Skip to main content

Full text of "Modeling ordnance movements into the Asian Pacific Theater"

See other formats


Calhoun 

iniQiuiic^iul Ar{hiv« of tilt Mil vdl Poii^roduiit School 


Calhoun: The NPS Institutional Archive 
□Space Repository 



Theses and Dissertations 


1. Thesis and Dissertation Collection, all items 


2009-03 

Modeling ordnance movements into the Asian 
Pacific Theater 

Almanza, Cielo I. 

Monterey, California. Naval Postgraduate School 


http://hdl.handle.net/10945/4832 


Downloaded from NPS Archive: Calhoun 



DUDLEY 

KNOX 

LIBRARY 


htt p://w ww. n ps. e du/l ib ra ry 


Caflwuo is the Naval Postgraduate School's public access digital repository for 
research mate rials and institutiional putilicatiiaos created by the NPS community. 
Calhoun is named for Professor of Mathematics Guy K. Caftiouo, NPS's first 
appointed — and putJlished — schoteily author. 

Dudley Knox Library / Naval Postgraduate School 
411 Dyer Road / 1 University Circle 
Monterey, California USA 93943 







NAVAL 

POSTGRADUATE 

SCHOOL 

MONTEREY, CALIFORNIA 


THESIS 


MODELING ORDNANCE MOVEMENTS INTO THE 
ASIAN PACIFIC THEATER 


by 


Cielo I. Almanza 

Thesis Advisor: 
Co-Advisor: 
Second Reader: 

March 2009 

Thomas W. Lucas 

W. David Kelton 
Keebom Kang 


Approved for public release; distribution is unlimited 




THIS PAGE INTENTIONALLY LEET BLANK 



REPORT DOCUMENTATION PAGE 


Form Approved 0MB 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. 

I. AGENCY USE ONLY (Leave blank) 2. REPORT DATE 3. REPORT TYPE AND DATES COVERED 

March 2009 Master’s Thesis 

4. TITLE AND SUBTITLE Modeling Ordnance Movements into the Asian Pacific 5. EUNDING NUMBERS 
Theater 

6. author" 

7. PEREORMING ORGANIZATION NAME(S) AND ADDRESS(ES) 8. PEREORMING ORGANIZATION 

Naval Postgraduate School REPORT NUMBER 

Monterey, CA 93943-5000 

9. SPONSORING /MONITORING AGENCY NAME(S) AND ADDRESS(ES) 10. SPONSORING/MONITORING 

Director of the Strategic Mobility and Comhat Logistics Division for the Chief of AGENCY REPORT NUMBER 
Naval Operations (OPNAV N421, CAPT Sean Geaney, Bldg: Crystal Square 2, 

Suite: 1002, 1550 Crystal Drive, Arlington, VA 22202) 

II. 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 / AVAII.ABIITTY STATEMENT I2b. DISTRIBUTION CODE 

Approved for public release; distribution is unlimited A 

13. ABSTRACT (maximum 200 words) 

This thesis explores the capabilities of ordnance movements into the Asian Pacific Theater. Through 
simulation, logistics modeling, and data analysis, this thesis identifies critical factors and capabilities that are 
important to the effective movement of ordnance by combat logistics ships through Guam during a military 
contingency. The experimental design incorporates the effects of competing requirements on the ordnance resupply 
process in Guam. The objective is to facilitate an evaluation of systems, identify possible improvements to fully 
exploit capabilities, and gain insights into the process methodology. Results indicate that the inclusion of competing 
requirements to the system degrades both Auxiliary Dry Cargo/Ammunition Ship (T-AKE) service level and the 
overall throughput of the system by nearly 25%. Analysis of critical factors contributing to this degradation indicates 
that the T-AKE arrival cycle is the largest contributing factor to the system’s effectiveness. The results also indicate 
that competition is a contributor to the effects on the system, but is never the most influential aspect, and the decision 
of where to process ordnance is significant for the best-performing scenarios in the experiments. Lastly, the analysis 
clearly shows that improving the system’s performance is not dependent on the distance of ordnance storage facilities 
from the wharf. 

14. SUBJECT TERMS Ordnance movement, Asian Pacific Theater, T-AKE, ARENA, Entity-based 15. NUMBER OE 

modeling, SEED Center, Logistics, Simulation, Design of Experiments PAGES 

_MT_ 

16. PRICE CODE 


17. SECURITY 

18. SECURITY 

19. SECURITY 

20. LIMITATION OE 

CLASSIEICATION OE 

CLASSIEICATION OE THIS 

CLASSIEICATION OE 

ABSTRACT 

REPORT 

PAGE 

ABSTRACT 


Unclassified 

Unclassified 

Unclassified 

UU 


NSN 7540-01-280-5500 Standard Form 298 (Rev. 2-89) 

Prescribed by ANSI Std. 239-18 


1 




























THIS PAGE INTENTIONALLY LEET BLANK 


11 



Approved for public release; distribution is unlimited 


Author: 


Approved by: 


MODELING ORDNANCE MOVEMENTS INTO THE 
ASIAN PACIFIC THEATER 

Cielo 1. Almanza 

Lieutenant Commander, United States Navy 
B.S., Seattle University, 1996 


Submitted in partial fulfillment of the 
requirements for the degree of 


MASTER OF SCIENCE IN OPERATIONS RESEARCH 


from the 


NAVAL POSTGRADUATE SCHOOL 
March 2009 


Cielo 1. Almanza 


Thomas W. Lucas 
Thesis Advisor 


W. David Kelton 
Co-Advisor 


Keebom Kang 
Second Reader 


Robert F. Dell 

Chairman, Department of Operations Research 



THIS PAGE INTENTIONALLY LEET BLANK 


IV 



ABSTRACT 


This thesis explores the capabilities of ordnance movements into the Asian Pacific 
Theater. Through simulation, logistics modeling, and data analysis, this thesis identifies 
critical factors and capabilities that are important to the effective movement of ordnance 
by combat logistics ships through Guam during a military contingency. The 
experimental design incorporates the effects of competing requirements on the ordnance 
resupply process in Guam. The objective is to facilitate an evaluation of systems, 
identify possible improvements to fully exploit capabilities, and gain insights into the 
process methodology. Results indicate that the inclusion of competing requirements to 
the system degrades both Auxiliary Dry Cargo/Ammunition Ship (T-AKE) service level 
and the overall throughput of the system by nearly 25%. Analysis of critical factors 
contributing to this degradation indicates that the T-AKE arrival cycle is the largest 
contributing factor to the system’s effectiveness. The results also indicate that 
competition is a contributor to the effects on the system, but is never the most influential 
aspect, and the decision of where to process ordnance is significant for the 
best-performing scenarios in the experiments. Eastly, the analysis clearly shows that 
improving the system’s performance is not dependent on the distance of ordnance storage 
facilities from the wharf. 


V 



THIS PAGE INTENTIONALLY LEET BLANK 


VI 



THESIS DISCLAIMER 


The reader is cautioned that the computer programs utilized in this research may 
not have been exercised for all cases of interest. While every effort has been made, 
within the time available, to ensure that the programs are free of computational and 
logical errors, they cannot be considered validated. Any application of these programs 
without additional verification is at the risk of the user. 



THIS PAGE INTENTIONALLY LEET BLANK 



TABLE OF CONTENTS 

I. INTRODUCTION.I 

A. BACKGROUND AND MOTIVATION.I 

B. OBJECTIVES.3 

C. BENEFITS OF THE STUDY.3 

D. METHODOLOGY.4 

II. SCENARIO DEVELOPMENT.7 

A. INTRODUCTION.7 

B. WHAT ARE ORDNANCE OPERATIONS?.7 

1. Overview.7 

2. Study Area Location and Facilities.7 

a. Kilo Wharf. . 8 

b. Buoy 702 . 9 

c. Ordnance Annex . 9 

d. Ordnance Handling Pad . 10 

3. Operations.II 

a. Offload . 12 

b. Handling (Unstuffing) . 13 

c. Moving Ordnance . 14 

d. Stowage . 15 

e. Onload . 15 

C. SCENARIO DESCRIPTION.15 

1. Overview.15 

2. General Situation.16 

D. THE ARENA SIMULATION TOOL.17 

1. Why Arena?.18 

2. Characteristics of the Arena Simulation Environment.18 

E. CHARACTERISTICS OF THE SIMULATION MODEL.19 

1. Goals and Measures of Effectiveness (MOEs).19 

2. Conceptual Model.20 

3. Key Components of the Model.20 

a. Entities . 21 

b. Queues . 22 

c. Resources . 23 

d. Stations . 24 

e. Transporters .25 

4. Arena Simulation Time.26 

5. Summary.27 

III. MODEL IMPLEMENTATION AND EXPERIMENTAL DESIGN.29 

A. INTRODUCTION.29 

B. PRIMARY ENTITIES AND ATTRIBUTES.29 

C. PRIMARY RESOURCES.35 

ix 












































1. Space Resources.36 

a. Kilo Berth . 36 

b. Buoy 702 . 36 

c. Pier-side Staging Space . 36 

d. Unstuffing Space . 36 

e. Ordnance Annex Magazine Storage . 36 

f Container Truck Loading Space . 37 

2 . Equipment Resources.38 

a. Crane . 38 

b. Ordnance Forklifts . 38 

3. Personnel Resources.38 

a. Ordnance Inspectors . 38 

b. Block and Brace Crew . 39 

D. PRIMARY PROCESSES.39 

E. PRIMARY TRANSPORTERS.42 

F. VARIABLES OF INTEREST.43 

1. Controllable Factors.44 

a. Ordnance Container Ship Arrival Cycle 

(v_OCS_Arr_Cycle) . 44 

b. T-AKE Arrival Cycle (v_TAKE_Arr_Cycle) . 45 

c. Number of Containers Offloaded per OCS Inport Period 

(vjContjjcrjOCS) . 45 

d. Percent Unstuffed Pierside (v_^ercent_unstuffed_pier) . 45 

e. Ordnance Inspector Capacity . 45 

f Unstuffing Space Capacity . 46 

g. Ordnance Forklifts . 46 

2. Uncontrollable Factors.46 

a. Competing Ship Arrival Time (v_CS_Arr_Time) . 46 

b. Percent of Containers for United States Navy 

(v _percent_Navy_Cont) . 47 

3. Other Variables of Interest.47 

a. Universal Stream Indicator (v_Univ_Stream) . 47 

b. Initial Inventory (v_Initial_Inventory) . 48 

G. THE EXPERIMENT.48 

1. Scenario Set I - The Baseline Model.49 

2. Scenario Set 2 - Addition of Competing Requirements.50 

3. Scenario Set 3 - Simulating Completion of the New Magazine 

on Orote Baseline.51 

4. Scenario Set 4 - Simulating Completion of New Magazine on 

Orote.51 

5. Scenario Set 5 - Exploratory Set.51 

6. Simulation Runs and Replications.53 

IV. DATA ANALYSIS.55 

A. DATA COLLECTION AND POST PROCESSING.55 

B. INSIGHTS INTO RESEARCH QUESTIONS.56 


X 











































1. MOE Correlation Analysis.57 

2. Analysis of Scenario Set 1 - The Baseline Model.59 

3. Analysis of Scenario Set 2 - Addition of Competing 

Requirements.59 

4. Analysis of Scenario Set 3 - Simulating New Magazine 

Baseline.68 

5. Analysis of Scenario Set 4 - Simulating Completion of 

Magazine on Orote.70 

6. Analysis of Scenario Set 5 - Exploratory Set.80 

a. Process Analyzer Results . 87 

V. CONCLUSIONS.91 

A. THESIS SUMMARY.91 

B. THESIS QUESTIONS.91 

1. Effect of Competing Requirements.91 

2. Critical Factors.92 

C. ADDITIONAL INSIGHTS.92 

1. Initial Inventory.92 

2. Operational Capacity.93 

3. Theater Challenges.93 

D. RECOMMENDATIONS.94 

E. FOLLOW-ON WORK.95 

APPENDIX: COMPONENT AND MODULE SPECIFICATION FOR THE 
MODELING ORDNANCE MOVEMENTS INTO THE ASIAN PACIFIC 
THEATER.97 

LIST OF REFERENCES.109 

INITIAL DISTRIBUTION LIST.Ill 


XI 
























THIS PAGE INTENTIONALLY LEET BLANK 



LIST OF FIGURES 


Figure 1. AO and Flow Paths of Inbound Ordnance. (After: Helber, Hastert & Fee, 

2003).xxii 

Figure 2. Map of Orote Peninsula Area, Guam (From: Goode & Smith, 2007).xxiii 

Figure 3. CONOPS for Battle Group Replenishment (From: Markle & Wileman, 

2001).2 

Figure 4. Ordnance Operations Flow Path Diagram.7 

Figures. Map of Orote Peninsula Area, Guam (From: Goode & Smith, 2007).8 

Figure 6. Aerial View of Kilo Wharf (From: MSDDC, 2008).9 

Figure 7. Igloo Storage Magazine at Ordnance Annex (From: MSDDC, 2008).10 

Figure 8. Ordnance Handling Pad at Orote Peninsula (From: MSDDC, 2008).11 

Figure 9. Ordnance Flow from Container Ship to a T-AKE Including the Dry Stores 

Component of any Replenishment Period (From: Goode & Smith, 2007)....12 
Figure 10. Area of Operations and Flow Paths of Inbound Ordnance. Note: The flow 
paths of incoming ordnance are based on the sources of ordnance supply 

(After: Helber, Hastert, & Fee, 2003).17 

Figure 11. The Create Module and Create GUI in Arena.21 

Figure 12. The Assign Module and Assign GUI in Arena.22 

Figure 13. The Process Module and associated Queue GUI in Arena.22 

Figure 14. The Station Module and Associated Station GUI in Arena.25 

Figure 15. The Run Setup Menu in Arena.27 

Figure 16. The First Separate Module for an OCS and Associated Separate GUI.32 

Figure 17. The Container Generation Segment.32 

Figure 18. The Assign Module for a Container.33 

Figure 19. The Create Module for Initialization.34 

Figure 20. The Assign Module for Pallets at the Ordnance Annex.35 

Figure 21. The GUI associated with an Arena Process Module.40 

Figure 22. Container Capable For kl ifts (From: MSDDC, 2008).43 

Figure 23. Arena Input Analyzer output for CS Interarrival Time.47 

Figure 24. The Models Output Segment.56 

Figure 25. Correlation and Scatterplot Matrix for MOEs.57 

Eigure 26. Correlation and Scatterplot Matrix for MOEs and T-AKE Arrival 

Erequency.58 

Eigure 27. Scenario Set 1 - The Baseline Model MOE Distributions.59 

Eigure 28. Scenario Sets 1 and 2 T-AKE SE Distribution Comparisons.60 

Eigure 29. Scenario Sets 1 and 2 Pallets Out Distribution Comparisons.61 

Eigure 30. Scenarios 1 and 2 MOE Annual Average Value Comparisons.62 

Eigure 31. Stepwise Regression Analysis of Scenario Set 2.62 

Eigure 32. Regression Analysis of Scenario Set 2.64 

Eigure 33. RSquare Plot for Scenario Set 2 Partitions.65 

Eigure 34. Partition and Column Contribution of T-AKE SE in Scenario 2.66 

Eigure 35. Partition and Column Contribution of Pallets Out in Scenario 2.67 

Eigure 36. Scenario Sets 1 and 3 - Direct Comparisons of MOE Distributions.68 

xiii 






































Figure 37. Scenarios 1 and 3 MOE Annual Average Value Comparisons.69 

Figure 38. Scenario Sets 3 and 4 T-AKE SE Distribution Comparisons.70 

Figure 39. Scenario Sets 3 and 4 Pallets Out Distribution Comparisons.71 

Figure 40. Scenarios 3 and 4 MOE Annual Average Value Comparisons.71 

Figure 41. Scenario Sets 2 and 4 - MOE Distributions.72 

Figure 42. Scenarios 2 and 4 MOE Annual Average Value Comparisons.73 

Figure 43. Stepwise Regression Analysis of Scenario Set 4.74 

Figure 44. Regression Analysis of Scenario 4.76 

Figure 45. RSquare Plot for Scenario Set 4 Partitions.77 

Figure 46. Partition and Column Contribution of T-AKE SF in Scenario 4.78 

Figure 47. Partition and Column Contribution of Pallets Out in Scenario 4.79 

Figure 48. Scenario Set 5 - The Exploratory Model MOE Distributions.80 

Figure 49. Scenario Sets 1, 2, and 5 T-AKE Distributions.81 

Figure 50. Scenarios 1, 2, and 5 MOE Annual Average Value Comparisons.82 

Figure 51. Scenario Set 5 Main Effects Regression.83 

Figure 52. RSquare Plot for Scenario Set 5 Partitions.84 

Figure 53. Partition and Column Contribution of T-AKE SF in Scenario 5.85 

Figure 54. Partition and Column Contribution of Pallets Out in Scenario 5.86 

Figure 55. Pallet Out Best Scenario in Scenario Set 5.87 

Figure 56. Model Structure Overview.98 























LIST OF TABLES 


Table 1. Offload Rates (From: Goode & Smith, 2007, and MSDDC, 2008).13 

Table 2. Unstuffing Rates (From: MSDDC, 2008).14 

Table 3. Movement Distance (in statute miles) and Travel Times (From: MSDDC, 

2008).15 

Table 4. The Resource Data Module in Arena.24 

Table 5. The Transporter Data Module in Arena.26 

Table 6. The Decision Factors and Noise Factors.44 

Table 7. Input Parameters For Scenario Set 1 - Baseline.49 

Table 8. Input Parameters for Scenario Set 2 - Competing Requirements.50 

Table 9. Input Parameters by Scenario for Scenario Set 5.52 

Table 10. Average Annual Pallet Throughput Calculations.73 

Table 11. Scenario Set 5 “Best” Input Parameters.88 


XV 














THIS PAGE INTENTIONALLY LEET BLANK 


XVI 



LIST OF KEY WORDS, SYMBOLS, ACRONYMS, 

AND ABBREVIATIONS 

(a_Arrival_Time_to_System) Attribute for Entity Arrival Time to System 

(a_Num_Containers) 

Attribute for Number of Containers per Container Ships 

(a_Pallets_Needed) 

Attribute for Pallets Needed by T-AKE 

(a_Ship_Type) 

Attribute for Determining Ship Type 

AAFB 

Andersen Air Force Base 

AF 

Air Force 

AO 

Area of Operation 

AOE-1 

Fast Combat Support Ship, USS SACRAMENTO 

AOE-6 

Fast Combat Support Ship, USS SUPPFY 

CCF 

Container Capable Forklift 

CCT 

Container Capable Truck 

CFF 

Combat Fogistics Force 

CNA 

Center for Naval Analyses 

CNO 

Chief of Naval Operations 

COMNAVMARIANAS 

Commander, U.S. Naval Forces Marianas 

COMPACFFT 

Commander, U.S. Pacific Fleet 

CONOPS 

Concept of Operations 

CONUS 

Continental United States 

CRM 

Center for Naval Analyses Research Memorandum 

CS 

Competing Ship 

CVBG 

Carrier Battle Group 

Det 

Detachment 

DoD 

Department of Defense 

DOE 

Design of Experiments 

DZSP 21 

Day and Zimmerman Services Inc., SKE Support Services 

Inc., and Parsons Infrastructure and Technology Group Inc. 

ESQD 

Explosives Safety Quantity Distance 

xvii 



GUI 

Graphical User Interface 

MEB 

Marine Expeditionary Brigade 

MHE 

Material Handling Equipment 

MIECON 

Military Construction 

MOE 

Measure of Effectiveness 

MPE 

Maritime Prepositioning Eorce 

MSDDC 

Military Surface Deployment and Distribution Command 

NAVBASE GUAM 

Naval Base, Guam 

NAVEACMARIANAS 

Naval Eacilities Engineering Command, Marianas 

NAVSEA 

Naval Sea Systems Command 

NEW 

Net Explosive Weight 

NMC 

Naval Munitions Command 

NOEH 

Nearly Orthogonal Eatin Hypercube 

NPS 

Naval Postgraduate School 

NSWC 

Naval Surface Warfare Center 

OCS 

Ordnance Container Ship 

OEH 

Orthogonal Eatin Hypercube 

OPEAN 

Operational Plan 

OPNAV N42 

Director of the Strategic Mobility and Combat Eogistics 

Division for the Chief of Naval Operations 

PTT 

Pallet Transport Truck 

PWRMS 

Prepositioned War Reserve Material Stock 

SEED 

Simulation Experiments and Efficient Designs 

SE 

Service Eevel 

T-AE 

Auxiliary Ammunition Ship 

T-AES 

Auxiliary Combat Stores Ships 

T-AO 

Auxiliary Eleet Replenishment Oiler 

T-AKE 

Auxiliary Dry Cargo/Ammunition Ship 

TEU 

Twenty-Eoot Equivalent Unit 

TEIN 

Simulation Einish Time 


xviii 



TNOW 

Current Simulation Time 

TRIA 

Triangular Distribution 

UNIF 

Uniform Distribution 

USAF 

United States Air Force 

USMC 

United States Marine Corps 

USN 

United States Navy 

(v_Cont_per_OCS) 

Variable for Containers per Container Ship 

(v_C S_Arr_T ime) 

Variable for Competing Ship Arrival Time 

(v_Initial_Inventory) 

Variable for Initial Pallet Inventory 

(v_OCS_Arr_Cycle) 

Variable for Container Ship Arrival Cycle 

(v_percent_N avy_Cont) 

Variable for Percent of Containers, Navy 

(v_percent_unstuffed_pier) 

Variable for Percent of Containers Unstuffed Pierside 

(v_TAKE_Arr_Cycle) 

Variable for T-AKE Arrival Cycle 

(v_Univ_Stream) 

Variable for Setting Random Seed to a Universal Stream 


XIX 



THIS PAGE INTENTIONALLY LEET BLANK 


XX 



EXECUTIVE SUMMARY 


As military campaigns evolve, there are a limited number of constants that hold 
true. One of these constants is the importance of sustainment and logistics capabilities. 
During a military crisis in which direct engagement is required, the ability to put 
ordnance on a target is vital to the concept of power projection. The underlying ability to 
sustain these operations is a logistics problem, which includes the continuous flow of 
commodities, such as ordnance, to the area of operation (AO). This logistics problem is 
equally important as the tactical problem, but often not as explored as the tactical 
application of targeting the ordnance. The responsibility then falls on the military 
logistician to study and explore the current and future possibilities of sustaining 
military actions. 

Consider if there were a modeling tool that exercised the possible options when 
such an event arises. Then, the decision maker has a tool capable of guiding his or her 
decision, with respect to resource allocation, in order to effectively move ordnance 
through ports into an AO. The question then becomes how to provide our Auxiliary Dry 
Cargo/Ammunition Ships (T-AKEs) with the resources required to efficiently service our 
combatant ships. The forward-most port at which resources are received from the 
Continental United States (CONUS) and then readied for transfer becomes a key part of 
the answer to this question. This thesis studies this link—the logistic capabilities of 
moving ordnance into the Asian Pacific Theater—and provides a modeling tool to assist 
decision makers involved in ordnance operations. 

The Asian Pacific Theater is a vast area and presents many logistics challenges. 
Moving ordnance into this area depends on three major evolutions. The first of these 
evolutions is the movement of ordnance from CONUS to a forward logistics base. This 
movement is done by large container ships originating from a handful of possible 
sources. Figure 1 illustrates the general flow of ordnance into the Asian Pacific Theater 
via Guam and displays how the movement of ordnance into the Asian Pacific is funneled 
into and through a single point of entry into the AO. 


XXI 




Alaska 


KOREA 

CWohM 


CHINA 


S»Mtjo* JAPAN 
• 0fcin<w4 / 


rMAlLANO 


PHILIPPINES « 


GUAM, 


AUSTRALIA 


Pacific Ocean 


Indian Ocean 


Figure 1. AO and Flow Paths of Inbound Ordnance. 

(After: Helber, Hastert & Fee, 2003). 

The second evolution is the processing of the containerized ordnance into 
palletized ordnance at the forward logistics base. These ordnance operations serve to 
process the ordnance for delivery to combatants. The ordnance operations for the AO of 
concern here are conducted on Guam. Guam’s location is significant, as it is the 
western-most U.S. territory with the physical facilities capable of offloading, storing, and 
loading large amounts of ordnance. If Guam is eliminated as an ordnance operation 
resource, the Navy’s next western-most capable facilities are in Hawaii, which results in 
a 3,320-nautical-mile difference in forward presence. Figure 2 provides an aerial view of 
the thesis study area and locations of interest for ordnance operations. 


xxii 








Figure 2. Map of Orote Peninsula Area, Guam (From: Goode & Smith, 2007). 

