Skip to main content

Full text of "Evaluating alternative network configurations and resource allocations for deployed Marine Corps aviation logistics units"

See other formats


Calhoun 

Iniiiiuiiortfl Arthivcof (he Navjl Pwigndualt School 


Calhoun: The NPS Institutional Archive 
DSpace Repository 



Theses and Dissertations 


1. Thesis and Dissertation Collection, all items 


2009-06 

Evaluating alternative network configurations 
and resource allocations for deployed Marine 
Corps aviation logistics units 


Jabin, Joshua M. 

Monterey, California: Naval Postgraduate School 


http://hdl.handle.net/10945/4780 


Downloaded from NPS Archive: Calhoun 



DUDLEY 

KNOX 

LIBRARY 


Calhoun is the Naval Postgraduate School's public access digitaI repository for 
research materials and institutional publications created by the NPS community. 
Calhoun is named for Professor of Mathematics Guy K. Calhoun, NPS's first 
appointed —and published —scholarly author. 

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


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







NAVAL 

POSTGRADUATE 

SCHOOL 

MONTEREY, CALIFORNIA 


THESIS 


EVALUATING ALTERNATIVE NETWORK 
CONFIGURATIONS AND RESOURCE ALLOCATIONS 
FOR DEPLOYED MARINE CORPS 
AVIATION LOGISTICS UNITS 

by 

Joshua M. Jabin 
June 2009 

Thesis Advisor: Moshe Kress 

Second Reader: David Alderson 


Approved for public release; distribution is unlimited 




THIS PAGE INTENTIONALLY LEFT BLANK 



REPORT DOCUMENTATION PAGE 


Form Approved OMB No. 0704-0188 

Public reporting burden for this collection of information is estimated to average 1 hour per response, including the time for reviewing instruction, 
searching existing data sources, gathering and maintaining the data needed, and completing and reviewing the collection of information. Send 
comments regarding this burden estimate or any other aspect of this collection of information, including suggestions for reducing this burden, to 
Washington headquarters Services, Directorate for Information Operations and Reports, 1215 Jefferson Davis Highway, Suite 1204, Arlington, VA 
22202-4302, and to the Office of Management and Budget, Paperwork Reduction Project (0704-0188) Washington DC 20503. _ 

2. REPORT DATE 

June 2009 


6. AUTHOR(S) Joshua M. Jabin 


11. SUPPLEMENTARY NOTES The views expressed in this thesis are those of the author and do not reflect the official policy 
or position of the Department of Defense or the U.S. Government. 


13. ABSTRACT (maximum 200 words) 

This thesis develops a model and performs analysis to estimate the operational effectiveness of the Marine Aviation 
Logistics Support Program II (MALSPII) under different system configurations and resource allocation policies. 
MALSP II is designed to protect the aviation logistics system from uncertain, possibly high variance, demand that 
could have a significant detrimental impact on the material readiness of deployed aircraft. Although an MALSP II 
pilot program has produced positive results since 2005, the overall design of the logistical support network has not yet 
been evaluated. We develop an inter-temporal network simulation model that measures the operational effectiveness 
of the network—with and without an additional level of supply called an Enroute Support Base—using four inventory 
buffer sizing policies. We use two measures of effectiveness (MOE): PackUp Effectiveness and PartShort. Packup 
Effectiveness is the current metric used by the Marine Corps to evaluate aviation logistics performance in a deployed 
setting. It represents the percentage of demands satisfied on the day demanded. PartShort, which is a new MOE 
proposed in this thesis, represents the magnitude and duration of unsatisfied demands during a certain finite time 
horizon. For different levels of acceptable risk, we provide recommendations for network configurations and 
inventory buffer levels. These results can help operational planners improve the efficiency of available resources and 
maximize the effectiveness of logistical support to deployed bases._ 


16. PRICE CODE 


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

Prescribed by ANSI Std. 239-18 


20. LIMITATION OF 
ABSTRACT 


15. NUMBER OF 
PAGES 

85 


14. SUBJECT TERMS MALSP II, inter-temporal network simulation model, simulation model. 
Marine Corps aviation logistics, deployed aviation logistics 


18. SECURITY 
CLASSIFICATION OF THIS 
PAGE 

Unclassified 


19. SECURITY 
CLASSIFICATION OF 
ABSTRACT 

Unclassified 


17. SECURITY 
CLASSIFICATION OF 
REPORT 

Unclassified 


12b. DISTRIBUTION CODE 


12a. DISTRIBUTION / AVAILABILITY STATEMENT 

Approved for public release; distribution is unlimited 


7. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES) 

Naval Postgraduate School 
Monterey, CA 93943-5000 

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

Naval Air Systems Command 
Norfolk, Virginia 


5. FUNDING NUMBERS 


8. PERFORMING ORGANIZATION 
REPORT NUMBER 


10. SPONSORING/MONITORING 
AGENCY REPORT NUMBER 


4. TITLE AND SUBTITLE Evaluating Alternative Network Configurations and 
Resource Allocations for Deployed Marine Corps Aviation Logistics Units 


3. REPORT TYPE AND DATES COVERED 

Master’s Thesis 


1. AGENCY USE ONLY (Leave blank) 


1 




























THIS PAGE INTENTIONALLY LEFT BLANK 


11 



Approved for public release; distribution is unlimited 


EVALUATING ALTERNATIVE NETWORK CONFIGURATIONS 
AND RESOURCE ALLOCATIONS FOR DEPLOYED MARINE CORPS 
AVIATION LOGISTICS UNITS 


Joshua M. Jabin 

Captain, United States Marine Corps 
B.S., United States Naval Academy, 2001 


Submitted in partial fulfillment of the 
requirements for the degree of 


MASTER OF SCIENCE IN OPERATIONS RESEARCH 

from the 


NAVAL POSTGRADUATE SCHOOL 
June 2009 


Author: 


Joshua M. Jabin 


Approved by: Moshe Kress 

Advisor 


David Alderson 
Second Reader 


Robert Dell 

Chairman, Department of Operations and Information Sciences 
iii 



THIS PAGE INTENTIONALLY LEFT BLANK 


IV 



ABSTRACT 


This thesis develops a model and performs analysis to estimate the operational 
effectiveness of the Marine Aviation Logistics Support Program II (MALSP II) under 
different system configurations and resource allocation policies. MALSP II is designed 
to protect the aviation logistics system from uncertain, possibly high variance, demand 
that could have a significant detrimental impact on the material readiness of deployed 
aircraft. Although an MALSP II pilot program has produced positive results since 2005, 
the overall design of the logistical support network has not yet been evaluated. We 
develop an inter-temporal network simulation model that measures the operational 
effectiveness of the network—with and without an additional level of supply called an 
Enroute Support Base—using four inventory buffer sizing policies. We use two 
measures of effectiveness (MOE): PackUp Effectiveness and PartShort. Packup 
Effectiveness is the current metric used by the Marine Corps to evaluate aviation logistics 
performance in a deployed setting. It represents the percentage of demands satisfied on 
the day demanded. PartShort, which is a new MOE proposed in this thesis, represents the 
magnitude and duration of unsatisfied demands during a certain finite time horizon. For 
different levels of acceptable risk, we provide recommendations for network 
configurations and inventory buffer levels. These results can help operational planners 
improve the efficiency of available resources and maximize the effectiveness of logistical 
support to deployed bases. 


v 



THIS PAGE INTENTIONALLY LEFT BLANK 


vi 



TABLE OF CONTENTS 


I. INTRODUCTION.1 

A. THESIS PURPOSE: EVALUATE MALSP II SUPPORT NETWORK....1 

B. MALSP USED FOR THE PAST 20 YEARS.1 

C. MALSP II WILL REPLACE MALSP BUT HAS NOT BEEN FULLY 

EVALUATED YET.4 

D. THESIS SCOPE: BUILD A MODEL THAT USES DEMAND 
PATTERN TO DESIGN AND EVALUATE SUPPORT NETWORK 8 

II. LITERATURE REVIEW.11 

A. BACKGROUND ON LEGACY MALSP.11 

B. MALSP II IS DESIGNED SPECIFICALLY FOR DEPLOYED 

OPERATIONAL LOGISTICS ENVIRONMENT.11 

III. ITN SIMULATION MODEL.15 

A. MODEL DEVELOPMENT.15 

B. RULES USED IN THE ITN SIMULATION MODEL.18 

C. MODEL FORMULATION.19 

1. Indices and Sets.19 

2. Data.19 

3. State Variables.20 

4. Algorithm.20 

5. Measures of Effectiveness.24 

D. EXAMPLES OF MEASURES OF EFFECTIVENESS.24 

E. EXAMPLE OUTPUT.29 

IV. RESULTS AND ANALYSIS.31 

A. SUMMARY OF RESULTS.31 

B. THE IMPACT OF SELECTING A RISK LEVEL.39 

C. NETWORK DESIGN USING NO-RISK BUFFER SIZING.46 

D. NETWORK DESIGN USING BUFFER SIZING WITH RISK.51 

V. CONCLUSIONS.61 

A. THESIS OBJECTIVES.61 

B. EVALUATING ALTERNATIVE NETWORK CONFIGURATIONS 

AND ALLOCATING SPARE PARTS.61 

C. FUTURE EXTENSIONS.62 

LIST OF REFERENCES.63 

INITIAL DISTRIBUTION LIST.65 


vii 
































THIS PAGE INTENTIONALLY LEFT BLANK 



LIST OF FIGURES 


Figure 1: Intermediate Maintenance Activity operating in expeditionary mode in 

Middle East (From ASL 2004).2 

Figure 2: Overview and key questions about the MAFSP II logistics support 

network.5 

Figure 3: MAFSP II Visual Network.16 

Figure 4: Overall average ranks based on PartShort using the four network 

configurations and four inventory buffer sizing risk levels.33 

Figure 5: Overall average ranks based on PackUp Effectiveness using the four 

network configurations and four inventory buffer sizing risk levels.33 

Figure 6: Frequency of ranks using no-risk inventory buffer sizes based on PartShort..35 

Figure 7: Frequency of ranks using low-risk inventory buffer sizes based on 

PartShort.36 

Figure 8: Frequency of ranks using medium-risk inventory buffer sizes based on 

PartShort.36 

Figure 9: Frequency of ranks using high-risk inventory buffer sizes based on 

PartShort.37 

Figure 10: Frequency of ranks using no-risk inventory buffer sizes based on PackUp 

Effectiveness.37 

Figure 11: Frequency of ranks using low-risk inventory buffer sizes based on PackUp 

Effectiveness.38 

Figure 12: Frequency of ranks using medium-risk inventory buffer sizes based on 

PackUp Effectiveness.38 

Figure 13: Frequency of ranks using high-risk inventory buffer sizes based on 

PackUp Effectiveness.39 

Figure 14: Overall average ranks for no-risk inventory buffers sizes based on network 

configuration.48 

Figure 15: Average ranks for no-risk inventory buffers based on high-demand parts.49 

Figure 16: Average ranks for no-risk inventory buffers based on low-demand parts.50 

Figure 17: Average ranks for each network configuration based on risk (low, 

medium, high).52 

Figure 18: Average ranks based on PackUp Effectiveness for each configuration 

using risk level (low, medium, high).54 

Figure 19: Average ranks based on PartShort using risk level (low, medium, high) 

based for high-demand parts.55 

Figure 20: Average ranks based on PartShort using risk level (low, medium, high) 

based for low-demand part.55 

Figure 21: Average ranks based on PackUp Effectiveness using risk level (low, 

medium, high) for high-demand parts.56 

Figure 22: Average ranks based on PackUp Effectivness using risk level (low, 

medium, high) for low-demand parts.57 


IX 
























THIS PAGE INTENTIONALLY LEFT BLANK 


x 



LIST OF TABLES 


Table 1: FOB Cumulative Distribution Function for TRR=1 day.8 

Table 2: TRR between nodes for each network configuration.17 

Table 3: NUN: 014290072; ESBO: FOB TRR=3, MOB TRR=10.29 

Table 4: No-risk inventory buffers for gearbox assembly representing low-demand 

part.41 

Table 5: Low-risk inventory buffers for gearbox assembly representing low- 

demand part.42 

Table 6: Medium-risk inventory buffers for gearbox assembly representing low- 

demand part.42 

Table 7: High-risk inventory buffers for gearbox assembly representing low- 

demand part.42 

Table 8: Alternative inventory buffers for gearbox assembly representing low- 

demand part.43 

Table 9: Alternative inventory buffers for gearbox assembly representing low- 

demand part.43 

Table 10: No-risk inventory buffers for gyroscope representing high-demand part.44 

Table 11: Low-risk inventory buffers for gyroscope representing high-demand part ....44 

Table 12: Medium-risk inventory buffers for gyroscope representing high-demand 

part.45 


Table 13: High-risk inventory buffers for gyroscope representing high-demand part ....45 

Table 14: Alternative inventory buffers for gyroscope representing high-demand part..46 

Table 15: Alternative inventory buffers for gyroscope representing high-demand part..46 

Table 16: Overall average ranks for each configuration. 20 spare parts were analyzed 

using results from the ITN model for no-risk inventory buffers. ESBO is 
the most efficient configuration, followed by ESB1 then ESB3 then ESB5. ..48 
Table 17: Average ranks of each configuration for parts with high demand. 20 spare 

parts were analyzed using results from the ITN model for no-risk 
inventory buffers. ESBO is the most efficient configuration, followed by 


ESB1 then ESB3 then ESB5.49 

Table 18: Average ranks of each configuration for parts with low demand. 20 spare 

parts were analyzed using results from the ITN model for no-risk 
inventory buffers. ESBO is the most efficient configuration, followed by 

ESB1 then ESB3 then ESB5.50 

Table 19: No-risk inventory buffer for gyroscope representing high-demand part.51 

Table 20: No-risk inventory buffer for gearbox assembly representing low-demand 

part.51 

Table 21: Overall average ranks based on PartShort for low-, medium- and high-risk 

inventory buffers.53 

Table 22: Overall average ranks based on PackUp Effectiveness for low-, medium- 

and high-risk inventory buffers.54 

Table 23: Average ranks, isolated for high- and low-demand parts, based on 

PartShort using low-, medium- and high-risk inventory buffers.56 


xi 
























Table 24: Average ranks, isolated for high- and low-demand parts, based on PackUp 

Effectiveness using low-, medium- and high-risk inventory buffers.57 

Table 25: Low-risk buffer sizing for gyroscope representing high-demand part.58 

Table 26: Medium-risk buffer sizing for gyroscope representing high-demand part.58 

Table 27: High-risk buffer sizing for gyroscope representing high-demand part.58 

Table 28: Low-risk buffer sizing for gearbox assembly representing low-demand 

part.59 

Table 29: Medium-risk buffer sizing for gearbox assembly representing low-demand 

part.59 

Table 30: High-risk buffer sizing for gearbox assembly representing low-demand 

part.59 


xii 











EXECUTIVE SUMMARY 


Marine Corps aviation logistics is in the process of transforming the model used 
to support deployed operations. For 20 years, the Marine Corps has used the Marine 
Aviation Logistics Support Program (MALSP)—a planning tool that dictates the 
necessary people, equipment, spare parts, and mobile containers to deploy in support of 
contingency operations. MALSP is a “push” system based on periodic fixed-size batches 
of supply. Due to increased operational requirements in Operation Iraqi Freedom and 
Operation Enduring Freedom, the Marine Corps has initiated the development of MALSP 
II, which is designed to protect the logistical system from uncertain, possibly high 
variance, demand that could have a significant detrimental impact on the material 
readiness of deployed aircraft. Since 2005, a MALSP II pilot program has produced 
positive results, but the overall design of the MALSP II logistical support network has 
not yet been evaluated. 

