rrange_bd=(randn*sigmarange_bd);
rdefl_bd=(randn*sigmadeflection_bd);

%Places the top left cell of the target matrix using a normal distribution
%of random numbers (rounded in order to ensure whole numbers for the matrix cell

s_range=round(rrange_aim+rrange_bd+meanrange_aim+meanrange_bd);
s_deflect=round(rdefl_aim+rdefl_bd+meandeflection_aim+meandeflection_bd);

k=1; %Counter for the target row currently being filled

%-----------------------------------------------------------------------------------------

for j=s_range:s_range+size_t_range-1 %loop to work through rows in weapon matrix

98
d=dmatrix(j-(floor(size_t_range/2)),:); % d is defined as the current row in the weapon we are looking at

for i=1:size_t_deflection %takes current target row and fills it with appropriate values of weapon row d
t(k,i)=d(s_deflect+i-1-(floor(size_t_range/2)));
i=i+1;

if i>size_t_range
break
end
end %end of i loop

k=k+1;
j=j+1;
end %end of j loop
t;

%Loop to sum target elements

t_sum=0;
for p=1:(size_t_range*size_t_deflection)
t_sum=t(p)+t_sum;
p=p+1;
end %end of t loop

%Calculates the average cell value for a single iteration

t_average=t_sum/(size_t_range*size_t_deflection);

%Counter to keep track of all of the calculated cell values from above
tfinalaverage=t_average+tfinalaverage;

z=z+1;
end %end of z loop

%Takes the average of all of the averages obatined from the trials
EFD=tfinalaverage/maxiter

%**************************************************************************
%**************************************************************************

%Rectangular Cookie Cutter Model

elseif model==2

%Find MAEF by summing elements of weapon matrix

weapon_sum=0;

for j=1:(weapon_size_x*weapon_size_y)
weapon_sum=weaponmatrix(j)+weapon_sum;
j=j+1;
end %end of t loop

MAEF=weapon_sum

%Define ratio of WRr/WRd

a=WRr/WRd;

%Define LET and WET

LET=sqrt(MAEF*a);
WET=LET/a;

%Pd=1 in the rectangle centered in the ground plane with dimensions given
%by WRr WRd
dmatrix=zeros(matrixsize);

for i=0:LET-1
for n=0:WET-1
dmatrix(i+offset-(floor(LET/2)),n+offset-(floor(WET/2)))=1;
n=n+1;
99
 end %end of n loop
i=i+1;
end %end of i loop

%Initialize final sum of target cells

tfinalaverage=0;

%Trial loop
for z=1:maxiter

rrange_aim=(randn*sigmarange_aim);
rdefl_aim=(randn*sigmadeflection_aim);

rrange_bd=(randn*sigmarange_bd);
rdefl_bd=(randn*sigmadeflection_bd);

%Places the top left cell of the target matrix using a normal distribution
%of random numbers (rounded in order to ensure whole numbers for the matrix cell

s_range=round(rrange_aim+rrange_bd+meanrange_aim+meanrange_bd);
s_deflect=round(rdefl_aim+rdefl_bd+meandeflection_aim+meandeflection_bd);

k=1; %Counter for the target row currently being filled

%-----------------------------------------------------------------------------------------

for j=s_range:s_range+size_t_range-1 %loop to work through rows in weapon matrix

d=dmatrix(j-(floor(size_t_range/2)),:); % d is defined as the current row in the weapon we are looking at

for i=1:size_t_deflection %takes current target row and fills it with appropriate values of weapon row d
t(k,i)=d(s_deflect+i-1-(floor(size_t_range/2)));
i=i+1;

if i>size_t_range
break
end
end %end of i loop

k=k+1;
j=j+1;
end %end of j loop
t;

%Loop to sum target elements

t_sum=0;
for p=1:(size_t_range*size_t_deflection)
t_sum=t(p)+t_sum;
p=p+1;
end %end of t loop

%Calculates the average cell value for a single iteration

t_average=t_sum/(size_t_range*size_t_deflection);

%Counter to keep track of all of the calculated cell values from above
tfinalaverage=t_average+tfinalaverage;

z=z+1;
end %end of z loop

%Takes the average of all of the averages obatined from the trials
EFD=tfinalaverage/maxiter

%**************************************************************************
100
%**************************************************************************

%JMEMs technique Chapter 8 - one weapon aginst area of targets - option #3

elseif model==3

%Define reliabilty of weapon

R=1;

