The Dynamics of Multi-Tier, Multi-C hannel
Supply Chains for High-Value Government Aviation Parts
Abstract
Multi-tier, multi-channel supply chains are now common in many industries including
aviation. Such supply chains provide high-value aviation parts to the Government, and
many have been plagued recently by shortages. A system dynamics model has been
developed of an aviation supply chain producing a major sub-assembly composed of
eight components, each component coming from a three tier supply chain. These
components are used in new production as well as overhaul of damaged parts. It was
found that in the face of varying demands substantial bullwhip was produced and that it
became especially pronounced at the lower levels of the supply chain. Moreover, it was
shown that the government ordering process is extremely sensitive to common data
errors such as the production lead-time and that production constraints, not included in
the ordering algorithms, created deep and prolonged shortages. Ongoing research is
developing improvements to the formulation of the ordering process and developing
optimum inventory strategies for creating push-pull boundaries within the manufacturing
process.
Introduction
Manufacturing has changed. Companies that were once known as automakers or
aircraft manufacturers are now more properly viewed as integrators or assemblers.
Parts and major sub-assemblies are now out-sourced and are planned to arrive just in
time at the assembly plant for integration into cars, airplanes and other major products.
Consider, for example, the new Boeing 787 Dreamliner. The wing comes from J apan,
the movable trailing edge of the wing is produced in Australia, the fixed and movable
leading edge of the wing is produced in Oklahoma, the wing tips are produced in Korea,
the center fuselage is made in Italy, the landing gear is made in the UK, and the landing
gear doors are made in Canada. (Avery, 2007) Automakers and electronic equipment
manufacturers have similar extended supply chains. Sub-assemblies and major
components come from a vast geographic network that is both broad and deep.
In these supply chains, major sub-assemblies are shipped to the OEM by hundreds of
first tier suppliers, but these first-tier companies are just the tip of the supply chain
iceberg. For each major component or sub-assembly, there is a multi-tier supply chain
that may extend back, for example, from a first tier precision machining company, to a
second tier casting company to a third tier raw material provider. Moreover, each major
sub-assembly such as a transmission or landing gear is made from multiple parts, each
provided by a separate channel through a multi-tiered supply chain. Thus, most, if not
all, major sub-assemblies are the product of a multi-tier, multi-channel supply chain. It
is important to note that in the supply chains for government aviation parts, overhaul is a
major source of supply. When damaged parts are returned for overhaul, they require
some of the components from the multi-channel, multi-tiered supply chains. Overhaul
thus creates demands in addition to those of the new production process. Shortages of
components thus affect both new production and overhaul of high-value aviation parts.
Performance problems often arise in the lower tiers of these supply chains. For
example, during 2004, the lead-time for both aerospace steels and titanium grew from
roughly three months to over a year. Lead-time for titanium continued to grow and
reached roughly seventy weeks in 2005 and 2006. These developments threw the
supply chains for aviation assemblies such as transmissions, landing gears, etc. into
disarray. Somewhat similarly, the resurgence of the aviation industry has led to growth
in orders that exceeded the production capacity of many lower tier suppliers, and
backorders are often common. For example, demands for aerospace fasteners today
exceed production capacities. As a result of raw material delays and capacity
constraints, inventories of many high-value spare parts for government aviation have
declined to very low levels and have had difficulty recovering. Similarly, supply chain
issues of one type or another have delayed both the Airbus 380 and the Boeing 787.
While numerous studies have suggested a reformation of the Government supply
process that was implemented decades ago, Gansler and Luby (2003), Abramson and
Harris (2004), and Folkeson and Brauner (2005), the same underlying process and
associated problems tend to plague the system in place today.
A research program was initiated to investigate the dynamics of multi-tier, multi-channel
supply chains providing high value aviation parts to the Government. The objectives
were to: examine the impacts of the Government ordering process under a variety of
time-varying demand conditions; assess the impacts on supply chain performance of
inaccurate data in the calculation of the recommended buys and overhaul; examine the
bull-whip effect in the multi-tiered, multi-channel supply chain; assess the potential for
cross-coupling of problems among the multiple channels; and examine supply chain
performance in the face of production capacity constraints not included in the supply
requirements determination process of the government.
Analytical Approach
System Dynamics is an appropriate technique for analyzing complex multi-tier, multi-
channel supply chains. System Dynamics has been used to analyze supply chains from
its very beginning as a modeling and simulation tool for policy analysis. Forrester's
(1958) groundbreaking article in the Harvard Business Review demonstrated
fundamental supply chain dynamic behavior such as how small changes in retail sales
and promotional activity can lead to large swings in factory production, i.e., the so-called
bullwhip or Forrester effect. Forrester (1961) also included a supply chain model and
demonstrated various modes of behavior. Forrester’s models included factory,
distribution and retail tiers in the supply chain but no suppliers to the factory. More
recently, Sterman (2000) has addressed supply chains with several models and case
studies. Again these are forward looking supply chains from factory to customer with
perhaps a single supplier. Huang and Wang (2007) addressed the bullwhip effect in a
closed loop supply chain using a simple model based on Sterman’s (2000) structure.
Simchi-Levi (2008) and Lee (1997) address bullwhip from an analytical perspective.
Schroeter and Spengler (2005) addressed the strategic management of spare parts in
closed-loop supply chains. Angerhofer (2000) presents a thorough discussion of
system dynamics modeling in supply chain management. Killingsworth, Chavez, and
Martin (2008) address the government ordering process within a system dynamics
model but does not include an extended supply chain. The intent of the current research
is to capture the actual algorithms of a government procurement process, embed this
procurement or ordering process within a system dynamics supply chain model
incorporating multiple tiers of suppliers and multiple channels of components, and
assess the impacts and performance of the extended enterprise supply chain.