The third evolution is the delivery of break-bulk ordnance to combatants at the 
forward edge of the battle. This task is carried out by Combat Logistics Force (CLF) 
ships. The specific CLF ship used in this thesis is the United States Navy’s newest class 
of underway replenishment ships, the Auxiliary Dry Cargo/Ammunition Ship, Lewis and 
Clark Class (T-AKE). The T-AKE is designed to deliver ammunition, stores, and fuel to 
carrier and expeditionary strike groups. These new ships keep combatant ships at sea, on 
station, and combat-ready in any scenario. 

The scenario established in this thesis is that the United States has become 
involved in a major military contingency in Asia and that T-AKEs are supporting the 
sea-based operation of a Maritime Prepositioned Force (MPF) squadron and its Marine 
Expeditionary Brigade (MEB) ashore. During such a contingency, the flow of supplies 
through Guam to forces at sea, or forces supported from the sea, is of critical importance. 
In order for T-AKEs to support only the sea-based operation of an MPE squadron and its 

MEB ashore, earlier studies have estimated how often they might have to go to port for 

xxiii 



resupply. In the case of a major military contingency, T-AKEs would also be supporting 
Carrier Strike Groups and other naval units. This translates to increased traffic intensity 
seen by the resupply port supporting the T-AKEs. 

Given the scenario, a systems analysis of the major forward ordnance supply node 
of Asian Pacific Theater operations is conducted in an effort to answer these questions: 

• How will introducing the competing requirements affect the predicted 
capabilities of the ordnance operations in Guam? 

• What are the critical factors in the ordnance operations process? 

Specifically, how do the competing requirements on the ordnance resupply 
process in Guam relate to other Department of Defense (DoD) needs to utilize the 
ordnance wharf, as well as their increase in ordnance requirements? 

To answer these questions, the system is modeled using the discrete-event 
simulation package from Rockwell Software, Arena version 10.00. The focal point of the 
model structure is on the ordnance operations (specifically at the ordnance pier. Kilo 
Wharf) on Orote Peninsula, Guam. The available resources are varied within the 
simulation to account for differences in processing performance characteristics and 
operations. The Arena modeling environment is a powerful modeling tool that enables 
the creation and running of experiments on models of systems. An Arena simulation has 
a framework that consists of an entity-based simulation that can be data farmed within a 
design of experiments (DOE) environment. This allows for the simultaneous 
examination of multiple factors and explores the high-dimensional relationships of these 
factors. Through the use of an interchangeable, component-based architecture, the 
simulation provides the user with extensive capabilities to modify entities, configurations, 
simulation parameters, and select data output collected. Arena, using a low-resolution 
approach, runs fast and is easy to set up, which is advantageous in performing many 
analytical runs for comparison and exploration of the landscape of possible outcomes. 

Use of the DOE approach to support the analysis of forward logistic capabilities 
provides data upon which quantitative analysis of the model is conducted, specifically 


XXIV 



looking at the effects of multidimensional, variable changes in an effort to estimate the 
effect on the frequency with which the T-AKEs could reload in Guam and the overall 
throughput of ordnance in the system. 

The experimental design includes five scenario sets: two of which are baselines, 
while three are built with the DOE approach. The simulation model built in Arena 
contains the flexibility to accommodate a number of scenarios using the same general 
framework for all the previously mentioned scenarios. Adjustments are either made in 
the Process Analyzer, through the matrix of input parameters, and/or directly in the 
model itself. 

The simulation experiment results show that introducing two forms of viable 
competition, based on previous years’ data and projected demands to the system, has a 
significant effect on both the T-AKE service level (the ratio of T-AKEs that leave the 
system to those that enter the system) and pallet throughput of the system. The impact of 
these effects holds true for the current system and the system that includes the new 
magazine on Orote Peninsula. T-AKE service level in the current system is reduced by 
an average value of 26% reduction in service level with a maximum value of 52%. This 
means that on average 1 of every 4 T-AKEs that enter the system is not serviced by the 
system. The T-AKEs not serviced at the end of the simulation time are left in queue. 
Pallet throughput is reduced by a maximum of 41,167 pallets and an average of 13,555 
pallets. This reduction in pallet output is equivalent to approximately four T-AKEs’ 
worth of ordnance that is not delivered to the forward edge of the contingency. 

Regression analysis and partition tree analysis are used to analyze the simulation 
experiment results. Across the current and new systems, the primary critical factor for 
both is the T-AKE arrival cycle. A greater T-AKE arrival cycle input (less frequent 
arrivals) consistently causes the system to see a reduction in pallet throughput. The 
analytical results also suggest that setting the arrival cycle of the T-AKE and the 
Ordnance Container Ship (OCS) to the same interval, but with sufficient offset, reduces 
the impact of the competing requirements introduced to the system. The trade-offs to the 


XXV 



optimal setting of the OCS and T-AKE arrival cycle are an increase in the number of 
containers offloaded from an OCS and a significant reduction in the number of containers 
unstuffed at Kilo Wharf. 

Both competing requirements are contributors to the effects on the system, but 
never the most influential. The impact from competing ships was more often seen 
affecting the T-AKE service level, whereas competition for ordnance affected the overall 
pallet throughput. The analytical results suggest that, during a time of contingency, T- 
AKE service level is improved by implementing policies that result in the mean arrival 
rate of competing ships by more than one arrival every 30 days. It also suggests that 
keeping the competition for ordnance under 26% of the total containers offloaded 
improves pallet throughput. 

Eastly, the analysis clearly shows that improving the system’s performance is not 
dependent on the distance of ordnance storage facilities from the wharf, but rather in the 
volumetric capability of the system, as defined by available resources and specific 
policies. The results for the new magazine are not practically significant enough in the 
model, as compared to the current system, to justify a large infrastructure investment 
alone. However, safety requirements to the general public and our forces, with respect to 
ordnance on Guam, are factors not considered in this model, but are actually influence 
investment decisions. 


XXVI 



ACKNOWLEDGMENTS 


These last nine months of thesis work have been one of the more challenging and 
interesting periods of my life. First, I would like to thank God for this opportunity. I also 
thank God for blessing me with my wife, Michelle, and daughter, Natalie, who supported 
me during this mentally and emotionally challenging time. If not for your patience and 
unwavering support, this thesis may never have been completed. 

Secondly, I would also like to acknowledge those who I most closely worked with 
during this thesis. Without the guidance, patience, technical expertise, and 
encouragement of my advisors. Professor Tom Lucas and Professor David Kelton, as 
well as my second reader. Professor Keebom Kang, I would not have completed such an 
undertaking. And last, but never least, I thank Colonel Ed Lesnowicz, USMC (Ret.) for 
his “motivational” words of encouragement. I would also like to acknowledge the 
students and faculty of the SEED Center. The support and flexibility you all showed in 
support of this thesis is greatly appreciated. 

There are many people outside of the Naval Postgraduate School who supported 
me. To the all folks in Guam who provided much of the data, your support was vital to 
accomplishing this task. A special acknowledgement goes to Rey Valdez from 
NAVEACMARIANAS for his time and contributions during my experience tour. 

Einally, I cannot finish without acknowledging the guys in my Cohort, the “Seven 
Deuce.” Gentlemen, you made this a fun ride! 


xxvii 



THIS PAGE INTENTIONALLY LEET BLANK 



I. INTRODUCTION 


My logisticians are a humorless lot... they know if my campaign fails, they 
are the first ones I will slay. 

Alexander 


It is in Asia where the United States will face its largest geopolitical 
challenges in the years ahead. 

Representative James Leach 
Former Chairman of the House Subcommittee on Asia and the Pacific 

September 21, 2006 
(Vaughn, 2007) 

As military campaigns evolve, there are a limited number of constants that hold 
true. One of these constants is the importance of sustainment and logistics capabilities. 
During a military crisis in which direct engagement is required, the ability to put 
ordnance on a target is vital to the concept of power projection. The underlying ability to 
sustain these operations is a logistics problem, which includes the continuous flow of 
commodities, such as ordnance, to the area of operation (AO). This logistics problem is 
equally important as the tactical problem, but often not as explored as the tactical 
application of targeting the ordnance. The responsibility then falls on the military 
logistician to study and explore the current and future possibilities of sustaining military 
actions. Consider if there were a modeling tool that exercised the possible options when 
such an event arises. Then, the decision maker would have a tool that could guide 
decision making, with respect to resource allocation, to effectively move ordnance 
through ports into an AO. This thesis studies the logistic capabilities of moving ordnance 
into the Asian Pacific Theater. 

A. BACKGROUND AND MOTIVATION 

In 2006, the United States Navy introduced its newest class of underway 
replenishment ships, the Auxiliary Dry Cargo/Ammunition Ship, Lewis and Clark Class 


1 



(T-AKE), to replace the aging combat stores and ammunition ships. This Combat 
Logistics Force (CLF) asset is designed to deliver ammunition, stores, and fuel to carrier 
and expeditionary strike groups (General Dynamics/NASSCO, 2007). These new ships 
will keep combatant ships at sea, on station, and combat-ready by providing a one-stop 
shopping source for replenishment. The combat logistics power of dry cargo/ammunition 
ships allows the United States Navy to provide critical logistics capabilities in today’s 
dynamic maritime environment. The United States’ ability to remain the preeminent 
naval power is enabled by our forward presence—our combat logistics ships are critical 
to this capability. The concept of operations (CONOPS) of this capability is illustrated in 
Figure 3. Shuttle ships cycle from resupply port to station ships, which serve as on-site 
logistic ships for a battle group. This CONOPS allows the battle group to travel freely, 
while maintaining a logistic line of communication for resupply. 



Figure 3. CONOPS for Battle Group Replenishment 
(From: Markle & Wileman, 2001). 


The question then becomes how to provide these vessels with the resources 
required to efficiently service our combatant ships. The forward-most port at which 
resources are received from the Continental United States (CONUS) and then readied for 
transfer becomes a key part of the answer to this question. Since forward presence and 
naval strength is predicated on the ability to put ordnance on target, this thesis focuses on 

2 





the movement of ordnance into a theater of concern. The insights gained from this thesis 
may identify factors that improve the ordnance operations performance in ports of 
interest and provide indications of where equipment, personnel, and processes could be 
improved. 

B. OBJECTIVES 

The primary objective of this thesis is to conduct a thorough analysis, by use of 
simulation, into the capabilities and critical path scenarios of resupplying T-AKEs in 
Guam during a military contingency. It is done in partnership with Naval Surface 
Warfare Center, Carderock Division (NSWC Carderock) and the Director of the Strategic 
Mobility and Combat Logistics Division for the Chief of Naval Operations 
(OPNAV N42). In addition to the above, this thesis intends to provide recommendations 
for resource allocation and system flow path changes. A secondary objective is to 
provide a model that can be utilized in future analysis as a template for any given port. 

C. BENEFITS OF THE STUDY 

This thesis supports the Navy by conducting systems analysis of the major 
forward ordnance supply node of Asian Pacific Theater operations in an effort to answer 
the question: “How will introducing the competing requirements affect the predicted 
capabilities of the ordnance operations in Guam?” and “What are the critical factors in 
the ordnance operations process?” It incorporates the effects of competing requirements 
on the ordnance resupply process in Guam, specifically related to other Department of 
Defense (DoD) needs to utilize the ordnance wharf, as well as their increase in ordnance 
requirements. Previous studies have analyzed the capabilities of the island transportation 
infrastructure (Military Surface Deployment and Distribution Command [MSDDC], 
2008), port operations pier-side (Goode & Smith, 2007), and optimization of the combat 
logistics force (Brown & Carlyle, 2007). This thesis combines some of the methods used 
in these approaches in an effort to provide a comprehensive model that moves ordnance 
from CONUS locations to the theater of interest. Additionally, this thesis produces a tool 
capable of being applied to other theaters of interest and future capability gap studies. 


3 



Current areas of interest for this type of research include infrastructure development, 
resource procurement and allocation, and policy decision-making processes. 

D. METHODOLOGY 

This thesis uses the discrete-event simulation package from Rockwell Software, 
named Arena, to model the port operations in support of resupplying T-AKEs in response 
to a military contingency in the Asian Pacific Theater. The focal point of the model 
structure is on the ordnance operations (specifically at the ordnance pier. Kilo Wharf) on 
Orote Peninsula, Guam. The available resources are varied within the simulation to 
account for differences in process performance characteristics and operations. The Arena 
modeling environment is a powerful modeling tool that enables the creation and running 
of experiments on models of systems. An Arena simulation has a framework that 
consists of an entity-based simulation that can be data farmed within a design of 
experiments (DOE) environment. This allows the simultaneous examination of multiple 
factors and explores the high-dimensional relationships of these factors. Through the use 
of an interchangeable, component-based architecture, the simulation provides the user 
with extensive capabilities to modify entities, configurations, simulation parameters, and 
data output. Arena, using a low-resolution approach, runs fast and is easy to set up. The 
Arena model can perform many analytical runs for comparison of more possible mixes. 

This thesis uses a DOE approach to support the Navy analysis of forward logistic 
capabilities and provide quantitative analysis of problem feasibility. Use of the model 
provides data upon which analysis of the model is conducted, specifically looking at the 
effects of multidimensional, variable changes in an effort to estimate the impact on the 
frequency with which the CEE ships, particularly T-AKEs, could reload in Guam. 

The DOE approach allows the user to vary a large number of factors 
simultaneously, and thus gain insight into the drivers of T-AKE resupply effectiveness 
and overall ordnance throughput. This enables the researcher to identify, compare, and 
contrast current methods and viable optional methods to optimize T-AKE reloading times 
and/or ordnance throughput, given a multitude of variable settings. 

The flow of this thesis is as follows. Chapter II explains the model development 

and the assumptions used in the model. This includes introducing the scenario used in 

4 



the model. Chapter III introduces the supporting data and methodology of the analysis 
applied to the simulation. Chapter IV presents the analysis and resulting insights. 
Chapter V provides conclusions and recommendations based on the analysis. 


5 



THIS PAGE INTENTIONALLY LEET BLANK 


6 



II. SCENARIO DEVELOPMENT 


A. INTRODUCTION 

In order to provide military relevance to the analysis, a plausible scenario is 
explored. This chapter provides a brief introduction to ordnance operations, to include 
the offload, storage, and onload of ordnance. The scenario that follows is developed 
through a combination of previous studies and plausible forecasting. Following the 
scenario development is a description of the Arena simulation tool that is used to model 
and analyze the scenario. This chapter concludes with a detailed description of the 
behavior of the simulation model. 


B. WHAT ARE ORDNANCE OPERATIONS? 
I. Overview 


In Chapter I, the term “ordnance operations” is introduced. As used in this thesis, 
ordnance operations describe a process of moving ordnance from CONUS to a theater of 
concern. In Figure 4, ordnance operations are simply illustrated as a flow path of 
processes. All of the terms below are thoroughly described in this chapter. 


Arriving 

Ordnance 


Offload 


Unstuff 


Stowage 


Onload 


Departing 

Ordnance 


Figure 4. Ordnance Operations Flow Path Diagram. 

2. Study Area Location and Facilities 

Guam’s location is significant as the western-most U.S. territory with the physical 
facilities capable of offloading, storing, and loading large amounts of ammunition. 
Figure 5 provides an aerial view of the thesis study area. 


7 







Figure 5. Map of Orote Peninsula Area, Guam (From: Goode & Smith, 2007). 


This thesis foeuses on the ordnanee operations that oeeur at the United States 
Navy Base, Guam. The base sits on Orote Peninsula near the mouth of Apra Harbor and 
includes the Kilo Wharf, Buoy 702, and the Ordnance Handling pad. The Ordnance 
Annex, another important location in the ordnance operation, is located on the south 
central part of the island, southeast of the Apra Harbor Naval Complex. This section will 
describe both locations and their roles in ordnance operations on Guam. 

a. Kilo Wharf 

Kilo Wharf is located at the entrance to Apra Harbor on the north side of 
Orote peninsula. It is the primary facility for ordnance loading and unloading. The 
wharf is able to accommodate a single ship at any given time. Ships carrying or handling 
large amounts of ordnance, such as CLF ships and aircraft carriers, must use Kilo Wharf 
because of the Net Explosive Weight (NEW) limit of three million pounds and 


8 



Explosives Safety Quantity Distance (EQSD) of 7,210 feet (MSDDC, 2008). Eigure 6 is 
an aerial view of Kilo Wharf from a northeast perspective. 



Eigure 6. Aerial View of Kilo Wharf (Erom: MSDDC, 2008). 

b. Buoy 702 

Buoy 702, at the northern edge of outer Apra Harbor, is the designated 
anchorage for fuel and ordnance-laden vessels waiting to dock at Kilo Wharf. This 
anchorage serves as the only standby location for vessels with more than 25 short tons of 
explosives. If there is a requirement for immediate berthing of a vessel with more NEW 
than allowable in the inner harbor, then Naval Munitions Command (NMC) East Asia 
Division, Detachment (Det) Guam must request a waiver. Accumulation of these waivers 
is not considered good practice (Naval Message, 2007). 

c. Ordnance Annex 

The Ordnance Annex is approximately 8,800 acres, making it the major 
ammunition magazine on Guam. The annex is also the location of NMC East Asia 
Division, Det Guam, and the joint venture formed by Day and Z immerman Services Inc., 
SKE Support Services Inc., and Parsons Infrastructure and Technology Group Inc. 
(DZSP) 21. NMC East Asia Division, Det Guam is the command responsible for 
ordnance operations on Guam, and DZSP 21 is the service contractor that provides 
ordnance management services to NMC East Asia Division, Det Guam. The annex has 

9 




over 100 storage magazines, providing a total NEW eapaeity of greater than 57 million 
pounds. The annex also has 10 open storage/staging areas eapable of handling 
approximately 725,000 pounds of ordnanee. It is important to note that the travel route to 
and from Kilo Wharf is along publie roads, and passes near residential areas and an 
elementary school. This exposes the local community to a portion of the ordnance 
operations—an inherently dangerous undertaking. Figure 7 is an example of the 
magazine facilities that are found at the Ordnance Annex. 



Figure 7. Igloo Storage Magazine at Ordnance Annex 
(From: MSDDC, 2008). 

d. Ordnance Handling Pad 

The Ordnance Handling Pad is located approximately one-half mile from 
Kilo Wharf. Its purpose is to serve as an area to relieve the constraints of unstuffing on 
Kilo Wharf itself. The 40,000-square foot concrete pad, constructed with lightning 
protection, is capable of holding 30 to 35 Twenty-foot Equivalent Unit (TEU) containers 
when being used to unstuff containers. When used strictly as a storage space for 
overflow containers, the pad stores up to 200 TEUs. Figure 8 is an image of the 
Ordnance Handling Pad that includes one of the corner posts used to elevate exposed 
cables for lighting protection. 


10 







Figure 8. Ordnance Handling Pad at Orote Peninsula 
(From: MSDDC, 2008). 


3. Operations 

Ordnance operations begin with incoming containerized ordnance arriving on 
container ships from CONUS. The next stage in the process is the subsequent unloading 
of the containers to the pier, at a wharf that is qualified to handle ordnance. After the 
containers are unloaded, they are subject to a number of processes. One of these 
processes is simply transportation to an end destination, where they complete the portion 
of the process that this thesis covers. All Navy CLF vessels are designed as break-bulk 
ships carrying only palletized material. Therefore, the Navy does not deliver 
containerized ordnance to the combatants and all containers must be open and emptied 
(Goode & Smith, 2007). This process is known as unstuffing. This process may occur at 
the pier itself or at another location after a container has been transported to an authorized 
location. Once the unstuffing process is complete, the next process is stowage. In order 
to reach a stowage location, the palletized ordnance must again be moved to the stowage 
facility. Stowage is simply the retention of palletized ordnance in an authorized space. 
The last process of ordnance operations is the loading of the palletized ordnance, often 
referred to as the “onload.” This occurs at the ordnance wharf, and involves loading the 
palletized ordnance onto a CLF vessel for delivery to combatants in the AOR. Figure 9 
demonstrates the general flow process of ordnance operations. This thesis excludes the 


11 









palletized stores portion of Figure 9, based on the requirement that no other operations 
oeeur during ordnanee operations. Sinee all ordnanee operations oeeur during the day, 
the result is that the stores loading operations take plaee at night. Therefore, the stores 
operation is assumed to not interfere with the ordnanee operations and is thus outside the 
bounds of this thesis. 


Container ship 


a 


T-AKE 


IF 


Kilo wharf_ 





linctjffng 

Pa «tzed 
adnaoc* 


Containers 


Lead 

Ira e'B 



1 ID 


Load . , Load 

Kilo wharf 


Pa «tzad 

• 

• 

• 

• 

• 

Pa atzad 


a 

• 

• 

• 

a 

sio'as 


T'an*- 

OO'l 


T-ant- 
00 “1 




O 

O 

Ordnance 

Lead 


a 


Annex 

ra fi 


a 

a 

a 





a 

a 

a 



Daytime operations 


Tra'is- 

OO'l 


Load 
t'a 6'$ 


U I 


DC'Oj’.on 

Oaool Qja’n Va'aias 
(DOOM) 
I^djft'a 
Sjoood Can*.*' 
FISC) 


Nighttime operations 


Figure 9. Ordnanee Flow from Container Ship to a T-AKE Including the Dry Stores 

Component of any Replenishment Period 
(From: Goode & Smith, 2007). 


a. Offload 

The offload process commences when an ordnance-laden container ship 
arrives at the berth at Kilo Wharf. An ordnance-laden container ship is capable of 
carrying thousands of TEU containers. Each of these containers is estimated to carry 
between 12 and 14 pallets, which is equivalent to a standard ordnance load for a TEU of 
13.9 short tons (Goode & Smith, 2007). The Kilo Wharf does not have an organic 
container crane, so arriving vessels are required to have their own crane(s) for offloading 
containers. Once pier-side, the containers are offloaded to the pier. At that point, 
container disposition could be conducted using one of three options. Eirst, the container 


12 


















































is moved onto the pier to an adjacent area for unstuffing. Second, the container is moved 
to a nearby ordnance-handling pad. Third, the container is moved to the Ordnance Annex 
for handling. Lastly, the container is delivered to Andersen Air Force Base (AAFB), 
located on the north end of Guam. Table 1 shows offload rate data (Goode & Smith, 
2007 and MSDDC, 2008). 

Table 1. Offload Rates (From: Goode & Smith, 2007, and MSDDC, 2008). 



Offload rate 

Days to offload' 

Estimate based on: 

Containers/day 

stons/day 


2003 study 

45 

625.5 

6 

TurboCADS 05 

51 

708.9 

5 

PacFIt planning factor'^ 

75 

1042.5 

4 

MSDDC Guam Ammunition Distribution Study‘s 

95 

1226.1 

3^ 


a. Rounded up 

b. Used to determine personnel augmentation from Expeditionary Logistic Support Force (ELSF) 

c. Determined by simulation 

d. Extrapolated using CNA report data 


b. Handling (Unstuffing) 

Unstuffing is done in conjunction with an inspection and inventory of the 
ordnance that is removed from each container. The ordnance units that are removed from 
the containers are in pallets. This thesis only considers ordnance to the smallest unit of 
pallet. In this thesis, all Navy containers of ordnance will be unstuffed, and all pallets 
inspected and inventoried as a part of the handling process. The inventory and inspection 
can only be done by qualified personnel. This adds a constraint to the palletized 
ordnance process flow. 

Handling done at AAFB is considered outside of the bounds of this thesis, 
but will be modeled for continuity and accountability of all incoming ordnance. 
Containers designated for the Air Force are moved to AAFB and complete their flow 
path. The only impact to the ordnance operations caused by these containers is the 
amount of resources required to transport the containers. See Table 2 for the container 
unstuffing rates and days to unstuff 3,450 short tons (one T-AKE load equivalent) 
(Goode & Smith, 2007). 


13 




Table 2. 


Unstuffing Rates (From: MSDDC, 2008). 


Number of containers 

unstuffed 

simultaneously 

Unstuffing rates 

Inspectors per Containers 

container per day 

Stons/day 

Days to 

unstuff 

1 

1 

6 

83.4 

43 

1 

9 

24 

333.6 

11 

2 

9 

48 

667.2 

6 

a. Rounded up 






c. Moving Ordnance 

Both containerized and palletized ordnance must be moved to a storage 
location at particular points in the ordnance operation process. Containerized movement 
requires both a Container Capable Forklift (CCF) and a Container Capable Truck (CCT). 
The CCF is required for movement on the Kilo Wharf and for loading to the CCT for 
transport. Currently, there are two operational CCFs available for ordnance operations at 
Kilo Wharf. Containerized movements occur from the wharf to the annex and to AAFB. 
Movements to AAFB can occur in three possible routes, whereas movements to the 
annex are by a single route. Palletized movement requires that the ordnance must be 
secured to a Pallet Transport Truck (PTT) by building a frame around the pallets, also 
known as block and brace loading (Goode & Smith, 2007). Palletized movements are 
generally between the wharf and the annex in both directions. Table 3 provides the 
distance set and estimated travel times. 


14 






Table 3. Movement Distance (in statute miles) and Travel Times 

(From: MSDDC, 2008). 