% Calculate and combine BD and aiming error

x_br=0.6745*SR*sigma_b/(1000*sin(I_radians));
x_bd=0.6745*SR*sigma_b/1000;

% RSS error probables delivery accuracies

REP_dash=sqrt((REP*REP)+(x_br*x_br));
DEP_dash=sqrt((DEP*DEP)+(x_bd*x_bd));

sigma_range=REP_dash/.6745;
sigma_deflection=DEP_dash/.6745;

%Define LET and WET

LET=sqrt(MAEF*a);
WET=LET/a;

%Define LEP and WEP

LEP=max(LET,size_t_range);
WEP=max(WET,size_t_deflection);

%---------------------------------------
%Fractional coverage in range

s_range=(LEP+size_t_range)/2
t_range=(LEP-size_t_range)/2

%Solving piecewise integral for range

I1=((normcdf(t_range,0,sigma_range))-(normcdf(-t_range,0,sigma_range)))
I2=((LEP+size_t_range)/(2*size_t_range))*(normcdf(-t_range,0,sigma_range)-normcdf(-s_range,0,sigma_range))
I3=((LEP+size_t_range)/(2*size_t_range))*(normcdf(s_range,0,sigma_range)-normcdf(t_range,0,sigma_range))

a_range=s_range/(sigma_range*sqrt(2))
b_range=t_range/(sigma_range*sqrt(2))

I4minusI5=((2*sigma_range)/(size_t_range*sqrt(2*pi)))*(exp(-(a_range*a_range))-exp(-(b_range*b_range)))

%Fractional coverage in range

EFRange=I1+I2+I3+I4minusI5

%Fractional coverage in deflection

s_deflection=(WEP+size_t_deflection)/2
t_deflection=(WEP-size_t_deflection)/2

%Solving piecewise integral for deflection

I1=(normcdf(t_deflection,0,sigma_deflection)-normcdf(-t_deflection,0,sigma_deflection))
I2=((WEP+size_t_deflection)/(2*size_t_deflection))*(normcdf(-t_deflection,0,sigma_deflection)-normcdf(-
s_deflection,0,sigma_deflection))
I3=((WEP+size_t_deflection)/(2*size_t_deflection))*(normcdf(s_deflection,0,sigma_deflection)-
normcdf(t_deflection,0,sigma_deflection))

a_deflection=s_deflection/(sigma_deflection*sqrt(2))
b_deflection=t_deflection/(sigma_deflection*sqrt(2))
101
I4minusI5=((2*sigma_deflection)/(size_t_deflection*sqrt(2*pi)))*(exp(-(a_deflection*a_deflection))-exp(-
(b_deflection*b_deflection)))

%Fractional coverage in deflection

EFDeflection=I1+I2+I3+I4minusI5

%-------------------------------------------------------
%Compute composite fractional coverage
EFC=EFRange*EFDeflection;

%Compute fractional damage

EFD=EFC*R*LET*WET/(LEP*WEP)

%**************************************************************************
%**************************************************************************
%Option 4
%JMEMs technique Chapter 7 -one weapon against a unitary target
else

%RSS error probables delivery accuracies

REP_dash=sqrt((REP*REP)+(x_br*x_br));
DEP_dash=sqrt((DEP*DEP)+(x_bd*x_bd));

LETprime=1.128*(sqrt(MAEF*a));
WETprime=LETprime/a;

SSPD=(LETprime*WETprime)/(sqrt(((17.6*REP_dash*REP_dash)+(LETprime*LETprime))*((17.6*DEP_dash*DEP_dash)
+(WETprime*WETprime))))

end

102
 APPENDIX B

%Computes and compares the EFD of Stick Deliveries based on 2 options

% 1)Lethal area matrices filled using the Carleton Damage function for each
% weapon in stick.
% 2)LAM method with rectangular cookie cutter for each weapon
% 3)WIN-JMEM fractional coverage of each weapon independently method
% 4)Ch 9 stick calculating methods

function [x]=stick(model)

%Define number of total trials

maxiter=100000

%-------------------------------------------------------------------------
%USER INPUTS
REP=50; %range error probable
DEP=50; %deflection error probable
SR= 10000; %slant range
I=45; %impact angle degrees
sigma_b=5; %Ballistic dispersion in mrads, standard deviation
MAEF=1000;
I_radians=I*pi/180;
a=max(1-.8*cos(I_radians),.3) %Aspect ratio of weapon radii