Model Description
The overall supply chain system providing high-value aviation spare parts is shown in
overview in Figure 1. This supply chain extends from raw material to final customer.
Demand arises from aircraft located in four regions of the world. Demand in each
region is driven by the number of aircraft in the region, monthly flight hours, and failure
rate per part per flight hour. Each region has an inventory of key spare parts, and these
inventories are replenished from a central distribution inventory. Supply of parts comes
from three sources: production of new items, commercial overhaul of damaged parts,
and government depot overhaul of damaged parts. Each type of production requires
that a number of parts be integrated into the major sub-assembly. In general, the
overhaul process requires fewer component parts than new part production. The
i Demand Info
peee eee e eee ee Ordering Process |gi= 2-22 2L-- 4
Supplier Tiers ' a H
3.2 1 ’ — ‘ '
Se ¥ C Tavern Parts in| ‘
Component One { HH Ly, Benet >| Use at |p: i
LI New N Region A] + H
Component Two | A Ly} Spares > T p :
— ‘ invents Parts in| +
Component Three | 4 >| Ls > | seat et cae!
— Commercial Regmae)
componenttour {HH Qvertaul |-——>
— N Parts in | “J
Component Five | inn >| V [pf Inventory} —_» | Use at |p!
-— Depot E c Region C] +
Componentsix | i Lp Overhaul A ‘
— 0 Parts in] |
Component Seven ; ili Rd nventory| _» | use at |»!
— Y D RegionD] +
Componenteight i }» 1 |
Returns for Overhaul
Figure 1: Overview of the Multi-Tiered, Multi-C hanneled Supply Chain Model
component parts are each produced through a three-tier supply chain. Each of these
chains typically has a different manufacturing time at each tier, and each channel has a
different total production time. The overall supply process is managed in a feedback
fashion by the government's ordering or requirements determination process. This
process is at the heart of many government and defense supply chains for high-value
parts. (Rosenman, 1964) This computerized process is used to determine the
recommended buys for new parts and the recommended number of parts to undergo
repair and overhaul. The supply chain control system compares current levels of
inventory, including due-ins and due-outs, with anticipated needs to calculate
recommended buys and repairs. Since the procurement of new spares and the
overhaul of damaged spares leads over time to changes in inventory, the system truly
functions in a feedback control fashion to manage the supply chain. (Killingsworth,
Chavez, and Martin, 2008)
Figure 2 provides a more detailed view of the flows present in the model. It is important
to note that many, if not the vast majority, of these aviation supply chains for high-value
government spare parts operate in a sequential fashion, with little information sharing
and little risk taking within the supply chains. For example, the government will request
a proposal from the OEM to provide a certain number of the major assemblies. The
OEM will respond with a proposal, and after negotiation, will be awarded a contract.
The OEM will then request a proposal from the first tier suppliers to provide their
components. A proposal will be submitted, negotiated, and the OEM will award a
contract to the first tier suppliers. These first tier suppliers will then turn to the second
tiers and repeat the same process. The second tiers will only then place an order with
the third tiers for, in many cases, the necessary raw material. Hence many months can
go by before the order for raw materials is placed, and recently, many months then go
by before the raw material is received. This sequential structure is built within the
model. Moreover, the first, second and third tier suppliers are very risk averse and
maintain essentially zero inventory of both their inputs and outputs. Purchasing of
inputs and production of output only occurs in the presence of a contract or purchase
order.
Several levels of calculation are incorporated into the Supply Chain Control Center to
determine recommended buys and repairs. (Killingsworth, Chavez, and Martin, 2008)
This computerized requirements determination process is embedded in many
government supply databases. Within the determination process, the recommended
procurement action for new spare parts is calculated by taking the difference between
the procurement reorder point and the total available net assets, and then adding the
procurement cycle requirement, the inventory necessary to meet demands until the next
scheduled order (see Figure 3). Total net assets are calculated from due-ins from
procurement and repair plus inventories, less due-outs. The procurement reorder point
is based on reserves and safety levels. Orders that are placed with the OEM enter
production subject to a maximum production rate and availability of all of the required
components. Production is completed after a manufacturing lead time. These parts
then flow into serviceable inventory.
PROGRAM DEMANDS
Component One Supply Chain supply (js —“—<
Component cuem——— pl SHAN ge ae > |
Ter —— 4 CENTER rot
Orders
[eee ry
— Aon ano
Component Two Supply Chain rl i WV 4 REGION A rot
Component 2_| 1. i 1 > i
[res (Pie Len] ad 18 1 5 Inventory inal Aircraft 4
Ores = | ! i} | Ma Ca
| 1 f I 1
‘Component Three Supply Chain x I Orders 1 all I 1
ee ee) eee 1 re recon |!
Mem [8] components Gee i |!
lt \e ‘a Inventory bl Alrcraft t 1
A i) —| !
Component Four Supply Chain = -{ _ Newspares a
Component |] Orders for Pitt Da
| Components itl i 1
oes | 0 1 rial! REGION C yy
Ml I owes | || est ii
6 1 r/o: toventory Ho[Aircrat iy |
Component 5 | i F| le T
NT components tty r{t
Orders | N ial lb 1
er eee i]t
Orders for r vial REGION D !
Componente ]P| Components | pleco pl i
Lees fel Keen =a lal | ordess ! |p) invention, eater
P| 4-| !