Distance 

_ Movement _ Min _ Max _ Average Travei Time 

Kiio Wharf-AAFB 25.1 29.1 50- 75 minutes 

Kiio Wharf - Ordnance Annex _73__ 25 - 35 minutes 

a. No min and max because only one possible route. 

d. Stowage 

Ordnance stowage is carried out primarily at the Ordnance Annex. 
Ordnance is occasionally stowed on the Kilo Wharf or at the Ordnance Handling Pad 
while awaiting handling. All ordnance stowage is constrained by NEW limits 
and EQSDs. 


e. Onload 

Once pallets are delivered to the wharf from their stowage location, the 
loading process, called the “onload,” begins. Onload requires ordnance material handling 
equipment (MHE) to load ordnance onto the T-AKE. The average onload rate calculated 
by the Center for Naval Analyses (CNA) in previous studies was 69.25 pallets per hour. 
Using a nominal weight of one ton per pallet, the rate for a 12-hour workday would be 
about 831 short tons of ordnance (Goode & Smith, 2007). Without access to the data 
used to make these calculations, the standard deviations are not available; therefore, these 
values are used in the model with a uniform distribution that varies slightly from the 
estimated rates. 

C. SCENARIO DESCRIPTION 
I. Overview 

When conducting a simulation study, it is imperative to use realistic scenarios that 
allow the analyst to measure factors of interest in a way that is sensible to decision 
makers. Eogistics planning is often done in advance of any known military contingency. 
This is done to ensure logistic capability gaps are discovered prior to any action. In order 
to draw on a plausible scenario, the basic outline for the scenario was obtained from two 

15 



previous studies. The first was done by the CNA in 2007, at the request of the Director 
of the Strategic Mobility and Combat Logistics Division, to estimate the flow rate of 
supplies, with emphasis on ordnance, through Guam in surge conditions. The second 
study was conducted by Military Surface Deployment and Distribution Command 
(MSDDC), Transportation Engineering Agency in 2008. The purpose of their study was 
to conduct an assessment of Guam’s transportation infrastructure and the ordnance 
operations in Guam under surge conditions. The purpose of this section of the thesis is to 
relate that scenario to potential consumers of this research. This provides a strong 
foundation for why this thesis is applicable to the Navy. The following is a brief 
synopsis of the scenario that forms the basis of the simulation model. 

2. General Situation 

The scenario established in this thesis is that the United States has become 
involved in a major military contingency in Asia and that T-AKEs are supporting the 
sea-based operation of a Maritime Prepositioned Eorce (MPE) squadron and its 
Marine Expeditionary Brigade (MEB) ashore. During such a contingency, the flow of 
supplies through Guam to forces at sea, or forces supported from the sea, is of critical 
importance. In order for T-AKEs to support only the sea-based operation of an MPE 
squadron and its MEB ashore, earlier studies have estimated how often they might have 
to go to port for resupply (Goode & Smith, 2007). In the case of a major military 
contingency, T-AKEs would also be supporting Carrier Strike Groups and other naval 
units. This translates to increased traffic intensity at the resupply port supporting 
the T-AKEs. 

This increased traffic intensity is driven by the increase in demand for logistic 
support of ordnance. In order to meet this demand, more material must be shipped from 
CONUS to Guam to replenish the stock on Guam that diminishes as the demand of the 
combatants is met. The ordnance shipped from CONUS is delivered to the berth at 
Kilo Wharf in Guam. Once delivered, the ordnance is unloaded and processed pier-side. 
Occurring in the same period, the T-AKEs are arriving at Kilo Wharf to pick up ordnance 
for deliver to meet combatants demand. This again increases the traffic intensity seen by 

the forward logistics port. To complicate the scenario, yet also add a realistic approach to 

16 



it, this thesis includes the competing requirements for use of the wharf by vessels other 
than the ordnance container ships and T-AKEs. Figure 10 illustrates the general flow of 
ordnance into the Asian Pacific Theater via Guam. 


100 E 


95 W 



CONUS 


CHINA 


• Yokotui 
JAPAN y. 


• Okinawa 


INDIA 


THAILAND 


PHILIPPINES ♦ 


GUAM, 


SbQapora 


AUSTRALIA 


Figure 10. Area of Operations and Flow Paths of Inbound Ordnance. Note: The flow 
paths of incoming ordnance are based on the sources of ordnance supply 
(After: Helber, Hastert, & Fee, 2003). 


D. THE ARENA SIMULATION TOOL 

Now that the scenario has been described, this section describes the Arena 
modeling and simulation environment, a tool for creating entity-based, process-driven 
simulations, and why it was chosen. In Chapter III, the implementation of the scenario in 
Arena is covered. Readers interested in a detailed technical description of the software 
should consult the textbook Simulation with Arena (4^’’ Ed.) by Kelton, Sadowski, & 
Sturrock (2007) or the user’s manual, which can be downloaded from the Rockwell 
Automation Website at http://www.arenasimulation.com/. 


17 







1. Why Arena? 

Arena is the modeling environment selected for the development of the logistics 
process used in this thesis. The Arena modeling and simulation tool was chosen because 
of its focus on process improvement, ease of use, and applicability to logistics problems. 
It is a commercial product based on the SIMAN simulation language developed in 1983 
by Systems Modeling Corporation, who also developed Arena in the mid-1990s. 
Systems Modeling was acquired by Rockwell Software in 2001 and they still support and 
develop Arena. Arena is simple in design; thus, any process that can be described by 
means of a flowchart can be simulated with Arena. As a modeling tool, it is very 
effective when analyzing manufacturing processes or flows. Arena was also chosen 
because it provides 2-D model animation. This feature is instrumental in the 
demonstration of the model in the debugging process. Providing visual support of 
process flow modeled in the simulation enhances credibility and ease of understanding 
for decision makers. 

The Arena software lends itself to modeling a variety of scenarios involving 
queuing processes. Recent applications of the Arena software include Naval 
Postgraduate School (NPS) theses and projects that analyze real-world applications, such 
as homeland defense research, unmanned aerial vehicle material reliability, and maritime 
interdiction operations. Contact information for readers interested in more information 
regarding Arena is found at http://www.arenasimulation.com/support. 

2. Characteristics of the Arena Simulation Environment 

Arena is a discrete event-driven, entity-based simulation environment that 
provides an intuitive, flowchart-style environment for building an “as-is” model of a 
process (Rockwell Automation Inc., 2005a). Arena simulation software is an effective 
modeling tool when analyzing complex, medium- to large-scale projects involving 
logistics, distribution, warehousing, and service systems. Arena provides the user with 
the ability to create custom templates for complex, repetitive logic; to simplify model 
development; and reduce model development time. In addition. Arena is used to create 
customized simulation modeling templates focused on specific applications or industries 


18 



(Rockwell Automation Inc., 2009). Arena is also easily capable of performing data 
farming techniques, which give it the ability to explore many input parameters. 

E. CHARACTERISTICS OF THE SIMULATION MODEL 

This section describes the basic characteristics of the Arena simulation model 
developed for this thesis. It starts with a description of the simulation’s goal, followed by 
an overview of the model at a conceptual level. Following the conceptual description are 
detailed descriptions of the component modules in the model. A detailed breakdown of 
the functional specifications of the model is contained in the Appendix, 
Functional Specification. 

1. Goals and Measures of Effectiveness (MOEs) 

The simulation models the military contingency scenario in this chapter as the 
sustainment of one year of operations. The length of the simulation run is easily 
adjustable for modeling longer or shorter periods of sustained operations. The ultimate 
goal of the simulation is measuring the impact of competing requirements on the 
effectiveness of ordnance operations in Guam. These competing requirements come in 
two forms: 

• Competition for the wharf space by vessels not engaged in either the 
offload or onload of ordnance. 

• Competition for the ordnance offloaded from the Ordnance Container 
Ships (OCSs). The Air Force will need to replenish their diminished 
munitions as well. Therefore, as an approximation, a percentage of the 
incoming ordnance loads will begin to be diverted to the Air Force. The 
Air Force requirements in the model are also a proxy for all other DoD 
requirements the system could possibly face. 

The MOEs that are used in this thesis are T-AKE Service Eevel (the ratio of 
T-AKEs that enter the system to those successfully served by the system) and overall 
ordnance throughput (measured as the number of pallets that leave the system). These 
MOEs directly relate to the combat effectiveness of the combatants because as the 
customer they dictate the operational demand for ordnance. Other measures of interest 
include, but are not limited to, the following: time in queue for entities, number of 


19 



containers of ordnance processed, number of pallets of ordnance processed out, 
equipment utilization, and resource utilization. Using data farming techniques allows for 
analysis of these and other factors. 

2. Conceptual Model 

The overall concept of the simulation model is represented as an inventory 
queuing model. A basic queuing model consists of customers who arrive for a service, 
servers who provide the service, an inventory available to the servers, and a warehouse 
where additional inventory is stored. In this model, the customers are OCSs, T-AKEs, 
and Competing Ships (CSs). The model considers the OCSs and the T-AKEs as primary 
customers because impacts to their operations will directly affect the combat capability of 
the fleet combatants. Although they are secondary customers, the CSs are not ignored 
because they are a realistic component in the model. The service that all vessels require 
is twofold. The required primary service is use of the wharf. The required secondary 
service is based on customer (ship) type. Successful service of an OCS is complete 
delivery of its ordnance load. This will increase the inventory level maintained at the 
Ordnance Annex (warehouse). Eikewise, a T-AKE that receives its requested ordnance 
load is a successful service. This will decrement the inventory maintained at the annex. 
This thesis considers maximum service of T-AKEs as optimal. Successful CS service is 
simply usage of the wharf and departure. The server is a combination of the Kilo Wharf 
and the ordnance operations required by the particular ship at the server. An effective 
service is considered to be a vessel served and, therefore, that MOE is the number of a 
particular vessel type served, divided by the total number of vessels to enter the system. 
In other words, the effectiveness of the process in its entirety is measured by how well its 
primary customers are served. 

3. Key Components of the Model 

This section describes some of the key components found in Arena simulation 
models, with emphasis on components widely used in this thesis. 


20 



a. 


Entities 


Entities represent the objects moving through the system. Entities are 
built into the system using the Create Module. Each entity has its own characteristics, 
referred to as attributes. An entity is assigned as many attributes as required for the 
different types of entities in the system. Each individual entity in the system has its own 
values of these attributes; these may be assigned at the various processes it encounters 
(Rockwell Automation Inc., 2005b). The assignment of attributes for entities is 
accomplished through an Assign Module. Eigure 11 represents a Create Module and 
associated graphic user interface (GUI), which allows for specific entities to be created 
and enter the system. 




Eigure 11. The Create Module and Create GUI in Arena. 

Eor example, all ships entering the model in this thesis are immediately 
given a minimum of two attributes. The first attribute is to indicate the time they entered 
the system, a_Arrival_Time_to_System with the current value of “time now” (TNOW), 
the current simulation time. The second attribute is a type identifier, a_Ship_Type, which 
simply indicates the type of ship entering the system. These attributes are later used by 
the model as a part of the process logic. Eigure 12 represents the Assign Module and 
Assign GUI that allows for specific entities to be assigned attributes that they carry 
through the system. 


21 



























Figure 12. The Assign Module and Assign GUI in Arena. 


b. Queues 


The primary purpose of a queue is to provide a waiting space for entities 
whose movement through the model has been suspended due to the system status (e.g., a 
busy resource). Queues are passive in nature; entities enter the queue and are removed 
from it based on the change in state of the system element associated with the queue (e.g., 
a resource) (Rockwell Automation Inc., 2005b). An example of a queue in this thesis is 
the one that is formed when the Kilo Berth resource is occupied. Figure 13 represents a 
Process Module and its associated Queue GUI. The process module is where queues are 
generated to indicate where an entity will wait, if required, for resources to complete the 
defined process. 


Process 'fetjetiff 



Idenlifief: 


Type 
r' Point 
(• Line 


OK 


~I\ 


F 

f Rotate 

r 


Cancel 


Colof... 


Help 


Figure 13. The Process Module and associated Queue GUI in Arena. 


22 











































There are two types of queues used by Arena. Individual queues have a 
symbolic name, a ranking rule, and a specific capacity. Entities in these queues may be 
displayed in the animation; statistics may be collected on them; they may be ranked using 
a flexible ranking rule mechanism; they may be collected into sets; and, when used with 
resources, they may be shared among modules (Rockwell Automation Inc., 2005b). 
Internal queues provide a basic first-in, first-out container for entities at a particular 
activity (module), but do not provide animation, statistics, or ranking mechanisms 
(Rockwell Automation Inc., 2005b). The queue of interest in this thesis is the queue for 
entities entering the system and is unlimited. This queue is discussed further in Chapter 
III. 


c. Resources 

Resources are stationary elements of a system that can be allocated to 
entities. They have a specified capacity (at any point in time) and a set of states (e.g., 
busy, idle, inactive, or failed) that they transition between during a simulation run. 
Resources may be used to represent people, machines, or even space in a storage area. In 
this thesis, resources include all three of the possibilities mentioned; ordnance inspectors, 
cranes, and storage and processing space. The terminology associated with resources is 
as follows: when an entity requires a resource, it seizes the resource; and when an entity 
no longer requires a resource, the entity releases it so that it is available to be seized by 
other entities. A resource has an associated queue to hold entities that try to seize the 
resource when it is unavailable (Rockwell Automation Inc., 2005b). An entity in the 
queue waiting for a resource will immediately seize the resource once available. Any 
transitional delays in resource seizure are accounted for in the process delays. Resource 
information is maintained in a data module as seen in Table 4. This data table allows the 
user to define the type and capacity of any given resource in the system. 


23 



Table 4. 


The Resource Data Module in Arena. 


Resource ■ Basic Process 



Name 

1 Tyire 

Capacity 

Busy Hour 

Idle Houi 

Pei Use 


f<llo Berth 

▼ | Fixed Capacity 

1 

0.0 

0.0 

0.0 

2 

Buoy 702 

Fixed Capacity 

1 

0.0 

0.0 

0.0 

3 

Ordnance Annex Magazine Storage 

Fixed Capacity 

99999999 

0.0 

0.0 

0.0 

4 

Crane 

Fixed Capacity 

2 

0.0 

0.0 

0.0 

5 

Pierside Staging Space 

Fixed Capacity 

2 

0.0 

0.0 

0.0 

6 

Container Truck Loading Space 

Fixed Capacity 

2 

0.0 

0.0 

0.0 

7 

Ordnance Inspector 

Fixed Capacity 

18 

0.0 

0.0 

0.0 

8 

Unstuffing Space 

Fixed Capacity 

120 

0.0 

0.0 

0.0 

9 

Block and Brace Crew 

Fixed Capacity 

10 

0.0 

0.0 

0.0 

10 

Ordnance Forklifts 

Fixed Capacity 

20 

0.0 

0.0 

0.0 


Double-click here to artd a new row. 


The capacity of a resource limits the number of entities that may seize it at 
any point in time. For instance, the wharf is a resource in the model that can only 
accommodate one ship. It is represented by a resource called Kilo Berth, having a 
capacity of one. An entity that seizes a resource is referred to as seizing a unit from its 
total capacity. Entities can seize and release multiple units of capacity (Rockwell 
Automation Inc., 2005b). 

d. Stations 

Systems typically have natural boundaries that suggest a systematic 
segmentation approach in forming their representation. For example, a manufacturing 
system is usually composed of a set of distinct workstations. Multiple workstations may 
then form a manufacturing line, and multiple lines form a manufacturing site (Rockwell 
Automation Inc., 2005b). 

Arena allows you to represent systems by first dividing them into the 
physical subsystems, referred to as stations, where the actual processing takes place. 
Thus, for example, each workstation in a manufacturing model can be represented by a 
station in Arena (Rockwell Automation Inc., 2005b). Figure 14 represents the Station 
Module, which provides the method for defining physical subsystems and process 
boundaries. 


24 


















station 1 



Figure 14. The Station Module and Associated Station GUI in Arena. 


e. Transporters 

Transporters are one type of device that moves entities through the 
system. They can be used to represent material-handling or transfer devices, such as fork 
trucks or delivery vehicles. Transporters can also be used to model personnel whose 
movement is important to modeling a system, such as a nurse or a food server. When 
transporters are used, you provide information defining the transporter’s speed and the 
travel distances between stations served by the transporter (Rockwell Automation Inc., 
2005b). 

The terminology associated with transporters is as follows: When an 
entity requires a transporter, it requests the transporter; then it is transported to its 
destination station (both transporter and entity move to the station together, and the entity 
enters the model at the module containing the destination station); and when the entity no 
longer requires a transporter, ii frees the transporter (Rockwell Automation Inc., 2005b). 

Animation transporter pictures show the movement of free-path 
transporters from station to station or of guided transporters from intersection to 
intersection. All transporters in this thesis are free-path transporters. Transporters can be 
idle, busy, or inactive, with different pictures for each state. Movement of free-path 

25 




















transporters occurs only between defined distances connecting stations (Rockwell 
Automation Inc., 2005b). Table 5 is an example of the transporter data module that 
defines the number, type, and velocity of each specific transporter. 

Table 5. The Transporter Data Module in Arena. 



Double-cick here to add a new row. 

4. Arena Simulation Time 

Since Arena is an entity-based simulation model, time advances only as directed 
by the entities as they encounter the models component modules. For example, if a 
process is defined to take a certain amount of delay to be complete, then the simulation 
will advance time when activated by an entity. That entity completes the process and 
moves on when that specific delay completes. A simple way to describe this type of 
modeling is to imagine walking the path of the process that is model. If along the path 
you encounter a process module that takes one day to complete, you will stay at that 
module for one day. Thankfully, Arena is able to advance time rapidly in its simulation 
process and thereby move a multitude of entities through a variety of processes that 
mirror real time delays. 

The Run Setup mode, as seen in Figure 15, provides a variety of setting options 
for application to the user’s specific system such as project parameters, run speed, 
replication parameters, run control, array sizes, and reports defined in the model. 


26 











Figure 15. The Run Setup Menu in Arena. 

5. Summary 

This thesis uses the Arena simulation tool to model realistic ordnance movements 
into an AOR through a forward supply node. The scenario used in this model was chosen 
because of its high visibility among logistic planners and because it is logistically 
challenging. The resulting model captures the essential components of ordnance 
movement and the operations necessary to gain insight into the effectiveness of the 
system when affected by competing requirements. 


27 
















































THIS PAGE INTENTIONALLY LEET BLANK 


28 



III. MODEL IMPLEMENTATION AND EXPERIMENTAL 

DESIGN 

A. INTRODUCTION 

This thesis makes use of a technique known as data farming, which was 
developed and used by the Simulation, Experiments, & Efficient Designs (SEED) Center 
at the Naval Postgraduate School (see http://harvest.nps.edu/). This technique provides 
the analyst with methods to explore the possible inputs in a more efficient manner. 
Specifically, the technique involves taking a simulation and running it many times, while 
simultaneously changing the input parameters. As the number of input parameters in the 
simulation increases, the analyst becomes challenged with the “curse of dimensionality,” 
a term coined by renowned applied mathematician Richard Bellman. This term describes 
the problem caused by the exponential increase in volume associated with adding extra 
dimensions to a (mathematical) space (Bellman, 1957). Data farming acknowledges this 
challenge. Instead of attempting to examine all the possibilities, data farming provides an 
output data set that allows the analyst to explore more of the landscape of possible 
outcomes in a mathematically intelligent fashion. This exploration leads to a better 
understanding of the initial problem and provides insight into which input factors, if any, 
have significant effects. 

This chapter starts by outlining the primary entities involved in the simulation and 
their assigned attributes. This is followed by describing the resources of interest and the 
variables chosen as input parameters for the simulation experiment in this thesis. Einally, 
this chapter describes experimental designs used to generate the data used to understand 
more completely the effects of competition on moving ordnance into the Asian Pacific 
Theater through Guam. 

B. PRIMARY ENTITIES AND ATTRIBUTES 

The entity is the primary participant in an Arena simulation. The entity is what 
travels through the simulated process and utilizes resources available in the system. An 
entity receives its identity through the process of attribute assignment. The naming 


29 



convention used in this thesis is that all attribute names begin with “a_” followed by the 
attribute name. This section describes each of these entity types and the attributes 
associated with each type. This thesis uses a total of five major entity types throughout 
the model: the Ordnance Container Ship (OCS), the TAKE Ship (spelled without the 
hyphen as a function of Arena-allowable naming conventions), the Competing Ship (CS), 
the Ordnance Container, and the Pallet. There is also one minor entity type. Entity 1, 
which is used in the system at the simulation finish time (TEIN). This entity is defined as 
minor because its only purpose is to initiate the ReadWrite process that writes the defined 
outputs to an Excel spreadsheet. 

Of the five major entities, the three ship types of entities are the first active 
entities to enter the system. An initial inventory of pallets actually enters the system 
before any of these entities, but remains inactive until the first T-AKE arrival. Aside 
from the system initialization with an inventory of pallets, all other pallets are not created 
for direct input into the system. The Ordnance Container Entity and the Pallet Entity are 
both generated as entities that result from the arrival of an OCS Entity to the system. 

The OCS Entity, as well the two other ship-type entities, is created at what would 
be considered the beginning of the process. Upon creation, the OCS Entity is 
immediately assigned a set of attributes. 

• Number of Containers On Board {a_Num_Containers) —This is the 
number of Ordnance Containers carried by the OCS. As the primary 
source of ordnance supplies to the system, this is a vital attribute of the 
OCS. In reality, container ships are capable of carrying thousands of 
containers. This thesis makes the assumption that the OCS will unload 
approximately enough containers to supply a T-AKE. This number can be 
a predefined constant or a variable. Both methods of defining 
a_Num_Containers are used in this thesis. The constant method was used 
in the baseline scenario. The value assigned in the constant method is 
255, and is based on the assumption that a T-AKE full load has the value 
of 3,540 short tons of ordnance. This value is approximately 70% of the 
possible ordnance load that a T-AKE could carry and purposely high to 
match the scenario requirement of supporting engaged combatants. Since 
the standard ordnance load for a TEU is approximately 13.9 short tons 
(Headquarters, Department of the Army, 1997), the calculation for a 
containers per T-AKE using these assumptions results in 255 containers. 


30 



The variable method is used in all other scenarios. The variable method 
uses a variable, v_Cont_j)er_OCS, which is discussed further in Section F 
defining the model variables. 

• Arrival Time to the System {a_Arrival_Time_to_System )—This is a 
timestamp given to the OCS upon entering the system. Time-based 
statistics use this attribute assignment to calculate outputs, such as how 
long the entity is in the system. 

• Ship Type (a_Ship_Type )—This is an attribute used to identify the ship 
type numerically. All OCSs are given a_Ship_Type assignment values of 
one (1). 

The TAKE Ship Entity receives similar attribute assignments to the OCS in 
Arrival Time to the System and Ship Type. The Arrival Time to the System is 
entity-arrival dependent and the Ship Type value assigned to T-AKEs is two (2). The 
attribute of interest for the T-AKE is: 

• The Number of Pallets Needed (a_Pallets_Needed )—This attribute is 
what defines the demand of the combatants involved in the contingency. 
The value of this attribute is the integer value of a triangular distribution 
with parameters, TRIA (3315, 3500, 3570). This distribution is based on 
the calculations for T-AKE load capacity and the likely load size 
assumption used to calculate a_Num_Containers. As T-AKEs are 
employed to the Elect, better data for actual load size carried can be 
obtained and this distribution can be adjusted. 

The CS Entity requires no distinctive attributes because it does nothing other than 
vie for a limited number of resources that the other ship entities require as well. Thus, 
the CS receives the attributes of Arrival Time to the System and Ship Type. The Arrival 
Time to the System is entity-arrival dependent and the Ship Type value assigned to CS is 
three (3). 

The Container Entity is a product of the OCS and is generated by separating the 
containers from the OCS and then assigning them attributes specific to containers. These 
entities are not “created” like the ship entities. The Separate Module in Arena provides 
the mechanism for generating entities from a higher level entity. In this case, the OCS is 
the higher level entity from which containers are generated. Eigure 16 shows the first 
step in this entity generation. The original OCS and a duplicate are separated, but the 
duplicate “inherits” the same attribute values of the original OCS. The duplicate is 


31 



routed along a different path and eventually departs the system when it has completed all 
required processes. This duplicate entity embodies the OCS, which moors at Kilo Wharf 
to unload ordnance containers. 



Pv 


Separate Ship 
from All Its 
Containers 


Original 


Duplicate 


Figure 16. The First Separate Module for an OCS and Associated Separate GUI. 


The next step in generating containers is to again use a separate module to 
duplicate the OCS. Before this happens, the intermediate OCS entity receives a variable 
assignment necessary to count the number of containers removed from the OCS. Once 
the variable assignment is made, the original OCS is separated into “a_Num_Containers 
- i” containers and one original. The number of containers is decremented by one 
because the original and duplicates will both be given container attributes and sent along 
the same process path. Figure 17 shows the separation process used to generate 
Ordnance Container Entities from an OCS Entity in this thesis. 