R=1; %Reliabilty of weapon

%Calculate weapon radii and blast radius

WRr=sqrt(MAEF*a/pi)
WRd=WRr/a
RBl=0;
%-------------------------------------------------------------------------
%Create Ground plane grid
matrixsize=1001;
dmatrix=zeros(matrixsize);

%Define size of weapon matrix

weapon_size_x=round(4*WRr)+1;
weapon_size_y=round(4*WRd)+1;
weaponmatrix=zeros(weapon_size_x,weapon_size_y);

%Initialize target matrix

size_t_range=50;
size_t_deflection=50;
t=zeros(size_t_range, size_t_deflection);
%--------------------------------------------------------------------------

% Calculate the ballistic dispersion errors in ft in the ground plane

x_br=SR*sigma_b/(1000*sin(I_radians));
x_bd=SR*sigma_b/1000;

%Stick User Inputs

totalweapons=4; %total number of weapons released
np=2; %Number of weapons released per pulse
dt=.2; %intervalometer setting (seconds)
Ws=100; %stick width
Ls=100; %stick length
PCD=1; %conditional probability of damage
%-----------------------------------------------------------------------

if model==1
%Option 1 -- Carleton Damage Function methods

%Define mean values and standard deviations of the range and deflection
103
%for...

% Aiming Error
meanrange_aim=0;
meandeflection_aim=0;
sigmarange_aim=REP/0.6745;
sigmadeflection_aim=DEP/0.6745;

%Ballistic Dispersion
meanrange_bd=0;
meandeflection_bd=0;
sigmarange_bd=x_br;
sigmadeflection_bd=x_bd;

%Initialize final sum of target cells

tfinalaverage=0;

%Trial loop
for z=1:maxiter

%Places the top left cell of the target matrix using a normal distribution
%of random numbers (rounded in order to ensure whole numbers for the matrix cell
rrange_aim=(randn*sigmarange_aim);
rdefl_aim=(randn*sigmadeflection_aim);

%Top left cell of target matrix influenced by aiming error of stick

s_range=round(rrange_aim+meanrange_aim);
s_deflect=round(rdefl_aim+meandeflection_aim);

%Place individual impact points including ballistic dispersion errors on

%each weapon
weapon1_range=-50+(randn*sigmarange_bd);
weapon1_deflection=-50+(randn*sigmadeflection_bd);
weapon2_range=-50+(randn*sigmarange_bd);
weapon2_deflection=50+(randn*sigmadeflection_bd);
weapon3_range=50+(randn*sigmarange_bd);
weapon3_deflection=-50+(randn*sigmadeflection_bd);
weapon4_range=50+(randn*sigmarange_bd);
weapon4_deflection=50+(randn*sigmadeflection_bd);

k=0;
i=0;

for k=0:size_t_range-1 % Row counter

for i=0:size_t_deflection-1 % Column counter

Pd1=exp(-((((weapon1_range-(s_range-k))^2)/(WRr*WRr))+(((weapon1_deflection-(s_deflect+i))^2)/(WRd*WRd))));
Pd2=exp(-((((weapon2_range-(s_range-k))^2)/(WRr*WRr))+(((weapon2_deflection-(s_deflect+i))^2)/(WRd*WRd))));
Pd3=exp(-((((weapon3_range-(s_range-k))^2)/(WRr*WRr))+(((weapon3_deflection-(s_deflect+i))^2)/(WRd*WRd))));
Pd4=exp(-((((weapon4_range-(s_range-k))^2)/(WRr*WRr))+(((weapon4_deflection-(s_deflect+i))^2)/(WRd*WRd))));

t(k+1,i+1)=1-((1-Pd1)*(1-Pd2)*(1-Pd3)*(1-Pd4));

i=i+1;

end
k=k+1;
i=0;
end

%Loop to sum target elements

t_sum=0;
for p=1:(size_t_range*size_t_deflection)
t_sum=t(p)+t_sum;
p=p+1;
end %end of t loop
104
%Calculates the average cell value for a single iteration
t_average=t_sum/(size_t_range*size_t_deflection);

%Counter to keep track of all of the calculated cell values from above
tfinalaverage=t_average+tfinalaverage;

z=z+1;
end %end of z loop

%Takes the average of all of the averages obatined from the trials
EFD=tfinalaverage/maxiter
t;

%**************************************************************************
%**************************************************************************
elseif model==2

%Option 2
%Rectangular cookie cutter method

offset=500;