Orders T 1 viel /
Si rR
‘Component Seven Supply Chain Companenes: Hl rt! 1
Component ie! fs] 1
== == {_Depotovernau f opt | | 1
Orders Orders for — f i
Components i
Component Eight Supply Chain »
Component Wholesale
= = = > Lqg-{| Maintenance He] Programs
Ter a =) fee = = pena
Return to Overhaul
Figure 2: Detailed Overview of Model Flows
New Spares [ug
wip. Production
New Spares
Orders
Awaiting
iProduction)*
Start
Completion Rate Store nate
—* Procurement
ae
_—
Total Net
—
Assets
Recommended
Procurement Action
~—
Procurement
Reorder Point
—— Due In From serviceable
Inventory
Unserviceable
Inventory
__Due In From
Repair
Due Out
Reserves
~~ Below Depot
—Requirement
__ Repair Cycle
Requirements
/
Safety Level
Requirements
Production
Availability of All Input
Lead Time i i
Components at Prime
Maximum
Production Rate
ALT
Requirements
PLT
Requirements
Procurement Cycle
Requirement
Assumed
Production Lead
Time
Figure 3: Recommended New Spares Procurement Action
In a separate calculation, the recommended repair action is determined. It must be
noted that repair and overhaul can only be conducted if there is a damaged part
available to be overhauled. The maximum recommended repair action is calculated by
subtracting the assets available for repair, including overhaul and procurement work-in-
progress less due-outs, from the repair action point, calculated with reserve levels and
safety requirements. This repair action point is largely driven by historical demands.
The maximum recommended repair action, however, is then limited by the
unserviceable inventory on hand (see Figure 4). The potentially constrained repair
order is allocated between government depot and commercial overhaul according to
capacity levels at each location. The overhaul rates may be limited by production
capacity levels. As inventory is repaired, it is shipped to serviceable inventory available
for issue.
Once orders are placed for overhaul and new production, orders are then placed with
the first tier suppliers for the components necessary to assemble the final product.
Figure 5 illustrates the model structure for commercial overhaul. Similar structures exist
for depot overhaul and new spare production. It is important to note that the overhaul
process can only begin if necessary components parts are on hand.
4
Comm Ovhi Orders
Depot ovnt orders "iaiing start
‘waiting Start
a
| Depot Ovht
wie
“ALT ses
Figure 5: Order Placement Process to Supply Chain Tiers
The orders that originate in Figure 5 at the OEM or overhaul sites flow to the first tier
suppliers. These suppliers then place orders with the second tier suppliers, who, in
turn, then place orders with the third tier suppliers as shown in Figure 6. Second tier
production can only begin if there is inventory available from the first tier such as the
raw material. Similarly, production at the first tier can only begin if there is inventory of
output from the second tier. Each tier is dependent upon the previous one to complete
the process, and each level may be limited by a production lead time of another
supplier.
As the components are shipped to new production, commercial overhaul or government
depot overhaul, they flow into inventories at these sites. The total availability of
components at these locations is determined by the minimum inventory level (see
Figure 7). This availability value then becomes a factor in the determination of the
production and overhaul start rates, seen previously in Figures 3 and 5.
Upon completion of orders through the overhaul and procurement processes, the
products are shipped to the central inventory site and then to one of four regional
inventories. The parts are pulled from these inventories and placed into service on an
aircraft. The damaged or worn parts that are removed are returned for repair, less a
percentage that are scrapped or not returned by the field. The returned parts enter the
unserviceable inventory on-hand and then enter the overhaul process.
Analysis and Simulation Results
Key objectives of the analysis were to: (i) assess the performance of the government's
requirements determination process in the presence of a multi-channel, multi-tier supply
chain; (ii) evaluate the likelihood of the bull-whip effect being produced in the supply
chain and the impact on lower tier suppliers; (iii) determine the sensitivity of the supply
control to inaccurate data and (iv) evaluate impacts of real-world production and
overhaul capacity constraints. The model described has been parameterized for
specific high value parts and has been used to simulate the behavior and performance
of the requirements determination process and the supply chain for these particular
parts. The following cases are presented with a simulation time covering 2001-2012:
Case 1: Constant demand;
Case 2: Step up in demand in 2003;
Case 3: Oscillating demand;
Case 4: Error in PLT resulting from an increase in PLT at a raw material supplier.
Data error persists for one year; and
Case 5: Production constraint limits production at a first tier supplier.
Cases 1 and 2 were used both in the validation of the model and to verify that the
requirements determination process generated appropriate new procurement and
overhaul orders in response to constant demand and a step-up in demand. These two
cases also enabled assessment of the behavior of the ordering and production process
in the multi-tier, multi-channel supplier network. Case 3 was conducted to determine
whether the governmental computerized ordering process and related supply chain
exhibited the bullwhip effect and to examine the impacts on the lower tier suppliers.
Because numerous reports, for example, GAO (1981) and GAO (July 2007), have
indicated that certain data, such as production lead-time, used in the ordering process is
often incorrect, Case 4 investigates the impact of incorrect production lead-time on the
ordering process and supply chain performance. Moreover, because the governmental
3UC WEUOCMdWOD JO} sie1 UIeEYD Ajddns :9 aun6i4
sonia
wou nor2)o0 aw
a ee PH perouersuupr
smoiciag” ——> 802474 sweccie
aay eeu) ————_ N\ —
5 sathossa eae
20002 ca oN
neuro
sora
15124 30) ——~ sup sey ue ‘ on aire er
epscietngg:-— een Pan] sever en vemsum agp ester 2p0 voxauey one a
= suopsauisig —_sau.pu0295 2409 Jars e| “eu pu029SsiC0 Iwommi men aD a seam] aewutor apo
ee al > # veumien wey aucD
sends way 7 =
aves —.
sien wey 2009
-2j9u, 6500
/
\ | /
it! \ \
ioupumee cba eH 9] Sa. eve
aeyou neve areas] nan] onan one
= Jonenan ye
ope fo 884s 109
DyRIRRH Me
oni Buus
Shipping Delay C1 Commovht
to CommOvhi Prod Start
) [Taventory |
C1 Parts to of C1
Commovhi > Receipt Rate of C1 | Inputs at /—use Rate of Cl
Inputs at Commovhi
Commovhi
Shipping Delay C2 Prod Start
i
to CommOvhl
Taventory
C2 Parts te f Hie
arts to_, Receipt Rate of C2 | inputs at
Commovh!