Eigure 17. The Container Generation Segment. 


Eigure 18 shows the Assign Module used to give all of the newly generated 
containers their initial attributes. The entity picture is assigned to differentiate this entity 
type in the model animation. The entity type is assigned as ""Container.” 


32 





























































Figure 18. The Assign Module for a Container. 

Containers receive four other attribute assignments types while in the system. 
The first is an assignment of ownership. One of the competing requirements used in this 
model is achieved here by giving each container a type of property stamp. The container 
can either be marked for the United States Navy (USN) or it can be marked for the 
United States Air Force (USAF). This assignment is determined by an input parameter 
V _percent_Navy_Cont that is described in Section F, defining the model variables. The 
attribute name used in this assignment is a_Switch. The attribute name is generic because 
it is used solely as a switch in a subsequent Decide Module to direct traffic. 

Once the containers path is determined by a_Switch, the container can then be 
assigned a destination designator attribute, a_Destination_ldentifier. This attribute has a 
value of either 111, designated to Andersen AFB, or 999, designated to the Ordnance 
Annex. All containers that are assigned to the USAF by a_Switch also receive 
a_Destination_Identifier value of 111. All containers assigned to the USN must 
encounter another attribute assignment before receiving their a_Destination_Identifier. 
This other attribute assignment given to containers is used similarly to the a_Switch 
attribute just described. In fact, because it performs a similar function to containers that 
have previously received an attribute named a_Switch, but at a different point in the 
process, it uses the same attribute name a_Switch. This particular a_Switch assignment is 
determined by an input parameter v jjercentjunstujfed_j)ier that is described in Section 


33 




















F, defining the model variables. This parameter determines where a container is 
unstuffed. Once this attribute is assigned, then the a_Destination_Identifier for this 
container is assigned a value of 999. 

As mentioned earlier, the Pallet Entity is the very first type of entity created in 
this system. Figure 19 shows this inventory initialization done by a Create Module. 
These pallets have the same attributes as other pallets that are generated in the system. 



Figure 19. The Create Module for Initialization. 

In order to establish an initial inventory on hand at the onset of the simulation, a 
Create Module is used once to generate an initial inventory defined as a variable, 
v_lnitial_lnventory. A minimum v_Initial_Inventory value of 75,000 pallets is used in 
this model for any scenario that involves competing requirements. Reasons for this 
minimum value setting are discussed in Chapter IV. This initial inventory represents a 
portion of the ordnance Prepositioned War Reserve Material Stock (PWRMS) located at 
the Ordnance Annex. As the PWRMS depletes, a safety level is required to keep the 
system from experiencing shortages; this is what v_lnitial_Inventory represents. 

All other Pallet Entities are generated in a manner very similar to Container 
generation from an OCS entity. The biggest difference is that Pallets are generated from 
Containers and thus receive a different set of attributes. Figure 20 shows the Assign 
Module used to give the newly generated pallets at the Ordnance Annex their initial 
attributes. 


34 
























Figure 20. The Assign Module for Pallets at the Ordnance Annex. 

The entity picture is assigned to differentiate this entity type in the model 
animation. The entity type is assigned as ""Pallet.” Pallet Entities are also generated at 
Kilo Wharf in the Unstuffing Area. These pallets are assigned the same entity type and 
picture. 


C. PRIMARY RESOURCES 

Resources are used to represent people, machines, or even space in a storage area. 
This section describes the resources used in this thesis, which include all three of the 
possibilities mentioned: ordnance inspectors (people), cranes (machines), and storage 
and processing space. In Arena, the capacity of a resource is a constant that cannot be 
changed unless using the Process Analyzer function. For this reason, this model was 
built with constant capacity values, based on existing resources. Changes to resource 
capacities are done by using the Control portion of the Process Analyzer. This technique 
is described in Section G of this chapter. A majority of the resources studied in this 
thesis are of the space variety. All space resources are described first, followed by 
personnel, and then equipment. 


35 




















1. Space Resources 

a. Kilo Berth 

Kilo Berth is the primary single server berth used by all ship-type entities 
in this thesis. The capacity of this resource is defined as a constant, value of one (1), 
throughout the thesis. 

b. Buoy 702 

Buoy 702 is the single server standby location (anchorage) for all ship- 
type entities awaiting the opportunity to berth at Kilo Berth. The capacity of this resource 
is defined as a constant, value of one (1), throughout the thesis. 

c. Pier-side Staging Space 

Pier-side Staging Space is the space located directly on the pier that is 
used to place Ordnance Containers as they are offloaded from the OCSs. This resource is 
important to the process in that, if it is busy, containers cannot be offloaded from the 
OCS. The capacity of this resource is defined as a constant, value of two (2), throughout 
the thesis. 


d. Unstuffing Space 

Unstuffing Space is the space located at the Kilo Wharf adjacent to the 
pier and at the Ordnance Handling Pad. Combined, the two sites provide a constant 
capacity of approximately 120 spaces before unworkable. This approximation is based 
on the 100 spaces available at Kilo and the 30 to 35 available at the Handling Pad. 
Reducing the number by 10 to 15 leaves appropriate space for moving containers and 
pallets while unstuffing occurs. 

e. Ordnance Annex Magazine Storage 

Ordnance Annex Magazine Storage is the space available at the 
Ordnance Annex available for pallet storage. The capacity of this resource was set to be 
essentially unlimited for the purpose of this thesis. The model is built with the capacity 

of this resource defined as a constant, value of 99,999,999, throughout the thesis. In 

36 



doing so, the initializing inventory level of pallets is used to determine if the current 
available space is sufficient to handle this type of contingency. If the initializing 
inventory required to run the model for the specified time period exceeds the current (and 
planned) storage capabilities, then this explicitly shows an infeasibility issue with the 
system as a whole. 

For purposes of model flexibility, the Ordnance Annex Magazine Storage 
resource is also used to model the new magazine being built on Orote Peninsula in the 
last scenario set of the experiment. The assumption is that the new magazine will be the 
primary transition point for the inbound and outbound ordnance supported by the Annex 
located further away. By assuming the same properties as the Annex for the new 
magazine, the only factor that changes in this scenario set is the distance between the 
Kilo Wharf and the then Annex and now magazine. This is admittedly a generous 
assumption, but it follows the same reasoning used in the NAVBASE GUAM FY 2008 
Military Construction Program Project P-425 document DD Form 1391, dated 01 August 
2005, that identifies the requirement to build the magazine. 

IMPACT IF NOT PROVIDED 

Without construction of magazines on Orote Peninsula, the safe and 
efficient pre-positioning of ammunition near Kilo Wharf cannot be 
accommodated. As a result, the level of throughput envisioned for Guam 
will not be achieved. Whether ordnance arrives via container or break- 
bulk, the materials will need to be hauled to the Ordnance Annex for 
temporary storage, and transported back to Kilo Wharf for the next T-AE 
upload. The need to haul ordnance between the two locations constrains 
throughput operations and the efficient delivery of ordnance to the fleet. 
Anticipated increases in the operational tempo in the Pacific and Indian 
Ocean theaters will exacerbate the problem. (NAVBASE GUAM DD 
Eorm 1392, 2005) 

/. Container Truck Loading Space 

Container Truck Loading Space is the space located on the pier that is 
used to load Ordnance Containers to Container Capable Trucks for transport to either the 
Annex or AAEB. The capacity of this resource is defined as a constant, value of two (2), 
throughout the thesis. 


37 





2. Equipment Resources 

a. Crane 

The Crane is an equipment resource that is inherent to OCSs arriving for 
ordnance offload. Currently there are no Cranes organic to Guam that can safely and 
efficiently offload ordnance. The capacity of this resource is defined as a constant, value 
of two (2), throughout the thesis. 

b. Ordnance Forklifts 

Ordnance Forklifts are equipment resources that are part of the T-AKE 
loading process. Although these forklifts could technically be considered free-path 
transporters, they are modeled as resources in this thesis because their negligible distance 
traveled is between a loading spot on the pier and one of 100 possible unstuffing spaces. 
Modeling these as transporters would require approximately 100^ = 10,000 paths to be 
built into the model for such small distances. Instead, the forklifts are built into the 
model as resources that incur a delay that accounts for distance traveled when seized by 
pallets that are loaded to the T-AKE. The capacity of this resource is defined as a 
constant, value of 20, throughout the thesis. This value assumes a slight increase in the 
assets listed in the CNA Report CRM D0017313.A1 (Goode & Smith, 2007), based on 
the scenario from 14 to 20 forklifts. 

3. Personnel Resources 

a. Ordnance Inspectors 

Ordnance Inspectors are personnel resources instrumental to the 
unstuffing process. Ordnance Inspectors inventory and inspect all pallets of ordnance 
unstuffed from a container. Delays incurred by the inventory and inspection process are 
built into the Ordnance Inspectors. The capacity of this resource is defined as a constant, 
value of 18, throughout the thesis (Goode & Smith, 2007). 


38 



b. Block and Brace Crew 

Block and Brace Crews are personnel resources instrumental to the pallet 
transport process. Block and Brace Crews ensure load stability of pallets transported by 
building a frame around the pallets. Delays incurred by the block and brace process are 
built into the Block and Brace Crews. The capacity of this resource is defined as a 
constant, value of 10, throughout the thesis. This value is set at just higher than 80% of 
the number of trucks able to transport pallets. This is done to ensure that a block and 
brace crew is available for pallets that are ready to be loaded, while the other 20% of the 
pallet trucks are in transit. This is also a generous assumption, but very feasible to 
achieve. 

D. PRIMARY PROCESSES 

A process in this thesis describes the action taken by an entity throughout the 
system. These processes are all directly related to the resources just described in Section 
C of this chapter. Figure 21 represents the GUI associated with an Arena Basic Process 
Module and displays the four types of action that a Basic Process can perform: Delay, 
Seize Delay, Seize Delay Release, and Delay Release. Advanced Processes are also 
available for use in the model. These consist of the individual actions listed in the Basic 
Process Module, except as the separate modules: Seize, Delay, and Release. This section 
describes and explains the major processes built into the thesis model. The processes are 
divided into categories based on the entity that is carrying out the identified process. 


39 




Figure 21. The GUI assoeiated with an Arena Process Module. 

The first Entity type explained is the ship type, to include OCS, T-AKE, and CS 
Entities. All ships entering the system attempt to Seize Kilo. If Kilo is unavailable, then 
the ship attempts to Seize Buoy 702. Any ship that enters the system and is denied either 
of these processes is held in the Seize Buoy 702 Queue. If the ship is not able to 
Seize Kilo, but is able to Seize Buoy 702, then its next process is to Seize Kilo from Buoy 
when Kilo becomes idle. Erom this point on in the model, the processes are dependent on 
the Entity Type. 

OCSs are held at Kilo until completely unloaded and then perform the 
OCS Release Kilo Berth process. This action releases the Kilo Berth resource and moves 
the OCS on to exit the system. 

T-AKEs are held at Kilo until completely loaded and then perform the 
TAKE Release Kilo process. This action releases the Kilo Berth resource and moves the 
T-AKE on to exit the system. 

CSs simply perform the Basic Process of delay and release at Kilo. The CS Delay 
and Release Kilo process uses an Expression to account for the delay of the Kilo Berth 

40 


























resource. This delay is a random distribution that was determined from historical data 
provided by the NEW Reports (Rivera, 2008). The historical data from the NEW Report 
was processed using the Input Analyzer tool in Arena. This tool provides the user with 
the ability to fit distributions quickly to the given input data. By accumulating the length 
of stay for CSs from 2003 through 2008, the data fits to a Uniform distribution, UNIE 
(4.01, 7) days. This action completes the process when it releases the Kilo Berth resource 
and moves the CS on to exit the system. 

The next entity type explained is the Container. The first thing a Container must 
do is to seize a crane for movement off the OCS. The Crane Moves Container from Ship 
to Pier is a Basic Process that uses a Seize Delay action to perform container offloading. 
This action requires both a crane and a staging space at different lengths of usage time. 
The delay time for the crane in this action is an assumed random distribution that was 
calculated by converting the daily offload rates found in the CNA Report CRM 
D0017313.A1 (Goode & Smith, 2007), to an hourly rate per container, based on the range 
of containers offloaded in a day. The resulting calculated range is used in a Uniform 
distribution, UNIE (0.00735, 0.01225) hours, for lack of better data on high-volume 
offload rates. 

Containers then Release Pierside Staging Space For Ord Annex Container or 
Release Pierside Staging Space when their appropriate destination is determined. Erom 
this point, the Container performs actions appropriate to their location. Containers that 
are unstuffed at the Ordnance Annex perform the action. Ordnance Inspection at 
Ordnance Annex, a standard Seize Delay Release action. One Ordnance Inspector per 
container is seized for the inspection and inventory during unstuffing. The delay in this 
action is assumed to be a Uniform distribution, whose range is based on historical 
unstuffing delays from the CNA Report CRM D0017313.A1 (Goode & Smith, 2007). 
The resulting calculated range is used in a Uniform distribution, UNIE (0.13333, 
0.16667) hours, for lack of better data on high-volume offload rates. 


41 



Containers that are unstuffed at Kilo perform the action Ordnance Inspection at 
Kilo, a standard Seize Delay Release action. Seize and delay actions are assumed to the 
same as those for Ordnance Inspection at Ordnance Annex because it is the same process 
carried out at a different location. 

The last entity type explained is Pallets. Pallets major process is the Load Pallets 
to TAKE process. This is a standard Seize Delay Release action where a pallet seizes an 
Ordnance Forklift and is loaded to the T-AKE with a delay of UNIF (2, 5) minutes. The 
Uniform distribution was based on the load times from the CNA Report CRM 
D0017313.A1 (Goode & Smith, 2007). 

The processes described above do not include the entirety of processes in the 
system. All other processes can be found in Appendix A, Component and Module 
Specification for Modeling Ordnance Movements into the Asian Pacific Theater. 

E. PRIMARY TRANSPORTERS 

The primary transporters in the model are Container Capable Forklifts, Container 
Trucks, and Pallet Transport Trucks. This section provides a description of each of these 
transporter types. 

There are two Container Capable Forklift free-path transporters in the model. 
These forklifts are top-handling (25 ton) container forklifts and assigned a velocity of 
26,400 feet per hour. Since Arena does not allow for fractional velocities, the velocity of 
the Container Capable Forklift had to be converted to feet per hour. This value equates 
to five miles per hour, a reasonable estimate for an average velocity of Container 
Capable Forklifts. Figure 22 is a picture of the two currently available Container 
Capable Forklifts located on Guam. 


42 




Figure 22. Container Capable Forklifts (From: MSDDC, 2008). 

There are eight Container Truck free-path transporters in the model (Goode & 
Smith, 2007). The Container Truck has an average veloeity of 12 miles per hour. This 
value is based on the travel times ealculated in the MSDDC Guam Ammunition 
Distribution Study (MSDDC, 2008). 

There are 12 Pallet Transport Truck free-path transporters in the model (Goode & 
Smith, 2007). The Pallet Transport Truck has an average velocity of 12 miles per hour. 
This value is also based on the travel times calculated in the MSDDC Guam Ammunition 
Distribution Study (MSDDC, 2008). 

F. VARIABLES OF INTEREST 

This section describes the simulation parameters, or factors, that were chosen for 
the experiment. Factors are defined by the decision maker’s ability to control them. A 
factor that is controllable by the decision maker in the real world is considered a decision 
factor. Uncontrollable factors are those beyond the decision maker’s control, e.g., 
weather delays, ship repair requiring in port periods, or competing requirements. These 
uncontrollable factors are often referred to as noise factors. In this thesis, the factors that 
are controlled by the Navy are considered controllable factors, to include arrival cycles, 
supply and demand quantities, and processing policies. The noise factors in this thesis 
are factors such as the arrival of CSs and AF ordnance requirements. Table 6 provides 
the variable simulation parameters, for both decision and noise factors, and their 
associated ranges, used in the experimental designs. 


43 









Table 6. 


The Decision Factors and Noise Factors. 


Factors 

Range 

Explanation 

Low 

High 

v_OCS_Arr_Cycle 

9 

13 

This cycle is directly related to v_TAKE_Arr_Cycle. The estimated replenishment cycle of the T- 
AKE based on operational demand requires an OCS approximately every eleven days. The range 
selected allow for minor variances in the arrival policy while maintaining sufficient supply of 
ordnance into the system. 

v_TAKE_Arr_Cycle 

10 

20 

This cycle is defined by the estimated ordnance sustainment requirements of a Marine 
Expeditionary Brigade (MEB) ashore. The range selected allows for an increase or decrease in 
arrival policy based on demand. 

v_CS_Arr_Time* 

23 

60 

This is the mean interarrival time for CS arrivals based on an exponential distribution. The range 
selected starts at the current level and explores the possibilities involved with a policy that limits 
CS entry into the system. 

v_Cont_per_OCS 

200 

300 

This is the number of containers offloaded from an OCS. The range explores changes in the 
current offload amount of approximately 255 containers. This range decision is possible because 
OCSs carry multiple loads (in reality) and can accommodate offloading more or less than 
currently prescribed. 

v_percent_Navy_Cont* 

0.7 

0.9999 

This is the direct competition for ordnance. The range defines a reduction in 100 percent 
ordnance supply by up to as much as 30 percent. The range is estimated on reasonable 
requirements during the contingency. 

v_percent_unstuffed_pier 

0.4 

0.9999 

This range represents the possibilities of the policy that defines where unstuffing occurs. 

Previous studies have suggested a change from 100 percent pierside unstuffing may increase 
throughput. The range was selected to exceed a a change by 50 percent to explore the policy 
possibilities. 

Ordnance Inspector 

18 

27 

This resource capacity range is for the number of qualified ordnance inspectors. The range 
selected represents the current to a 50 percent increase in personnel. 


The * indicates the competing requirements/noise factors. 


1. Controllable Factors 

The following factors were chosen to explore the effect of competing 
requirements on the ordnance operations on Guam under a variety of possible 
support aspects. 


a. Ordnance Container Ship Arrival Cycle (v_OCS_Arr_Cycle) 

This is defined as the arrival cycle for OCSs to Guam. This cycle time is 
directly related to v_TAKE_Arr_Cycle and represents the supply required for combatant 
demand. Since the OCS carries more ordnance than is removed at one time, this model 
represents this by utilizing a rate proportionally higher to indicate the OCS delivering a 
partial load, going out to sea to loiter, and then returning to deliver another load. The 
estimated replenishment cycle of the T-AKE, based on operational demand, requires an 


44 





































OCS arrival approximately every 11 days. The range of 9 to 13 allows for minor 
variances in the arrival policy, while maintaining sufficient supply of ordnance into the 
system. 


b. T-AKE Arrival Cycle (v_TAKE_Arr_Cycle) 

This cycle time is defined by the estimated ordnance sustainment 
requirement of an MEB ashore, which is every 16 days. The range of 10 to 20 allows for 
an increase or decrease in arrival policy, based on demand or available assets. Combined 
with other factors, the range may provide insights into how this policy affects throughput. 

c. Number of Containers Offloaded per OCS Inport Period 
(vjContjjerjOCS) 

This is the number of containers offloaded from an OCS. The range 
explores changes in the current offload amount of approximately 255 containers. This 
range of 200 to 300 is possible because, in reality, OCSs carry thousands of containers 
and can accommodate offloading more or less than the 255 containers currently 
prescribed depending on the policy of ordnance operations. 

d. Percent Unstuffed Pierside (v_j)ercent_unstuffed_pier) 

This is the policy that determines where containers are unstuffed. This 
range represents the possibilities of changes in this policy. Previous studies have 
suggested that a change from 100% pierside unstuffing may increase throughput. The 
range was selected to exceed a change by 50% to explore the policy possibilities. During 
noncontingency times, pierside unstuffing cannot be determined to be the optimal policy, 
although it is the one most often used. 

e. Ordnance Inspector Capacity 

This is the number of qualified ordnance inspectors available to inventory 
and inspect pallets of ordnance during the unstuffing process. During a contingency, this 
number may be increased from the current availability to meet the operational tempo. 


45 



The ease of changing this number in reality, and the ordnance inspector’s significant role 
in ordnance operations, makes this a good factor to explore. The range selected 
represents the current availability to a 50% increase in personnel. 

/. Unstuffing Space Capacity 

This resource capacity range is for the number of actual physically 
available spaces to unstuff ordnance. Previous studies have suggested that an increase in 
this resource availability would increase ordnance operational efficiency. Selection of 
this decision factor explores those suggestions in an effort to provide quantitative analysis 
of the improved efficiency. This range represents the current amount of space up to a 
25% increase. Amounts larger than the selected high level would require an infeasible 
amount of physical space. 

g. Ordnance Forklifts 

This resource capacity range is for the number of available ordnance 
forklifts available for the loading of ordnance to T-AKEs. The range represents a 50% 
increase in the resource capacity from the current level. This explores the possible 
impact of a relatively inexpensive increase in resources on ordnance operations on Guam. 

2. Uncontrollable Factors 

The noise factors, generally comprised of the competitive requirements, are used 
to ensure that conclusions drawn from this thesis are reflective of the broad exploration of 
competing requirement effects. These are the factors that the thesis sponsor, OPNAV 
N421, wants explored in this thesis. 

a. Competing Ship Arrival Time (v_CS_Arr_Time) 

This is the mean interarrival time for CSs. This random interarrival time 
required a suitable distribution. Real-world data, gathered by NMC Guam for Kilo 
Wharf occupancy during a five-year period from 2003 to 2008, provides the distribution 
for v_CS_Arr_Time. The distribution of v_CS_Arr_Time is shown in Figure 23 and is 
defined by the expression, v_CS_Arr_Time = -0.001 -i- EXPO (23.7), where the value 


46 



23.7 is the average interarrival time for competing ships. The range selected starts at the 
current level and explores the possibilities involved, with a policy that limits CS entry 
into the system. 



Figure 23. Arena Input Analyzer output for CS Interarrival Time. 


b. Percent of Containers for United States Navy 
(v _percent_Navy_Cont) 

This factor defines the competing requirement for ordnance by the Air 
Force. A larger competing requirement decreases v _percent_Navy_Cont. Although the 
forces would be operating in a joint effort during a military contingency, the Navy does 
not have control of AF requirements for ordnance. The range defines a reduction in 
100% ordnance supply by up to as much as 30%. The range is estimated based on 
reasonable requirements during the contingency. 

3. Other Variables of Interest 

a. Universal Stream Indicator (v_Univ_Stream) 

The universal stream indicator is a variable that is attached to every 
expression in the model that uses the random number seed. By attaching the universal 
stream indicator, the model then produces a set of replications using the same random 
number stream. This is critical to using the Process Analyzer in Arena in conjunction 
with DOE. When the set of replication (a run) is completed, the model moves to a new 
set of input parameters. The universal stream indicator applies a new random number 


47 


















stream to the subsequent run, thus producing runs that are independent of each other. 
This random-number-stream allocation ensures independence not only with scenarios, but 
also across them as well. 

b. Initial Inventory (v_Initial_Inventory) 

In this model, the initial inventory variable is used for two purposes. The 
first is to build the initial starting condition inventory. This developed into its second 
purpose through the debugging process and second scenario of the experiment. Once 
competition is added into the system, a much higher initial inventory is required for the 
model to successfully complete the minimum simulation requirement of one year of 
operating time. Therefore, the second purpose became a test for the starting condition 
feasibility. This secondary purpose is further discussed in Section C of Chapter V. 

G. THE EXPERIMENT 

This thesis uses five scenario sets to conduct the experiment. The first is a 
baseline scenario that uses a combination of the ordnance operations process observed in 
reality, and input parameters from the previous studies in the simulation model, in an 
effort to establish a verifiable baseline. Validation indicates that the model used in this 
thesis models the process flow as a close as possible to reality based on the previous 
studies of CNA and MSDDC. This scenario will act as the control scenario. 

The second scenario is the initial introduction to competing requirements to the 
system. The third scenario is another baseline scenario in which the model represents a 
physical change to the system. The fourth scenario set is similar to the second in that it 
introduces competing requirements to the new system baseline in the third scenario. The 
fifth scenario set is the experimentation set, where the landscape of possible outcomes is 
explored. Orthagonal Latin Hypercubes (OLH) and the Nearly Orthogonal Hypercubes 
(NOLH) are the primary method of exploring the factor space for insights in this thesis 
(Cioppa, 2002). Both methods of DOE were used, based on the number of input 
parameters required in the scenario. The OLH and NOLH are quickly developed by 
using the automated versions of the OLH and NOLH (Sanchez, 2005) found on the 
SEED Center Website, http://harvest.nps.edu. 