%Creates matrix filled by Carleton damage function

h=1;
for i=0:weapon_size_x-1 %------------looping through the columns
for n=0:weapon_size_y-1 %------------looping through the rowa

g=exp(-(((i-floor(weapon_size_x/2))*(i-floor(weapon_size_x/2))/(WRr*WRr))+(((n-floor(weapon_size_y/2))*(n-
floor(weapon_size_y/2)))/(WRd*WRd))));
r=sqrt(((i-floor(weapon_size_x/2))^2)+((n-floor(weapon_size_y/2))^2));

if (r<=RBl)
g=1;
end
dmatrix(i+offset-(floor(weapon_size_x/2)),n+offset-(floor(weapon_size_y/2)))=g;

%Fills in wepaon
weaponmatrix(i+1,n+1)=g;

h=h+1;
n=n+1;
end %end of n loop
i=i+1;
end %end of i loop

%Find MAEF by summing elements of weapon matrix

weapon_sum=0;

for j=1:(weapon_size_x*weapon_size_y)
weapon_sum=weaponmatrix(j)+weapon_sum;
j=j+1;
end %end of t loop

%Define LET and WET

LET=sqrt(MAEF*a)
WET=LET/a

%Define mean values and standard deviations of the range and deflection
%for...

% Aiming Error
meanrange_aim=0;
105
meandeflection_aim=0;
sigmarange_aim=REP/0.6745;
sigmadeflection_aim=DEP/0.6745;

%Ballistic Dispersion
meanrange_bd=0;
meandeflection_bd=0;
sigmarange_bd=x_br;
sigmadeflection_bd=x_bd;

%Initialize final sum of target cells

tfinalaverage=0;

%Trial loop
for z=1:maxiter

%Places the top left cell of the target matrix using a normal distribution
%of random numbers (rounded in order to ensure whole numbers for the matrix cell
rrange_aim=(randn*sigmarange_aim);
rdefl_aim=(randn*sigmadeflection_aim);

%Top left cell of target matrix influenced by aiming error of stick

s_range=round(rrange_aim+meanrange_aim);
s_deflect=round(rdefl_aim+meandeflection_aim);

%Place individual impact points including ballistic dispersion errors on

%each weapon
weapon1_range=-50+(randn*sigmarange_bd);
weapon1_deflection=-50+(randn*sigmadeflection_bd);
weapon2_range=-50+(randn*sigmarange_bd);
weapon2_deflection=50+(randn*sigmadeflection_bd);
weapon3_range=50+(randn*sigmarange_bd);
weapon3_deflection=-50+(randn*sigmadeflection_bd);
weapon4_range=50+(randn*sigmarange_bd);
weapon4_deflection=50+(randn*sigmadeflection_bd);

%Pd=1 in the rectangle centered in the ground plane with dimensions given
%by WRr WRd

for r=1:size_t_range
for d=1:size_t_deflection

if (abs(r+weapon1_range+s_range)<=LET/2 &
abs(d+weapon1_deflection+s_deflect)<=WET/2)|(abs(r+weapon2_range+s_range)<=LET/2 &
abs(d+weapon2_deflection+s_deflect)<=WET/2)|(abs(r+weapon3_range+s_range)<=LET/2 &
abs(d+weapon3_deflection+s_deflect)<=WET/2)|(abs(r+weapon4_range+s_range)<=LET/2 &
abs(d+weapon4_deflection+s_deflect)<=WET/2)
t(r,d)=1;
else
t(r,d)=0;
end

d=d+1;
end
r=r+1;
d=0;
end

%Loop to sum target elements

t_sum=0;
for p=1:(size_t_range*size_t_deflection)
t_sum=t(p)+t_sum;
p=p+1;
end %end of t loop

%Calculates the average cell value for a single iteration

106
t_average=t_sum/(size_t_range*size_t_deflection);

%Counter to keep track of all of the calculated cell values from above
tfinalaverage=t_average+tfinalaverage;

z=z+1;
end %end of z loop

%Takes the average of all of the averages obatined from the trials
EFD=tfinalaverage/maxiter

%**************************************************************************
%**************************************************************************

elseif model==3

%Option 3
%WIN-JEM calculation method

%Ballistic Dispersion Effects

LET=sqrt(MAEF*a)
WET=LET/a

%Define ballistic dispersion as standard deviations in the ground plane

sigma_br=SR*(sigma_b)/(1000*sin(I_radians))
sigma_bd=SR*(sigma_b)/1000

%Define LB and WB
LB=sqrt(LET^2+(8*((sigma_br)^2)))
WB=sqrt(WET^2+(8*((sigma_bd)^2)))