Inputs at Commovh! [CommOvh!
Shipping Delay C3 Commovhl
to CommOvhl Prod Start
} Taventory
th of C3
Inputs at
caParsto_, “Receipthate of Ce [eta
CommoOvhl "inputs at Commovhi Inputs at
commovht
Shipping Delay Ca ape
to CommOvhi
Taventory Availability of All Input
of C4 ——e aie Components at
C4 Parts to, ~Receipt Rate of ca | MPULS at TTS ate once CommOvh!
Commovhl~™ inputs at CommOvh!
Inputs at CommOvhl
Commovhl
Prod Start
Shipping Delay C5
to CommOvhl
Tnventory |
C5 Parts t etcess
arts to Receipt Rate of C5 | Inputs at | Use nate of eo
CommOvh! Inputs at CommovalC2mrOvh!
Shipping Delay C6 Commovhi
to Commovhl Prod Start
Taventory
of C6
C6 Parts to__) ‘Receipt Rate of C6 | Inputs at,
CommOvhl inputs at Commovhi-C2mmOw
Shipping Delay C7
to CommOvhl
Taventory
ofc?
C7 Parts to_, Receipt Rate of C7 | Inputs at
CommOvhi Inputs at Commovni [CommOvht
Shipping Delay C8
to CommOvhl
Thventory
CB Parts to of C8
—m Receipt Rate of C8 | Inputs at
Use Rate of CB
Inputs at Commovhi [CommOvhl
Inputs at CommOvh!
Commovhl
Figure 7: Availability of Components for Assembly
10
Failure Rate Per
Monthly Useage
Months of Average Region A
Rec Demands
Monthly Useage
Per Part Region A
Recurring
Demands Region A
Average Recurring
Demand Region A
Backorders
| Desired inventory Tepe
Coverage Region A
| vesired inventory Region A rise —
| Regio A Inventory] installRate in. [\"A°"| Removal Rate
| j Region | Region
| | Region A Inv Adil / Proc/Transit Time \
\ | Factor ll arrivat hate w— 2 Region A \
\ neat a si svoN enon
Inventory and ~~ rogram Demands
Backorder Correction Shipping Rate to Region A
Region A Region A /
| wae
Desired [rransit tof~#—¥—=)
Shipments to Regen a] SR to Region
Region A A
Be ‘Average Program
Months of Average ———____ Demand Region A
Program Demands — =
Figure 8: Region A Demands and Returns
ordering process does not include the potential for production capacity constraints,
Case 5 examines the behavior of the supply control process and the ability of the
system to meet rising demand in the presence of capacity constraints. Finally, Case 6
examines a “real world” scenario involving shifting demand, production constraints, and
data errors.
Case 1 assumes constant demand of 14 units a month (divided between the four
regions - Region A (7 units/mo), Region B (5 units/mo), Region C (1 unit/mo), and
Region D (1 unit/mo)) as depicted in Figure 9. Other key assumptions include no limit
on production or overhaul rates, an overall production lead time of 22 months, and a
repair lead time of 11 months. The overall production lead time is calculated as the
maximum lead time of the eight components plus the production lead time and
administrative lead time at the primary supplier. The lead time of each component, that
is, of each channel, is determined to be the sum of the shipping delays, the manufacture
lead times, and the administrative lead times for each tier. The first four cases assume
that four components are used for new spare production only and have a common
overall lead time of 12.2 months. The other four components are used for both overhaul
and new spare production and have a common overall lead time of 8.2 months. For
new spare production, the OEM requires 9.8 months for assembly and integration
resulting in the 22 month overall PLT used in the requirements determination process.
For overhaul, the depot and commercial overhaul facility require 2.8 months for
integration and assembly yielding the 11 month overall RLT used in the requirements
determination process. The assumed PLT and RLT for the ordering determination
process are equivalent to the overall actual values in this case.
D1,
The simulation output from Case 1 is presented in Figures 9-15. Figure 9 shows the
constant input demands. Figure 10 shows that the Central and Regional Inventory
levels remain constant. Figure 11 shows that removals, shipments to regional
inventories and production and overhaul rates are constant. The system establishes an
equilibrium that is maintained throughout the simulation. Figure 12 presents the
availability of component inventory at the OEM and the two overhaul sites. The
somewhat surprising oscillation in the available components at Prime stems from the
procurement process. Figure 13 shows that when the Total Net Assets dip below the
Procurement Reorder Point, a procurement action occurs for a period of time. The
procurement action ceases because the procurement action leads to orders that create
due-ins increasing Total Net Assets. Once a procurement action is initiated, an order to
the supply chain tiers follows as shown in Figure 14. Figure 15 shows how these
pulsing orders thus create highly variable input inventory at the OEM’s even in the face
of constant demands. The repair process is more stable because repair actions are
limited by the requirement for damaged or worn parts that maintain a constant flow due
to constant removals. This stability is shown in Figure 14.