MALSP II is a pull system where supplies are ordered, as needed, following 
consumption. Inventory buffers at the various nodes of the logistics network are 
determined by an information tool called the Enterprise Logistics Analysis Tool (ELAT). 
ELAT gives users four options—no-risk, low-risk, medium-risk and high-risk—to 
determine inventory buffers that correspond to the 100 th , 97 th , 94.7 th and 87 th percentile, 
respectively, of observed demand distribution during the logistical lead time—called time 
to reliably replenish (TRR)—specific to the operation supported. In this thesis we 
evaluate four logistics network configurations and attempt to answer the following 
questions: (1) at what distance—if at all—does an additional level of supply called an 
Enroute Support Base (ESB) improve network efficiency and (2) what impact do the four 
inventory buffer sizing risk levels used by ELAT have on the estimated measures of 
effectiveness. To answer these questions, we develop an inter-temporal network (ITN) 
simulation model that measures the operational effectiveness of the network—with and 
without the ESB—using the four inventory buffer sizing methods used in ELAT. 

In the analysis we use two measures of effectiveness (MOE): PackUp 

Effectiveness and PartShort. Packup Effectiveness is the current metric used by the 

xiii 



Marine Corps to evaluate aviation logistics performance in a deployed setting. It 
represents the percentage of demands satisfied on the day demanded. PartShort, which is 
a new MOE proposed in this thesis, represents the magnitude and duration of unsatisfied 
demands during a certain finite time horizon. 

The model compares four network configurations. The first configuration, called 
ESBO, represents the network without an ESB. The other three configurations, labeled 
ESB5, ESB3 and ESB1, assume the existence of an ESB and they differ in terms of 
distance, measured by TRR, from the Parent Marine Aviation Logistics Squadron 
(PMALS) and the Main Operating Base (MOB): ESB5 is 5 days from the MOBs and 6 
days from the PMALS, ESB3 is 3 and 8 days, respectively, and ESB1 is 1 day from the 
MOBs and 10 days from the PMALS. Each configuration is modeled using each one of 
the four risk levels—no-risk, low-risk, medium-risk, high-risk—used in ELAT to 
determine inventory buffers. Each instance of the simulation runs for 360 days and is 
replicated 20 times. The model produces expected operational effectiveness measured by 
the two aforementioned MOEs for the varying levels of input determined by the selected 
risk level. Together, these parameters—inputs and MOEs—are used for determining 
efficiency. 

To compare the alternative network configurations, for each risk level, the total 
number of spare parts in the various node inventory buffers is fixed for all four 
configurations. This constitutes a common denominator for comparison. First, the four 
inventory buffer sizes—corresponding to the four risk levels—are used as a basis to 
determine the total number of spare parts dedicated to the system. Next, parts are re¬ 
allocated to the various nodes in an effective way such that the selected MOEs are 
optimized or nearly optimized. 

The appropriate network configuration depends on the selected risk level, which 
is affected by desirable network properties. If the objective is a high level of attainability 
at each node—measured by 100% PackUp Effectiveness and 0 PartShort—then no-risk 
inventory buffer sizing is required. No-risk inventory buffers will have a significantly 
greater range—the number of line items of spare parts in the logistical support package— 
and depth—the quantity of each spare part in the logistical support package—than other 



risk level inventory buffers. Using this risk level, the most efficient network design omits 
the proposed ESB. If operational flexibility, defined as a system’s ability to quickly and 
effectively respond to changes, or survivability is a high priority, planners should be 
willing to accept less than maximum attainability to reduce the logistical support 
footprint at each node. In this situation, low-, medium- or high-risk level buffer sizing is 
appropriate. 

PackUp Effectiveness and PartShort may not improve in tandem as a result of 
changing the design of the support network. A system that increases the percentage of 
demands satisfied on the day demanded does not necessarily improve the response time 
for unsatisfied demand. These two MOEs may require different network configurations 
to perform most effectively. 

As a result of our modeling and analysis, we come to the following conclusions: 

(1) For no-risk inventory buffer sizes, ESBO dominates all other alternative 
network configurations. 

(2) For all non-zero risk levels, if the objective is to maximize PackUp 
Effectiveness, ESBO is still the dominating alternative. 

(3) For all non-zero risk levels, if the objective is to minimize PartShort, ESB 1 is 
the dominating alternative. Also, any network with an ESB less than five days 
TRR from the demand nodes will outperform ESBO. 

(4) If the ESB is included, performance improves as the TRR between the ESB 
and the demand nodes is minimized. 


xv 



THIS PAGE INTENTIONALLY LEFT BLANK 


xvi 



LIST OF ACRONYMS AND ABBREVIATIONS 


ACE 

Air Combat Element 

AOR 

Area of Responsibility 

CCSP 

Common Contingency Support Package 

ELAT 

Enterprise Logistics Analysis Tool 

ESB 

Enroute Support Base 

FISP 

Fly-in Support Package 

FOB 

Forward Operating Base 

FOSP 

Follow-on Support Package 

FY 

Fiscal Year 

IMA 

Intermediate Maintenance Activity 

MAGTF 

Marine Air Ground Task Force 

MALS 

Marine Aviation Logistics Squadron 

MALSP 

Marine Aviation Logistics Support Program 

MOB 

Main Operating Base 

NUN 

National Item Identification Number 

OEF-Afghanistan 

Operation Enduring Freedom-Afghanistan 

OEF-HOA 

Operation Enduring Freedom-Horn of Africa 

OIF 

Operation Iraqi Freedom 

OPLOG 

Operational Logistics 

PCSP 

Peculiar Contingency Support Package 

PMALS 

Parent Marine Aviation Logistics Squadron 

SAMMS 

Stand-alone Material Management System 

T-AVB 

Marine Corps Aviation Logistics Ship 

TRR 

Time to Reliably Replenish 


XVII 



THIS PAGE INTENTIONALLY LEFT BLANK 



I. INTRODUCTION 

A. THESIS PURPOSE: EVALUATE MALSP II SUPPORT NETWORK 

Due to increased operational requirements in Operation Iraqi Freedom and 
Operation Enduring Freedom, Marine Corps aviation logistics planners must find 
innovative ways to improve efficiency of resources. The Marine Aviation Logistics 
Support Program (MALSP) II is designed specifically to protect the logistical system 
from uncertain, possibly high variance, demand that could have a significant detrimental 
impact on the material readiness of deployed aircraft. MALSP II is a pull-based 
replenishment system that supports packages of spare aircraft parts—called inventory 
buffers —that are determined using demand patterns specific to the operation supported 
during the time to reliably replenish (TRR). TRR is determined by the elapsed time from 
when a spare part is issued from an inventory buffer until the part is replaced in the 
inventory buffer. The MALSP II system includes an additional level of supply, which is 
designated to improve the responsiveness of the logistic system. Since 2005, a MALSP 
II pilot program has produced positive results; however, the overall design of the 
logistical support has not yet been evaluated. Using MALSP II conditions, this thesis 
develops an inter-temporal simulation model that measures the operational effectiveness 
of four network configurations using four inventory buffer sizing methods. The results of 
this model and subsequent analysis will assist operational planners select the system 
configurations and policies that improve the efficiency of available resources and 
maximize the effectiveness of logistical support to deployed bases. 

B. MALSP USED FOR THE PAST 20 YEARS 

Marine Corps aviation logistics is in the process of transforming the model used 
to support deployed operations. For 20 years, the Marine Corps has used MALSP—a 
model that dictates the necessary people, equipment, spare parts, and mobile containers to 
deploy in support of contingency operations. Based on monthly averages, MALSP 
separates 30 days of supply into mobile containers that deploy with the organizational 
level flying squadron. During those 30 days, the Marine Aviation Logistics Squadron 


1 



(MALS) deploys on a support ship (T-AVB) and brings an additional 60 days of 
logistical support. The end result is an entire Intermediate Maintenance Activity (IMA) 
with 90 days of supply and repair capability in theatre with a long replenishment chain. 


v 



Figure 1: Intermediate Maintenance Activity operating in expeditionary 

mode in Middle East (From ASL 2004) 


MALSP measures effectiveness using PackUp Effectiveness. This metric 
calculates the number of demands satisfied at the time demanded as a percentage of 
overall demand. It does not measure the response time of unsatisfied demands. 

Aviation Logistics is comprised of three maintenance levels defined in the Naval 
Aviation Maintenance Program OPNAVINST 4790.2. The first level is the 
organizational level. This level performs preventative and planned maintenance on 
aircraft at regular intervals based on flight hours. Further, the organizational level 
inspects planes for unplanned maintenance problems. If a problem is identified, the 
organizational level maintainers remove repairable components and order replacements 
from the intermediate level. When the repairable component is removed from the 

2 







aircraft, it creates a “hole” in the aircraft that makes it non-mission or partial-mission 
capable until that “hole” is filled. The intermediate level is organized as a MALS 
comprised of both supply, responsible for maintaining a warehouse of replacement spare 
parts, and an IMA, responsible for repairing aircraft components. The MALS is 
responsible for replacing repairable components either by issuing a replacement part or 
fixing the broken part. If the MALS is unable to repair components, they are sent to the 
third level—the depot level—for “overhaul” maintenance or disposal. 

Prior to MALSP, the Marine Corps did not have a standardized means of 
organizing and deploying aviation logistics. Developed in 1989 during the Cold War, 
MALSP was designed to enable the Marine Air Ground Task Force (MAGTF) Air 
Combat Element (ACE) to deploy with sufficient support while the MALS transported 
necessary resources to set up a forward logistics base (Delaporte 2007, 4). It is a concept 
based on deploying in layers and sustaining operations for a prolonged time. Using 
multiple building blocks, MALSP enabled peace-time aviation logistics units to quickly 
organize and deploy in support of contingency operations. The mission of MALSP is to 
identify and integrate necessary spare parts, support equipment, and people to support all 
aircraft types that could comprise a MAGTL ACE. 

MALSP is a push system consisting of four types of standardized support 
packages intended to support any contingency operation. The Lly-in Support Package 
(LISP) is designed to support the fly-in echelon aircraft—those that deploy in the initial 
phase—of the MAGTL ACE for the first 30 days. The LISP is a standalone package 
intended to sustain the ACE while the intermediate maintenance capability is transported 
by the aviation logistics support ships. The Common Contingency Support Package 
(CCSP), Peculiar Contingency Support Package (PCSP), and Lollow-on Support Package 
(LOSP) are designed to provide an additional 60 days of spare parts, mobile facilities 
with test benches, intermediate maintenance repair capability and support equipment to 
sustain operations (Delaporte 2007, 4). The CCSP contains aircraft components used on 
multiple types of aircraft while the PCSP is intended for a single type of aircraft. The 
LOSP contains support not included in either the CCSP or PCSP but necessary for 
deployments of longer duration. These packages are pre-determined without knowledge 

3 



of the environmental conditions in the theatre, operational tempo or other tactical or 
operational factors. Inventories of spare parts are determined using past average monthly 
demand and subject matter expert opinion. Unplanned requirements and re-supply rely 
on a supply chain with high variation and are dependent on available transportation. 

Support packages are designed for each aircraft platform and fit together like 
building blocks. In garrison, Marine Aircraft Groups are organized by peculiar aircraft; 
however, contingency operations require composite squadrons with multiple types of 
aircraft. For example, a Marine Expeditionary Force ACE comprised of CH-46, CH-53, 
AH-1W and UH-1N helicopters deploys with one CCSP, three FISPs and three PCSPs 
totaling 95 pallets that weigh over 4800 tons (Delaporte 2007, 5). MALSP results in a 
large, immobile footprint at the deployed sites to sustain operations. 

C. MALSP II WILL REPLACE MALSP BUT HAS NOT BEEN FULLY 

EVALUATED YET 

MALSP II is designed to protect the logistical support system from uncertain, 
possibly high variance, demand for spare parts. Further, MALSP II aims to reduce the 
maintenance repair equipment, mobile facilities and people at the forward operating bases 
required by legacy MALSP. MALSP II changes business rules and designs the logistical 
support network to buffer the system from variation and improve the efficiency of using 
resources. It develops support packages that can be quickly organized to meet specific 
demand patterns of varying configurations of aircraft for unknown durations. 


4 



At what distance—if at 
all—does an additional 
level of supply (ESB) 
improve network 
efficiency? 




? 


MOEs: 

1) PackUp Effectiveness 

2) PartShort (?) 



~ * 

> MOB 


-► 

TRR=3 


FOB 


Parent Marine Aviation Logistics Squadron (PMALS) 
Enroute Support Base (ESB) 

Main Operating Base (MOB) 

Forward Operating Base (FOB) 

Time to Reliably Replenish (TRR) 


What impact do the 
starting buffers have on 
effectiveness? 


Figure 2: Overview and key questions about the MALSP II logistics support 

network 

MALSP II currently uses the same measure of effectiveness —PackUp 
Effectiveness —as legacy MALSP. Since the goal of MALSP II is to be flexible and 
responsive, this thesis suggests a new measure of effectiveness— PartShort —that counts 
the number of unsatisfied daily demands to assess overall responsiveness of the system 
(ASL 2004). 

MALSP II dictates a change in business rules. Departing from MALSP’s reliance 
on monthly averages, MALSP II determines the quantity of spare parts—called an 
inventory buffer —allocated to each node using the demand pattern within the re-supply 

5 


































lead-time called Time to Reliably Replenish (TRR). TRR begins when a spare part is 
issued for consumption at a maintenance facility, henceforth called demand nodes, and 
ends when the part is replaced in the inventory buffer at the demand node. Since a shorter 
TRR reduces the time periods in which demand may occur, one objective of MALSP II is 
to minimize TRR. Legacy MALSP defines a re-order objective, the maximum inventory 
at a site, and a re-order point that signals when a re-order is placed. The re-order point is 
not necessarily reached immediately after each demand. The Marine Corps Aviation 
Supply Desktop Procedures (MCO P4400.177E) do not explicitly define the frequency 
the parent node re-supplies the child node. MALSP II requires immediate requests for re¬ 
supply from the demand node and daily replenishments from the supply node (Garant 
2006, 12). Changing behavior is a necessary part of minimizing TRR, but it also requires 
a synchronized support network. 