48 



The purpose of this thesis is not to predict outcomes, but rather, to provide 
insights into the effect of new factors introduced to the system (competing requirements) 
and the variability of existing factors. Therefore, the response data in each experimental 
scenario set is twofold. The first response measures T-AKE service level by calculating 
the ratio of T-AKEs that are served by the system, to those who enter the system. By 
using an unprioritized queue in the model, this information is captured easily. The 
second response is the total number of pallets that are processed out. Although these two 
measures appear correlated quite closely, it is important to measure the service level to 
the combatants in this manner. For all intents and purposes, the combatants do not care if 
the ordnance operations cannot service 100% of T-AKEs that enter the system, as much 
as they care about receiving a sufficient amount of ordnance; in the case of this thesis, 
pallets of ordnance. 

1. Scenario Set 1 - The Baseline Model 

The baseline model is a representation of the system as it exists at present. The 
two primary competing requirement factors are built into the baseline, but their input 
values are set to model no competition. CSs are not introduced to the system and the 
Navy receives 100% of all containers that arrive on an OCS. This setting is a direct 
comparison to the previous study done by CNA. The input parameters in the baseline are 
set to constant values and the variability of the model response after 100 replications is 
caused by the inherent variability some of the processes possess. Table 7 displays the 
input parameters for the baseline. These parameters mirror current operating policy, 
physical reality, and the input parameters of previous studies. 


Table 7. 


Input Parameters For Scenario Set 1 - Baseline. 


d' 


(O' 

o 


I 


S 


£ 

t' 


(0 

o 

<b 

«. 

c 

d 


c 

d 

t' 

c 

§ 

QJ 

Q. 


to 

C 

S 

c 

§ 

Qj 

Q. 


o 

QJ 

i 


C 

# 


s 

s. 


*r 






11 

16 

0 

255 

0.9999 

0.9999 

18 

120 

118 


49 











The responses for this model are the standard for all scenarios run in this thesis: 
T-AKE service level and pallet throughput. The baseline model is also the model used to 
debug the simulation before proceeding with further experimentation. 

2. Scenario Set 2 - Addition of Competing Requirements 

Scenario Set 2 is the first experiment conducted on the system. This scenario 
examines the effect of adding the competing requirements to the system. Differences 
from the baseline model include using the CS interarrival rate determined by historical 
data and the possibility for the Navy to receive less than 100% of the ordnance entering 
the system. The effect of competing scenarios is seen in comparison to the baseline 
scenario. Table 8 displays the input parameters for Scenario Set 2. 


Table 8. Input Parameters for Scenario Set 2 - Competing Requirements. 



10 

20 

25 

238 

0.78 

0.96 

25 

143 

101 

9 

13 

23 

256 

0.7 

0.59 

19 

145 

102 

10 

14 

24 

225 

0.89 

0.89 

24 

160 

103 

10 

16 

24 

300 

0.87 

0.48 

17 

150 

104 

12 

19 

28 

213 

0.79 

0.4 

16 

153 

105 

13 

13 

30 

281 

0.72 

0.85 

24 

155 

106 

12 

12 

27 

231 

0.96 

0.66 

20 

158 

107 

11 

19 

27 

294 

0.94 

0.78 

22 

148 

108 

11 

15 

27 

250 

0.85 

0.7 

21 

140 

109 

12 

10 

28 

263 

0.93 

0.44 

17 

138 

110 

13 

18 

30 

244 

0.9999 

0.81 

23 

135 

111 

13 

16 

29 

275 

0.81 

0.51 

18 

120 

112 

12 

14 

29 

200 

0.83 

0.93 

25 

130 

113 

10 

11 

25 

288 

0.91 

0.9999 

26 

128 

114 

9 

17 

23 

219 

0.98 

0.55 

19 

125 

115 

11 

18 

26 

269 

0.74 

0.74 

22 

123 

116 

11 

11 

26 

206 

0.76 

0.63 

20 

133 

117 


50 



























The responses for this model are used to quantify the effect of competing 
requirements and for analysis into which of the input parameters have a significant effect 
on the system. 


3. Scenario Set 3 - Simulating Completion of the New Magazine on 
Orote Baseline 

Scenario Set 3 is the baseline for an experiment to explore the effect of the 
NAVFACMARIANAS Project, P-425, on the system. This project is building a new 
magazine located on the Orote Peninsula in an effort to increase safety and reduce the 
amount of transit time to and from the Ordnance Annex. This scenario uses the original 
baseline model scenario setup, with the exception of an adjusted distance set to account 
for the new magazine. The model assumption generously gives the new magazine the 
same capacity as the Ordnance Annex. By doing this, the original model is easily altered 
to reflect a closer facility, changing the distance from Kilo Wharf to the Annex from 
seven miles down to one mile. The remaining input parameters remain the same as those 
seen in Table 7 - Input Parameters for Scenario Set 1 - Baseline. 

The responses in this model are used in comparison to the original baseline and to 
the next Scenario Set. These comparisons show both the effect of the new magazine to 
the existing system and the impact of competing requirements in the new system. 

4. Scenario Set 4 - Simulating Completion of New Magazine on Orote 

Scenario Set 4 is the experiment that introduces the competing requirements to the 
new magazine baseline set up in Scenario Set 3. The purpose of this scenario is to 
explore the impact of competing requirements on the system with the new magazine. 
The same input parameters used in Scenario Set 2 are used to evaluate the system in this 
experiment. This provides a method for not only comparing the responses of this 
scenario to its baseline scenario, but also for comparison to Scenario Set 2. 

5. Scenario Set 5 - Exploratory Set 

Scenario Set 5 is the experiment that uses all the input parameters listed in Table 


6 to explore a broad landscape of possibilites. The purpose of this experiment to evaluate 

51 



which input parameter has a signifieant effeet on the response. The nine input parameters 
used in this experiment all relate to viable changes that ean be implemented in the 
system. Insights from the analysis of this experiment provide a basis for 
recommendations regarding ehanges to the system. These ehanges ean be represented in 
either policy changes or resource allotments in the system. Table 9 represents the nine 
input parameters and the universal stream indieator variable. 

Table 9. Input Parameters by Scenario for Scenario Set 5. 





low level 

9 

10 

23 

200 

0.7 

0.4 

18 

120 

8 

- 

high level 

13 

20 

60 

300 

0.9999 

0.9999 

27 

160 

12 

- 

decimals 

0 

0 

0 

0 

4 

4 

0 

0 

0 



13 

11 

39 

219 

0.9624 

0.7749 

22 

160 

11 

101 

13 

20 

28 

238 

0.8406 

0.5125 

21 

156 

11 

102 

13 

14 

57 

216 

0.7094 

0.7562 

18 

133 

11 

103 

11 

19 

60 

241 

0.9812 

0.4937 

19 

138 

11 

104 

13 

10 

40 

222 

0.9062 

0.8312 

23 

123 

10 

105 

13 

19 

35 

228 

0.8312 

0.5312 

26 

121 

9 

106 

12 

15 

59 

225 

0.7 

0.7937 

26 

155 

10 

107 

11 

17 

58 

234 

0.9718 

0.55 

27 

141 

9 

108 

12 

13 

31 

253 

0.9156 

0.5875 

20 

144 

8 

109 

12 

17 

33 

269 

0.7656 

0.7187 

21 

154 

8 

no 

12 

12 

51 

297 

0.8031 

0.4375 

19 

135 

9 

111 

12 

17 

47 

294 

0.9249 

0.9812 

22 

129 

10 

112 

11 

12 

30 

256 

0.8781 

0.475 

25 

134 

12 

113 

12 

16 

37 

288 

0.7469 

0.7374 

24 

130 

12 

114 

12 

12 

54 

291 

0.8125 

0.4 

25 

149 

11 

115 

12 

16 

45 

300 

0.9437 

0.9437 

24 

153 

10 

116 

11 

15 

42 

250 

0.85 

0.7 

23 

140 

10 

117 

9 

19 

44 

281 

0.7375 

0.625 

23 

120 

9 

118 

9 

10 

55 

263 

0.8593 

0.8874 

24 

124 

9 

119 

10 

16 

26 

284 

0.9905 

0.6437 

27 

148 

9 

120 

11 

11 

23 

259 

0.7187 

0.9062 

26 

143 

9 

121 

9 

20 

43 

278 

0.7937 

0.5687 

22 

158 

10 

122 

9 

11 

48 

272 

0.8687 

0.8687 

19 

159 

11 

123 

10 

15 

24 

275 

0.9999 

0.6062 

19 

125 

11 

124 

11 

13 

25 

266 

0.7281 

0.8499 

18 

139 

11 

125 

10 

18 

52 

247 

0.7843 

0.8124 

25 

136 

12 

126 

10 

13 

50 

231 

0.9343 

0.6812 

24 

126 

12 

127 

10 

18 

32 

203 

0.8968 

0.9624 

26 

145 

12 

128 

10 

13 

36 

206 

0.775 

0.4187 

23 

151 

10 

129 

11 

18 

53 

244 

0.8218 

0.9249 

20 

146 

8 

130 

10 

14 

46 

213 

0.953 

0.6625 

21 

150 

8 

131 

11 

18 

29 

209 

0.8874 

0.9999 

20 

131 

9 

132 

10 

14 

38 

200 

0.7562 

0.4562 

21 

128 

10 

133 


52 

































































A secondary purpose of this experiment is to develop a set of observations that 
can be used for future research. By providing the decision maker with information about 
which factors have significance in the model, future research can be used to investigate 
these factors even further. 

6. Simulation Runs and Replications 

Each of the design points in the Scenario Sets was replicated 100 times, with a 
total run time of approximately 8 to 10 minutes per design point. This provides adequate 
precision to resolve differences in statistically significant ways, while at the same time 
proving workable in terms of computing time. The Process Analyzer in Arena allows for 
the selection of all design points in an experiment and running them consecutively. The 
universal stream indicator is used as an input into the Process Analyzer. This applies a 
new random number stream to the subsequent run, thus producing runs that are 
independent of each other. This random-number-stream allocation ensures independence 
not only with scenarios, but also across them as well. This simplified the 
experimentation by allowing the analyst to start an experiment in the morning and return 
in the afternoon to a completed run of the experiment. 


53 



THIS PAGE INTENTIONALLY LEET BLANK 


54 



IV. DATA ANALYSIS 


The experimental scenario sets described in Chapter III provide an opportunity to 
generate a significant amount of data for analysis. In this chapter, the focus is on 
discovering insights into the movement of ordnance into the Asian Pacific Theater. In an 
effort to address the thesis questions presented in Chapter I, the analysis is centered on 
the MOEs of interest, T-AKE Service Eevel (SE), the ratio of T-AKEs serviced by the 
system, and Pallet Throughput (Pallets Out). 

This chapter begins with a brief description of the data collection and post¬ 
processing methods. Eollowing a detailed scenario-by-scenario analysis of the data, the 
thesis presents insights and conclusions drawn from the analysis. 

A. DATA COLLECTION AND POST PROCESSING 

Data collection, using the Process Analyzer in Arena, is a very simple process. 
The Process Analyzer gathers response data, defined by the analyst as the statistical 
averages of replications in each run. Although useful for looking at system performance 
averages, the run average data does not allow analysts to look at the landscape of possible 
outcomes in a refined manner. To do this, the individual output from each replication is 
required. In order to gather the response data from individual replications, the response 
data is an intermediate step required during data collection. Response data from 
individual replications are passed to an Excel spreadsheet via the Output to a spreadsheet 
segment in the model. At the time that a replication reaches the finishing time for the 
simulation, tfin, the model creates an entity that directs the output of statistics gathered 
during the simulation to write out to a specified file. Eigure 24 is the Output to a 
spreadsheet segment, and the associated GUI that is used to identify which statistics are 
sent to the output file. 


55 



ReadWrite 


Output to a spreadsheet 


Write Out Stet 


Name: 

|Write Out Stat 




Arena File Name: 


[Write to File 
Recordset ID: 


^[FiTi 

Record Number: 




[Recordset 2 
Assignments: 


■^r 


Variable, v_Univ_Stteam ^ 

Other, NC(Recotd QCS Uiru Kilo) 

Other, NC(RecotdQCS In] 

Other, NC(Recotd Number of TAKE thru Ki 
Other, NC(Recotd TAKE In) 

Other, NC(Recoid Competing Ship at Kilo) 

Other, NC(Recotd CS In) 

nther WrrBpmrHTaiCFPalUM ' 


OK 


Figure 24. The IVIodels Output Segment. 

The flexibility of this feature in the model allows the analyst to define many 
different statistics gathered by Arena, or those defined in the model by the analyst. 

Once the output data are written to the Excel spreadsheet, they are ready for post 
processing. In this thesis, post processing primarily consists of merging the columns of 
input data into the output file and conversion of the output data into the T-AKE SE. The 
MOE, T-AKE SE, is calculated by simply dividing the number of T-AKEs that enter the 
system by the number of T-AKEs that leave the system. This ratio provides an MOE 
bounded by 0 and 1. This ratio is presented as a percentage, where bigger values equal 
higher service levels. Therefore, a perfectly running system will not have anyone in 
queue and have a SE of 1. The Pallets Out MOE is simply a tally statistic that is gathered 
within the model and is automatically reported as output response data. Once the output 
response data is processed in Excel, it is imported into IMP Statistical Discovery 
Software version 7.0, which is the primary tool used for the remaining post processing 
and analysis. 


B. 


INSIGHTS INTO RESEARCH QUESTIONS 


Recall from Chapter I the two general questions about the movement of ordnance 
into the Asian Pacific Theater that this thesis sets out to answer. 


What is the impact on competing requirements to the movement of 
ordnance into the Asian Pacific Theater? 

What, if any, are the critical factors related to providing maximum T-AKE 
SEs and Ordnance Pallet throughput? 


56 


































These questions are directly addressed through data analysis in the following section. 

1. MOE Correlation Analysis 

The first step in the analysis is to validate the need to analyze both MOEs. The 
initial hypothesis is that T-AKE SE and Pallets Out are correlated. The Correlations 
Multivariate option in IMP gives the Correlations table, which is a matrix of correlation 
coefficients that summarizes the strength of the linear relationships between each pair of 
response (Y) variables. This correlation matrix only uses the observations that have 
nonmissing values for all variables in the analysis (SAS Institute Inc., 2007). Eigure 25 is 
the correlation matrix and scatterplot for the chosen MOEs in Scenario Sets 1 and 2. 



Eigure 25. Correlation and Scatterplot Matrix for MOEs. 

As observed in Eigure 25, the MOEs are not strongly correlated. An explanation 
for the lack of correlation is that T-AKE SE is a ratio of T-AKEs that enter the system to 
those that leave the system. This implies that the closer the cycle of T-AKEs, the higher 
the likelihood of congestion with other vessels at Kilo Wharf. Therefore, T-AKE SE is 
more likely to be correlated with T-AKE arrival frequency. Eigure 26 shows a stronger 


57 




















correlation between T-AKE SL and T-AKE arrival frequeney. It also shows that Pallets 
Out has a strong negative eorrelation (-0.74) to T-AKE arrival frequeney. This is a 
sensible result because a lower T-AKE arrival frequeney means less overall opportunity 
for suceessful services. Thus, fewer pallets are drawn from the system, sinee Pallets Out 
is a function of the T-AKE demand. 



Figure 26. Correlation and Seatterplot Matrix for MOEs and T-AKE Arrival Frequency. 


With no strong correlation between the MOEs, the analysis in this seetion is 
focused on both MOEs as separate measures of the impact of competing requirements 
and parameter variability. 


58 















































2. Analysis of Scenario Set 1 - The Baseline Model 

Anchoring the experimental design of this thesis is a reliable baseline. This 
section analyzes the baseline data and, by using the input parameters defined in the CNA 
and MSDDC studies, determines if the baseline is feasible. Figure 27 shows the 
distributions of the Scenario Set 1 MOEs. 


Distiibiitioiis 



Quantiles ] 
▼ Moments 


Mean 0.9629644 

StdDev 0.0209124 

Std Err Mean 0.0020912 
upper 95% Mean 0.9671139 
lower 95% Mean 0.9588149 
N 100 


Distill)lit!Oils 



^ 1 Quantiles | 

Moments | 

Mean 71458.33 

StdDev 1570.5245 

Std Err Mean 157 .05245 
upper 95% Mean 71769.956 
lower 95%Mean 71146.704 
N 100 


Figure 27. Scenario Set 1 - The Baseline Model MOE Distributions. 


The baseline scenario produced a mean T-AKE SE of 96.30%, with a 95% 
confidence interval of (95.88, 96.71). It also produced a mean Pallets Out value of 71458 
pallets, with a 95% confidence interval of (71147, 71770). By using the input parameters 
recommended by the previous studies mentioned, the baseline is feasible and operates at 
a high T-AKE SE and produces a throughput of pallets sufficient to meet the minimum 
requirement for T-AKEs supporting an MEB ashore. 


3. Analysis of Scenario Set 2 - Addition of Competing Requirements 


Upon the addition of competing requirements to the system, quantitative 
measurement of the impact on the system is measured. Comparing this scenario against 


59 









































the baseline seenario shows the immediate quantitative results of eompeting 
requirements. Figure 28 shows the eomparison of the distributions of Seenario Sets 1 and 
2 T-AKE SL. 



Figure 28. Seenario Sets 1 and 2 T-AKE SE Distribution Comparisons. 


Considering that there is no overlap of the T-AKE SE confidenee intervals 
between Seenarios 1 and 2, the impact of competing requirements on the system is 
significant and not attributable to the model variance. The Competing Requirements 
scenario produced a mean T-AKE SE of 70.97%, with a 95% confidence interval of 
(70.49, 71.45). When compared to the baseline, T-AKE SE sees an impact of 25.33% 
reduction in expected service level. 

The specific design points within the scenario set that performed best and worst 
are indicated in Eigure 28. Eor the design points that performed well, the only 
commonalities seen in the inputs are a higher number of v_Cont_per_OCS and higher 


60 















































percentages of v _j)ercent_Navy_Cont. As for the poorest performing design point, in 
contrast to the better performers, it has lower values for both vjCont_per_OCS and 
V _percent_Navy_Cont. This insight is analyzed further later in this section. 

The Competing Requirements scenario produced a mean Pallets Out value of 
579034 pallets, with a 95% confidence interval of (57437, 58370). Concurrently, the 
mean of pallet output is reduced by 13,554 pallets annually. As a percentage of reduction 
in pallet throughput, competing requirements reduce the system by approximately 
18.97%. A comparison of Pallets Out is seen in Figure 29. 



Figure 29. Scenario Sets 1 and 2 Pallets Out Distribution Comparisons. 


The specific design points within the scenario set that performed best and worst 
are indicated in Figure 29. For the design points that performed well, the only 
commonalities seen in the inputs are a lower v_TAKE_Arr_Cycle and higher percentages 
of V _percent_Navy_Cont. As for the poorest performing design point, in contrast to the 
better performers, it has higher values for v_TAKE_Arr_Cycle and lower values for 
V _percent_Navy_Cont. This insight is analyzed further later in this section. 


61 





















































Looking at system performance in broader terms, Figure 30 shows the comparison 
of the annual mean values for the MOEs and their measurable differences. 


I 

OS 

Oi 

04 

0 ^ 

o 


T-AKE Service Level Com pari son 


Average Pallet Output Comparison 



0»flcrt-»tec In T-AKf M i'SiiS 


8000D 

/LUUJ 

GOOOD 

5000D 

3000D 

2LIUUJ 

UKKU 

0 



Figure 30. Scenarios 1 and 2 MOE Annual Average Value Comparisons. 


By quantifying the significant effect of competing requirements on the system, the 
next step in analysis of this scenario is to explore the factors in the model that are 
contibutors to this effect. In order to identify these possible significant factors, both 
regression analysis and the nonparametric method of regression tree partitioning are used 
to see if any particular factors in the model are significant. 

In the Step History table, a stepwise regression analysis of both Scenario Set 2 
MOEs indicates the order in which the terms entered the model and shows the effect, as 


reflected by RSquare. The significant factors in the set are v_OCS_Arr_Cycle, 
v_TAKE_Arr_Cycle,v_Cont_per_OCS, and v_percent_Navy_Cont. Figure 31 is the IMP 
output for a stepwise regression analysis of Scenario Set 2. 



Figure 31. Stepwise Regression Analysis of Scenario Set 2. 

62 














































Using this analysis, v_TAKE_Arr_Cycle is the largest contributing factor for both 
MOEs. In the case of T-AKE SE, for every additional day added to the 
v_TAKE_Arr_Cycle interval, the service level increases by approximately 2%. This 
result makes sense in that as the number of T-AKEs that enter the system goes down, the 
traffic intensity seen at Kilo Wharf decreases, and allows for fewer ships in the queue. 
Eewer ships in the queue translates into increased chances of reaching Kilo Wharf and 
completing service. In the case of Pallets Out, for every additional day added to the 
v_TAKE_Arr_Cycle interval, the number of Pallets Out decreases by approximately 2,400 
pallets. This result also makes sense. As fewer T-AKEs enter the system, the 
opportunity for T-AKEs to load pallets also decreases. 

The factor, v_OCS_Arr_Cycle, contributes to the T-AKE SE with the same logic 
as v_TAKE_Arr_Cycle. More OCSs equates to more chances of waiting in the queue and 
less chance of being served. However, when considering Pallets Out, v_OCS_Arr_Cycle 
has a reciprocal effect. As the arrivals of OCSs becomes more spread out, more T-AKEs 
are able to be served and therefore. Pallet Out increases. 

Scenario Set 2 main effects regression analysis of both MOEs indicates by a 
Prob>ltl that the significant factors in the model are v_OCS_Arr_Cycle, 
v_TAKE_Arr_Cycle, v_Cont_j)er_OCS, and v _percent_Navy_Cont. Prob>ltl is the 
probability of getting an even greater t-statistic (in absolute value), given the hypothesis 
that the parameter is zero. This is the two-tailed test against the alternatives in each 
direction. Probabilities less than 0.05 are often considered as significant evidence that 
the parameter is not zero (SAS Institute Inc., 2007). Eigure 32 is the IMP output for a 
main effects regression analysis of Scenario Set 2. 


63 




Figure 32. Regression Analysis of Scenario Set 2. 

The results of the regression analysis direct the focus of the nonparametric 
analysis that follows. Before arbitrarily partitioning the data, a decision is required to 
determine the approriate number of partitions used in the analysis. Deciding the 
approriate number of partitions is accomplished by plotting the RSquare values by 
partition to find a point of diminishing returns. RSquare estimates the proportion of the 
variation in the response around the mean that can be attributed to terms in the model, 
rather than to error (SAS Institute Inc., 2007). An initial number of 10 partitions is used 
to evaluate the RSquare. Figure 33 shows the RSquare plot for Scenario Set 2 partitions 
and indicates where the diminishing returns are observed for further partitions. 


64 






























RSquare Analysis 



- Scenario Set 2 SL 

- Scenario Set 2 PO 


Figure 33. RSquare Plot for Scenario Set 2 Partitions. 


By evaluating the 10 partitions, the bend in the curve for both MOEs occurs 
between the fourth and fifth split for each MOE. Using this information, each MOE is 
evaluated through the fifth partition. Using the regression analysis previously conducted, 
along with the partition trees, provides insights into how the significant factors involve 
themselves in the system under certain conditions. 

The Partition platform in JMP 7.0 recursively partitions data according to a 
relationship between the X and Y values, creating a tree of partitions. It finds a set of 
cuts, or groupings, of X values that best predict a Y value. It does this by exhaustively 
searching all possible cuts or groupings. These splits (or partitions) of the data are done 
recursively, forming a tree of decision rules until the desired fit is reached (SAS Institute 
Inc., 2007). Figure 34 displays the partitioning and column contributions of the factors. 


65 






















Figure 34. Partition and Column Contribution of T-AKE SL in Scenario 2. 

The largest contributor through five partitioning splits is the decision factor, 
v_TAKE_Arr_Cycle. This is expected because a well-timed arrival rate for T-AKEs 
minimizes the number of T-AKEs that are held in queue. This results in maximizing the 
number of T-AKEs that are served, thus an increase in T-AKE SE. The only competing 
requirement shown as a critical factor is v _percent_Navy_Cont. The only factor 
indicated as having significance in the regression analysis, that does not appear in the 
first five partitions, is the factor v_OCS_Arr_Cycle. 

The partition tree also provides insights into situational influences of the input 
parameters on the system. Eor example, the first split is on v_TAKE_Arr_Cycle greater 
than or equal to 16. Eollowing the split to the right, the next influencing parameter is the 
competing requirement, v _percent_Navy_Cont. Therefore, in a situation in which the 
T-AKE cycle is greater than 16 days, the best policy for maximizing T-AKE SE is to 


66 






























have 100% Navy ordnance. If this is not possible, the mean T-AKE SL will be 74%. 
This type of “If-Then” analysis is useful to the decision maker when faced with 
situational decision making. 