Constant Demand
8
4
0
2001 2003 ©2005 = 2007.-—S «2009S «2011S 2013
Year
iii ena .-—____. Region C Demands. —————______—_
Region B Demands. ————________ Region D Demands. ——________
Figure 9: Constant Demand Levels
Inventories with Constant Demand
80
40
0
2001 2003 2005 2007 2009 2011 2013
Year
Unserviceable Inventory. 4 Region B Inventory,
Serviceable Inventory. ————————__ Region C Inventory
Region A Inventory, ———4H4HHH4H4H4H4H Region D Inventory
Figure 10: Inventories with Constant Demand
12
Key Rates with Constant Demand
20
10
0
2001 2003 2005 2007 2009 2011 2013
Year
Total Removals
Depot Overhaul Completion Rate
Commerial Overhaul Completion Rate
Prime New Spares Completion Rate
Total Shipment Rate to Regions
Figure 11: Key Rates with Constant Demand
Availability of Components with Constant Demand
20
10
PAPAL PIP APL PP PPPPPPYI PAD AAA
EAYAADDADDRADADA ADDADRDDAADINVV
0
2001 2003 2005 2007 2009 2011 2013
Year
Availability of All Input Components at Commmercial Overhaul
Availability of All Input Components at Depot
Availability of All Input Components at Prime
Figure 12: Availability of Components for the Overhaul and Production Processes
with Constant Demand
Procurement Action with Constant Demand
400
200
OOD T PADI
2001 2003 2005 2007 2009 2011 2013
Year
Total Net Assets
Procurement Reorder Point
Procurement A ction
Figure 13: Procurement Action with Constant Demand
13
Order Rate to Suppliers with Constant Demand
40
0 RECS
ATAVAVAUIAVEY! LVATAUAUAVAVAUAUEY! MJ vat
2001 2003 2005 2007 2009 2011 2013
Year
Commercial Overhaul Onder Rate to Suppliers
Depot Overhaul Order Rate to Suppliers
Prime Order Rate to Suppliers
Figure 14: Order Rate of Components for Overhaul and Procurement Processes
with Constant Demand
Inventory of Component One at Prime
20
10
0
2001 2003 2005 2007 2009 2011 2013
Inventory of Component One at Prime
Receipt Rate of Component One at Prime
Use Rate of Component One at Prime
Figure 15: Inventory of Component 1 at OEM with Constant Demand
Case 2 assumes a Step increase in demand from 14 to 18 parts a month in 2003 (each
region increasing demand by 1 unit per month), while holding all other assumptions the
same as the steady state case. This case illustrates the typical growth in demand that
has been seen for aviation parts within the past five years and also provides a basic
determination of the recovery time of the system after a disturbance from equilibrium
(Forrester, 1961). Figure 16 illustrates the input demands, and Figure 17 presents the
serious impacts on the central and regional inventories. As may be seen, the central
inventory is depleted for a period of nearly two years. This shortage occurs because of
the long lead times in production and overhaul of high-value aviation spare parts. As
may be seen in Figure 18, the new production rate increases in response to the higher
demands, but overhaul rates are constrained by the number of unserviceable items on
hand. Component availability presented in Figure 19 slowly grows to support higher
levels of procurement and overhaul but does not grow fast enough to enable production
to halt the depletion of inventory. Since the Procurement Reorder Point is largely
14
determined by a twenty-four month average demand, the procurement process does not
respond to the higher demand for at least a year. As may be seen in Figures 20 and
21, as the Reorder Point begins to increase in 2004 and 2005, it causes larger and
more frequent orders for new spares. Since raw material is ordered last but used first in
the supply chain, the increase in orders causes the raw material inventory at tier two to
be quickly depleted and remain at very low levels (see Figure 22). The repair process is
once again limited by the unserviceable inventory on-hand, as depicted in Figure 23, in
which the repair action is equal to the unserviceable inventory, while the desired
maximum repair is significantly higher. This constraint on the repair action therefore
dampens the number of components utilized in overhaul and keeps this process
relatively stable.
Demand with Step Increase in 2003
10
—|
5 |
0 |!