%Define conditional probability of damage from enlarged effected area

PCD1=PCD*LET*WET/(LB*WB)

%Convert REP and DEP into sigmas

sigma_range=(REP/.6745)
sigma_deflection=(DEP/.6745)

%Define LEP and WEP

LEP=max(LB,size_t_range)
WEP=max(WB,size_t_deflection)

%Fractional coverage in range

s_range=(LEP+size_t_range)/2
t_range=(LEP-size_t_range)/2

%Solving piecewise integral for range

I1=((normcdf(t_range,50,sigma_range))-(normcdf(-t_range,50,sigma_range)))
I2=((LEP+size_t_range)/(2*size_t_range))*(normcdf(-t_range,50,sigma_range)-normcdf(-s_range,50,sigma_range))
I3=((LEP+size_t_range)/(2*size_t_range))*(normcdf(s_range,50,sigma_range)-normcdf(t_range,50,sigma_range))

I4=((-sigma_range)/(size_t_range*sqrt(2*pi)))*((exp(-((-t_range-50)^2)/(2*(sigma_range^2))))-(exp(-((-s_range-
50)^2)/(2*(sigma_range^2)))))+((50/size_t_range)*(normcdf(-t_range,50,sigma_range)-normcdf(-
s_range,50,sigma_range)))

I5=((-sigma_range)/(size_t_range*sqrt(2*pi)))*((exp(-((s_range-50)^2)/(2*(sigma_range^2))))-(exp(-((t_range-
50)^2)/(2*(sigma_range^2)))))+((50/size_t_range)*(normcdf(s_range,50,sigma_range)-normcdf(t_range,50,sigma_range)))

%Fractional coverage in range

EFRange=I1+I2+I3+I4-I5

%Fractional coverage in deflection

s_deflection=(WEP+size_t_deflection)/2
107
t_deflection=(WEP-size_t_deflection)/2

%Solving piecewise integral for deflection

I1=(normcdf(t_deflection,50,sigma_deflection)-normcdf(-t_deflection,50,sigma_deflection))
I2=((WEP+size_t_deflection)/(2*size_t_deflection))*(normcdf(-t_deflection,50,sigma_deflection)-normcdf(-
s_deflection,50,sigma_deflection))
I3=((WEP+size_t_deflection)/(2*size_t_deflection))*(normcdf(s_deflection,50,sigma_deflection)-
normcdf(t_deflection,50,sigma_deflection))

I4=((-sigma_deflection)/(size_t_deflection*sqrt(2*pi)))*((exp(-((-t_deflection-50)^2)/(2*(sigma_deflection^2))))-(exp(-((-
s_deflection-50)^2)/(2*(sigma_deflection^2)))))+((50/size_t_deflection)*(normcdf(-t_deflection,50,sigma_deflection)-
normcdf(-s_deflection,50,sigma_deflection)))

I5=((-sigma_deflection)/(size_t_deflection*sqrt(2*pi)))*((exp(-((s_deflection-50)^2)/(2*(sigma_deflection^2))))-(exp(-
((t_deflection-50)^2)/(2*(sigma_deflection^2)))))+((50/size_t_deflection)*(normcdf(s_deflection,50,sigma_deflection)-
normcdf(t_deflection,50,sigma_deflection)))

%Fractional coverage in deflection

EFDeflection=I1+I2+I3+I4-I5

%-------------------------------------------------------
%Compute composite fractional coverage
EFC=EFRange*EFDeflection

%Compute fractional damage from one weapon

singleEFD=PCD1*EFC*R

%Total EFD from 4 weapons

EFD=1-((1-singleEFD)*(1-singleEFD)*(1-singleEFD)*(1-singleEFD))

%**************************************************************************
%**************************************************************************

else

%Option 4
%Old JMEM Method

%Define LET and WET

LET=sqrt(MAEF*a);
WET=LET/a;

%Define ballistic dispersion as standard deviations in the ground plane

sigma_br=SR*(sigma_b)/(1000*sin(I_radians))
sigma_bd=SR*(sigma_b)/1000

%Define LB and WB
LB=sqrt(LET^2+(8*(sigma_br)^2))
WB=sqrt(WET^2+(8*(sigma_bd)^2))

%Define conditional probability of damage from enlarged effected area

PCD1=PCD*LET*WET/(LB*WB);

% Calculate delivery accuracy errors

sigma_range=REP/.6745;
sigma_deflection=DEP/.6745;