MALSP II creates four logistic levels to buffer the system from variation in 
demands inherent in deployed military operations. The P-level, comprised of the Parent 
Marine Aviation Logistics Squadron (PMALS), is the highest level. It contains spare 
parts repair capability and receives spare parts directly from wholesale supply, depots, 
and the original equipment manufacturers. The next level of supply is the E-level 
comprised of the Enroute Support Base (ESB). The ESB manages an inventory buffer 
for the forward deployed nodes, but does not have local demand. The ESB contains 
limited or no spare parts repair capability. The next level of supply is the M-level 
comprised of the Main Operating Base (MOB). The MOB is located in the theater of 
operations, known as the Area of Operations (AOR), with the organizational level flying 
squadrons. It contains limited or no spare parts repair capability. It contains an inventory 
buffer both to satisfy local demand and to re-supply the next level of supply, the F-level 
which is comprised of the Forward Operating Base (FOB). The FOB is also located in 
the theater of operations with a detachment of planes from the organizational flying 
squadron. For example, the FOB at Operation Enduring Freedom-Horn of Africa (OEF- 
HOA) supported a 4 plane detachment of CH-53 helicopters during FY-08. The FOB has 
an inventory buffer to satisfy local demand only. The four levels of nodes are based on a 


6 



hierarchical structure with increased inventory of spare parts and repair capability at the 
higher echelons and smaller footprints at the lower echelons (Steward 2008, 41). 

MALSP II builds the support network to improve the responsiveness of the 
system and reduce the footprint at the forward demand nodes. An additional level of 
supply—the ESB—is added to the network to reduce the long supply chain between the 
supply nodes and the demand nodes. The ESB is intended to reduce TRR to the M-level 
thereby reducing the inventory buffer at the demand nodes for planned requirements. 
Further, the ESB adds an additional supply buffer to the network designed to increase 
responsiveness for unplanned requirements (Steward 2008, 42). Though historical 
demand data is useful for planning material requirements, it does not perfectly predict 
future requirements. By adding this additional inventory buffer, the downstream nodes 
may have an extra layer of protection from prolonged supply shortages. 

MALSP II is a pull system that uses demand patterns to allocate spare parts rather 
than averages used in traditional push systems such as MALSP. While averages may be 
useful in steady state low-variance environments, combat operations are characterized by 
higher uncertainty due to high operational and environmental variability. The following 
describes the method used by the Enterprise Logistics Analysis Tool (ELAT), developed 
by Colonel Laurin Eck, USMC (Ret.), to determine inventory buffer sizes (Eck, 2009). 
The first step is to estimate the probability mass functions (PMF) and cumulative 
distribution functions (CDF) of demand during various lengths of TRR. The estimates 
are based on observed in-context data from the various theatres of operation. An 
example of PMF and CDF is presented in Table 1. From the CDF, four initial levels of 
spare parts are determined that corresponds to levels of risk accepted by the commander. 
If a commander is not willing to accept any risk, the maximum observed demand (100 th 
percentile) during the TRR is used as an initial inventory buffer. In Table 1, the no risk 
inventory buffer is 8 spare parts. The low risk inventory buffer is calculated using the 
97 th percentile of total demands during TRR. In Table 1, the low risk inventory buffer is 
6 spare parts. The medium risk inventory buffer is calculated using the 94.7 th percentile 
and the high risk inventory buffer is calculated using the 87 th percentile. The medium 


7 



risk and high risk inventory buffers in Table 1 are 5 and 3, respectively. The demand 
pattern for initial allowances for spare parts depends on specific contingency conditions. 


Demand 

PMF 

CDF 

0 

0.589 

0.000 

1 

0.106 

0.589 

2 

0.172 

0.694 

3 

0.006 

0.867 

4 

0.067 

0.872 

5 

0.011 

0.939 

6 

0.028 

0.950 

7 

0.017 

0.978 

8 

0.006 

0.994 



1.000 


Table 1: FOB Cumulative Distribution Function for TRR=1 day 


MALSP II designs the network to improve the efficiency of available resources 
such as transportation, repair capability, spare parts, and people. Using the MALSP II 
methodology, some nodes will require less spare parts than legacy MALSP. However, 
spare parts comprise a relatively small proportion of the overall support package and 
reducing spare parts will only have a small effect on overall footprint. More importantly, 
MALSP II aims to reduce the amount of repair equipment, mobile facilities and people 
needed at the deployed bases thereby reducing total footprint forward. By applying new 
logistical practices and better coordinating the logistical support network, MALSP II 
aims to improve the effectiveness of support to multiple, simultaneous contingencies. 

D. THESIS SCOPE: BUILD A MODEL THAT USES DEMAND PATTERN 

TO DESIGN AND EVALUATE SUPPORT NETWORK 

The Deputy Commandant of Aviation Logistics, Support Branch implemented a 
MALSP II pilot program in 2005 with the intention to fully implement MALSP II by 
2015 (Delaporte 2007, 6). This implementation requires operational planners to select 
the system configuration and inventory buffer sizing policy that most efficiently uses 
available resources to maximize the effectiveness of logistical support to deployed bases. 

This thesis develops an inter-temporal simulation model that measures the 
operational effectiveness of four network configurations using the four inventory buffer 

sizing methods used in ELAT. It compares results using the traditional goal of PackUp 

8 




Effectiveness and a new measure of effectiveness—PartShort—that tracks the number 
and duration of unsatisfied demands referred to as holes in aircraft. Further, the model 
quantifies the impact on operational effectiveness of the risk levels used by ELAT to 
develop inventory buffers. It does not address allocation of other resources such as 
maintenance repair equipment, transportation assets or personnel. 


9 



THIS PAGE INTENTIONALLY LEFT BLANK 


10 



II. LITERATURE REVIEW 


A. BACKGROUND ON LEGACY MALSP 

Legacy MALSP follows the traditional Economic Lot Size Model introduced by 
Ford W. Harris in 1915 (Simchi-Levi 2000, 43). According to this model, demand occurs 
at a known fixed rate and lead time is considered negligible. The goal of this model is to 
minimize a cost function to achieve maximum profit. 

Traditional economic models assume that demand follows a normal distribution. 
By making this assumption, inventory sizes are determined using average daily demand 
and standard deviation (Simchi-Levi 2000, 52). By contrast, the MALSP II method of 
determining inventory buffers does not assume that demand for spare parts follow a 
normal distribution. Rather than using average demand, the MALSP II method of 
determining inventory buffers uses demand patterns over time. Though averages may be 
appropriate in long-term steady-state environments, they do not protect against demand 
spikes that may prove critical during shorter term military operations. 

Legacy MALSP is a push-based supply chain while MALSP II is a pull-based 
supply chain. Push-based systems rely on long-term forecasts and can lead to an inability 
to meet changing demand patterns due to increased variability (Simchi-Levi 2000, 118). 
To offset that risk, push-based systems add a safety level that result in larger inventories. 
Pull-based systems are demand driven and decrease variability thereby decreasing 
necessary inventory levels. However, pull-based systems are difficult to implement when 
lead times are too long to respond quickly to demand signals. Therefore pull-based 
systems require additional transportation considerations to ensure responsive lead times. 

B. MALSP II IS DESIGNED SPECIFICALLY FOR DEPLOYED 

OPERATIONAL LOGISTICS ENVIRONMENT 

Logistical planning factors are different in a military setting and a business 
setting. The logistical support network can be specifically designed to accomplish 
military objectives. Operational logistics (OpLog) is defined as “a collection of means, 
resources, and organizations and processes that share the common goal of sustaining 


11 



campaigns and large-scale military operations. OpLog is designated to sustain battles 
that are distributed in time and space” (Kress 2002, 40). 

Military costs are different than civilian economic costs. Unlike the business 
logistics model—Economic Order Quantity—the cost function to be minimized in a 
deployed military setting is nebulous. In the military, cost is anything that limits a unit’s 
fighting ability. For example, a unit with a large logistics tail loses mobility or requires 
additional transportation capacity, which may hinder operational agility (Kress 2002, 42). 

The relative stability of a business logistics flow is not always present in an 
OpLog system. Uncertainty in military operations is inherent and may lead to large 
variances in consumption rates. Variability is caused by several factors. The changing 
combat situation may increase or decrease tempo over time which may affect demand for 
parts. Further, individual combat units, with differing maintenance practices, are rotated 
at regular intervals (Kress 2002, 136). 

Military planners have three logistics options: obtain the needed resources at the 
battlefield, deploy resources with the troops prior to operations and employ necessary 
resources to the battlefield over time (Kress 2002, 10). With improvements in 
transportation and communication during the past century, the third option has become 
dominant. However, planners must carefully balance both deployment and employment 
to ensure resources are available when needed without creating an unnecessarily large 
logistics tail. Forward deployed logistics units require sustainment as well which causes 
an increasing cycle of personnel and supplies at the forward deployed site (Kress 2002, 
13). 

There are many desirable properties of an OpLog support network. Flexibility is 
defined as a system’s ability to quickly and effectively respond to changes in a system. 
In the context of current operations, these changes include operational tempo and 
location. As operational tempo increases, demand for parts is likely to increase therefore 
it is important to have a logistical support network that is responsive to a changing 
demand pattern. Further, force size increases and decreases by location. A flexible 
system has the ability to quickly determine the logistical support needed at a location and 
have spare parts available to deploy. 


12 



Attainability is defined as a node’s ability to independently satisfy demand. 
Higher attainability allows a node to remain self-sufficient for a longer period of time 
(Kress 2002, 63). Larger deployments of spare parts increase a node’s attainability. 

Survivability describes the degree that logistical assets are vulnerable to enemy’s 
hostile action (Kress 2002, 63). In today’s operational environment, logistical support 
and the people required to maintain support have a higher degree of survivability at nodes 
other than the demand nodes. Future conflicts may have lower degrees of survivability at 
the demand nodes and in the transportation system that may affect the desired ratio of 
deployment and employment. 

Logistics efficiency measures the ratio between inputs invested in logistical 
capability and the estimated operational effectiveness—anything that leads to mission 
accomplishment (Kress 2002, 42). The objective at the tactical level is to minimize two 
gaps: the quantity gap and time gap (Kress 2002, 74). When logistical support is 
synchronized by quantity and time, units will receive sufficient supplies without causing 
an avoidable loss of fighting ability or impeding another unit’s effectiveness (Kress 2002, 
68-69). 

Achieving these properties requires making tradeoffs between time and quantity. 
Decreasing response time enables the node to decrease the quantity pre-positioned 
thereby improving flexibility without a significant loss of attainability (Kress 2002, 68). 

Lead-time and uncertainty are important considerations when modeling the 
OpLog system. Due to large variances caused by uncertainty, mean values are 
inappropriate planning values at the operational level. Instead, demand pattern must be 
considered with respect to time. Logistical demand and the available resources that 
determine lead-time must be coordinated (Kress 2002, 46). Structuring the size and 
location of rear and forward intermediate nodes are based on mitigating uncertainty 
caused by demand pattern and lead-time (Kress 2002, 51). 

Kress introduces the Logistics Inter-Temporal Network (ITN) Optimization 
Model that represents the importance of time dependence on logistical support in a 
deployed environment. This model is specifically designed to account for uncertainty in 
military operations and the importance that lead-time, defined in MALSP II as TRR, 


13 



plays in mitigating that uncertainty. The ITN model determines the deployment and 
employment of resources by calculating measures of effectiveness. Specifically, the ITN 
determines the quantity of assets to place at forward nodes and the structure of the 
supporting network needed to provide sufficient distribution (Kress 2002, 219). 


14 



III. ITN SIMULATION MODEL 


A. MODEL DEVELOPMENT 

This thesis builds an ITN simulation model based on the Logistics ITN 
Optimization Model developed by Kress (Kress 2002, 219). The model evaluates the 
performance of the logistics system under four different network configurations with 
varying resource allocations using MALSP II design and business rules. It varies the 
number of echelons of supply nodes, TRR between nodes and starting inventories at each 
node and then evaluates performance using two measures of effectiveness, PackUp 
Effectiveness and PartShort. 

The network structure of the ITN model is loosely based on current deployed 
operations—Operation Iraqi Freedom, Operation Enduring Freedom-Afghanistan and 
Operation Enduring Freedom-Trans Sahara (Horn of Africa)—requiring aviation 
logistical support. Figure 3 presents the network containing three areas of responsibility 
(AOR): one MOB supporting one FOB, one MOB supporting two FOBs and a unique 
node with FOB demand but TRR of a MOB. Four logistics levels are defined: P-level 
includes PMAFS, E-level includes ESB, M-level includes MOBs, and F-level includes 
FOBs. Though multiple Marine Aviation Fogistics Squadrons may provide support, the 
model simplifies to assume one PMAFS supports the network with an unlimited supply 
of spare parts. 


15 



P-level 


E-level 


M-level 


F-level 



Figure 3: MALSP II Visual Network 

The TRR figures among the various logistics levels are based on values used 
during fiscal year (FY) 2008 in the MALSP II pilot program. Without an ESB, the TRR 
between the PMALS and the MOB is 10 days. The TRR between the MOB in A1 Asad, 
Iraq and the FOB in A1 Taqadum, Iraq was 3 days. It is assumed that adding an 
additional node between the PMALS and the MOB will add an extra day of handling 
time therefore the total TRR from the PMALS to the ESB and from the ESB to the MOB 
equals 11 days; however, the exact TRR from the PMALS to the ESB and from the ESB 
to the MOB will vary in our analysis to find the values with best results. It is assumed 
that the TRR between any MOB and its associated FOB is fixed at 3 days. 

The model considers one aircraft part at a time. For each part, four network 
configurations are modeled. ESBO: the network has no ESB and therefore the TRR from 
the PMALS to a MOB is 10 days. ESB5: the ESB is added at a distance of 5 days TRR 
to any MOB and 6 days TRR from the PMALS. ESB3: the ESB is added at a distance of 


16 














3 days TRR to any MOB and 8 days TRR from the PMALS. ESB1: the ESB is added at 
a distance of 1 day TRR to any MOB and 10 days TRR from the PMALS. Twenty 
aircraft parts with varying demand patterns are analyzed. 



TRR FOB1 

TRR FOB2 

TRR FOB 3 

TRR FOB4 

TRR MOB 1 

TRR MOB2 

TRR ESB 

ESB0 

3 

3 

3 

10 

10 

10 

N/A 

ESB5 

3 

3 

3 

5 

5 

5 

6 

ESB3 

3 

3 

3 

3 

3 

3 

8 

ESB1 

3 

3 

3 

1 

1 

1 

10 


Table 2: TRR between nodes for each network configuration 


We obtain demand data for the model from recent operations. By design, the 
MOB supports more aircraft than the FOB and therefore the former will typically have a 
higher demand pattern than the latter. This thesis uses 12 months of CH-53 demand data. 
MOB demand data captures a full squadron participating in Operation Iraqi Freedom 
during FY 2008 obtained from ELAT. FOB demand data captures a four plane 
detachment participating in Operation Enduring Freedom-Horn of Africa during FY 2008 
obtained from the Standalone Material Management System (SAMMS). 