2001 2003 2005 2007 2009 2011 2013
Year
ies -———__. Region C Demands. —————______—_
Region B Demands. ——————————_ Region D Demands
Figure 16: Demand Steps Up in 2003
Inventories w/Increase in Demand in 2003
80
40
0
2001 2003 2005 2007 2009 2011 2013
Year
Unserviceable Inventory. 4 Region B Inventory,
Serviceable Inventory. ———4H4HHH44H4H4HH Region C Inventory
Region A Inventory. ———44H4HH4H4H4H4HM Region D Inventory
Figure 17: Inventories with a Step Increase in Demand
15
Key Rates w/Increase in Demand in 2003
40
20
| |
0
2001 2003 2005 2007 2009 2011 2013
Year
Total Removals
Depot Overhaul Completion Rate
Commercial Overhaul Completion Rate
Prime New Spares Completion Rate
Total Shipment Rate to Regions
Figure 18: Key Rates with a Step Increase in Demand
Availability of Components w/Increase in Demand
20
10
oN RADI PIN
0
2001 2003 2005 2007 2009 2011 2013
Year
Availability of All Input Components at Commmercial Overhaul
Availability of All Input Components at Depot
Availability of All Input Components at Prime
Figure 19: Availability of Components for the Overhaul and Production Processes
with a Step Increase in Demand
Procurement Action w/Increase in Demand in 2003
400 Ll ae |
200
CO Ce
2001 2003 2005 2007 2009 2011 2013
Year
Total Net Assets
Procurement Reorder Point
Procurement A ction
Figure 20: Procurement Action with a Step Increase in Demand
16
Order Rate to Suppliers w/Increase in Demand
40
20
Se TOE
9 u NAVAUAVAURVATAVAUAURUAURUINAY
2001 2003 2005 2007 2009 2011 2013
Year
Commercial Overhaul Onder Rate to Suppliers
Depot Overhaul Order Rate to Suppliers
Prime Order Rate to Suppliers
Figure 21: Order Rate of Components for Overhaul and Procurement Processes
with a Step Increase in Demand
Inventory of Raw Material at Tier 2
20
10
0
2001 2003 2005 2007 2009 2011 2013
Year
Inventory of Raw Material for Component One
Receipt Rate of Raw Material for Component One
Use Rate of Raw Material for Component One
Figure 22: Inventory of Raw Material for Component One used in Overhaul and
Procurement Processes with a Step Increase in Demand
Repair Action w/Increase in Demand in 2003
80
40
0)
2001 2003 2005 2007 2009 2011 2013
Year
RepairA ction
Unserviceable Inventory
Max Repair Action
Figure 23: Repair Action with a Step Increase in Demand
17
To examine the potential for bullwhip effect in this extended supply chain, Case 3
assumes a 420% sinusoidal oscillation in demand over a four year period. All the other
key assumptions from Case 1 remain the same. Figure 24 presents the input demand
assumptions and Figure 25 shows the considerable variation in inventories. The
production and overhaul rates fluctuate, but the bullwhip is not severe. This is because
of the sequential nature of the ordering process which does not include amplification but
simply passing orders along the chain and because overhaul variation is limited by
unserviceable inventory on-hand (see Figure 26). Availability of components is also
affected by these fluctuations in demands, producing spikes in inventory levels that vary
from the initial inventory level by as much as 90% (see Figure 27). This volatility is a
result of the irregular procurement action, shown in Figure 28, which then causes the
order rate of components to suppliers to fluctuate (see Figure 29). The order rate to
suppliers also affects the inventory of materials in the supply chain tiers. Since the raw
material supplier is the last to receive the order, this supplier is affected the most by the
instability of the orders, causing the inventory levels at the second tier to suffer.
Inventory levels grow as orders are placed, but almost completely diminish as orders
slow down and materials are used; as a result, significant fluctuations ranging by as
much as 100% of the initial inventory levels ensue (see Figure 30). Meanwhile, tiers
closer to the customer are able to maintain some level of inventory at all times, but this
level is still extremely unstable, varying again by about 100% for some short durations
(see Figure 31). This volatile nature is also present in the repair process, although the
extent of the variation in inventory levels is limited by the unserviceable inventory on-
hand. In this case, however, the repair action is affected by both the limited
unserviceable inventory at times and the recommended repair action at other times (see
Figure 32). The bullwhip effect is therefore still quite evident in the supply chain tiers,
as indicated in the component inventory levels at commercial overhaul in Figure 33.
Demand with +/-20% Oscillation over Four Y ears
NL NON
5 an ee
0 eee HS I
2001 2003 2005 2007 2009 2011 2013
Year
Region A Demands. —————________ Region C Demands
Region B Demands. ————_________ Region D Demands
Figure 24: Demand with a 20% Oscillation and a 4-year Oscillation Period
18
Inventories with Oscillating Demand
80
40
gam eS 7,
= Al ALO ALU
7
2001 2003 2005 2007 2009 2011 2013
Year
Unserviceable Inventory, —————____— Region B Inventory
Serviceable Inventory. —————_____— Region C Inventory
Region A Inventory, —————______— Region D Inventory
Figure 25: Inventories with a 20% Oscillation in Demand, 4-year Oscillation Period
Key Rates with Oscillating Demand
20
10
| ee
0)
2001 2003 2005 2007 2009 2011 2013
Year
Total Removals
Depot Overhaul Completion Rate
Commercial Overhaul Completion Rate
Prime New Spares Completion Rate
Total Shipment Rate to Regions
Figure 26: Key Rates with a 20% Oscillation in Demand, 4-year Oscillation Period
Availability of Components with Oscillating Demand
40
20
0
2001 2003 2005 2007 2009 2011 2013
Year
Availability of All Input Components at Commmercial Overhaul
Availability of All Input Components at Depot
Availability of All Input Components at Prime
Figure 27: Availability of Components for the Overhaul and Production Processes
with a 20% Oscillation in Demand and a 4-year Oscillation Period
19
Procurement Action with Oscillating Demand
400
panna nn Raph Pune. pp hh ind
200
O Lnnoodt | Done neoot Pepiontnd |
2001 2003 2005 2007 2009 2011 2013
Year
Total Net Assets
Procurement Reorder Point
Procurement A ction
Figure 28: Procurement Action with a 20% Oscillation in Demand, 4-year Period
Order Rate to Suppliers with Oscillating Demand
40
20
0 ARBRE
2001 2003 2005 2007 2009 2011 2013
Year
Commercial Overhaul Onder Rate to Suppliers
Depot Overhaul Order Rate to Suppliers
Prime Order Rate to Suppliers
Figure 29: Order Rate of Components for Overhaul and Procurement Processes
with a 20% Oscillation in Demand and a 4-year Oscillation Period
Inventory of Raw Material at Tier 2
20
10
0
2001 2003 2005 2007 2009 2011 2013
Year
Inventory of Raw Material for Component One
Receipt Rate of Raw Material for Component One
Use Rate of Raw Material for Component One
Figure 30: Inventory of Raw Material Component One used in Overhaul and
Procurement Processes with a 20% Oscillation in Demand and a 4-year
Oscillation Period
20
Inventory of Component One at Prime
40
20
0
2001 2003 2005 2007 2009 2011 2013
Year
Inventory of Component One at Prime
Receipt Rate of Component One at Prime
Use Rate of Component One at Prime
Figure 31: Inventory of Material used Procurement Process with a 20% Oscillation
in Demand and a 4-year Oscillation Period
Repair Action with Oscillating Demand
60
30
0
2001 2003 2005 2007 2009 2011 2013
Year
Repair A ction
Unserviceable Inventory
Max Repair Action
Figure 32: Repair Action with a 20% Oscillation in Demand 4-year Period
Inventory of Component Eight at Comm Ovhl
20
10
0
2001 2003 2005 2007 2009 2011 2013
Year
Inventory of Component Eight at Commercial Overhaul
Receipt Rate of Component Eight at Commercial Overhaul
Use Rate of Component Eight at Commercial Overhaul
Figure 33: Inventory of Material used in Overhaul Process with a 20% Oscillation
in Demand and a 4-year Oscillation Period
21
Assuming alternative periods for the oscillation in demand greatly affects the supply
chain performance (Figure 34). Increasing the period from 2 years to 4 years to 8 years
amplifies the bullwhip effect in serviceable inventory, as shown in Figure 35.