%Define Pattern Dimensions

Lp=Ls+LB;
Wp=Ws+WB;

%Overlap
nr=totalweapons/np %Number of intervalometer pulses

%Define Degree of overlap in the deflection direction

nod=np*WB/Wp;
108
%Define the conditional probability of damage in the deflection direction
if nod>=1
PCDd=1-(1-PCD1)^nod;
else
PCDd=np*PCD1*WB/Wp;
end

%Define Degree of overlap in the range direction

nor=nr*LB/Lp;

%Calculate the conditional probability for the whole pattern

if nor>=1
PCDS=1-(1-PCDd)^nor;
else
PCDS=nr*PCDd*LB/Lp;
end

%---------------------------------------
%Define LEP and WEP
LEP=max(Lp,size_t_range);
WEP=max(Wp,size_t_deflection);

%Fractional coverage in range

s_range=(LEP+size_t_range)/2;
t_range=(LEP-size_t_range)/2;

%Solving piecewise integral for range

I1=((normcdf(t_range,0,sigma_range))-(normcdf(-t_range,0,sigma_range)));
I2=((LEP+size_t_range)/(2*size_t_range))*(normcdf(-t_range,0,sigma_range)-normcdf(-s_range,0,sigma_range));
I3=((LEP+size_t_range)/(2*size_t_range))*(normcdf(s_range,0,sigma_range)-normcdf(t_range,0,sigma_range));

a_range=s_range/(sigma_range*sqrt(2));
b_range=t_range/(sigma_range*sqrt(2));

I4minusI5=((2*sigma_range)/(size_t_range*sqrt(2*pi)))*(exp(-(a_range*a_range))-exp(-(b_range*b_range)));

%Fractional coverage in range

EFRange=I1+I2+I3+I4minusI5

%Fractional coverage in deflection

s_deflection=(WEP+size_t_deflection)/2;
t_deflection=(WEP-size_t_deflection)/2;

%Solving piecewise integral for deflection

I1=(normcdf(t_deflection,0,sigma_deflection)-normcdf(-t_deflection,0,sigma_deflection));
I2=((WEP+size_t_deflection)/(2*size_t_deflection))*(normcdf(-t_deflection,0,sigma_deflection)-normcdf(-
s_deflection,0,sigma_deflection));
I3=((WEP+size_t_deflection)/(2*size_t_deflection))*(normcdf(s_deflection,0,sigma_deflection)-
normcdf(t_deflection,0,sigma_deflection));

a_deflection=s_deflection/(sigma_deflection*sqrt(2));
b_deflection=t_deflection/(sigma_deflection*sqrt(2));

I4minusI5=((2*sigma_deflection)/(size_t_deflection*sqrt(2*pi)))*(exp(-(a_deflection*a_deflection))-exp(-
(b_deflection*b_deflection)));

%Fractional coverage in deflection

EFDeflection=I1+I2+I3+I4minusI5

%Compute composite fractional coverage

EFC=EFRange*EFDeflection;

%Compute fractional damage

EFD=EFC*Lp*Wp*R*PCDS/(LEP*WEP)

end %end of model if/else statement

109
THIS PAGE INTENTIONALLY LEFT BLANK

110
 LIST OF REFERENCES

1. Driels, M., Weaponeering, AAIA, 2004.

2. Joint Technical Coordination Group for Munitions Effectiveness, Derivation

of JMEM/AS Open-End Methods, Distribution authorized to U.S.
Government agencies and their contractors only, 2004.

111
THIS PAGE INTENTIONALLY LEFT BLANK

112
 INITIAL DISTRIBUTION LIST

1. Defense Technical Information Center

Ft. Belvoir, VA

2. Dudley Knox Library

Naval Postgraduate School
Monterey, CA

3. Mechanical Engineering Department Chairman, Code ME

Naval Postgraduate School
Monterey, CA

4. Naval/Mechanical Engineering Curriculum Code 34

Naval Postgraduate School
Monterey, CA

5. Professor Morris R. Driels, Code ME/Dr

Department of Mechanical Engineering
Naval Postgraduate School
Monterey, CA

6. Commander NAWC-WD
Attn: W. Tonkin 4J6000D/W
China Lake, CA

7. Director US ANSAA
Attn: AMXSY-SA (R. Chandler)
Aberdeen Proving Ground, MD

8. ASC/XREW
Attn: Carolyn Holland
Eglin AFB, FL

113