Combinations of initial inventory levels for each node in the E-, M-, and F-levels 
are used to evaluate the network. Using the MALSP II methodology described in 
Chapter I, the initial inventory associated with each risk level (no, low, medium, and 
high) is calculated for each node using the TRR values associated with each network 
configuration described in Table 2. The model evaluates performance measures of the 
network for each allocation of spare parts. 

We calculate two measures of effectiveness (MOE)—PackUp Effectiveness and 
PartShort. PackUp Effectiveness is the current metric used by the Marine Corps to 
evaluate the effectiveness of aviation logistical support in a deployed setting. It 
calculates the percentage of demands over a given time period (weekly, monthly, etc) that 
a node is able to satisfy demand for spare parts from its inventory buffer at the time when 
the order is placed. If the node is unable to satisfy demand, the requirement will exist 
until parts are received from a higher logistics level to fill that hole. The value of PackUp 


17 




Effectiveness is computed as the ratio between the number of parts issued on the day 
ordered and the total number of parts demanded during the time period. For example, if 5 
parts are demanded in a certain day and only 3 parts are issued from the inventory buffer, 
PackUp Effectiveness for that day is 3/5 or 60%. PartShort is the summation of unfilled 
daily demands during the observed period. PartShort measures the response time for a 
node to recover from a supply shortage. If each part ordered is a missing part from an 
aircraft, which is therefore unable to fly, PartShort represents the magnitude and duration 
of holes in aircraft. For example, if 5 parts are demanded in a certain day and only 3 
parts are issued from the inventory buffer, PartShort for that day is 2. If no 
replenishments are received and no demands occur the next day, PartShort for the next 
day is 2. The total PartShort for the two days is 2 + 2 or 4. 

B. RULES USED IN THE ITN SIMULATION MODEL 

The ITN model uses the network topology depicted in Figure 3, and the supply 
rules applicable to MAFSP II to perform Monte Carlo simulation to evaluate the benefit 
of an ESB, and the effect of its relative location, with respect to four responsiveness risk 
levels. The model also evaluates the expected performance of different inventory 
allocations. Using historical demand data, the model first builds cumulative distribution 
functions of daily demands at the M-level and F-level nodes and then generates 
randomized daily demand values for the simulation. Initial inventory buffers for the 
various risk levels are determined based on EFAT’s methodology described in Chapter I 
and the TRRs between nodes. For example, if the risk level is low, and the TRR from a 
node to the higher logistics level is 5 days then the buffer size in that node is the 97 th 
percentile of the observed demand values during all 5 day time periods using 360 days of 
observed demand data. The simulation follows standard logistical rules, which prioritize 
how parts are consumed and distributed. The simulation proceeds as follows: spare parts 
are consumed as demanded (if available), immediately re-ordered and replenishment 
spare parts are received within each node’s TRR consistent with MAFSP II. A detailed 
description of the simulation is given below. Four network configurations are compared 
using the aforementioned two MOEs. Each AOR calculates the MOEs separately and 

then MOEs are aggregated to evaluate total system performance. 

18 



Since the MOB is responsible for satisfying local demands and re-supplying its 
child FOB(s), priorities are assigned that ensure spare parts are used to fill actual holes in 
aircraft before filling shortages in inventory buffers. Holes in aircraft at the FOBs take 
precedence over holes in aircraft at the MOBs. If the MOB and its child FOB have 
demands in the same day that they are unable to satisfy from their local inventory 
buffers—both have aircraft with holes that are unable to fly—the MOB ships spare parts 
to the FOB before filling internal demand. However, if the FOB satisfies demand from 
its inventory buffer or has sufficient spare parts in the shipping pipeline to satisfy supply 
shortages, the MOB satisfies MOB demand before sending inventory replenishment to 
the FOB. MOB2 prioritizes FOB2 over FOB3 (see Figure 3). The flow of spare parts 
only flows from the parent node to the child node. Once a part is sent downstream it does 
not go back to the parent. Nodes within the same level (M- or F-) do not send spare parts 
to each other. 

In cases where multiple AORs require parts, the ESB fills holes in aircraft before 
shortages in inventory buffers. If multiple AORs have holes in aircraft, the ESB will fill 
FOB4 requirements then MOB2 then MOB1. The ESB does not send parts directly to 
FOB1, FOB2, or FOB3. 

C. MODEL FORMULATION 

1. Indices and Sets 

t time periods, t e {0,1,2, ...,360} 

m MOB nodes in the network, me {1,2} 

f FOB nodes in the network, /e {1, 2,3,4} 

F (m) Set of FOBs that belong to MOB m 

2 . Data 

trr f Time to reliably replenish FOB /from its MOB 

trr m Time to reliably replenish MOB m from a higher echelon 

trr e Time to reliably replenish ESB from the PMAFS 

d f (t ) demand at FOB / at time 1 


19 



e J0 

demand at MOB m at time t 

Buffer f 

initial inventory buffer of FOB / 

Buffer m 

initial inventory buffer of MOB m 

Buffer e 

initial inventory buffer of ESB 


3. State Variables 


Y f (t) 

X m (t) 

Z e(0 

R f (t ) 

TR f (t) 

RJt ) 

TR m (t ) 

TR e (t) 

PaitShort, (f) 
PartS hort m (0 
issue f (0 
issue Jt) 


supply at FOB/at time t 
supply at MOB m at time 1 
supply at ESB at time t 

supply shipped to FOB/from its parent MOB m at time t 

total supply in pipeline to FOB/at time t 

supply shipped to MOB m from ESB or PMALS at time t 

total supply in pipeline to MOB m at time t 

supply shipped to ESB from PMALS at time t 

total supply in pipeline to ESB at time t 

number of parts short at FOB/at time t 

number of parts short at MOB m at time t 

number of parts issued at FOB / at time t 

number of parts issued at MOB m at time t 


4 . Algorithm 

The algorithm described below is accompanied with pseudo-code. During each 
time period, each node begins with supply from the end of the previous time period. The 
FOB can then receive replenishment—increase supply—and fill demands—decrease 
supply. The MOB can receive replenishment—increase supply—and fill local demand or 
send supply to the FOB—decrease supply. There are two reasons a parent ships supply 


20 



to its child: fill holes in aircraft and fill shortages in inventory buffer. Holes in aircraft 
have a higher priority than shortages in inventory buffer. If the supply at a node is less 
than 0, Y f {t)< 0 or X m (t)< 0, then there is a hole in aircraft. The sequence of actions 

simulated in our model is determined by priorities based on the needs at each node, as 
described in section B of this chapter. The simulation of ESBO—the network without an 
ESB—is described below using a verbal description and pseudo-code. 

(1) On dayl, supply at each node is set equal to its inventory buffer. Initially, there is no 
supply in the shipping pipeline to any node. 

x Jt) <- Buffer m 
Y f (t) <- Buffer f 
TR m it) <— 0 
TR f (t) <— 0 

(2) At the beginning of each time period t, the MOB and the FOB start with the supply at 
the end of time period t-1. The MOB and the FOB receive supply shipped from their 
parent exactly TRR days prior. The supply at each node is incremented and the total 
supply in the shipping pipeline to each node is decremented. If the time period is not 
greater than the node’s TRR or there was no supply shipped TRR days prior, the node 
does not receive replenishment during that time period. 


21 



If (t> trr m ) 

X(t) <- X(t- 1)+ R(t-trr ) 

TRJt) <— TR Jr -1) - R(t-trr ) 

End if 
If (t > /rr^) 

J 7 / (0 <— (f — 1) + ) 

TR f (t) <-TR f (t- 1) - R f (t-trr f ) 

End if 

(3) Next, the FOB's demand is subtracted from the FOB’s supply. If demand exceeds 
supply, the FOB’s supply reflects a negative number indicating holes in aircraft. If the 
FOB’s unfilled demand (the absolute value of the FOB’s negative supply) is greater than 
the supply already in the shipping pipeline to the FOB and the MOB has spare parts 
(supply at the MOB is greater than 0), the MOB ships parts to the FOB exclusively to fill 
holes in aircraft. Total supply in the shipping pipeline to the FOB is incremented. The 
quantity shipped from the MOB to the FOB is decremented from the MOB’s supply. 

Y f (t)<-Y f (t)-d f (t) 
lf( Y f (t)<0 and X m (t) >0) 

R f (0 <- Max [0, Min (Abs ( Y f (t) )- TR f (; t ), X m (t) )] 

TR f (t) <— TR f (t) + R f (t) 

X m (0 ^ X m(0 

End if 

(4) Next, the MOB’s demand is subtracted from the MOB’s supply. If demand exceeds 
supply, the MOB’s supply reflects a negative number indicating holes in aircraft. 

22 



(5) If the MOB’s supply is greater than 0, it ships the FOB inventory replenishment that 
is not already in shipping pipeline. Total supply in the shipping pipeline to the FOB is 
incremented. The quantity shipped from the MOB to the FOB is decremented from the 
MOB’s supply. 

If(X m (0>0) 

R f (0 <- R f it) +Min (( Buffer f - Y f (t) - TR f (t )), X m (t)) 

TR f (t) <- TR f (t)+ R f (t) 

X m( t ) 

End if 

(6) Last, if the supply at the MOB and the supply in the shipping pipeline to the MOB is 
less than the MOB’s inventory buffer or supply at the FOB and the supply in the shipping 
pipeline to the FOB is less than the FOB’s inventory buffer, the PMALS ships supply to 
the MOB. The total supply in the shipping pipeline to the MOB is incremented. 

RJt) <r- ( Buffer m -X m (t) - TR m (t))+(Buffer f -Y f (t)-TR f (t )) 

TR m (t)^TRJt)+R m (t) 

(7) At the end of each time period t, the simulation calculates variables used to compute 
MOEs. At this point, d f (t) is the demand that has already occurred at FOB / during time 

period t and Y f (t) is the resulting supply at the end of time period t. 

Description: PartShort 

PartShort is calculated each day and added to the previous day’s total. First, the variables 
are initialized to 0. If the supply at the FOB at the end of time t is negative, the number 
of holes in aircraft is equal to the absolute value of supply at the FOB. For example, if 
the FOB supply is -3 then there are 3 aircraft with holes. PartShort represents the number 
of aircraft with holes. 


23 



PartShort / (0 <—0 
lf(Y f (t)<0) 

PartShort / (?) <—abs (Y f (t) ) 

End if 

Description: Issue 

An issue is used to calculate PackUp Effectiveness. An issue occurs only if demand is 
filled during the time period demanded. Partial issues are allowed. If a squadron orders 
8 parts and the FOB fills 5 in that time period, issues for that time period equals 5. 

If ( d f (?) > 0) and ( Y f (?) >= 0) 
issue / (?) <— d f (t ) 

End if 

If ( d f (?) > 0) and (Y f (?) < 0) and (Abs ( Y f (?)) < d f ( t )) 
issue f ( t ) <— d f ( t ) - Abs ( Y f ( t )) 

End if 


5. Measures of Effectiveness 

Description: PartShort is the summation of parts short during each time period. 

(1) PartShort = ^ PartShort f (t) 

t 

Description: PackUp Effectiveness is the summation of all issues divided by the 
summation of all demands. 

(2) PackUp Effectiveness = ^ issue f ( t ) / ^ d f (?) (if 5X(')> o) 

t t t 

D. EXAMPLES OF MEASURES OF EFFECTIVENESS 

Table 3 depicts a CH-53 wheel (NUN: 014290072) for MOB1 and FOB1 using 
ESB0 for 30 days. MOEs for Dayl through Day6 are described below using ESB0: 


24 



Bujfer m = 5; trr m = 10 
Buffer f =\\ trr f =3 

TR f (t) = 0 
TR m (t) = 0 

Day 1 (/=!): On day 1, each node’s supply begins at full inventory buffer. No demand 
occurs. No replenishments are received because the time period is less than either node’s 
TRR. Each node’s supply does not change during day 1. PartShort = 0 and PackUp 
Effectiveness is N/A because ^ Demand is not greater than 0. 

Y f (l)= Buffer f = 1 
X m (l)= Buffer m = 5 
d f (t)=0 
e m (t)=0 

PartShort (1) = 0; ^ PartShort = 0 
Issues (1) = 0; ^Issues = 0 
Demand (1) = 0; V Demand = 0 

PackUp Effectiveness = N/A ( ^ Demand is not greater than 0) 

Day 2 (t=2): On day 2, each node’s supply begins at the same number it ended day 1. 
No demand occurs and no replenishments are received at the FOB. The MOB receives a 
demand for 1 part and its supply is updated. PartShort is unchanged. Issues and Total 
Demand are updated to reflect 1 each. PackUp Effectiveness is updated to reflect 1 
is sue/1 demand or 100%. 

>7(2)= >7(1) = i 
Xm(2)=X m (l) =5 
df( 2)=0 
e m (2)=l 
X m ( 2)=5-l = 4 

PartShort (2) = 0; J' i PartShort = 0 


25 



Issues (2) = 1; y Issues = 1 
Demand (2) =1; / Demand = 1 

Issues / y Demand = 1/1= 100% 

Day 3 (/=3): On day 3, each node’s supply begins at the same number it ended day 2. 
No demand occurs and no replenishments are received at either the FOB or the MOB. 
PackUp Effectiveness and PartShort are unchanged. 

>7(3)= Yf( 2) = 1 
X m (3)=X m (2) =4 
df( 3)=0 
e m (3)=0 

PartShort (3) = 0; y PartShort = 0 
Issues (3) = 0; V Issues = 1 
Demand (3) =0; y Demand = 1 

Issues / y Demand = 1/1= 100% 

Day 4 (/=4): On day 4, each node’s supply begins at the same number it ended day 3. 
No demand occurs and no replenishments are received at either the FOB or the MOB. 
PackUp Effectiveness and PartShort are unchanged. 

> 7 ( 4 )= > 7 ( 3 ) = 1 

X m (4)=Xm(3) =4 
df( 4)=0 
e m (4)=0 

PartShort (4) = 0; y PartShort = 0 
Issues (4) = 0; y Issues = 1 
Demand (4) =0; y Demand = 1 

Issues / y Demand = 1/1= 100% 


PackUp Effectiveness = y 


PackUp Effectiveness = y 


PackUp Effectiveness = y 


26 



Day 5 (t= 5): On day 5, each node’s supply begins at the same number it ended day 4. 
The FOB has a demand of 3 and issues 1 from its inventory buffer. The FOB has 2 
“holes” in aircraft. The MOB ships the FOB 3 parts to fill both “holes” in aircraft and 
shortage in inventory buffer. The MOB orders a replenishment of 3 parts from the 
PMALS. PartShort is updated to reflect 2 during this time period and a total of 2 during 
the simulation. Issues are updated to reflect 1 during this time period and a total of 2 
during the simulation. Demand during this time period is updated to reflect 3 and total 
during the simulation is updated to reflect 4. PackUp Effectiveness is updated to reflect 2 
issues/ 4 demands or 50%. 