Again here, the largest contributor through five partitioning splits is the decision 
factor, v_TAKE_Arr_Cycle. As discussed in the the T-AKE SL analysis, this result is 
expected. An unexpected difference in the partition tree analysis from the regression 
analysis is the contribution of v_CS_Arr_Time. In this partition tree analysis, both 
competing requirement factors, v _percent_Navy_Cont and v_CS_Arr_Time, are shown as 
critical factors. However, just as in the analysis on the T-AKE SL partition, 
V _percent_Navy_Cont is a larger contributor in the Pallets Out partition. Erom 
examination of the tree in Eigure 34, the number of Pallets Out is most affected when the 
competition for ordnance is greater than 9% in this Scenario Set. Eigure 35 displays the 
partitioning and column contributions of the factors. 



Number 


Term of Splits SS 


V OCS Arr Cycle 

0 

0 


V TAKE Arr Cycle 

2 

8.9234e+10 


y_CS_Arr_Rate (A) 

1 

1.2212e+10 

J 

y_Cont_per_OCS 

0 

0 

□ 

y_percent_Nayy_Cont 

2 

1.9133e+10 

y percent unstuffed pier 

0 

0 



Total 5 1.2058e+11 


Eigure 35. Partition and Column Contribution of Pallets Out in Scenario 2. 

67 




























In summary, Scenario Set 2 quantifies the effect of including competing 
requirements to the system as a reduction in mean T-AKE SL by 25.33% and mean 
Pallets Out by 18.97%. Scenario Set 2 also indicates that the significant factors in the 
model are v_OCS_Arr_Cycle, v_TAKE_Arr_Cycle, vjCont_per_OCS, and 
V _j)ercent_Navy_Cont. Of these, v_TAKE_Arr_Cycle appears as the strongest candidate 
of all inputs, and v _j>ercent_Navy_Cont appears as the strongest candidate of the 
competing requirements to have the greatest effect on the system. 

4. Analysis of Scenario Set 3 - Simulating New Magazine Baseline 


Scenario Set 3 is similar to Scenario Set 1, with the exception of the distance to 
the primary ordnance storage facility. The expeeted results are an increase in both 
MOEs, as compared to the initial baseline of Scenario Set 1. Figure 36 shows the 
distributions of Scenario Set I’s and 3’s MOEs. 



1.01 • 
1 • 
0.99- 
0.98- 
0.97- 
0.96- 
0.95- 
0.94- 
0.93- 
0.92- 
0.91 • 


H~l 

- 

1 

c2 

1 

_ 


Oudiitiles 


^ Moments 

Mean 0.9629644 

StdDev 0.0209124 

Std Err Mean 0.0020912 
upper 95% Mean 0.9671139 
lower 95% Mean 0.9588149 
N 100 


0.98- 
0.97- 
0.96- 
0.95- 
0.94- 
0.93- 
0.92- 
0.91 ■ 


I 1 

- 


c2 

1 



OiMiitiles 


^ Moments 

Mean 
Std Dev 
Std Err Mean 
upper 95% Mean 0.9676111 
lower 95% Mean 0.9591874 
N 100 


0.9633992 

0.0212267 

0.0021227 



^ Oiicintiles | 


Moments 


Mean 71458.33 

Std Dev 1570.5245 

Std Err Mean 157.05245 
upper 95% Mean 71769.956 
lower 95% Mean 71146.704 
N 100 



Moments 


Oiiontiles 


Mean 71488.89 

Std Dev 1590.5134 

Std Err Mean 159.05134 
upper 95% Mean 71804.482 
lower 95% Mean 71173.298 
N 100 


Figure 36. Scenario Sets 1 and 3 - Direct Comparisons of MOE Distributions. 


The New Magazine Baseline scenario produced a mean T-AKE SE of 96.34%, 
with a 95% confidence interval of (95.92, 96.76). It also produced a mean Pallets Out 
value of 71488.89 pallets, with a 95% confidence interval of (71,173.30, 71,804.48). 
When compared to the Scenario Set 1 baseline, T-AKE SE experiences a 0.04% increase 
in expected service level. Concurrently, the expected value of Pallet Output is increased 

68 












































































by 73 pallets annually. As a percentage of pallet throughputs, the new magazine 
positively affects the system annually by approximately 0.1%. Considering that there is 
overlap of the MOE confidence intervals between Scenarios 1 and 3, the effect of the new 
magazine on the system is insignificant and possibly attributable to the model variance. 

The lack of significant difference in outcomes between Scenario Sets 1 and 3 
indicates that simply changing the distance that either containerized or break-bulk has to 
travel does not produce a noticeable effect in the MOEs. Eooking at system performance 
in broader terms, Eigure 37 shows the comparison of the annual expected values for the 
MOEs and their measurable differences. 



Eigure 37. Scenarios 1 and 3 MOE Annual Average Value Comparisons. 

In summary. Scenario Set 3 shows very little difference from Scenario Set 1 in 
either MOE. This result provides insights into the efficiency of the operation. The 
distance travelled in the process does not appear to have a significant effect on the 
efficiency of the process. This is an important finding because it shows that even in the 
best-case scenario of no competing requirements, changing the distance ordnance has to 
travel is not a critical path to improving either T-AKE SE or Pallets Out. Eliminating 
distance as a factor, only leaves the volume of ordnance operations capable as an area of 
interest. Specifically, in Scenario Set 5, this thesis looks into the available resource 
aspect of the problem. 


69 



































5. Analysis of Scenario Set 4 - Simulating Completion of Magazine on 
Orote 

This section begins with a comparison of Scenario Sets 3 and 4. Upon the 
addition of competing requirements to the system, quantitative measurement of the effect 
on the system is measured. Comparing this scenario against the baseline scenario shows 
the immediate quantitative results of competing requirements. Figure 38 shows the 
comparison of the distributions of Scenario Sets 3 and 4 T-AKE SL. 



Figure 38. Scenario Sets 3 and 4 T-AKE SL Distribution Comparisons. 


Considering that there is no overlap of the T-AKE SL confidence intervals 
between Scenarios 1 and 2, the impact of competing requirements on the system is 
significant and not attributable to the model variance. The New Magazine with 
Competing Requirements scenario produced a mean T-AKE SL of 71.02%, with a 95% 
confidence interval of (70.53, 71.49). When compared to the baseline, T-AKE SL sees 
an effect of 25.33% reduction in expected service level. 

The New Magazine with Competing Requirements scenario also produced a mean 
Pallets Out value of 57910.70 pallets, with a 95% confidence interval of (57445.53, 
58375.86). When compared to the baseline, T-AKE SL sees an effect of 25.32% 
reduction in expected service level. Concurrently, the expected value of pallet output is 


70 




































reduced by 13,621 pallets annually. As a percentage of reduction in pallet throughput, 
competing requirements influence the system by approximately 19.04%. A comparison 
of Pallets Out is seen in Figure 39. 



Figure 39. Scenario Sets 3 and 4 Pallets Out Distribution Comparisons. 

Looking at system performance in broader terms. Figure 40 shows the comparison 
of the annual expected values for the MOEs and their measurable differences. 



Figure 40. Scenarios 3 and 4 MOE Annual Average Value Comparisons. 


71 


























































Considering that there is no overlap of the MOE confidence intervals between 
Scenarios 3 and 4, as seen in Figures 38 and 39, the impact of competing requirements on 
the system is significant and not attributable to model variance. 

The next step in the analysis is to compare Scenario Sets 4’s MOEs to those in 
Scenario Set 2. Scenario Set 4 mirrors Scenario Set 2 as the baseline comparisons did in 
the previous analysis. This gives a comparison of the current system and the system that 
will exist when the new magazine construction is completed. Figure 41 shows the 
distributions of the Scenario Set 2’s and 4’s MOEs. 



Figure 41. Scenario Sets 2 and 4 - MOE Distributions. 


The lack of significant difference in outcomes between the Scenario Sets 2 and 4 
indicates that simply changing the distance that either containerized or break-bulk has to 
travel does not produce a noticeable effect in the MOEs. These results are very similar to 
the results comparing Scenario Sets 1 and 3. Eooking at system performance in broader 
terms, Figure 42 shows the comparison of the annual expected values for the MOEs and 
their measurable differences. 


72 


























































































T-AKESL Comparison Current and 
Future System 

1.00 
0.80 
O.CO 
0.40 
0.20 
U.1NJ 

DiHormco In T AKF SI = 0.0S% 



Average Comparison Current and 
Future System 


60000 


50000 
•I0000 
30000 
20000 
KKKIO 
0 



Pifte'eoce in -»3iiet Oitput = / 


Figure 42. Scenarios 2 and 4 MOE Annual Average Value Comparisons. 

This is an extremely insightful finding when considering the cost of building the 
new magazine. Estimating costs for the magazine are stated at $76M (NAVBASE 
GUAM DD Form 1392, 2005). Using these estimates and the results of this thesis, the 
projected cost is $1.04M per pallet improvement. These calculations do not consider the 
return on investment over time, but they do suggest that further simulation and modeling 
of the system and its infrastructure are required before further capital invenstment occurs. 
These calculations also do not consider the explosive safety issues that are considered 
when making infrastructure investments of this size. Setting these exceptions aside, the 
results still provide a strong argument for using simulation and modeling to assist in the 
decision-making process. Table 10 calculates Average Pallet Output difference between 
Scenario Sets land 3, as well as Scenario Sets 2 and 4. 

Table 10. Average Annual Pallet Throughput Calculations. 


Future with Closer Magazine 
- Current System 

Baselines SS 1 & 3 

Competing Requirements SS 2 & 4 

71532 

57911 

71458 

57904 

Average Pallet Output 

73 

7 


By quantifying the significant impact of competing requirements on the system, 
the next step in analysis of this scenario is to explore the factors in the model that are 
possible contibutors to this effect. In order to identify these possible significant factors. 


73 





























both regression analysis and the nonparametrie method of regression tree partitioning is 
used to see if any partieular faetors in the model are signifieant. Eaeh MOE is evaluated 
using this method. 

In the Step History table, a stepwise regression analysis of both Seenario Set 4 
MOEs indieates the order in whieh the terms entered the model and shows the effeet, as 
reflected by RSquare. The significant factors in the set are v_TAKE_Arr_Cycle, 
V_percent_unstujfed_pier, v_OCS_Arr_Cycle, and v _percent_Navy_Cont. Eigure 43 is 
the IMP output for a stepwise regression analysis of Scenario Set 2. 




^ ^ Stepwise Fit 

Response: T-AKE SL 


Stepwise Retjiessloii Control 


Prob to Enter 
Prob to Leave 
Direction Forward v ( KenTovt 

I Go I I Stop I I Step I [ Make Model | 


Ciineiit Estimates 


A 






^ Stepwise Fit 


Response; Pellets Out 

^ Stepwise Regiession Contiol 

Prob to Enter 

0.010 


Prob to Leave 

0.100 


Direction: Forward v 

1 Remove All | 


I Go 11 Stop ] [ Step I [ Make Model 


Ciirieiit Estimates 



SSE 

DFE MSE RScitiare 

RSquare Ailj 

Cp 

AlC 




SSE 

DFE MSE RSquare RSquare Adj 

Cp 

AlC 



7.3789106 

1695 0.0043533 

0.5684 

0.5674 

11.732327 

-9237.59 



4.227e+10 

1695 24935205 

0.7398 0.7392 

12.576368 

28959.04 



Lock Entered Parameter 

Estimate nDF 

ss 

T Ratio" 

•T»rob>F" 


Lock Entered Parameter 

Estimate nDF 

SS 

T Ratio" 

•T>rob>F" 




Intercept 

-0.1970707 1 

0 

0.000 

1.0000 




Intercept 

42308.4642 1 

0 

0.000 

1.0000 


n 

Rl 

V OCS Arr Cvde 

0.0228816 1 

1.379038 

316.777 

0.0000 


n 

FI 

v_OCS_Arr_Cyde 

1811.10144 1 

8.64e*9 

346.478 

0.0000 


11 


V TAKE Arr Cycle 

0.0193878 1 

5.846092 

1342.898 

0.0000 


11 

FI 

v_TAKE_Arr_Cycl8 

-2285.4276 1 

8.12e+10 

3257.848 

0.0000 


n 

n 

V Cont oer OCS 


0 1 

0.01012 

2.327 

0.1274 


11 

1 1 

v_ContjDer_OCS 

0 1 

1.603e+8 

6.448 

0.0112 


n 

FI 

V oercent Navv Cont 

0.19570009 1 

0.54114 

124.304 

0.0000 


□ 

FI 

V percent Navy Cont 

16718.0272 1 

3.949e-<-g 

158.374 

0.0000 


n 

FI 

vj5ercent_unstuffedj3ier 0.26668231 1 

3.335614 

766.220 

0.0000 


u 

bd 

vjDercent_unstuffed_pier 21489.3743 1 

2.17e+10 

868.604 

0.0000 


11 

1 1 

Ordnance Inspector 


0 1 

0.000571 

0.131 

0.7173 


11 

1 1 

Ordnance Inspector 

0 1 

6307748 

0.253 

0.6151 


n 

n 

Urrstuffing Space 


0 1 

0.000706 

0.162 

0.6873 


11 

1 1 

Unstuffing Space 

0 1 

18833759 

0.755 

0.3850 


u 

u 

v_CS_Arr_Rate 


0 1 

0.021074 

4.852 

0.0278 


u 

U 

v_CS_Arr_Rate (A) 

0 1 

49414472 

1.983 

0.1593 


^ Step History 

^ Step Histoiy 

Step 

Parameter 

Action 

••Sig Prob" 

Seq SS 

RSquare 

Cp 

l> 

Step 

Pat ametei 

Action "Sig Prob" 

Seq SS 

RSquare 

Cp 

P 


1 

V TAKE Arr Cycle 

Entered 

0.0000 

5.104379 

0.2986 

1069.6 

2 


1 

v_TAKE_Arr_Cyde 

Entered 0.0000 

9e+10 

0.5538 

1223.9 

2 


2 

V oercent unstuffed oier Entered 

0.0000 

2.669734 

0.4547 

455.94 

3 


2 

vj5ercent_unstuffedjDier Entered 0.0000 

1.75e+10 

0.6614 

522.03 

3 


3 

V OCS Arr Cvcie 

Entered 

0.0000 

1.402331 

0.5367 

134.53 

4 


3 

V OCS Arr Cycle 

Entered 0.0000 

8.797e+9 

0.7155 

169.66 

4 


4 

vj5ercent_N8vy_Cont 

Entered 

0.0000 

0.54114 

0.5684 

11.732 

5 


4 

V _percent_Mavy_Cont 

Entered 0.0000 

3.949e+9 

0.7398 

12.576 

S 


Eigure 43. Stepwise Regression Analysis of Scenario Set 4. 

Using this analysis, v_TAKE_Arr_Cycle is the largest contributing factor for both 
MOEs. In the case of T-AKE SE, for every additional day added to the 
v_TAKE_ArrjCycle interval, the service level increases by approximately 2%. This 
result makes sense, in that, as the number of T-AKEs that enter the system goes down, 
the traffic intensity seen at Kilo Wharf decreases and allows for fewer ships in the queue. 
Eewer ships in the queue translates into increased chances of reaching Kilo Wharf and 
completing service. In the case of Pallets Out, for every additional day added to the 
v_TAKE_Arr_Cycle interval, the number of Pallets Out decreases by approximately 


74 




































2,300 pallets. This result also makes sense. As fewer T-AKEs enter the system, the 
opportunity for T-AKEs to load pallets also decreases. These results are very similar to 
those seen in Scenario Set 2. 

The factor, v_OCS_Arr_Cycle, contributes to the T-AKE SE, with the same logic 
as v_TAKE_Arr_Cycle. More OCSs equates to more chances of waiting in the queue and 
less chance of being served. However, when considering Pallets Out, v_OCS_Arr_Cycle 
has a reciprocal effect. As the arrivals of OCSs becomes more spread out, more T-AKEs 
are able to be served and, therefore. Pallet Out increases. 

The analysis differs from Scenario Set 2 in that the factor 
V _j)ercent_unstujfed_j)ier has replaced v_Cont_j)er_OCS as a contributing factor. This is 
is very interesting, considering the specific scenario. In this scenario, the magazine is 
closer and yet there appears to be a benefit to increasing the amount of ordnance that is 
processed pier-side. 

Eigure 44 is the regression analysis of Scenario Set 4’s MOEs. The analysis of 
Scenario Set 4 T-AKE SE indicates by a Prob>ltl that the significant factors in the model 
are v_OCS_Arr_Cycle, v_TAKE_Arr_Cycle, v_CS_Arr_Time, v_percent_Navy_Cont, and 
V _j)ercent_unstujfed_j)ier. Similarly, regression analysis of Scenario Set 4 MOE Pallets 
Out indicates by a Prob>ltl that the significant factors in the model are 
v_OCS_Arr_Cycle, v_TAKE_Arr_Cycle, v_CS_Arr_Time, v_Cont_per_OCS, 

V_percent_Navy_Cont, and v _percent_unstuffed_pier. Each of these regressions tells a 
story about the system and the influence of the identified factors. Eor example, every 
unit percent increase in v _percent_unstujfed_pier positively influences the system by 
0.266 in service level. Therefore, this analysis indicates that unstuffing pier-side in this 
scenario is an efficient process that increases T-AKE SE. Another example is for every 
v_TAKE_Arr_Cycle unit added, the Pallets Out is influenced negatively by 
2273.359 pallets. Eogically, this makes sense, in that the further apart arrivals are to the 
wharf, the fewer pallets are able to leave the system. Therefore, an ideal cycle time for 
T-AKEs will limit congestion, while maximizing pallet output. 


75 




Figure 44. Regression Analysis of Seenario 4. 

The results of the regression analysis direet the focus of the nonparametric 
analysis that follows. Using a method similar to that used in Scenario Set 2 analysis, a 
decision to determine the approriate number of partitions is accomplished by plotting the 
RSquare values by partition to find a point of diminishing returns. An initial number of 
10 partitions is used to evaluate the RSquare. Figure 45 shows the RSquare plot for 
Scenario Set 4 partitions and indicates where the diminishing returns are observed for 
further partitions. 


76 



































RSquare Analysis 



■- Scenario Set 4 SL 

- Scenario Set 4 PO 


Figure 45. RSquare Plot for Scenario Set 4 Partitions. 


By evaluating the 10 partitions, the bend in the curve for both MOEs occurs 
between the fourth and fifth split for each MOE. Using this information, each MOE is 
evaluated through the fifth partition. Using the regression analysis previously conducted, 
along with the partition trees, provides insights into how the significant factors involve 
themselves in the system under certain conditions. Eigure 46 displays the partitioning 
and column contributions of the factors. 


77 























Number 


Term of Splits SS 

v_OCS_Arr_Cycle 1 0.50090228 _ 

v_TAKE_Arr_Cycle 1 5.4524101 _ 

v_Cont_per_OCS 0 0 

v_percent_Navy_Cont 1 0.94452705 

v_percent_unstuffed_pier 1 0.92805747 

Ordnance Inspector 0 0 

Unstuffing Space 0 0 

v_CS_Arr_Rate (A) 1 1.59146405 H. 


Total 5 9.41736095 


Figure 46. Partition and Column Contribution of T-AKE SL in Scenario 4. 


The largest contributor through five partitioning splits is the decision factor, 
v_TAKE_Arr_Cycle. The next largest contributor is now the competing requirement, 
v_CS_Arr_Time. Both of these factors are expected to influence T-AKE SE because 
larger spacing of interarrival times creates a decrease in traffic intensity. A decrease in 
traffic intensity gives the server more opportunity to serve each arrival to the system. 

The competing requirement, v _percent_Navy_Cont, does appear again as a 
contributor, but interestingly, v _percent_unstujfed_pier is a larger contributor. This is an 
interesting insight because the splits where v _j>ercent_unstuffed_j)ier appear are based on 
the v_TAKE_Arr_Cycle and are far apart. Eor v_TAKE_Arr_Cycle less than every 15 
days, the split for v _percent_unstujfed_pier occurs at 66%. On the other hand, for a 
v_TAKE_Arr_Cycle greater than or equal to 15 days and v_CS_Arr_Time less than 30, 
the split for v _percent_unstujfed_j)ier occurs at 93%. By definition of the contingency 


78 






























that establishes the seenario, this would limit T-AKE interarrival times, while enduring 
CS requirements near what they have been historieally. Otherwise, onee 
v_TAKE_ArrjCycle is less than 15, the results indieate that the eurrent praetiee of 
unstuffing as elose to 100% of eontainers possible pier-side may not be the best praetiee. 


Figure 47 displays the partitioning and column contributions of the factors. 



Number 

Term of Splits SS 

v_OCS_Arr_Cycle 0 0 

v_TAKE_Arr_Cycle 2 8.9321 e+10 

v_Cont_per_OCS 0 0 

v_percent_Navy_Cont 0 0 

v_percent_unstuffed_pier 2 2.7715e+10 

Ordnance Inspector 0 0 

Unstuffing Space 0 0 

v_CS_Arr_Rate (A) 1 3475125828 

Total 5 1.2051e+11 


Figure 47. Partition and Column Contribution of Pallets Out in Scenario 4. 


Again, the largest contributor through five partitioning splits is the decision 
factor, v_TAKE_Arr_Cycle. In this MOE analysis of the competing requirement factors, 
v_CS_Arr_Time, is the only one shown as a contributing factor. Just as in the analysis on 
the T-AKE SL partition, v _percent_unstujfed_pier is a large contributor in the Pallets Out 
partition. 


79 































In summary, Scenario Set 4 quantifies the impaet of ineluding eompeting 
requirements to the system as a reduetion in T-AKE SL by 25.32% and Pallets Out by 
19.04%. Seenario Set 4 also indieates that v _j)ercent_unstujfedjjier is the strongest 
eandidate as the eritieal faetor to have the greatest effect on the system. 

6. Analysis of Scenario Set 5 - Exploratory Set 

Scenario Set 5 begins the exploration of the system beyond the competing 
requirements examined in the previous scenarios. Introducing a few resource eapaeities, 
this scenario primarily focuses more on the possible outcomes, with the acquisition of 
resources. Upon the addition of these new input parameters to the system, quantitative 
measurement of effect on the system is calculated. Figure 48 shows the distributions of 
the Scenario Set 5’s MOEs. 



Figure 48. Scenario Set 5 - The Exploratory Model MOE Distributions. 

The Exploratory seenario produced a mean T-AKE SE of 74.55%, with a 95% 
confidence interval of (74.17, 74.93). It also produced a mean Pallets Out value of 
60,807 pallets, with a 95% eonfidence interval of (60,472, 61143). When compared to 
the baseline, T-AKE SE sees an effect of 21.75% reduction in expected service level. 
Coneurrently, the expected value of pallet output is redueed by 10,651 pallets annually. 
As a pereentage of reduetion in pallet throughput, competing requirements negatively 

80 








































influence the system by approximately 14.91%. These results are slight increases in 
comparison to the results seen in Scenario Set 2 which had fewer variable input 
parameters. This result is expected because allowing for additional resources available to 
the system provides the system with more possible configurations in which to operate as 
best as possible. Figure 49 shows the MOE comparisons between Scenario Sets 1, 2, and 
5. 



Figure 49. Scenario Sets 1, 2, and 5 T-AKE Distributions. 


Considering that there is no overlap of the T-AKE SE confidence intervals 
between Scenario Sets 1 and 5, the effect of competing requirements on the system is 
significant and not attributable to the model variance. When considering the differences 
between Scenario Sets 2 and 5, the initial overall indication is that there is some positive 
effect from the additional resources available to the system. Figure 50, which shows this 
slight increase, displays the mean MOEs. 


81 






























































T-AKE Service Level CR and Exploratory 
Comparison 





ScendrioSe: 1 


Scena'ioS«t 2 


Sc€ndricS«t £ 


Average Pallet Output CR and 
Exploratory Comparison 


80000 - 
/ww ■ 

MIUUU 
SOOOO 
40000 - 
30000 ■ 
2U0UU ' 
lUOUU 


— 


SrwariciSpt 1 SrenarinSet ? 


SceivtrioSet S 


Figure 50. Scenarios 1, 2, and 5 MOE Annual Average Value Comparisons. 


However, the lack of a practical, significant difference in outcomes between 
Scenario Sets 2 and 5 indicates that simply adding more resources does not produce a 
statistically significant effect in the MOEs. Conversely, this experiment limited the 
number of resources possible in the system. This provides a great opportunity for future 
studies to explore the bounds of resource allocation limits and their effects on the system. 
Possible candidates for this research may come to light later in this chapter, when critical 
factors are identified through regression analysis and partition trees. 

By quantifying the significant effect of competing requirements on the system, the 
next step in the analysis of this scenario is to explore the factors in the model that are 
possible contibutors to this effect. In order to identify these possible significant factors, 
both regression analysis and the nonparametric method of regression tree partitioning are 
used to see if any particular factors in the model are significant. Each MOE is evaluated 
using this method. Eigure 51 is the IMP output for the regression analysis of Scenario 
Set 5. 