Demand with +/-20% Oscillation
20
PPS NN OES
WIN BAI SAASSST
10
0
2001 2003 2005 2007 2009 2011 2013
Year
‘Steady State Demands. ——H444H4H4H4H4H 4 Yr Period Demands
2 Yr Period Demands ————————__ 8 Yr Period Demands
Figure 34: Demand with a 20% Oscillation and a 2, 4, and 8-year Period
Total Serviceable Inventory with +/-20% Oscillation
200
tL ES
il aa
0
2001 2003 2005 2007 2009 2011 2013
Year
Steady State Serviceable Inventory
2Yr Period Serviceable Inventory
4 Yr Period Sevviceable Inventory
8 Yr Period Serviceable Inventory
Figure 35: Total Serviceable Inventory with a 20% Oscillation in Demand and a 2,
4, and 8-year Period
A series of simulations were conducted to develop a comparison of the extent of the
bullwhip effect on inventory levels, both at the prime supplier as well as the 3” tier raw
material supplier. The impacts of two variables were examined: the period of the
oscillation in demands and the averaging time used to calculate expected demands to
reflect future orders (see Table 1). The bullwhip effect is clearly evident from the
degree of the error in inventory on hand between the constant case and the oscillating
demand case in the raw material tier. The amount of the error in serviceable inventory
levels between these two cases is dependent upon the months of average demand.
The government standard of 24 months is most volatile with the longer oscillation
periods.
22
NO PRODUCTION CONSTRAINTS
20% variance in demand Months of Avg Demand
Period 6 12 24 P .
2 23% 23% 17% ercent error in
4 53% 43% 40% Serviceable
6 52% 48% 52% Inventory between
8 42% 48% 57% constant case and
10 37% 47% 57% variable demand
2 100% 100% 19% Percent error in
4 100% 100% 100% Raw Material
6 100% 100% 100%
8 100% 100% 100% Inventory between
constant case and
variable demand
Hw
0 100% 100% 100%
Table 1: Results of Sensitivity Analysis Varying the Period of Demand Oscillation
and the Months of Averaging Demand
Case 4 examines a problem that is frequently occurs in government supply chains for
high-value aviation spare parts. This is an error in assumed Production Lead Time
(PLT). In Case 4, demand begins at the constant level of 14 parts per month. In 2003,
demand ramps up over six months to 18 parts a month, increasing 1 unit per month in
each of the four regions. It is then assumed that demand ramps down to the original
level over a two year period beginning in mid-2009. At the beginning of the simulation,
the actual and assumed PLT are both equal to 22 months; in 2004, however, the
queuing time for the component eight raw material increases by 10 months, from 2 to 12
months. Component Eight is a necessary component for both new production and for
the overhaul process. This assumption reflects circumstances that occurred in 2004 as
lead times for raw materials increased dramatically. In the Case 4 simulation, the
assumed PLT remains at 22 months and RLT at 11 months for a year before these
values are adjusted in the requirements determination process to the actual overall
values of 32 months and 21 months, respectively. As a result of this error in the
calculations for recommended new buys and for overhaul, inventory levels drop for over
three years, creating a significant problem within the supply chain process (see Figure
36). Because of the error, the control system is assuming that it will receive deliveries
much more rapidly than it will. In other words, the requirements determination is
ordering too little too late because of this error. This is reflected in the very slowly
growing Prime New Spares Completion Rate in Figure 37. The recommended
procurement action is shown in Figure 38. An interesting dynamic occurs here. The
Reorder Point begins to rise thus generating orders and a growing Total Net Assets.
However, when the PLT error is corrected during 2005, it causes the Reorder Point to
increase and effectively reduce the gap between itself and the Total Net Assets. The
result is a counter-intuitive reduction in the recommended orders. This is shown in
Figure 39. This prolongs the problems with available inventory. Although the
recommended repair action reflects the same increase as the production order rate,
overhaul is once again limited by the amount of unserviceable inventory on-hand (see
Figure 40), and repair actions are much less than the desired maximum repair action.