Yf( 5)= Yf( 4) = 1 
X m (5)=X m (4) =4 
df{ 5)=3 
Yf( 5)= 1-3= -2 
Rf( 5)=3 
TRf(5)=3 

X m (5)= 4 - Rf( 5)= 4-3 =1 

e m (5)=0 

X m ( 5)= 1-0= 1 

Rm( 5)=3 

77? m (5)=l+3=4 

PartShort (5) = 2; ^ PartShort = 2 
Issues (5) = 1; Y Issues = 2 
Demand (5) =3; Y, Demand = 4 

Issues / Y Demand = 2/4= 50% 

Day 6 (/=6): On day 6, each node’s supply begins at the same number it ended day 5. 
Neither the MOB nor the FOB receives replenishment or demand. PartShort reflects 2 
during this time period and a total of 4 during the simulation. PackUp Effectiveness is 
unchanged. 


PackUp Effectiveness = Y 


27 



Yf( 6)= Yf( 5) = -2 


X m (6)=X m (5) =1 
df ( 6)=0 
e m ( 6)=0 

PartShort (6) = 2; y Part Short = 4 
Issues (6) = 0; V Issues = 2 
Demand (6) =0; y Demand = 4 

Issues / V Demands = 2/4 = 50% 


PackUp Effectiveness= 


28 



E, 


EXAMPLE OUTPUT 


t 

Buffer } 

df(t) 

Yf(t) 

Rf(t) 

TRf(t) 

Buffer m 

e m ( t ) 

Xm(t) 

Rm(t) 

TRm(t) 

Part 

Short 

issues 

total 

demand 

P-Up 

Effect 

1 

1.00 

0.00 

1.00 

0.00 

0.00 

5.00 

0.00 

5.00 

0.00 

0.00 

0.00 

0.00 

0.00 

NA 

2 

1.00 

0.00 

1.00 

0.00 

0.00 

5.00 

1.00 

4.00 

1.00 

1.00 

0.00 

1.00 

1.00 

1.00 

3 

1.00 

0.00 

1.00 

0.00 

0.00 

5.00 

0.00 

4.00 

0.00 

1.00 

0.00 

1.00 

1.00 

1.00 

4 

1.00 

0.00 

1.00 

0.00 

0.00 

5.00 

0.00 

4.00 

0.00 

1.00 

0.00 

1.00 

1.00 

1.00 

5 

1.00 

3.00 

- 2.00 

3.00 

3.00 

5.00 

0.00 

1.00 

3.00 

4.00 

2.00 

2.00 

4.00 

0.50 

6 

1.00 

0.00 

- 2.00 

0.00 

3.00 

5.00 

0.00 

1.00 

0.00 

4.00 

4.00 

2.00 

4.00 

0.50 

7 

1.00 

0.00 

- 2.00 

0.00 

3.00 

5.00 

0.00 

1.00 

0.00 

4.00 

6.00 

2.00 

4.00 

0.50 

8 

1.00 

0.00 

1.00 

0.00 

0.00 

5.00 

0.00 

1.00 

0.00 

4.00 

6.00 

2.00 

4.00 

0.50 

9 

1.00 

0.00 

1.00 

0.00 

0.00 

5.00 

1.00 

0.00 

1.00 

5.00 

6.00 

3.00 

5.00 

0.60 

10 

1.00 

0.00 

1.00 

0.00 

0.00 

5.00 

0.00 

0.00 

0.00 

5.00 

6.00 

3.00 

5.00 

0.60 

11 

1.00 

1.00 

0.00 

0.00 

0.00 

5.00 

3.00 

- 3.00 

4.00 

9.00 

9.00 

4.00 

9.00 

0.44 

12 

1.00 

0.00 

0.00 

0.00 

0.00 

5.00 

0.00 

- 2.00 

0.00 

8.00 

11.00 

4.00 

9.00 

0.44 

13 

1.00 

0.00 

0.00 

0.00 

0.00 

5.00 

0.00 

- 2.00 

0.00 

8.00 

13.00 

4.00 

9.00 

0.44 

14 

1.00 

0.00 

0.00 

0.00 

0.00 

5.00 

0.00 

- 2.00 

0.00 

8.00 

15.00 

4.00 

9.00 

0.44 

15 

1.00 

0.00 

0.00 

0.00 

0.00 

5.00 

2.00 

- 1.00 

2.00 

7.00 

16.00 

5.00 

11.00 

0.45 

16 

1.00 

0.00 

0.00 

0.00 

0.00 

5.00 

0.00 

- 1.00 

0.00 

7.00 

17.00 

5.00 

11.00 

0.45 

17 

1.00 

1.00 

- 1.00 

0.00 

0.00 

5.00 

1.00 

- 2.00 

2.00 

9.00 

20.00 

5.00 

13.00 

0.38 

18 

1.00 

0.00 

- 1.00 

0.00 

0.00 

5.00 

1.00 

- 3.00 

1.00 

10.00 

24.00 

5.00 

14.00 

0.36 

19 

1.00 

0.00 

- 1.00 

1.00 

1.00 

5.00 

0.00 

- 3.00 

0.00 

9.00 

28.00 

5.00 

14.00 

0.36 

20 

1.00 

0.00 

- 1.00 

0.00 

1.00 

5.00 

1.00 

- 4.00 

1.00 

10.00 

33.00 

5.00 

15.00 

0.33 

21 

1.00 

0.00 

- 1.00 

0.00 

1.00 

5.00 

2.00 

- 2.00 

2.00 

8.00 

36.00 

5.00 

17.00 

0.29 

22 

1.00 

0.00 

0.00 

0.00 

0.00 

5.00 

0.00 

- 2.00 

0.00 

8.00 

38.00 

5.00 

17.00 

0.29 

23 

1.00 

0.00 

0.00 

0.00 

0.00 

5.00 

0.00 

- 2.00 

0.00 

8.00 

40.00 

5.00 

17.00 

0.29 

24 

1.00 

0.00 

0.00 

0.00 

0.00 

5.00 

0.00 

- 2.00 

0.00 

8.00 

42.00 

5.00 

17.00 

0.29 

25 

1.00 

0.00 

0.00 

0.00 

0.00 

5.00 

0.00 

0.00 

0.00 

6.00 

42.00 

5.00 

17.00 

0.29 

26 

1.00 

0.00 

0.00 

0.00 

0.00 

5.00 

1.00 

- 1.00 

1.00 

7.00 

43.00 

5.00 

18.00 

0.28 

27 

1.00 

0.00 

0.00 

1.00 

1.00 

5.00 

0.00 

0.00 

0.00 

5.00 

43.00 

5.00 

18.00 

0.28 

28 

1.00 

0.00 

0.00 

0.00 

1.00 

5.00 

0.00 

1.00 

0.00 

4.00 

43.00 

5.00 

18.00 

0.28 

29 

1.00 

0.00 

0.00 

0.00 

1.00 

5.00 

0.00 

1.00 

0.00 

4.00 

43.00 

5.00 

18.00 

0.28 

30 

1.00 

0.00 

1.00 

0.00 

0.00 

5.00 

0.00 

2.00 

0.00 

3.00 

43.00 

5.00 

18.00 

0.28 


Table 3: NUN: 014290072; ESBO: FOB TRR=3, MOB TRR=10 


29 




THIS PAGE INTENTIONALLY LEFT BLANK 


30 



IV. RESULTS AND ANALYSIS 


A. SUMMARY OF RESULTS 

Evaluating the design of the logistical support network and the allocation of 
resources to its various nodes requires prioritizing desirable network properties described 
in Chapter II. The results of this thesis focus on one property—efficiency—defined as 
the ratio between inputs invested in logistics capability and estimated operational 
effectiveness (Kress 2002, 42) measured, in our case, by two MOEs, described in Chapter 
III: (1) PackUp Effectiveness represents the percentage of demands satisfied on the day 
demanded and (2) PartShort represents the magnitude and duration of unsatisfied 
demands. The system’s goal is to produce the highest possible value of PackUp 
Effectiveness and the smallest possible value of PartShort, while constraining the 
logistical footprint. The model compares the four network configurations described in 
Chapter III: ESBO represents the network without an ESB. The other three 
configurations assume the existence of an ESB and they differ in terms of distance, 
measured by TRR, from the PMALS and the MOBs: ESB5 is 5 days from the MOBs and 
6 days from the PMALS, ESB3 is 3 and 8 days, respectively, and ESB1 is 1 day from the 
MOBs and 10 days from the PMALS. Each configuration is modeled using each one of 
the four risk levels—no-risk, low-risk, medium-risk, high-risk—used in ELAT to 
determine inventory buffers, described in Chapter I. Each instance of the simulation runs 
for 360 days and is replicated 20 times. The model computes the values of the two 
MOEs for each combination of network configuration and risk level. The model is 
applied to 20 different spare parts 

This thesis evaluates both the design of the logistical support network and the 
impact of selecting a certain risk level for inventory buffer sizing. Both MOEs—PackUp 
Effectiveness and PartShort—quantify the improvement gained by allocating additional 
spare parts to each node in the network. If operational planners prioritize efficiency 
above other desirable network properties, these results may help select the appropriate 
risk level to use when determining inventory buffers, as shown and discussed later on. 
For each one of the 20 spare parts, the analysis follows the following steps: 


31 



(1) For inventory buffer sizes determined by ELAT with respect to the four risk 
levels and the various values of TRR that correspond to the four network configurations 
we simulate demands and compute the corresponding values of PackUp Effectiveness 
and PartShort. Note that the total number of parts in all buffers may vary from one 
configuration to another. 

(2) To facilitate simple efficiency comparisons among the network configurations, 
for each risk level, the total number of spare parts in the various node inventory buffers is 
fixed for all four configurations. This constitutes a common denominator for 
comparison. First, the four inventory buffer sizes are used as a basis to determine the 
total number of spare parts dedicated to the system. Next, parts are re-allocated to the 
various nodes to find the inventory buffer sizes at each node that produce the highest 
values of the selected MOEs for the entire system using the ITN simulation model. Each 
network configuration is ranked from l(best) to 4(worst) for each spare part based on the 
selected MOE. Then, the ranking for all 20 parts are averaged to assign an overall 
average rank to each network configuration. Also, the frequency of ranks is presented for 
each network configuration and buffer size. In the case of a tie, the lower rank is 
assigned to each network configuration. 

The results of the analysis with respect to the 20 selected spare parts and the four 
risk levels indicate that no network configuration dominates the others throughout. The 
priority ranking of the four alternatives depend on the particular risk level used to 
determine inventory buffers and the selected MOE. The priority ranking is independent 
of the frequency of demand for individual spare parts analyzed. Average rankings are 
displayed in Figures 4 and 5 below. The frequency of ranks using each network 
configuration based on PartShort is displayed in Figures 6, 7, 8 and 9. The frequency of 
ranks using each network configuration based on PackUp Effectiveness is displayed in 
Figures 10, 11, 12 and 13. Using no-risk inventory buffer sizing, ESBO generally 
requires less total number of parts in the support network than the other configurations. 
Using any other risk level of buffer sizing—low, medium or high—results in different 
prioritizations based on the selected MOE. Using low, medium or high risk level 
inventory buffer sizing, ESBO produces the highest value of PackUp Effectiveness. 


32 



Using low, medium or high risk level inventory buffer sizing, ESB1 produces the least 
number of PartShort. Generally, if the ESB is used in the network, effectiveness 
decreases as the ESB moves farther from the demand nodes and closer to the PMALS. 
ESB3 has fewer PartShort than ESBO. ESB5 has approximately equal PartShort as 
ESBO. 



Figure 4: Overall average ranks based on PartShort using the four network 

configurations and four inventory buffer sizing risk levels 



Figure 5: Overall average ranks based on PackUp Effectiveness using the 

four network configurations and four inventory buffer sizing risk levels 


33 





































PackUp Effectiveness and PartShort do not necessarily improve in tandem as a 
result of changing the design of the support network. A system that increases the 
percentage of demands satisfied on the day demanded does not necessarily improve the 
response time for unsatisfied demand. Considering each one of these two MOEs may 
result in different network configurations. A simple example illustrates this finding. The 
logistical system has a choice between two network designs: ESBO deploys more parts to 
the MOB but has a 10 day TRR between the PMALS and the MOB while ESB3 deploys 
fewer parts to the MOB but has a 3 day TRR between the ESB and the MOB. Assume 
the MOB using both configurations has a demand of 10 parts. Using ESBO, the MOB 
issues 7 parts from its inventory buffer yielding 70% PackUp Effectiveness. However, 
each part not issued the day ordered takes 10 days to arrive resulting in PartShort of 
3*10=30. Using ESB3, the MOB only issues 4 parts from its inventory buffer yielding a 
40% PackUp Effectiveness. However, each part not issued the day ordered takes 3 days 
to arrive resulting in PartShort of 6*3=18. While ESBO has a significantly higher 
PackUp Effectiveness, ESB3 has fewer (better) PartShort. 

Twenty spare parts with varying demand patterns, categorized into three demand 
levels, are analyzed. Demand is categorized as low if the total demand during 360 days 
was three or less, medium if the total demand is between four and ten parts and high if 
the total demand is more than ten parts. Nine parts analyzed represent moderate or high 
demand at both the MOB and FOB. Nine parts analyzed represent moderate or low 
demand at both the MOB and FOB. Two parts represent high MOB demand but low 
FOB demand. 

MOEs are calculated for each AOR illustrated in Figure 3 in Chapter III. MOB 1 
and FOB1 produce one result for each MOE for each network configuration modeled. 
MOB2, FOB2, and FOB3 produce one result for each MOE for each network 
configuration modeled. FOB4 produces its own MOEs for each network configuration 
modeled. System performance is evaluated by computing the MOEs for all three AORs 
together. 

Low-demand parts and high-demand parts are separated to examine if both 
demand patterns perform the same using the four network configurations. There is not a 


34 



significant difference between results for low- and high-demand parts. The above 
summary of results applies to both high-demand and low-demand parts. 

Two spare parts are used to illustrate results for high- and low-demand aircraft 
parts. A gyroscope (NUN: 010632830) represents a part with high MOB demand, 
ordered 117 times at OIF during FY-08, and moderate FOB demand, ordered 7 times at 
OEF-HOA. A gearbox assembly (NUN: 014117040), which was the most expensive part 
ordered during either OIF or OEF-HOA, represents a part with low demand at both the 
MOB and FOB. The gearbox assembly was ordered 3 times during OIF during FY-08 
and once during OEF-HOA. 


Frequency of ranks using no-risk inventory 
buffer sizes based on PartShort 


20 

15 

10 

5 

0 

RANK 


1 



■ J _ 

-III 

1 

2 

3 

4 

■ ESB0 

15 

5 

0 

0 

■ ESB5 

1 

6 

7 

6 

■ ESB3 

1 

9 

10 

0 

■ ESB1 

9 

10 

0 

1 


Figure 6: Frequency of ranks using no-risk inventory buffer sizes based on 

PartShort 


35 



























Frequency of ranks using low-risk 
inventory buffer sizes based on PartShort 


30 

20 

10 

0 

RANK 


1 . 