82 






























i Tr'rrtlT— 


' Response TAKE SL 




^ Response Pollets Out 


▼ Actual by Predicted Plot 



TAKE SL Predicted P<.0001 
RSq=0.48 RMSE=0.0801 


Summary of Fit 

RSquare 0.48359 

RSquareAdJ 0.482177 

Root Mean Square Error 0.080072 

Mean o1 Resporree 0.7455 

Observations (or Sum Wgts) 3300 


' Aiidlvsis of Variance 



Source DF 

Sum of 
Squares 

Mean Square 

F Ratio 

Model 9 

19.753507 

2.19483 

342.3223 

Error 3290 

21.094166 

0.00641 

Piob >F 

C-Total 3299 

40.847673 


0.0000* 

’ Lack Of Fit 





Sum of F Ratio 


Source DF Squares Mean Square 83.5015 

Lack Of F« 23 7.809504 0.339544 Prob>F 

Pure Error 3267 13.284662 0.004066 < 0001* 

TotalError 3290 21.094166 MaxRSq 

0.6748 

^ Parameter Estimates 


^ Actual by Predicted Plot 



Pellets Out Predicted 
P<.0001 RSq=0.63RMSE=5970.7 


^ Summaiy of Fit 

RSquare 0.631662 

RSquare AcQ 0.630654 

Root Mean Square Error 5970.666 

Mean of Response 60807.26 

Observations (or Sum Wgts) 3300 

^ Analysis of Variance 


Sum of 

Source DF Squares Mean Square F Ratio 

Model 9 2.0113e+11 2.235e+10 626.8901 

Error 3290 1.1728e+11 35648847 Ptob ’ F 

C. Total 3299 3.1842e+11 0.0000* 

▼ Lack Of Fit _ 

Sum of F Ratio 

Source DF Squares Mean Square 70.4259 

Lack Of Fit 23 3.8876e+10 1.6902e+9 Ptob-F 

Pure Error 3267 7,8409e*10 24003326 <.0C»1* 

TotalError 3290 1.1728e*11 MaxRSq 

0.7538 

^ Parameter Estimates 


Term 

bitercept 

v_OCS_Arr_Cyde 
V_TAKE_Arr_Cycle 
v_CS_Arr_Rate (A) 
v_Contj3er_OCS 
vjDercent_N8vy_Cont 
v_percent_unstuffedjDief 
Ordnance inspectors 
Unstuffing Space 
Ordnarrce Forklifts 


Estimate Std Error 
-0.187758 0.033958 

0.0330044 0.00112 

0.0214056 0.00047 

0.0015853 0.0£M126 
3.7811e-5 0.001047 

0.1509555 0.015665 

5.0121e-5 0.007818 
-0.00287 0.00524 
0.002658 0.00117 

0.0012009 0.0112 


t Ratio Prob>|t| 
-5.53 <.0001* 

29.46 <.0001* 

45.55 0.0000* 

12.56 <.0001* 

0.81 0.4206 

9.64 <.0001* 

0.01 0.9949 

•0.55 0.5842 

2.26 0.0236* 

1.07 0.2836 


Term 

Intercept 

v_OCS_Arr_Cycle 
v_TAKE_Arr_Cycle 
v_CS_Arr_Rate (A) 
v_Cont_per_OCS 
v_percent_Navy_Cont 
v_percer4_unstuffed_pier 
Ordnance Inspectors 
Unstufling Space 
Ordnance Forkifts 


Estimate Std Error 

50161.464 2532.073 

2642.5635 83.54071 

-2264,672 35.0385 

154.52962 9.411181 

10.002524 3.500498 

9594.6071 1168.107 

-1296.016 582.9539 

48.046988 39.08692 

-11.30691 8.753526 

-20.44158 83.49913 


t Ratio Prob>ft| 

19.81 <.IXI01* 

31.63 <.(XI01* 

-64.63 0.0000* 

16.42 <.0001* 

2.86 0.CW43*_ 

8.21 <.0001* 
-2.22 0.0263* 

1.23 0.2191 

-1.29 0.1966 

-0.24 0.^66 


Figure 51. Scenario Set 5 Main Effects Regression. 


Regression analysis of Scenario Set 5 MOE T-AKE SL indicates by a Prob>ltl 
that the significant factors in the model are v_OCS_Arr_Cycle, v_TAKE_Arr_Cycle, 
v_CS_Arr_Time , v_Cont_per_OCS, v_percent_Navy_Cont, and Unstuffing Space. 
Similarly, regression analysis of Scenario Set 5 MOE Pallets Out indicates by a Prob>ltl 
that the significant factors in the model are v_OCS_Arr_Cycle, v_TAKE_Arr_Cycle, 
v_CS_Arr_Tinie, v_Cont_per_OCS, v _percent_Navy_Cont, and 

V jjercentjunstuffed_j)ier. Each of these regressions tells a story about the system and 
the influence of the identified factors. For example, for every v_TAKE_Arr_Cycle unit 
added, the T-AKE SL is influenced positively by 2.1%, and the Pallets Out is influenced 
negatively by 2265 pallets. 

The results of the regression analysis direct the focus of the nonparametric 
analysis that follows. Using a method similar to that used in Scenario Set 2 analysis, a 
decision to determine the approriate number of is accomplished by plotting the RSquare 


83 





































values by partition to find a point of diminishing returns. An initial number of 
10 partitions is used to evaluate the RSquare. Figure 52 shows the RSquare plot for 
Seenario Set 5 partitions and indieates where the diminishing returns are observed for 
further partitions. 



Figure 52. RSquare Plot for Seenario Set 5 Partitions. 


By evaluating the 10 partitions, the bend in the eurve for the Pallets Out MOE 
occurs between the fourth and fifth split. The RSquare values for T-AKE SE appear to 
have two breakpoints, where the first occurs after the second split and the next after the 
eigth split, with a significant slope increase at the fourth split. Using this information, 
each MOE is evaluated through the fifth partition. Using the regression analysis 
previously conducted, along with the partition trees, provides insights into how the 
significant factors involve themselves in the system under certain conditions. Eigure 53 
displays the partitioning and column contributions of the factors. 


84 






















Number 


Term 

of Splits 

ss 



V OCS Arr Cycle 

0 

0 



V TAKE Arr Cycle 

1 

11.7134192 



y_CS_Arr_Rate (A) 

1 

1.12985923 




y_Cont_per_OCS 

0 

0 




y percent Nayy Cent 

0 

0 




y percent unstuffed pier 

2 

7.13626954 




Ordnance Inspecters 

1 

0.68698527 




Unstuffing Space 

0 

0 




Ordnance Forklifts 

0 

0 




Total 

5 

20.6665333 



Figure 53. Partition and Column Contribution of T-AKE SL in Scenario 5. 

Through five partitioning splits for T-AKE SE, the major contributors are 
v_TAKE_Arr_Cycle, v_CS_Arr_Time, v _percent_unstujfed_pier, and Ordnance 
Inspectors. The largest contributor through five partitioning splits is the decision factor, 
v_TAKE_Arr_Cycle. The next largest contributor is v _percent_unstujfed_pier, followed 
by v_CS_Arr_Time and Ordnance Inspectors. The are many interesting insights found in 
this analysis. The first is that the only competing requirement to contribute at this point 
is v_CS_Arr_Time, and it does so fractionally, compared to other contributors. The 
second interesting insight is the influence of v _percent_unstujfed_pier. The regression 
analysis of T-AKE SE did not signify v _percent_unstujfed_pier as a significant factor. 

Another interesting insight is in the second tier of the partition tree. As the 
v_TAKE_Arr_Cycle gets larger (farther apart) the factor that contributes most to the next 
split is V _j)ercent_unstujfed_j)ier. On the other hand, when v_TAKE_Arr_Cycle gets 


85 
































smaller (closer together) the factor that contributes most to the next split is Ordnance 
Inspectors. This result makes sense, in that, when T-AKEs arrive at larger intervals there 
is less competition at the wharf; therefore, the most expeditious method of processing 
containers is best. Whereas, when they arrive at tighter intervals, the most expeditious 


method of unstuffing provides the best results because the containers are transformed into 
pallets and more readily available for the arriving T-AKEs. Figure 54 displays the 


partitioning and column contributions of the factors. 



Term 

v_OCS_A rr_Cy c le 
v_TAKE_Arr_Cycle 
v_CS_Arr_Rate (A) 
v_Cont_per_OCS 
v_percent_Navy_Cont 
v_percent_unstuffed_pier 
Ordnance Inspectors 
Unstuffing Space 
Ordnance Forklifts 
Total 


Number 
of Splits 

1 

2 

0 

1 

1 

0 

0 

0 

0 

5 


ss 

1.7648e+10 
1.5362e+11 
0 

1.9478e+10 

1.4953e+10 

0 

0 

0 

0 

2.0569e+11 


J 


Figure 54. Partition and Column Contribution of Pallets Out in Scenario 5. 


Through five partitioning splits for Pallets Out, the largest contributors are the 
decision factors, v_TAKE_Arr_Cycle, v_OCS_Arr_Cycle, vjCont_per_OCS, and 
V _percent_Navy_Cont. The results are sensible, in that, v_TAKE_Arr_Cycle relates to 
how often a T-AKE arrives to pick up pallets of ordnance. Further, both 


86 






























V _percent_Navy_Cont and v_Cont_per_OCS relate to the eongestion at the wharf. Notiee 
in the partition tree of Figure 54 that the system performs better when both of these 
factors are smaller. Granted, these factors also contribute to the number of containers and 
subsequent pallets that are available in the system, but because of the initial inventory 
carried they do not affect the system in this manner. Without this initial inventory, those 
factors would have substantial influence because they directly influence the supply 
coming into the system. 

a. Process Analyzer Results 

Using these insights, the decision factors are again analyzed by comparing 
the outputs of the independent input scenarios in Scenario Set 5 to the findings in the 
partition analysis. The Arena Process Analyzer provides response (MOE) charts 
identifying the “best” scenario within the set. Since T-AKE SE is calculated in the data 
post processing from the T-AKE In and TAKE Out responses, the chart directly from the 
Process Analyzer is unavailable. However, Pallets Out is readily available for analysis in 
the Process Analyzer. Therefore, considering that Pallets Out is the MOE that most 
directly relates to combat potential in the AOR, this thesis uses it as the MOE of interest 
in this section. Eigure 55 is a box and whisker chart that identifies Scenario 15, followed 
closely by Scenario 11 as the “best” scenarios to maximize the MOE in Scenario Set 5. 

Record TAKE Pallets by Scenario 

TAk'C BsllaVc 



■ Scenario Scenario 

Eigure 55. Pallet Out Best Scenario in Scenario Set 5. 

87 



























































To gain insights into why these scenarios were the top performers, the 
scenario inputs are examined. Table 11 extracts the scenarios of interest from the NOLH 
used in Scenario Set 5. 


Table 11. Scenario Set 5 “Best” Input Parameters. 


Scenario 

111 

115 


_QJ 

<J 



> 


12 

12 


0) 

■q. 


_QJ 

<J 

CT 

0) 

(/) 

u 

O, 

c 

o 

‘^1 

? 

ra 

•a 

iS 

3 

C 

o 

tj 

0) 

Q. 

0) 

<j 

ra 

n 


1 

ro 

ec 

1 

Z 

1 

3 

1 

C 

(/) 

o 

Lk 

UJ 

j 

0) 

Q. 

^1 

C 

c 

0) 

<J 

C 

0) 

0) 

<J 

c 

ra 

(20 

C 

3 

0) 

<J 

c 

ra 

< 


Q 

0) 


c 


c 

■"i 

<3, 

Q. 

1 

Q. 

1 

•a 

f/i 

C 

•a 

> 

> 

> 

> 

> 

O 

3 

O 

12 

51 

297 

0.80 

0.44 

19 

135 

9 

12 

54 

291 

0.81 

0.40 

25 

149 

11 


The parity between the scenarios is the first noticeable finding. By 
applying an appropriately offset matching cycle between OCSs and T-AKEs, the cycle 
can be as few as every 12 days, which is less than the requirement of every 16 days, at 
most, for T-AKEs involved in supporting forces in the contingency. To accomplish this, 
an increase in the time between CS arrivals is required in order to reduce the traffic 
intensity at the wharf. This would require a policy that involved using waivers to moor 
the CSs that are ordnance-laden at other piers not commonly used by this type of vessel. 
In order to accommodate the changes in OCS and T-AKE cycle times, the number of 
containers offloaded requires an increase of 16.5% over the 255 containers suggested in 
previous studies, pushing this value into the 290 range seen in Table 11. Of these 
containers offloaded, the Navy could support up to 20% competition from the Air Eorce 
for the ordnance coming into Guam. The biggest change in this scenario from current 
operating policy is in the amount unstuffed pier-side. By reducing this number by more 
than 50%, these results are acheivable under resource conditions very close to those that 
presently exist. The current capacities for the remaining resources in Scenario 11 vary 
slightly from their low level inputs and would all be feasible during a contingency. In 
Scenario 15, where the resource capacities are higher, they are proportionally higher in 


88 



relation to the increase in Unstuffing Space. More available Unstuffing Space can only 
provide positive effects to the system, if there are ordnance inpsectors to process the 
ordnance and forklifts to move the pallets. 

In summary, Scenario Set 5 quantifies the effect of including competing 
requirements to the system as a reduction in T-AKE SL by 25.33% and Pallets Out by 
18.97%. Scenario Set 5 also indicates that v _j)ercent_unstujfed_j)ier is the critical factor 
required to change the most in order to maximize pallet throughput. 


89 



THIS PAGE INTENTIONALLY LEET BLANK 


90 



V. CONCLUSIONS 


A. THESIS SUMMARY 

This thesis set out to explore the impact of competing requirements on the 
ordnance operations currently available in the Asian Pacific Theater. Through the 
combination of previous studies, the development of a realistic scenario, an Arena-based 
simulation model, and thorough experimentation and analysis; this thesis produced a 
quantitative analysis of the challenges involving the movement of ordnance into an AOR 
of concern. The simulated movement and ordnance operations generated by this thesis 
provide a strong argument for logistics, infrastructure, and resource allocation modeling 
in future decision-making processes. This thesis also provides a strong foundation for 
future study of the challenge of moving ordnance into the Asian Pacific Theater. 

B. THESIS QUESTIONS 

The goal of this thesis was to answer the following questions: 

• How will introducing the competing requirements affect the predicted 
capabilities of the ordnance operations in Guam? 

• What are the critical factors in the ordnance operations process? 

This section briefly summarizes the answers to these questions. 

I. Effect of Competing Requirements 

The simulation experiment results showed that introducing two forms of viable 
competition into the system has a statistically significant effect on both the T-AKE 
service level and pallet throughput of the system. The impact of these effects held true 
for the current system and the system that includes the new magazine on Orote Peninsula. 
T-AKE service level in the current system is reduced by an average value of 26% 
reduction in service level with a maximum value of 52%. This means that on average 1 
of every 4 T-AKEs that enter the system is not serviced by the system. The T-AKEs not 
serviced at the end of the simulation time are left in queue. Pallet throughput is reduced 


91 



by an average of 13,555 pallets and a maximum of 41,167 pallets. This reduction in 
pallet output is equivalent to approximately four T-AKEs’ worth of ordnance that is not 
delivered to the forward edge of the contingency. 

2. Critical Factors 

Regression analysis and partition tree analysis are used to analyze the simulation 
experiment results. Across the current and new systems, the primary critical factor for 
both is the arrival cycle of the T-AKE. A greater T-AKE arrival cycle, T-AKEs arriving 
further apart, consistently caused the system to see a reduction in pallet throughput. The 
analysis results also suggest that setting the arrival cycle of the T-AKE and the OCS to 
the same interval, but with sufficient offset, reduces the impact of the competing 
requirements introduced to the system. The trade-offs to the optimal setting of the OCS 
and T-AKE arrival cycle are an increase in the number of containers offloaded from an 
OCS and a significant reduction in the number of containers unstuffed at Kilo Wharf. 

Both competing requirements were found to have statistical significance across 
the different scenario sets, but in varying intensities. The impact from competing ships 
was seen more often affecting T-AKE service level, whereas competition for ordnance 
from the Air Eorce mostly affected the overall pallet throughput. The analysis results 
suggest that the T-AKE service level improves by implementing policies during a time of 
contingency that result in the mean arrival rate of competing ships to be greater than 30 
days. It also suggests that keeping the competition for ordnance under 26% improves 
pallet throughput. 

C. ADDITIONAL INSIGHTS 

This thesis discovered several additional insights during the course of the 
experimentation and analysis. The three most significant are summarized in this section. 

I. Initial Inventory 

During the course of debugging the model, the initializing inventory required 
75,000 pallets to keep the system from ever failing. The system is considered to fail 
when a T-AKE requests more pallets than in inventory. This translates to 150 million 


92 




pounds of ordnance. This value was chosen after several tests were run and the 
initializing inventory was raised by 5,000 pallets each time, until the simulation ran to 
completion with no failures in 100 replications. No further experimentation was done 
with this value, but this does suggest that a certain inventory safety level is required to 
support a contingency of the magnitude in this thesis. This also suggests that the current 
capacity of the Ordnance Annex, which is 58 million pounds of munitions, may be 
insufficient to handle the variability of the scenario if a lesser value of initial inventory is 
the expected starting point in a contingency. 

2. Operational Capacity 

The operational capacity of the ordnance operations on Guam has been studied 
from a variety of approaches. An additional insight came to light using the simulation 
and modeling approach when the new magazine was modeled in the system. The 
simulation results indicate that there is no statistically significant difference between the 
system with the new magazine and the current system. Therefore, this component of the 
ordnance operations system is not considered a critical path. Although it seems logical 
that reducing the distance that ordnance has to travel would improve overall efficiency 
and throughput, it did not. The explanation found in the analysis is rooted in other 
critical factors. This insight provides for justification into using simulation and modeling 
research to investigate process and infrastructure improvements as a method of validating 
assumptions prior to expending large amounts of military construction funds. 

3. Theater Challenges 

The previous CNA and MSDDC studies, as well as this thesis, all indicate serious 
challenges when faced with moving a significant amount of ordnance or material through 
Guam. As this thesis developed, it was realized that having a single transshipment point 
for ordnance into the Asian Pacific Theater may be a serious issue, if its ordnance 
operations were somehow affected other than in ways introduced by our own military 
requirements. Alternative facilities in the Asian Pacific Theater are severely limited and 
eliminating Guam results in Hawaii being the western-most U.S. forward logistics base. 
That is a 3,320-nautical mile difference in forward presence. 


93 




D. RECOMMENDATIONS 


Based on the conclusions of this thesis, several recommendations are made. 

• Quantifying the impact of competing requirements to this system strongly 
promotes further research in to how to maximize the efficiency and 
throughput of the system. This information also suggests that 
incorporating alternative planning measures into the logistics planning 
portion of any major contingency in the Asian Pacific Theater is 
imperative. Variability and competition in the system are inevitable; 
therefore, future research is recommended to assist the Navy in developing 
measures to reduce the effect. 

• The ordnance requirements at the forward battle edge will determine the 
T-AKE arrival cycle and are estimated based on the operational plan 
(OPLAN) used for the contingency. Pairing the information with the 
model in this thesis will provide decision makers with their best options 
for scheduling OCS arrivals and resource allocations at Guam. With 
limited T-AKEs in the Elect, only a portion is assigned to the Asian 
Pacific Theater at any given time. In order to support the given OPEAN, 
it is recommended to use this model to assist in determining whether 
T-AKEs from other theaters are required in order to successfully achieve 
the desired T-AKE cycle. 

• Dealing with the competing requirements primarily requires policy 
adjustments or joint coordination during the development of the OPEAN. 
By granting waivers and diverting competing ships to other wharfs, the 
Navy can achieve a mean CS interarrival time greater than 50 days and 
lessen the impact seen on ordnance operations at Kilo Wharf. It is also 
recommended that strategic coordination with the Air Eorce be carried out 
to ensure that their requirements are met, but do not exceed 20% of the 
incoming ordnance. 

• The results of this thesis indicate that, under certain conditions, some of 
the current policies, such as the percent of containers unstuffed pierside, 
should be more flexible in order to maximize performance. The partition 
tree analysis approach is recommended for developing situational 
operating procedures when the given conditions exist. Adding flexibility 
to the policies that ordnance operations use, while maintaining safety 
considerations, shows improved performance of the system. 

• The insights gained from this thesis have proven valuable to identifying 
system constraints and critical factors. Development of models similar to 
the one used in this thesis should be applied to other commodities vital to 
sustaining military contingencies. In particular, fuel requirements during a 
contingency display similar logistical challenges. 


94 



E. FOLLOW-ON WORK 

The following is a list of valuable follow-on research that could be accomplished 
using this work. 

• More detailed exploration and analysis into more robust input parameter 
ranges, to include realistically infeasible ranges in order to assess the cost 
of losing resource capacities, and the value of good policies. 

• More detailed exploration and analysis into the best mix of resources for 
optimal performance when faced with the current competing requirements. 

• More detailed exploration and analysis into best mix of resources for 
optimal performance, when faced with predetermined competing 
requirements. 

• Focused analysis over the key parameters and ranges identified, including 
further analysis of parameter interactions. 

• Analysis into the optimal level of initial inventory, to ensure a level of 
system viability when faced with the variability of the contingency. 
Essentially asking, “How low can the inventory be allowed to get before 
the system fails X percent of the time?” 

• Exploration and analysis of other possible sites in the region, using the 
model as a framework for ordnance operations ashore. 

• Analysis of the alternatives for a scenario that includes periods of 
unavailability to Kilo Wharf. 

The following is a list of examples for follow-on research stemming directly from 
this thesis and the model. 

• Analysis of new technology and resources on the ordnance operations 
process; specifically, analysis of the process with the proposed gantry 
crane on Kilo Wharf. 

• Analysis of the provided contingency scenario for both shorter and longer 
periods of time. This would include extending the current model to 
account for resource maintenance and failures. 

• Analysis and development of a recommended scheduling of vessel arrivals 
to optimize the throughput of the system, while providing for the ability to 
handle fluctuation of competing requirements. 

• Extension of the model to include a dynamic queue that removes 
competing ships from the queue after a specified wait time (also known as 
“reneging”), as well as prioritizes OCSs and T-AKEs 


95 



Of the follow-on research listed above, the two that would provide the most 
insight into the system are: 

New Technology and Resources —Guam’s location makes it a cornerstone to 
success for contingencies in the Asian Pacific Theater. By applying new technologies 
and the best mix of resources to the system in Guam, every effort can be made to 
maximize its usefulness despite its limitations. 

Dynamic Queue —The flexibility of United States forces has always played a 
hand in its military successes. Developing the current model into one that provides the 
decision maker with large-scale policy and resource flexibility by including a dynamic 
queue, will provide an entirely new dimension of analyzing this challenge. 


96 



APPENDIX: COMPONENT AND MODULE SPECIEICATION EOR 
THE MODELING ORDNANCE MOVEMENTS INTO THE ASIAN 

PACIEIC THEATER 


A. INTRODUCTION 

This specification is a document of the development and implementation of the 
simulation modeling necessary to address the existing and future ordnance operations 
systems at the Kilo Wharf on Orote Peninsula, Guam. 

I. Document Organization 

This document describes the model components and process modules used to 
simulate ordnance operations conducted at Kilo Wharf, and the proposed operations upon 
completion of the military construction (MILCON) project to build a magazine on the 
Orote Peninsula. The description includes most of the detail necessary to develop an 
Arena simulation model of the operations. 

This specification is divided into two sections. The first section defines the 
purpose of the document and the software and hardware required to run the Arena model. 
The second section describes the components and process modules used to build the 
Arena model. 


a. Purpose of the Functional Specification 

The purpose of this document is to describe the components and process 
modules used to build the Arena model at the level of detail required for modeling 
purposes. This provides documentation for interested readers to follow when examining 
the model in Arena. 

2. Hardware and Software Requirements 

The thesis is developed in the Microsoft Windows operating system environment. 
The software and hardware required to run the model include (Kelton, Sadowski, & 
Sturrock, 2007): 


97 



• Arena Standard Edition 10.0 or higher 

• Microsoft Windows (latest version available) 

• At least 30MB hard disk space 

B. MODEL DESCRIPTION 

The following sections define the model timeline and provide a “parts list” of 
components and modules used to build the model. All other amplifying information 
about the model development or modeling approach can be directed to the author or to 
the NPS SEED Center, http://harvest.nps.edu/. 

I. Model Timeline 


The model is able to simulate ordnance operations of different run lengths for 
different purposes. The base unit of time used in Arena will be one day and the standard 


run length is one year. Eigure 56 is an overview shot of the model structure. 



Eigure 56. Model Structure Overview. 


98 























2 . 


Model Components 


This section describes all model components and modules used in this thesis. 
Each process tab in Arena and its related data modules are separately described to 
provide an easier method of following the descriptions. 