- 1 J i li 

i 

2 

3 

4 

■ ESBO 

0 

5 

5 

10 

■ ESB5 

3 

1 

8 

8 

■ ESB3 

5 

13 

2 

0 

■ ESB1 

20 

0 

0 

0 


Figure 7: Frequency of ranks using low-risk inventory buffer sizes based on 

PartShort 


Frequency of ranks using medium-risk 
inventory buffer sizes based on PartShort 


30 

20 

10 

0 

RANK 


1 | 

-1 . 1 ■ 

1 

2 

3 

4 

■ ESBO 

2 

4 

4 

10 

■ ESB5 

1 

2 

9 

8 

■ ESB3 

2 

16 

2 

0 

■ ESB1 

20 

0 

0 

0 


Figure 8: Frequency of ranks using medium-risk inventory buffer sizes based 

on PartShort 


36 

















































Frequency of ranks using high-risk 
inventory buffer sizes based on PartShort 


20 

15 

10 

5 

0 

RANK 











■ 







■ ■ 

_ . 

1 

2 

3 

4 

■ ESBO 

1 

3 

3 

13 

■ ESB5 

0 

2 

16 

2 

■ ESB3 

0 

18 

2 

0 

■ ESB1 

17 

2 

1 

0 


Figure 9: Frequency of ranks using high-risk inventory buffer sizes based on 

PartShort 


Frequency of ranks using no-risk inventory 
buffer sizes based on PackUp 
Effectiveness 

&■ 

| 15 

O' 

&> 

£ 10 

5 


■ 

1 ■ ■ 1 . - 

1 ■ 1 _ 


u 

RANK 

1 

2 

3 

4 


■ ESBO 

15 

5 

0 

0 

■ ESB5 

1 

7 

6 

6 

■ ESB3 

1 

10 

9 

0 

■ ESB1 

9 

10 

0 

1 




PackUp Effectiveness 


[ on 


37 


























































Frequency of ranks using low-risk 
inventory buffer sizes based on Packllp 
Effectiveness 


>. 20 

U 

I 15 


10 


u 

RANK 

1 

2 

3 

4 

■ ESBO 

18 

1 

0 

1 

■ ESB5 

2 

9 

6 

3 

■ ESB3 

2 

10 

8 

0 

■ ESB1 

4 

13 

1 

2 


Figure 11: Frequency of ranks using low-risk inventory buffer sizes based on 

PackUp Effectiveness 


Frequency of ranks using medium-risk 
inventory buffer sizes based on PackUp 
Effectiveness 


30 

20 

10 

0 

RANK 



l | | j 

1 

2 

3 

4 

■ ESBO 

20 

0 

0 

0 

■ ESB5 

1 

14 

2 

3 

■ ESB3 

1 

15 

4 

0 

■ ESB1 

1 

18 

0 

1 


Figure 12: Frequency of ranks using medium-risk inventory buffer sizes based 

on PackUp Effectiveness 


38 












































Frequency of ranks using high-risk 
inventory buffer sizes based on Packllp 
Effectiveness 


> 

u 

c 

QJ 

3 

O’ 

QJ 


30 

20 

10 

0 

RANK 


l 1 1 


1 

2 

3 

4 

■ ESB0 

20 

0 

0 

0 

■ ESB5 

3 

11 

4 

2 

■ ESB3 

3 

12 

4 

1 

■ ESB1 

3 

16 

1 

0 


Figure 13: Frequency of ranks using high-risk inventory buffer sizes based on 

PackUp Effectiveness 


B. THE IMPACT OF SELECTING A RISK LEVEL 

The risk level selected to size inventory buffers has a significant impact on many 
desirable network properties such as attainability, flexibility, survivability and efficiency. 
Planners of the OPLOG system may prioritize a high level of attainability, defined in 
Chapter II as a node’s ability to independently satisfy demand, and choose no-risk buffer 
sizing to achieve 100% PackUp Effectiveness at each node. Alternatively, planners may 
give high priority to flexibility, defined in Chapter II as a system’s ability to quickly and 
effectively respond to changes in a system. Using no-risk inventory buffer sizing may 
lead to a loss of flexibility. By increasing the number of spare parts placed at each node, 
the system may lose the ability to quickly move to another location. Additionally, using 
no-risk buffer sizing may require more parts in the network than the support system has 
available thereby limiting the number of nodes that are allocated inventory buffers. If 
improving flexibility is given a high priority, assigning inventory buffers using low-, 
medium- or high-risk may be appropriate. Another important network property is 


39 






















survivability defined in Chapter II as a node’s vulnerability to enemy’s hostile action 
(Kress 2002, 66). Placing logistical support at demand nodes, which are typically hostile 
environments, may decrease survivability. Operational planners may prioritize 
efficiency—defined as the ratio between inputs invested in logistics capability and 
estimated operational effectiveness. Each risk level represents a certain level of input of 
spare parts to the system. The ITN simulation model produces expected operational 
effectiveness for these varying levels of input. Together, these parameters serve as a 
proxy for determining efficiency. 

In this section we first examine the estimated MOEs produced by the ITN model 
for each network configuration using inventory buffers, corresponding to the various risk 
levels, as determined by ELAT. Then we re-allocate the number of spare parts at each 
node and examine the resulting estimated MOEs. 

Allocation of spare parts involves two decisions concerning range and depth. The 
range refers to the number of line items of spare parts included in the logistical support 
package. The depth refers to the quantity of each spare part included in the logistical 
support package. Applying no-risk buffer sizing, the range of parts allocated to forward 
nodes would include all parts that were ordered at least once during the time frame of the 
observed demand data. This could significantly increase the range of parts at the demand 
nodes thereby increasing the footprint of those nodes. Using the low risk buffer sizing 
method described in Chapter I, only parts that have been ordered in at least 3% of the 
time periods considered will be added to the range of line items. Using higher risk levels 
to determine inventory buffers will increase the percentage of time periods a part will 
need to have been ordered during the time frame considered to be added to the support 
package of each node. Using low-, medium- or high-risk level inventory buffer sizing 
may exclude many parts from the demand nodes and lead to some loss of operational 
effectiveness. 

Selecting the risk level to use to determine inventory buffers for low-demand 
parts has a significant effect on the range of spare parts and the expected operational 
effectiveness. The range of spare parts dedicated to the OPLOG system should be 

determined considering the aggregated performance of the network. For example, Table 

40 



5 depicts that using ESB3 and low-risk inventory buffers removes all parts but 1 at the 
ESB. Though 6 nodes do not have parts allocated to their inventory buffer, their 
aggregated demand determines that 1 part is allocated to the ESB. Therefore, this low- 
demand part is included in the range of the network even though it is not included in the 
range of each node. 

Table 4 depicts no-risk inventory buffer sizing for a low-demand aircraft part that 
requires 8 total parts in the network for every node to achieve 100% PackUp 
Effectiveness and 0 PartShort using ESBO. All other configurations require 9 total parts 
in the network to achieve 100% PackUp Effectiveness and 0 PartShort. Tables 5, 6 and 7 
depict the distribution of spare parts and resulting measures of effectiveness using low-, 
medium- and high-risk level inventory buffer sizing. Using ESB3, no-risk buffer sizing 
allocates spare parts to every node in the network, but low-risk buffer sizing removes all 
spare parts but one at the ESB. Low-risk buffer sizing, using ESB3, reduces the total 
parts needed to support the network from 9 to 1, but also reduces PackUp Effectiveness 
from 100% to 0% and increases PartShort from 0 to 44. 


014117040 

Fobl 

Buffer 

Fob2 

Buffer 

Fob3 

Buffer 

Fob4 

Buffer 

Mobl 

Buffer 

Mob2 

Buffer 

ESB 

Buffer 

Total 

Buffer 

PackUp Effect 

PartShort 

No Risk 











ESBO 

1 

1 

1 

1 

2 

2 


8 

100% 

0 

ESB5 

1 

1 

1 

1 

1 

2 

2 

9 

100% 

0 

ESB3 

1 

1 

1 

1 

1 

2 

2 

9 

100% 

0 

ESB1 

1 

1 

1 

1 

1 

2 

2 

9 

100% 

0 


Table 4: No-risk inventory buffers for gearbox assembly representing low- 

demand part 


41 




014117040 

Fobi 

Buffer 

Fob2 

Buffer 

Fob3 

Buffer 

Fob4 

Buffer 

Mobf 

Buffer 

Mob2 

Buffer 

ESB 

Buffer 

Total 

Buffer 

PackUp Effect 

PartShort 

Low Risk 











ESBO 

0 

0 

0 

0 

1 

1 


2 

55% 

24 

ESB5 

0 

0 

0 

0 

0 

1 

1 

2 

11% 

32 

ESB3 

0 

0 

0 

0 

0 

0 

1 

1 

0% 

44 

ESB1 

0 

0 

0 

0 

0 

0 

1 

1 

0% 

26 


Table 5: Low-risk inventory buffers for gearbox assembly representing low- 

demand part 


014117040 

Fobi 

Buffer 

Fob2 

Buffer 

Fob3 

Buffer 

Fob4 

Buffer 

Mobf 

Buffer 

Mob2 

Buffer 

ESB 

Buffer 

Total 

Buffer 

PackUp Effect 

PartShort 

Medium Risk 











ESBO 

0 

0 

0 

0 

1 

1 


2 

55% 

24 

ESB5 

0 

0 

0 

0 

0 

0 

1 

1 

0% 

53 

ESB3 

0 

0 

0 

0 

0 

0 

1 

1 

0% 

44 

ESB1 

0 

0 

0 

0 

0 

0 

1 

1 

0% 

26 


Table 6: Medium-risk inventory buffers for gearbox assembly representing 

low-demand part 


014117040 

Fobi 

Buffer 

Fob2 

Buffer 

Fob3 

Buffer 

Fob4 

Buffer 

Mobf 

Buffer 

Mob2 

Buffer 

ESB 

Buffer 

Total 

Buffer 

PackUp Effect 

PartShort 

High Risk 











ESBO 

0 

0 

0 

0 

0 

0 


0 

0% 

100 

ESB5 

0 

0 

0 

0 

0 

0 

0 

0 

0% 

97 

ESB3 

0 

0 

0 

0 

0 

0 

0 

0 

0% 

102 

ESB1 

0 

0 

0 

0 

0 

0 

0 

0 

0% 

98 


Table 7: High-risk inventory buffers for gearbox assembly representing 

low-demand part 


The ITN model quantifies the increase in effectiveness that results from adding 
additional spare parts. Using ESB3, comparing Table 5 containing low-risk buffer sizes, 
and Table 8 containing an alternative allocation of spare parts, adding one spare part to 

42 




the network improves PackUp Effectiveness from 0% to 12% and reduces PartShort from 
44 to 27. Using ESB3, comparing Table 5 and Table 9, adding two additional spare parts 
to the network improves PackUp Effectiveness from 0% to 24% and reduces PartShort 
from 44 to 10. Using ESB3, comparing Table 9 and Table 4, reducing the quantity of 
spare parts in the network from 9 to 3 decreases PackUp Effectiveness from 100% to 
24% but only increases PartShort from 0 to 10. By quantifying expected MOEs, planners 
have the information necessary to make important decisions regarding allocation of spare 
parts for low-demand items. 


014117040 

Fobl 

Buffer 

Fob2 

Buffer 

Fob3 

Buffer 

Fob4 

Buffer 

Mobl 

Buffer 

Mob2 

Buffer 

ESB 

Buffer 

Total 

Buffer 

PackUp Effect 

PartShort 












ESB0 

0 

0 

0 

0 

1 

1 


2 

55% 

24 

ESB5 

0 

0 

0 

0 

0 

1 

1 

2 

11% 

32 

ESB3 

0 

0 

0 

0 

0 

1 

1 

2 

12% 

27 

ESB1 

0 

0 

0 

0 

0 

1 

1 

2 

12% 

17 


Table 8: Alternative inventory buffers for gearbox assembly representing 

low-demand part 


014117040 

Fobl 

Buffer 

Fob2 

Buffer 

Fob3 

Buffer 

Fob4 

Buffer 

Mobl 

Buffer 

Mob2 

Buffer 

ESB 

Buffer 

Total 

Buffer 

PackUp Effect 

PartShort 












ESB0 

0 

0 

0 

0 

1 

2 


3 

51% 

20 

ESB5 

0 

0 

0 

0 

1 

1 

1 

3 

21% 

11 

ESB3 

0 

0 

0 

0 

1 

1 

1 

3 

23% 

10 

ESB1 

0 

0 

0 

0 

1 

1 

1 

3 

25% 

9 


Table 9: Alternative inventory buffers for gearbox assembly representing 

low-demand part 


The risk level for high-demand parts has a significant effect on the depth of spare 
parts and the expected operational effectiveness. The ITN model quantifies that effect. 
Using the ITN model after determining inventory buffers in ELAT may help produce the 
allocation of spare parts with the highest MOEs. 


43 




Using no-risk inventory buffer sizing, depicted in Table 10, ESB3 requires 56 
total parts in the network to achieve 100% PackUp Effectiveness. Using low-risk 
inventory buffer sizing, depicted in Table 11, ESB3 decreases the quantity of spare parts 
required to support the network from 56 to 26 but also decreases PackUp Effectiveness 
from 100% to 59% and increases PartShort from 0 to 102. Applying higher levels of risk 
to determine inventory buffers, depicted in Tables 12 and 13, further reduces the number 
of spare parts needed to support the network and decreases the expected effectiveness of 
the system. 


010632830 

Fobl 

Buffer 

Fob2 

Buffer 

Fob3 

Buffer 

Fob4 

Buffer 

Mobl 

Buffer 

Mob2 

Buffer 

ESB 

Buffer 

Total 

Buffer 

PackUp Effect 

PartShort 

No Risk 











ESB0 

2 

2 

2 

2 

16 

18 


42 

100% 

0 

ESB5 

2 

2 

2 

2 

15 

15 

18 

56 

100% 

0 

ESB3 

2 

2 

2 

2 

15 

15 

18 

56 

100% 

0 

ESB1 

2 

2 

2 

2 

9 

9 

22 

48 

100% 

0 


Table 10: No-risk inventory buffers for gyroscope representing high-demand 

part 


010632830 

Fobl 

Buffer 

Fob2 

Buffer 

Fob3 

Buffer 

Fob4 

Buffer 

Mobl 

Buffer 

Mob2 

Buffer 

ESB 

Buffer 

Total 

Buffer 

PackUp Effect 

PartShort 

Low Risk 











ESB0 

0 

0 

0 

2 

14 

14 


30 

90% 

113 

ESB5 

0 

0 

0 

1 

9 

9 

12 

31 

64% 

75 

ESB3 

0 

0 

0 

0 

6 

6 

14 

26 

59% 

102 

ESB1 

0 

0 

0 

0 

3 

3 

16 

22 

55% 

114 


Table 11: Low-risk inventory buffers for gyroscope representing high- 

demand part 


44 




010632830 

Fobl 

Buffer 

Fob2 

Buffer 

Fob3 

Buffer 

Fob4 

Buffer 

Mobl 

Buffer 

Mob2 

Buffer 

ESB 

Buffer 

Total 

Buffer 

PackUp Effect 

PartShort 

Medium Risk 











ESBO 

0 

0 

0 

1 

11 

12 


24 

84% 

161 

ESB5 

0 

0 

0 

0 

6 

7 

10 

23 

59% 

134 

ESB3 

0 

0 

0 

0 

4 

5 

12 

21 

58% 

150 

ESB1 

0 

0 

0 

0 

2 

2 

14 

18 

44% 

172 


Table 12: Medium-risk inventory buffers for gyroscope representing high- 

demand part 


010632830 

Fobl 

Buffer 

Fob2 

Buffer 

Fob3 

Buffer 

Fob4 

Buffer 

Mobl 

Buffer 

Mob2 

Buffer 

ESB 

Buffer 

Total 

Buffer 

PackUp Effect 

PartShort 

High Risk 











ESBO 

0 

0 

0 

0 

7 

9 


16 

74% 

310 

ESB5 

0 

0 

0 

0 

4 

4 

8 

16 

45% 

320 

ESB3 

0 

0 

0 

0 

3 

3 

10 

16 

43% 

308 

ESB1 

0 

0 

0 

0 

0 

0 

12 

12 

0% 

450 


Table 13: High-risk inventory buffers for gyroscope representing high- 

demand part 


Comparing proposed low-risk inventory buffer sizes in Table 11 and an 
alternative allocation of spare parts that produce the highest MOEs using the ITN model 
in Table 14, ESB3 has less PartShort and equal PackUp Effectiveness using less total 
spare parts by removing parts from the ESB and adding parts to the buffer at the MOBs. 
Comparing proposed high-risk buffer sizes in Table 13 and an alternative allocation of 
spare parts that produce the highest MOEs using the ITN model in table 15, ESB5 
experiences less PartShort with equal total spare parts by removing parts from the ESB 
and adding parts to the buffer at the MOBs. Though inventory buffer sizing based on risk 
level is useful for allocating spare parts to a single node, network efficiency may be 
improved using the results from the ITN model. 


45 




010632830 

Fobl 

Buffer 

Fob2 

Buffer 

Fob3 

Buffer 

Fob4 

Buffer 

Mobl 

Buffer 

Mob2 

Buffer 

ESB 

Buffer 

Total 

Buffer 

PackUp Effect 

PartShort 












ESBO 

1 

1 

1 

1 

10 

10 


24 

87% 

134 

ESB5 

0 

0 

0 

0 

8 

8 

8 

24 

62% 

93 

ESB3 

0 

0 

0 

0 

8 

8 

8 

24 

60% 

91 

ESB1 

0 

0 

0 

0 

6 

6 

12 

24 

63% 

81 


Table 14: Alternative inventory buffers for gyroscope representing high- 

demand part 


010632830 

Fobl 

Buffer 

Fob2 

Buffer 

Fob3 

Buffer 

Fob4 

Buffer 

Mobl 

Buffer 

Mob2 

Buffer 

ESB 

Buffer 

Total 

Buffer 

PackUp Effect 

PartShort 












ESBO 

0 

0 

0 

0 

7 

9 


16 

74% 

310 

ESB5 

0 

0 

0 

0 

5 

5 

6 

16 

49% 

290 

ESB3 

0 

0 

0 

0 

5 

5 

6 

16 

48% 

216 

ESB1 

0 

0 

0 

0 

4 

4 

8 

16 

53% 

186 


Table 15: Alternative inventory buffers for gyroscope representing high- 

demand part 

The ITN model produces estimated MOEs that can assist planners select the risk 
level in ELAT that allocates the appropriate level of parts input to the system to achieve 
the desired effectiveness. Additionally, once total inventory buffers are determined, the 
ITN model can help planners re-allocate parts within the network to improve estimated 
MOEs. 


C. NETWORK DESIGN USING NO-RISK BUFFER SIZING 

In the previous section we established that selecting the appropriate risk level to 
use when developing inventory buffers is dependent on prioritizing desirable network 
properties. Similarly, selecting the design of the network—the number and placement of 
nodes—requires careful consideration of many network properties. The ITN model 
quantifies the impact of network design on achieving efficiency. 


46 




In this section, we rank the four network configurations using no-risk inventory 
buffers consistent with the method used in ELAT. Since no-risk inventory buffers 
produce 100% PackUp Effectiveness and 0 PartShort, we use the total spare parts 
dedicated to the system to compare the efficiency of each network configuration. Next, 
we distinguish rankings between low- and high-demand parts to analyze if the results are 
consistent for both demand patterns. Then, we present example results of a low-demand 
part and a high-demand part. 

The Borda Method of Marks is used to aggregate the performance evaluations of 
the 20 selected spare parts, using the four network configurations, to facilitate 
comparison (Cook 1992, 134). For each part and MOE, the four network configurations 
are ranked. For example, the configuration that produces the highest PackUp 
Effectiveness is assigned a “1”, the configuration that produces the next highest PackUp 
Effectiveness is assigned a “2”, etc. The same method is used for PartShort. The 
configuration that has the fewest PartShort is assigned a “1” and the configuration that 
has the most PartShort is assigned a “4”. The midpoint of ranks is used for ties. Ranks 
for each part are added together and divided by the number of parts analyzed to present 
an overall rank for each configuration. Rankings for low-demand and high-demand parts 
are separated to evaluate whether results differ based on frequency of demand. 

Overall, using no-risk buffer sizing, ESBO requires the least number of spare parts 
to achieve 100% PackUp Effectiveness and 0 PartShort. Comparing configurations with 
an ESB, the network generally requires less spare parts as the TRR decreases between the 
ESB and the demand nodes. 11 parts require the least total spare parts using ESBO, 5 
require the least spare parts using ESB1 and 4 require an equal number of spare parts 
using ESBO and ESB1. In all cases, ESBO requires less spare parts than ESB5 and ESB3 
to achieve 100% PackUp Effectiveness and 0 PartShort. 


47 



No-Risk Inventory buffers 


4.00 



ESBO ESB5 ESB3 ESB1 


■ No-Risk PartShort 


■ No Risk PackUp 
Effectiveness 


Figure 14: Overall average ranks for no-risk inventory buffers sizes based on 

network configuration 


PartShort 


OVERALL 

No Risk 

ESBO 

1.43 

ESB5 

3.43 

ESB3 

3.05 

ESB1 

2.1 


PackUp 

Effect 


OVERALL 

No Risk 

ESBO 

1.45 

ESB5 

3.4 

ESB3 

3.05 

ESB1 

2.15 


Table 16: Overall average ranks for each configuration. 20 spare parts were 

analyzed using results from the ITN model for no-risk inventory buffers. ESBO 
is the most efficient configuration, followed by ESB1 then ESB3 then ESB5. 


Parts with high demand and parts with low demand have similar results. Using 
no-risk inventory buffers, ESBO is generally more efficient than the other configurations 
with an ESB. Comparing the configurations with an ESB, results improve for high- 
demand parts, depicted in Table 17, as the TRR decreases between the ESB and the 
MOBs. However, results have less or no improvement for parts with low demand, 
depicted in Table 18, as the TRR decreases between the ESB and the MOBs. Parts with 
low demand generally require equal spare parts at each node to achieve approximately 
equal MOEs no matter where the ESB is placed. 


48 









































No-Risk inventory buffers based on 
PartShortand Packup Effectiveness 



Figure 15: Average ranks for no-risk inventory buffers based on high-demand 

parts 


PartShort 


HIGH DEM 

No Risk 

ESBO 

1.50 

ESB5 

3.78 

ESB3 

3.06 

ESB1 

1.67 


Packup Effect 


HIGH DEM 

No Risk 

ESBO 

1.50 

ESB5 

3.78 

ESB3 

3.06 

ESB1 

1.67 


Table 17: Average ranks of each configuration for parts with high demand. 

20 spare parts were analyzed using results from the ITN model for no-risk 
inventory buffers. ESBO is the most efficient configuration, followed by ESB1 

then ESB3 then ESB5. 


49 



































No-risk inventory buffers for low 
demand parts 



■ PartShort results for no- 
risk inventory buffers 
based on low demand 
parts 

■ PackUp Effectiveness 
results for no-risk 
inventory buffers based 
on low demand parts 


Figure 16: Average ranks for no-risk inventory buffers based on low-demand 

parts 


PackUp Effect 


LOW DEM 

No Risk 

ESBO 

1.39 

ESB5 

3 

ESB3 

3 

ESB1 

2.61 


PartShort 


LOW DEM 

No Risk 

ESBO 

1.33 

ESB5 

3.06 

ESB3 

3.06 

ESB1 

2.56 


Table 18: Average ranks of each configuration for parts with low demand. 20 

spare parts were analyzed using results from the ITN model for no-risk inventory 
buffers. ESBO is the most efficient configuration, followed by ESB1 then ESB3 

then ESB5. 


Results for the CH-53 gyroscope, representing high demand at the MOB and 
moderate demand at the FOB, are shown in Table 19. It contains no-risk inventory 
buffers at each node for each configuration. ESBO requires the least parts, 42, in the 
network to achieve 100% effectiveness. ESB1 requires the next least total parts, 48, and 
significantly fewer parts at the MOBs to obtain maximum effectiveness. ESB5 and ESB3 
also require fewer parts at the MOBs than ESBO but require more total parts supporting 
the network, 56, to obtain 100% effectiveness. 


50 




































010632830 

Fobl 

Buffer 

Fob2 

Buffer 

Fob3 

Buffer 

Fob4 

Buffer 

Mobl 

Buffer 

Mob2 

Buffer 

ESB 

Buffer 

Total 

Buffer 

PackUp 

Effect 

PartShort 

No Risk 











ESBO 

2 

2 

2 

2 

16 

18 


42 

100% 

0 

ESB5 

2 

2 

2 

2 

15 

15 

18 

56 

100% 

0 

ESB3 

2 

2 

2 

2 

15 

15 

18 

56 

100% 

0 

ESB1 

2 

2 

2 

2 

9 

9 

22 

48 

100% 

0 


Table 19: No-risk inventory buffer for gyroscope representing high-demand 

part 

The gearbox assembly, representing low demand at both the MOB and FOB, is 
shown in Table 20. It contains the no-risk inventory buffers at each node for each 
configuration. ESBO requires the least quantity of spare parts in the network using no- 
risk buffer sizing. All other configurations require the same total number of spare parts 
to achieve 100% effectiveness. 


014117040 

Fobl 

Buffer 

Fob2 

Buffer 

Fob3 

Buffer 

Fob4 

Buffer 

Mobl 

Buffer 

Mob2 

Buffer 

ESB 

Buffer 

Total 

Buffer 

PackUp 

Effect 

PartShort 

No Risk 











ESBO 

1 

1 

1 

1 

2 

2 


8 

100% 

0 

ESB5 

1 

1 

1 

1 

1 

2 

2 

9 

100% 

0 

ESB3 

1 

1 

1 

1 

1 

2 

2 

9 

100% 

0 

ESB1 

1 

1 

1 

1 

1 

2 

2 

9 

100% 

0 


Table 20: No-risk inventory buffer for gearbox assembly representing low- 

demand part 

D. NETWORK DESIGN USING BUFFER SIZING WITH RISK 

In this section, we present rankings to compare the efficiency of the four network 
configurations using low-, medium- and high-risk levels to determine inventory buffers. 
Spare parts at each node have been re-allocated from the number suggested by ELAT to 
use a common denominator of total spare parts in the network to compare alternative 
configurations in a proper way. The four inventory buffer sizing risk levels are used as a 
basis to determine total spare parts to dedicate to the system for each configuration. 
However, spare parts are re-allocated to find the inventory buffer sizes that produce the 

51 




highest values of the selected MOEs using the ITN simulation model. The rankings of 
low-demand parts and high-demand parts are separated to analyze if the results are 
consistent for both demand patterns. Then, we present example results of a low-demand 
part and a high-demand part. 

Planners must prioritize between the two MOEs—PackUp Effectiveness and 
PartShort—before designing the network using low-, medium- or high-risk inventory 
buffers. The two MOEs do not improve in tandem as a result of changing the design of 
the support network. A system that increases the percentage of demands satisfied on the 
day demanded does not necessarily improve the response time for unsatisfied demand. 
There is no configuration that provides the highest PackUp Effectiveness with fewest 
PartShort requiring the least spare parts. 

Table 21 displays overall results for PartShort using low-, medium-, and high-risk 
buffer sizes. ESB3 and ESB1 have fewer PartShort than ESBO and ESB5 with equal total 


spare parts in the network. Approximately equal quantities of parts have fewer PartShort 
using ESBO and ESB5. Increasing TRR between the ESB and the demand nodes 
increases PartShort. Results are consistent between low-, medium- and high-risk 


inventory buffers. 


Average Ranks based on PartShort for each 
configuration based on risk level (low, medium, 

high) 



ESBO ESB5 ESB3 ESB1 


■ Low-risk 

■ Medium-risk 

■ High-risk 


Figure 17: Average ranks for each network configuration based on risk (low, 

medium, high) 


52 












PartShort 




OVERALL 

Low Risk 

Med Risk 

High Risk 

ESBO 

3.40 

3.30 

3.50 

ESB5 

3.25 

3.35 

3.10 

ESB3 

2.15 

2.23 

2.15 

ESB1 

1.20 

1.13 

1.25 


Table 21: Overall average ranks based on PartShort for low-, medium- and 

high-risk inventory buffers 

Table 22 displays overall results for PackUp Effectiveness using low-, medium-, 
and high-risk buffer sizes. ESBO has higher PackUp Effectiveness than the other 
configurations for all risk levels. These results reflect that—using a common 
denominator of total spare parts dedicated to the logistical support network—placing 
parts at the ESB requires removing parts from the demand nodes thereby having fewer 
parts available on the day ordered. Once the ESB is added to the network, there is not a 
significant difference between configurations based on the distance between the ESB and 
the demand nodes. Using low-risk buffer sizing, ESB1 has higher PackUp Effectiveness 
than ESB5 and ESB3 for 7 out of 20 parts analyzed while 9 parts have approximately 
equal PackUp Effectiveness. Using medium-risk buffer sizing, ESB1 has higher PackUp 
Effectiveness than ESB5 and ESB3 for 4 out of 20 parts analyzed while 15 parts have 
approximately equal PackUp Effectiveness. Using high-risk buffer sizing, ESB1 
experiences higher PackUp Effectiveness than ESB5 and ESB3 for 4 out of 20 parts 
analyzed while 13 parts have approximately equal PackUp Effectiveness. 