Basic Processes 


Create Modules 

Description 

OCS Arrives 

An OCS Ship is first created at a time defined by the expression ANINT (UNIF 
(9,13, v_Univ_Stream)). Following OCS Ships are created with an interarrival 
defined by v_OCS_Arr_Cycle. 

TAKE Arrives 

A TAKE Ship is first created at a time defined by the expression ANINT (UNIF 
(10, 21, v_Univ_Stream)). Following TAKE Ships are created with an interarrival 
defined by v_TAKE_Arr_Cycle. 

Competing Ship Arrives 

A Competing Ship is first created at a time defined by the expression ANINT 
(UNIF (1, 30, v_Univ_Stream)). Following OCS Ships are created with an 
interarrival defined by =-0.001 + EXPO (23.7, v_Univ_Stream) or 
v_CS_Arr_Rate. 

Create Extra Pallets 

The initial inventory of pallets, v_lnitial_lnventory, is created at time 0.001 to 
preload the system. This is a onetime event for the model. 

Output Out 

This one-time entity is created at simulation end time (tfin) and enables the 
model to write output to a designated output file. 



Dispose Modules 

Description 

Container Ship Disposal 
Module 

This module disposes OCS Ship entities when they complete their respective 
processes in the system. 

Competing Ship Disposal 
Module 

This module disposes Competing Ship entities when they complete their 
respective processes in the system. 

Error Dispose Module 

This module disposes entities that fail to designate properly. This module is 
included as a debugging function. 

TAKE Disposal Module 

This module disposes TAKE Ship entities when they complete their respective 
processes in the system. 

Dispose Air Force(AF) 
Containers Module 

This module disposes Containers entities designated for the Air Force when 
they complete their respective processes in the system. 

Pallet Disposal Module 

This module disposes Pallet entities when they complete their respective 
processes in the system. 

Output Dispose 

This module disposes Entity 1 entities when they complete their respective 
processes in the system. 



Process Modules 

Description 

CS Delay and Release Kilo 

This process performs a Delay Release action on CSs for a delay period of 
UNIF(4.01, 7,v_Univ_Stream) days. 

Crane Moves Container 
from Ship to Pier 

This process performs a Seize Delay action on cranes for a delay period of 

UNIF(0.00735,0.01225,v_Univ_Stream) hours. 

Ordnance Inspection at 
Ordnance Annex 

This process performs a Seize Delay Release action on Ordnance Inspectors at 
the Ordnance Annex for a delay period of 

UNIF(0.13333,0.16667,v_Univ_Stream) hours. 

Ordnance Inspection at 

This process performs a Seize Delay Release action on Ordnance Inspectors at 


99 









Kilo 

the Kilo Wharf for a delay period of UNIF(0.13333,0.16667,v_Univ_Stream) 
hours. 

Load Pallets to TAKE 

This process performs a Seize Delay Release action on Ordnance Inspectors at 
the Kilo Wharf for a delay period of UNIF(2,5,v_Univ_Stream) minutes. 

Block and Brace 

This process performs a Seize Delay Release action on Ordnance Inspectors at 
the Kilo Wharf for a delay period of UNIF(0.25,0.5,v_Univ_Stream) hours. 

Seize Pallet Loading 
ResourcesZ 

This process performs a Seize Delay Release action on Ordnance Inspectors at 
the Kilo Wharf for a delay period of TRIA( 0.5,10,15 
,v_Univ_Stream)minutes. 

Seize Spot at Kilo to 

Unload 

This process performs a Seize Delay Release action on Ordnance Inspectors at 
the Kilo Wharf for a delay period of TRIA( 0.25,0.5,1 ,v_Univ_Stream) hours. 



Decide Modules 


Type 

Description 

2-way by Condition 

This is a 2-way by Condition decision module defined by a test of the 
expression: If MR(Kilo Berth) - NR(Kilo Berth) > 0, then TRUE 

2-way by Condition 

This is a 2-way by Condition decision module defined by a test of the 
expression: If the variable, v_Containers_Off == a_Num_Containers, then 

TRUE 

2-way by Condition 

This is a 2-way by Condition decision module defined by a test of the 
expression: If a_Switch == 1, then TRUE. 

2-way by Condition 

This is a 2-way by Condition decision module defined by a test of the 
expression: If a_Switch == 1, then TRUE. 

2-way by Condition 

This is a 2-way by Condition decision module defined by a test of the 
expression: If a_Destination_ldentifier == 999, then TRUE. 

N-way by Condition 

This is a N-way by Condition decision module which decides the entity path by 
entity type: OCS Ship OR Competing Ship OR TAKE Ship. 

2-way by Condition 

This is a 2-way by Condition decision module defined by a test of the 
expression: For a_Pallets_Needed, ABS (v_TAKE_Appetite_Now - 
v_Pallets_Loaded) <= v_Min_Batch_Size, then TRUE. 



Batch Processes 

Description 

Pallet Load From Kilo To 

Annex 

This is a temporary batching of size (e_Batch_Size) using Pallet as the 
representative entity type. 

Batch Pallets for 

Movement to Kilo 

This is a temporary batching of size (e_Batch_Size) using Pallet as the 
representative entity type. 



Separate Modules 

Description 

Bust Container Ship Into 
Containers 

This module separates a duplicate OCS into a_Num_Containers -1 duplicate. 

Separate Ship from All Its 
Containers 

This module separates an OCS into an OCS and a single duplicate. 

Containers to Pallets 

OrdAnnex 

This module separates containers into pallets with a value determined by the 
expression, DISC(0.33, 9, .67,10,1.0, 11) 

Containers To Pallets 

This module separates containers into pallets with a value determined by the 
expression, DISC(0.33, 9, .67,10,1.0, 11) 

Separate From Truckload 
To Pallets 

This module splits an existing batch of pallets loaded to a truck back into 
pallets. The member attributes retain their original entity values. 

Separate From Truck To 
Pallet at Kilo 

This module splits an existing batch of pallets loaded to a truck back into 
pallets. The member attributes retain their original entity values. 


100 







Assign Modules 

Description 

Assign OCS Attributes 

This module assigns the following attributes: 

Assign TAKE Attributes 

This module assigns the following attributes: 

Assign Competing Ship 
Attributes 

This module assigns the following attributes: 

Assign Container Picture 
and Entity Type 

This module assigns the following attributes: 

Zero Out Count of 
Containers Off This 
Container Ship 

This module assigns the following attributes: 

Increment Count of 

Containers Off This 
Container Ship 

This module assigns the following attributes: 

Designator for Andersen 
AFB 

This module assigns the following attributes: 

Designator for Ordnance 
Annex 

This module assigns the following attributes: 

Change Containers to 
Pallets at Ordnance Annex 

This module assigns the following attributes: 

Change Container Entity 
Type And Attribute To 
Pallets 

This module assigns the following attributes: 

Count Loaded Pallets 

This module assigns the following attributes: 

Zero Out Count of Pallets 

Loaded 

This module assigns the following attributes: 

Copy TAKE Appetite to 
Global Variable 

This module assigns the following attributes: 

Assign to USN or USAF 

This module assigns the following attributes: 

Assign Unstuff Pierside 

This module assigns the following attributes: 



Record Modules 

Description 

Record Competing Ship at 
Kilo 

This module records the number of Competing Ships through Kilo Wharf using 
the variable, v_CS_Out + 1 

Error counter 

This module records the number of entity errors that are disposed of in the 
system. A debugging function. 

Record AF Containers 

This module records the number of AF Containers through Andersen AFB. 

Record TAKE Pallets 

This module records the number of TAKE Pallets out using the variable, 
v_Pallets_Out + 1. 

Record Number of TAKE 

thru Kilo 

This modules records the number of TAKE Ships through Kilo Wharf using the 
variable, v_TAKE_Out + 1 

Pallets Counted At Annex 

This module records the number of Pallets Counted At Annex. 

Record Pallet Count at Kilo 

This module records the number of Pallets Counted At Kilo Wharf. 

Record OCS thru Kilo 

This module records the number of OCSs through Kilo Wharf using the 
variable, v_OCS_Out +1. 

Record OCS In 

This module records the number of OCS In using the variable, v_OCS_ln + 1. 

Record TAKE In 

This module records the number of TAKE In using the variable, v_TAKE_ln + 1. 

Record CS In 

This module records the number of CS In using the variable, v_CS_ln + 1. 


101 





Basic Process Data Modules 


Component Name 

Type 

Description 

OCS Ship 

Entity 

An Ordnance Container Ship with attributes: 

TAKE Ship 

Entity 

An Auxiliary Dry Cargo/Ammunition Ship with attributes: 

Competing Ship 

Entity 

A Competing Ship with attributes: 

Container 

Entity 

Container entities are cloned from OCSs, not created. Therefore, 
they retain the attributes of their respective OCSs. 

Pallet 

Entity 

Pallet entities are cloned from Container, not created, with the 
exception of the initializing inventory. 

Entity 1 

Entity 

Entity 1 is created at the end of the simulation to initiate the Write 
Out function in the model. 

Kilo Berth 

Resource 

This is a single server resource, capacity (1). 

Buoy 702 

Resource 

This is a single server resource, capacity (1). 

Ordnance Annex 

Magazine Storage 

Resource 

The Ordnance Magazine is given a capacity of (99999999) pallets to 
indicate an essentially infinite capacity at the magazine. 

Crane 

Resource 

Ship board cranes of capacity (2). 

Pierside Staging Space 

Resource 

The Staging Space is where containers coming off the OCS are set 
down. Capacity (2). 

Container Truck Loading 
Space 

Resource 

The Container Truck Loading Space is where containers are loaded 
for transport. Capacity (2). 

Ordnance Inspector 

Resource 

The Ordnance Inspectors are used in the unstuffing process to 
inspect and inventory the ordnance. Capacity (18). 

Unstuffing Space 

Resource 

The Ordnance Unstuffing Space is where containers are unstuffed. 
Capacity (120). This only applies to the unstuffing process. This 
space has a higher capacity when only used for container storage. 

Block and Brace Crew 

Resource 

The Block and Brace Crew are used to secure palletized ordnance for 
transit on a Pallet Transport Truck. Capacity (10). 

Ordnance Forklifts 

Resource 

The Ordnance Forklifts are used to move ordnance to facilitate the 
ordnance operations process. Capacity (20). 

v_Containers_Off 

Variable 

Used to count containers offloaded from OCS 

v_Pallet_Count 

Variable 

Used to count pallets processed at Kilo Wharf. 

v_Pallets_Loaded 

Variable 

Used to count pallets loaded to T-AKE. 

v_TAKE_Appetite_Now 

Variable 

Used to translate the T-AKE pallet requirement from an attribute to 
a variable. 

v_Min_Batch_Size 

Variable 

Used to determine a minimum batch size for pallet loads. 

v_lnitial_lnventory 

Variable 

Used to establish the number of pallets created at the beginning of 
the model as an initializing inventory. 

v_OCS_Arr_Cycle 

Variable 

Used as an input variable for OCS arrival cycles. 

v_TAKE_Arr_Cycle 

Variable 

Used as an input variable for T-AKE arrival cycles. 

v_CS_Arr_Rate 

Variable 

Used as an input variable for CS arrival rate, lambda. 

v_Cont_per_OCS 

Variable 

Used as an input variable that determines how many containers will 
be offloaded from an OCS. 

v_percent_Navy_Cont 

Variable 

Used as an input variable that determines how many containers will 
be sent to the Air Force. 

v_percent_unstuffed_pier 

Variable 

Used as an input variable that determines the percent of containers 
unstuffed at the Kilo Wharf. 

v_OCS_Out 

Variable 

Used to count the number of OCS that exit the system. 

v_TAKE_Out 

Variable 

Used to count the number of T-AKE that exit the system. 

v_CS_Out 

Variable 

Used to count the number of CS that exit the system. 


102 







v_Pallets_Out 

Variable 

Used to count the number of pallets that exit the system. 

v_Univ_Stream 

Variable 

Used to establish the random number stream for a scenario 
replication. 

v_OCS_ln 

Variable 

Used to count the number of OCS that enter the system. 

v_TAKE_ln 

Variable 

Used to count the number of T-AKE that enter the system. 

v_CS_ln 

Variable 

Used to count the number of CS that enter the system. 


103 




Advanced Processes 


Delay Modules 

Description 

Delay for Mooring 

This module delays ship entities for e_Mooring_Time (a_Ship_Type) hours. 

Delay To Load Container 
Truck to Ord Annex 

This module delays Containers entities for UNIF (6,14, v_Univ_Stream) 
minutes. 

Delay To Load Container 
Truck to AFB 

This module delays Containers entities for UNIF (6,14, v_Univ_Stream) 
minutes. 

Delay to Unload Pallets At 
Annex 

This module delays Pallet entities for UNIF (14, 21, v_Univ_Stream) minutes. 

Delay to Unload Pallets at 
Kilo 

This module delays Pallet entities for UNIF (14, 21, v_Univ_Stream) minutes. 



Hold Modules 

Description 

Wait for Empty Container 
Ship Signal 

This module holds Container Ships until the signal, 777, is received indicating 
that the Container Ship is empty. 

Wait for TAKE to complete 
pallet load 

This module holds T-AKEs until the signal, 567, is received indicating that the 
T-AKE has received it's pallet request. The condition is defined by 
a_Pallets_Needed == a_Pallets_Loaded. 

Wait for Signal from TAKE 

This module holds Pallets until the signal, 123, is received indicating that the T- 
AKE is ready to load. 



ReadWrite Modules 

Description 

Write Out Stat 

This module directs the output from the model to a Microsoft Excel (*.xls) file, 
H:\Thesis 2009\Output.xls\Run Tracker2.xls. 



Release Modules 

Description 

Release Buoy 702 

This module releases Buoy 702. 

OCS Release Kilo Berth 

This module releases the berth at Kilo. 

Release Ships Crane 

This module releases the Ships Crane. 

Release Pierside Staging 
Space For Ord Annex 
Container 

This module releases the Pierside Staging Space For Ord Annex Container. 

Release Pierside Staging 
Space 

This module releases the Pierside Staging Space. 

TAKE Release Kilo 

This module releases the berth at Kilo. 

Release Ordnance Annex 
Magazine Space 

This module releases the Ordnance Annex Magazine Space. 

Release Unstuffing Space 

This module releases the Unstuffing Space. 

Release Container Loading 
Space Annex 

This module releases the Container Loading Space at the Annex. 

Release Container Loading 
Space Andersen 

This module releases the berth at Kilo. 



Seize Modules 

Description 

Seize Kilo 

This module seizes Kilo. 

Seize Buoy 702 

This module seizes Buoy 702. 

Seize Kilo from Buoy 

This module seizes Kilo from Buoy 702. 

Seize Container Loading 

This module seizes a Container Loading Spot on Kilo. 


104 








Spot on Kilo 


Seize Spot in Unstuffing 
Area 

This module seizes a spot in the Unstuffing Area. 

Seize Ordnance Annex 
Magazine Space 

This module seizes an Ordnance Annex Magazine Space. 



Signal Modules 

Description 

Signal that Container Ship 
is Empty 

This module sends the signal, 777, indicating that the Container Ship is empty. 

TAKE Signal To Pallets 

This module sends the signal, 123, indicating that the T-AKE is ready to load. 

Signal That Pallet Load 
Complete 

This module sends the signal, 567, indicating that the T-AKE loading is 
complete. 



Store Modules 

Description 

Store for Mooring Delay at 
Kilo 

This module stores entities during their mooring delay at Kilo. 

Store for Delay at Kilo 

This module stores Competing Ship at Kilo during the standard delay at Kilo. 



Unstore Modules 

Description 

Unstore from Mooring 
Delay at Kilo 

This module unstores entities after their mooring delay at Kilo. 

Unstore from Delay at Kilo 

This module unstores Competing Ships at Kilo after the standard delay at Kilo. 


Advanced Processes Data Modules 


Expressions 

Description 

e_702_to_Kilo_Time 

This expression, UNIF(40, 56, v_Univ_Stream), is a distribution that determines 
the time it takes to get from Buoy 702 to Kilo. 

e_Mooring_Time 

This expression, TRIA(4,5,6,v_Univ_Stream), is a distribution that determines 
the time it takes to moor. 

e_Batch_Size 

This expression, DISC(0.25,7, 0.50, 8,1.00,9,v_Univ_Stream), is a distribution 
that determines pallet batch size. 

e_Container_Forklift_Rate 

This expression, (UNIF(100,250,v_Univ_Stream)/.00026), is a distribution that 
determines the Container Forklifts velocity . 


105 








Advanced Transfer Processes 


Enter Modules 

Description 

Enter Container Loading 
Spot Andersen AFB 

Station Module 

This Enter Module establishes the system boundary for the Container Loading 
Spots used by Container Capable Forklifts. 

Enter Ordnance Annex 

Station Module 

This Enter Module establishes the system boundary for Ordnance Annex 

Station with an additional delay of TRIA (25, 30, 35) minutes for the Container 
Trucks. 

Enter Container Loading 
Spot Annex Station 

Module 

This Enter Module establishes the system boundary for Container Loading 

Spot Station used by Container Capable Forklifts. 

Enter Andersen AFB 

Station Module 

This Enter Module establishes the system boundary for with an additional 
delay of TRIA(50,60,75) minutes for Container Trucks. 

Enter Unstuffing Area 

Enter Station Module 

This Enter Module establishes the system boundary for the Unstuffing Area 
Station used by Container Capable Forklifts. 

Pallet Enter Ordnance 

Annex Station Module 

This Enter Module establishes the system boundary for Ordnance Annex 

Station used by Pallet Transport Trucks. 

Enter Pallet Truck Unload 

Station at Kilo Module 

This Enter Module establishes the system boundary for the Pallet Truck 

Unload Station at Kilo Station used by Pallet Transport Trucks. 



Leave Modules 

Description 

Leave Pallets to Kilo 

This Leave Module establishes an outward boundary for Pallets leaving the 
Ordnance Annex headed to Kilo. 

Leave Request Pallet 

Truck to Annex 

This Leave Module establishes an outward boundary for Pallets leaving Kilo to 
the Ordnance Annex. 



Route Modules 

Description 

Steam to Kilo Berth 

This Route Module establishes the routing boundary for entities that transit 
from sea to the berth at Kilo. The route time is determined by the expression, 
e_702_to_Kilo_Time(a_Ship_Type) in hours. 

Steam to Buoy 702 

This Route Module establishes the routing boundary for entities that transit 
from sea to the berth at Buoy 702. The route time is determined by the 
expression, e_702_to_Kilo_Time(a_Ship_Type) in hours. 

Steam from Buoy 702 to 
Kilo 

This Route Module establishes the routing boundary for entities that transit 
from Buoy 702 to the berth at Kilo. The route time is determined by the 
expression, TRIA(0.5,1,1.5) in hours. 



Station Modules 

Description 

OCS Arrival Station Name 

This module establishes the system boundary for OCS Arrivals. 

TAKE Arrival Station Name 

This module establishes the system boundary for TAKE Arrivals. 

Kilo Berth Station Name 

This module establishes the system boundary for Kilo Berth Arrivals. 

Buoy 702 Station Name 

This module establishes the system boundary for Buoy 702 Arrivals. 

Competing Ship Arrival 
Station Name 

This module establishes the system boundary for Competing Ship Arrivals. 

Container Truck Loading 
Station Module 

This module establishes the system boundary for Container Truck Loading 
Arrivals. 

Pierside Staging Station 
Module 

This module establishes the system boundary for Pierside Staging Arrivals. 

Pallet Truck To Annex 

This module establishes the system boundary for Pallet Truck To Annex 


106 






Station Module 

Arrivals. 

Pallet Truck To Kilo Station 

Module 

This module establishes the system boundary for Pallet Truck To Kilo Arrivals. 



Request Modules 

Description 

Request CCForklift to 
Unstuffing 

This module requests a Container Capable Forklift for use in Unstuffing. 

Request CCForklift to 
Transport 

This module requests a Container Capable Forklift for use in transporting 
containers. 

Request Container Truck 
Kilo to Annex 

This module requests a ContainerTruck movement from Kilo to the Ordnance 
Annex. 

Request Container Truck 
Kilo to Andersen 

This module requests a ContainerTruck movement from Kilo to the Andersen 
Air Force Base. 


107 




Advanced Transfer Data Modules 


Transporters 

Description 

Container Capable Forklift 

These are 2 Free Path Transporters with velocity 26400 feet per hour. 

Container Truck 

These are 8 Free Path Transporters with velocity 12 miles per hour. 

Pallet Transport Truck 

These are 12 Free Path Transporters with velocity 12 miles per hour. 



Distance Module 

Description 

Container Capable 

Forklift.Distance 

This distance set establishes the set of distances for all origins and destinations 
travelled by Container Capable Forklifts. 

Container Truck.Distance 

This distance set establishes the set of distances for all origins and destinations 
travelled by Container Trucks. 

Pallet Transport 

Truck.Distance 

This distance set establishes the set of distances for all origins and destinations 
travelled by Pallet Transport Trucks. 


108 




LIST OF REFERENCES 


Bellman, R.E. (1957). Dynamic programming. Princeton, NJ: Princeton University 
Press. 

Brown, G.G., & Carlyle, W.M. (2008). Optimizing the U.S. Navy’s combat logistics 
force. Naval Research Logistics, 55, 800-810. 

Cioppa, T.M. (2002, September). Efficient nearly orthogonal and space-filling 

experimental designs for high-dimensional complex models. PhD Dissertation. 
Naval Postgraduate School, Monterey, CA. 

General Dynamics/NASSCO. (2007, January). Lewis & Clark (T-AKE 1) Class Dry 
Cargo/Ammunition Ship EACT SHEET. 

Goode, B., & Smith, M.W. (2007, December). Resupplying T-AKEs at Guam during a 
military contingency. Center for Naval Analyses, CNA Report CRM 
D0017313.A1,2007. 

Headquarters, Department of the Army. (1997, October 27). Transportation reference 
data. Eield Manual 55-15. Washington, D.C. 

Helber, Hastert, & Lee (Planners). (2003, May). COMNAVMARIANAS ordnance 

functional plan. Prepared for Commander, Naval Eorces Marianas and Pacific 
Division, Naval Eacilities Engineering Command. 

Kelton, W.D., Sadowski, R.P., & Sturrock, D.T. (2007). Simulation with Arena (4* ed.) 
New York, NY: McGraw-Hill. 

Markle, S., & Wileman, S. (2001, July 2). Lewis & Clark Class (T-AKE) Milestone C 
CAIG Briefing. NAVSEA PMS325 SUPPORT SHIPS, BOATS, and CRAET 
PROGRAM OEEICE. 

Military Surface Deployment and Distribution Command (MSDDC). (2008, June). Guam 
ammunition distribution study - Transportation infrastructure capability 
assessment. Transportation Engineering Agency. 

Naval Message. (2007, Eebruary 201600Z). MSGID/EQSD EVENT WAIVER. EM 
NAVBASE GU TO COMPACELT PEARL HARBOR//N42//. 

NAVBASE GUAM DD Eorm 1392. (2005, August I). EY2008 military construction 
program. 

Rivera, P.W. (2008). NEW reports Excel spreadsheet-explosives safety Guam. Navy 
Munitions Command East Asia Division Headquarters. 


109 



Rockwell Automation Inc. (2009) Arena product overview. Retrieved 31 January 2009 
from http://www.arenasimulation.com/products/default.asp 

Rockwell Automation Inc. (2005a, October). Arena basic user’s guide. Publication 
ARENAB-UMOOIF-EN-P. 

Rockwell Automation Inc. (2005b). Arena online help. Arena version 10.00 [computer 
software]. 

Sanchez, S.M. (2005). NOEHdesigns spreadsheet. Retrieved on 16 February 2009 from 
http://diana.cs.nps.navy.mil/SeedEab/ 

SAS Institute Inc. (2007). JMP 7.0 Help. IMP Statistical Discovery Software version 7.0 
[computer software] Cary, NC. 

Vaughn, B. (2007, January). U.S. Strategic and Defense Relationships in the Asia-Pacific 
Region. Congressional research service report for members and committees of 
congress. Washington, D.C. 


no 



INITIAL DISTRIBUTION LIST 


1. Defense Technical Information Center 
Ft. Belvoir, VA 

2. Dudley Knox Library 
Naval Postgraduate School 
Monterey, CA 

3. CAPT Sean Geaney 
OPNAV N421 
Arlington, VA 

4. Frank Leban 

NSWC Carderock Divison 
Bethesda MD 

5. Professor Thomas W. Lucas 
Naval Postgraduate School 
Monterey, CA 

6. Professor David Kelton 
University of Cincinnati 
Cincinnati, OH 

7. Professor Keebom Kang 
Naval Postgraduate School 
Monterey, CA 

8. Professor Donald Gaver 
Naval Postgraduate School 
Monterey, CA 

9. Jeffrey Uejio 

Naval Facilities Command, Pacific 
Pearl Harbor, HI 

10. LCDR Cielo I. Almanza 

Navy Operational Logistics Support Command 
Pearl Harbor, HI 


III