53 




Average Ranks based on PackUp Effectiveness 
for each configuration using risk level (low, 
medium, high) 



ESBO ESBS ESB3 ESB1 


■ Low-risk 

■ Medium-risk 

■ High-risk 


Figure 18: Average ranks based on PackUp Effectiveness for each 

configuration using risk level (low, medium, high) 


PackUp Effect 




OVERALL 

Low 

Med 

High 

ESBO 

1.30 

1.08 

1.23 

ESB5 

3.13 

3.15 

3.10 

ESB3 

2.98 

2.95 

2.98 

ESB1 

2.60 

2.83 

2.70 


Table 22: Overall average ranks based on PackUp Effectiveness for low-, 

medium- and high-risk inventory buffers 

Results, depicted in Table 23, were consistent between parts with low demand and 
parts with high demand. ESB3 and ESB1 have fewer PartShort than ESBO and ESB5 for 
both high-demand and low-demand parts. ESBO and ESB5 have mixed results compared 
with each other. High-demand parts perform better using ESB5 while low-demand parts 
perform better using ESBO. Using low risk, 4 low-demand parts have fewer PartShort 
using ESBO than ESB5, 3 low-demand parts have fewer PartShort using ESB5 than 
ESBO, and 2 low-demand parts have approximately equal shortages using both ESB5 and 
ESBO. Using low-risk buffer sizing, 6 high-demand parts have fewer PartShort using 
ESBO than ESB5 while 3 high-demand parts have fewer PartShort using ESB5 than 
ESBO. 


54 



















Average Ranks based on PartShort using risk 
level (low, medium, high) for high demand part 


c 

cz 

tL 

<D 

CXO 

c3 

i- 

<v 

> 

< 



■ Low-risk 

■ Medium-risk 

■ High-risk 


ESBO ESB5 ESB3 ESB1 


Figure 19: Average ranks based on PartShort using risk level (low, medium, 

high) based for high-demand parts 


Average Ranks based on PartShort using risk 
level (low, medium, high) for low demand part 



ESBO 


ESB5 


ESB3 


ESB1 


■ Low-risk 

■ Medium-risk 

■ High-risk 


Figure 20: Average ranks based on PartShort using risk level (low, medium, 

high) based for low-demand part 


55 













PartShort 




HIGH DEM 

Low Risk 

Med Risk 

High Risk 

ESBO 

3.61 

3.61 

3.89 

ESB5 

3.22 

3.17 

3.11 

ESB3 

2.06 

2.11 

2.00 

ESB1 

1.11 

1.11 

1.00 


PartShort 




LOW DEM 

Low Risk 

Med Risk 

High Risk 

ESBO 

3.17 

2.94 

3.22 

ESB5 

3.22 

3.50 

3.00 

ESB3 

2.28 

2.39 

2.22 

ESB1 

1.33 

1.17 

1.56 


Table 23: Average ranks, isolated for high- and low-demand parts, based on 

PartShort using low-, medium- and high-risk inventory buffers 


Table 24 depicts results based on PackUp Effectiveness for high-demand parts 
and low-demand parts in isolation. ESBO produces higher PackUp Effectiveness than the 
other configurations, regardless of demand pattern, using low-, medium- and high- 
demand inventory buffers. If the ESB is included in the network, parts with high demand 
experience higher PackUp Effectiveness with shorter TRR between the ESB and the 
demand nodes. However, parts with low demand experience approximately equal 


PackUp Effectiveness regardless of the TRR between the ESB and the demand nodes. 


Average Ranks based on PackUp Effectiveness 
using risk level (low, medium, high) for high 
demand part 


c 

rc 

a: 

OJ 

00 

fO 

QJ 

> 

< 



■ Low-risk 

■ Medium-risk 

■ High-risk 


ESBO ESB5 ESB3 ESB1 


Figure 21: Average ranks based on PackUp Effectiveness using risk level 

(low, medium, high) for high-demand parts 


56 
































Average Ranks based on PackUp Effectiveness 
using risk level (low, medium, high) for low 
demand part 


C 

TO 

C£ 

Ol 

00 

TO 

s- 

O) 

> 

< 



■ Low-risk 

■ Medium-risk 

■ High-risk 


ESBO ESB5 ESB3 ESB1 


Figure 22: Average ranks based on PackUp Effectivness using risk level (low, 

medium, high) for low-demand parts 


PackUp 

Effect 




HIGH DEM 

Low Risk 

Med Risk 

High Risk 

ESBO 

1.50 

1.00 

1.00 

ESB5 

3.28 

3.33 

3.28 

ESB3 

2.78 

2.89 

3.11 

ESB1 

2.44 

2.78 

2.61 


PackUp 

Effect 




LOW DEM 

Low Risk 

Med Risk 

High Risk 

ESBO 

1.17 

1.17 

1.50 

ESB5 

2.89 

2.94 

2.83 

ESB3 

3.06 

2.94 

2.83 

ESB1 

2.89 

2.94 

2.83 


Table 24: Average ranks, isolated for high- and low-demand parts, based on 

PackUp Effectiveness using low-, medium- and high-risk inventory buffers 

Tables 25, 26 and 27 contain results for the gyroscope representing a high- 
demand part. Using equivalent total buffers for the system, configurations with an ESB— 
ESB5, ESB3 and ESB1—result in fewer PartShort than ESBO; however ESBO produces 
higher PackUp Effectiveness than the other configurations. 


57 




























010632830 

Fobl 

Buffer 

Fob2 

Buffer 

Fob3 

Buffer 

Fob4 

Buffer 

Mobl 

Buffer 

Mob2 

Buffer 

ESB 

Buffer 

Total 

Buffer 

PackUp 

Effect 

PartShort 

Low Risk 











ESBO 

1 

1 

1 

1 

14 

14 


32 

93% 

73 

ESB5 

1 

1 

1 

1 

9 

9 

10 

32 

82% 

42 

ESB3 

1 

1 

1 

1 

9 

9 

10 

32 

85% 

36 

ESB1 

1 

1 

1 

1 

9 

9 

10 

32 

83% 

34 


Table 25: Low-risk buffer sizing for gyroscope representing high-demand 

part 


010632830 

Fobl 

Buffer 

Fob2 

Buffer 

Fob3 

Buffer 

Fob4 

Buffer 

Mobl 

Buffer 

Mob2 

Buffer 

ESB 

Buffer 

Total 

Buffer 

PackUp 

Effect 

PartShort 

Medium Risk 











ESBO 

1 

1 

1 

1 

10 

10 


24 

87% 

134 

ESB5 

0 

0 

0 

0 

8 

8 

8 

24 

62% 

93 

ESB3 

0 

0 

0 

0 

8 

8 

8 

24 

60% 

91 

ESB1 

0 

0 

0 

0 

6 

6 

12 

24 

63% 

81 


Table 26: Medium-risk buffer sizing for gyroscope representing high- 

demand part 


010632830 

Fobl 

Buffer 

Fob2 

Buffer 

Fob3 

Buffer 

Fob4 

Buffer 

Mobl 

Buffer 

Mob2 

Buffer 

ESB 

Buffer 

Total 

Buffer 

PackUp 

Effect 

PartShort 

High Risk 











ESBO 

0 

0 

0 

0 

7 

9 


16 

74% 

310 

ESB5 

0 

0 

0 

0 

5 

5 

6 

16 

49% 

290 

ESB3 

0 

0 

0 

0 

5 

5 

6 

16 

48% 

216 

ESB1 

0 

0 

0 

0 

4 

4 

8 

16 

53% 

186 


Table 27: High-risk buffer sizing for gyroscope representing high-demand 

part 


Tables 28, 29 and 30 contain results for the gearbox representing a low-demand 
part. It experiences approximately equal PartShort using ESB5, ESB3 and ESB1. ESBO 
experiences approximately twice as many PartShort as the other three configurations. 


58 




However, ESBO has significantly higher PackUp Effectiveness than the other three 
configurations. ESB5, ESB3 and ESB1 have approximately equal PackUp Effectiveness 
for each risk level. 


014117040 

Fobl 

Buffer 

Fob2 

Buffer 

Fob3 

Buffer 

Fob4 

Buffer 

Mobl 

Buffer 

Mob2 

Buffer 

ESB 

Buffer 

Total 

Buffer 

PackUp 

Effect 

PartShort 

Low Risk 











ESBO 

0 

0 

0 

0 

1 

2 


3 

51 % 

20 

ESB5 

0 

0 

0 

0 

1 

1 

1 

3 

21 % 

11 

ESB3 

0 

0 

0 

0 

1 

1 

1 

3 

23 % 

10 

ESB1 

0 

0 

0 

0 

1 

1 

1 

3 

25 % 

9 


Table 28: Low-risk buffer sizing for gearbox assembly representing low- 

demand part 


014117040 

Fobl 

Buffer 

Fob2 

Buffer 

Fob3 

Buffer 

Fob4 

Buffer 

Mobl 

Buffer 

Mob2 

Buffer 

ESB 

Buffer 

Total 

Buffer 

PackUp 

Effect 

PartShort 

Med Risk 











ESBO 

0 

0 

0 

0 

1 

1 


2 

55 % 

24 

ESB5 

0 

0 

0 

0 

0 

1 

1 

2 

11% 

32 

ESB3 

0 

0 

0 

0 

0 

1 

1 

2 

12 % 

27 

ESB1 

0 

0 

0 

0 

0 

1 

1 

2 

12 % 

17 


Table 29: Medium-risk buffer sizing for gearbox assembly representing low- 

demand part 


014117040 

Fobl 

Buffer 

Fob2 

Buffer 

Fob3 

Buffer 

Fob4 

Buffer 

Mobl 

Buffer 

Mob2 

Buffer 

ESB 

Buffer 

Total 

Buffer 

PackUp 

Effect 

PartShort 

High Risk 











ESBO 

0 

0 

0 

0 

0 

1 


1 

28 % 

112 

ESB5 

0 

0 

0 

0 

0 

0 

1 

1 

0% 

53 

ESB3 

0 

0 

0 

0 

0 

0 

1 

1 

0% 

44 

ESB1 

0 

0 

0 

0 

0 

0 

1 

1 

0% 

26 


Table 30: High-risk buffer sizing for gearbox assembly representing low- 

demand part 


59 




THIS PAGE INTENTIONALLY LEFT BLANK 


60 



V. CONCLUSIONS 


A. THESIS OBJECTIVES 

This thesis is aimed at assisting Marine Corps aviation logistical planners to 
efficiently design the support network and allocate resources for deployed operations. It 
develops an inter-temporal network (ITN) simulation model that measures the operational 
effectiveness of four network configurations with respect to four inventory buffer sizes 
obtained using the same methodology as ELAT. The buffer sizes correspond to four risk 
levels—no, low, medium and high. Two MOEs are used: PackUp Effectiveness, which 
represents the percentage of demands satisfied on the day demanded and PartShort, 
which represents the magnitude and duration of unsatisfied demands during a given 
period. 

B. EVALUATING ALTERNATIVE NETWORK CONFIGURATIONS AND 

ALLOCATING SPARE PARTS 

The main takeaways from this research are: 

(1) Operational planners must prioritize desirable network properties. There are 
no dominating network configurations for all risk levels with respect to the selected 
MOEs. 

(2) If the objective is a high level of attainability at each node—measured by 
100% PackUp Effectiveness and 0 PartShort—then no-risk inventory buffer sizing is 
required. No-risk inventory buffers will have a significantly greater range and depth than 
other risk level inventory buffers. For no-risk inventory buffer sizes, ESBO dominates all 
other alternative network configurations. 

(3) If operational flexibility or survivability, which seek smaller logistical 
footprint at the demand nodes, is a high priority, planners should be willing to accept less 
than maximum effectiveness to reduce the logistical support footprint at each node. In 
this situation, low-, medium- or high-risk level buffer sizing are appropriate. 


61 



(4) Using low-, medium- or high-risk buffer sizing levels, planners must prioritize 
MOEs before designing the support network. PackUp Effectiveness and PartShort do not 
necessarily improve in tandem as a result of changing the design of the support network. 
A system that increases the percentage of demands satisfied on the day demanded does 
not necessarily improve the response time for unsatisfied demand. Considering each one 
of these two MOEs may result in different network configurations. 

(5) For all non-zero risk levels, if the objective is to maximize PackUp 
Effectiveness, ESBO is the dominating alternative. 

(6) For all non-zero risk levels, if the objective is to minimize PartShort, ESB1 is 
the dominating alternative. Also, any network with an ESB less than five days TRR from 
the demand nodes will outperform ESBO. 

(7) If the ESB is included, performance improves as the TRR between the ESB 
and the demand nodes is minimized. 

C. FUTURE EXTENSIONS 

1. In this study we used a single value to estimate the TRR between nodes 
because TRR data is unavailable. In reality, TRR is a random variable that may be 
subject to significant variance. Once data is collected and becomes available, the TRR 
should be incorporated in the model along its estimated probability distribution. 

2. The model presented in this thesis was limited to spare parts. It can be 
extended to include other resources such as transportation, repair capability, mobile 
facilities, and manpower. 

3. The model used in this thesis considers spare parts individually. It can be 
extended to consider demands for multiple parts and model their interactions. 

4. The original ITN model is an optimization model. The model used in this 
thesis can be extended to produce the optimal allocation of spare parts that produce 
objective functions that maximize selected MOEs while meeting constraints on the total 
number of spare parts dedicated to the logistics support network. 


62 



LIST OF REFERENCES 


ASL, Headquarters, US Marine Corps. 2004. The Marine Aviation Logistics Support 
Program II Draft Concept Document. 


Cook, Wade and Moshe Kress. 1992. Ordinal information and preference structures. 
New Jersey: Prentice Hall Inc. 

Delaporte, Murielle, and Robbin Laird. 2007. Recrafting expeditionary logistics: USMC 
Aviation prepares for the future. Military Logistics International 2, 6: 4-7. 

Eck, Laurin. Discussion with the author. 6 February 2009. 

Garant, Pierre. 2006. The transformation of Marine Aviation Logistics. The TOC TIMES'. 
9-15. 


Kress, Moshe. 2002. Operational Logistics: The Art and Science of Sustaining Military 
Operations. Boston: Kluwer Academic Publishers. 

Office of the Chief of Naval Operations. 2001. OPNAVINST 4790.2H Volume 1. 

Simchi-Levi, David, Edith Simchi-Levi, and Philip Kaminsky. 2000. Designing and 
Managing the Supply Chain. San Francisco: Irwin McGraw-Hill. 

Steward, Douglas. 2008. Pushing a pull system: Transforming Marine Aviation Logistics. 
Defense AT&L: 40-43. 


63 



THIS PAGE INTENTIONALLY LEFT BLANK 


64 



INITIAL DISTRIBUTION LIST 


1. Defense Technical Information Center 
Ft. Bel voir, Virginia 

2. Dudley Knox Library 
Naval Postgraduate School 
Monterey, California 

3. Commanding General, Training and Education Command 
MCCDC, Code C46 

Quantico, Virginia 

4. Director, Marine Corps Research Center 
MCCDC, Code C40RC 

Quantico, Virginia 

5. Marine Corps Tactical Systems Support Activity 
(Attn: Operations Officer) 

Camp Pendleton, California 

6. Director, Operations Analysis Division 
Code Cl9, MCCDC 

Quantico, Virginia 

7. Marine Corps Representative 
Naval Postgraduate School 
Monterey, California 

8. MCCDC OAD Liaison to Operations Research Department 
Naval Postgraduate School 

Monterey, California 

9. Naval Air Systems Command 
Norfolk, Virginia 


65