23
Inventories with Error in PLT
100
50
ae
— a hd
0 are
2001 2003 2005 2007 2009 2011 2013
Year
Unserviceable Inventory
Serviceable Inventory
Region A Inventory
Region B Inventory
Region C Inventory
Region D Inventory
Figure 36: Inventories with a 22-32 Month Discrepancy in Assumed and Actual
PLT that Lasts for One Year
Key Rates with Error in PLT
40
20
|
ee Ol ce
0
2001 = =2003 =©2005 ~=2007- 3s 2009-—S 2011 2013
Year
Total Removals
Depot Overtiaul Completion Rate
Commercial Overhaul Completion Rate
Prime New Spares Completion Rate
Total Shipment Rate to Regions
Figure 37: Key Rates with a 22-32 Month Discrepancy in Assumed and Actual PLT
that Lasts for One Year
24
Procurement Action with Error in PLT
600
oo Perr eeeeeee
O [pono otooone onan edd inoiod i io
2001 2003 2005 2007 2009 2011 2013
Year
Total Net Assets
Procurement Reorder Point
Procurement A ction
Figure 38: Procurement Action with a 22-32 Month Discrepancy in Assumed and
Actual PLT that Lasts for One Year
Order Rate to Suppliers with Error in PLT
40
20
TOA
ON NAVATATAUAN Nata VAVATAULULUAU
2001 2003 2005 2007 2009 2011 2013
Commercial Overhaul Order Rate to Suppliers
Depot Overhaul Order Rate to Suppliers
Prime Order Rate to Suppliers
Figure 39: Order Rate of Components for Overhaul and Procurement Processes
with a 22-32 Month Discrepancy in Assumed and Actual PLT for One Year
Repair Action with Error in PLT
400
200
0
2001 2003 2005 2007 2009 2011 2013
Year
RepairA ction
Unserviceable Inventory
Max Repair Action
Figure 40: Repair Action with a 22-32 Month Discrepancy in Assumed and Actual
PLT that Lasts for One Year
25
Figures 41 and 42 show the sensitivity of the supply chain to changes in the queuing
time for raw material. The affects on serviceable inventory levels are notable. In
particular, the last case, in which the PLT increases by 9 months, causes almost a year
delay in the start of the recovery of the inventory levels in comparison to the case in
which there is no error. Similarly, and to a much greater extent, these delays affect the
raw material inventory levels at Tier 2. As may be seen in Figure 42, as would be
expected, the greater the PLT and the error, the longer the duration of depleted
inventory levels. During this time, backorders are growing at the tier 3 raw material
supplier, and inventory levels throughout the supply chain are greatly affected.
Serviceable Inventory with Error in PLT
200
100
0
2001 2003 2005 2007 2009 2011 2013
Year
No PLT Erro/Queuing Time 2 Months
PLT Enor of 3 Months/Queuing Time 5 Months
PLT Enor of 6 Months/Queing Time 8 Months
PLT Enorof9 Months/Queing Time 11 Months
Figure 41: Serviceable Inventory with Varying Errors in PLT Lasting for One Year
Inventory of Raw Material at Tier 2 with Error in PLT
20
10
0
2001 2003 2005 2007 2009 2011 2013
Year
No PLT Enor{Queuing Time 2 Months
PLT Enor of 3 Months/Queuing Time 5 Months
PLT Enor of 6 Months/Queing Time 8 Months
PLT Enor of 9 Months/Queing Time 11 Months
Figure 42: Inventory of Raw Material at Tier 2 with Varying Errors in PLT Lasting
for One Year
Case 5 examines another “real world” scenario that occurs frequently and that involves
production constraints. Production limitations or constraints are not included in the
algorithms of the requirements determination process used by many government
26
agencies. In reality, however, availability of tooling and labor do limit these processes.
Case 5 assumes a ramp up in demand in 2003 from 14 to 18 parts a month and a ramp
down in demand beginning in 2009 back to 14 parts a month. All other assumptions
from Case 1 remain the same, except however, component eight, which is necessary
for both overhaul and procurement processes, has a production constraint at the first
tier of 20 parts a month. Without this constraint, the first tier supplier of this component
should generally be producing up to 22 parts a month after the ramp up in demand.
Limiting this production by this small difference delays the start of the recovery of
serviceable inventory by about nine months, from late 2006 to mid-2007 (see Figure
43). Additionally, the inventory levels do not fully recover nearly as quickly as the case
in which there are no production constraints (see Case 4), and therefore do not reach
the level necessary to sustain the higher demand levels before demand ramps back
down.
Inventories with Production Constraints
80
40
6 S
2001 2003 2005 2007 2009 2011 2013
Year
Unserviceable Inventory. ————____—— Region B Inventory
Serviceable Inventory, ————————_— Region C Inventory
Region A Inventory, ————_______ Region D Inventory
Figure 43: Inventories with a 20/Month Production Limit on Component 8 at Tier 1
Figure 44 presents the results of a sensitivity analysis of the impacts of varying the
production limit on component 8 at Tier 1. In the unconstrained case, production is
roughly twenty-two units a month. Reducing the limitation to twenty or twenty-one
components per month causes a significant impact on the recovery of serviceable
inventory. Ata limit of 19 components a month, the serviceable inventory levels do not
begin to recover for approximately four years after the unconstrained simulation. Both
Case 4, incorporating the impacts of inaccurate data, and Case 5, assuming production
constraints, demonstrate the high sensitivity of this extended supply chain and the high
risk of performance problems. Many government supply chains balance on this knife-
edge where small changes can create major problems.
27
Serviceable Inventory
100
50
0 VIN
2001 2003 2005 2007 2009 2011 2013
Year
‘No Constraints
Constraint of 21 Component 8 Parts Per Month
Constraint of 20 Component 8 Parts Per Month
Constraint of 19 Component 8 Parts Per Month
Figure 44: Serviceable Inventory Levels with Varying Production Constraints on
Component 8 at Tier 1
Conclusions
Government supply chains for high-value aviation spare parts have experienced
considerable problems in assuring stable supply. Many of these problems are shown
to be the result of the government requirements determination process, the high
sensitivity to inaccurate data and production constraints, and the extended complexity of
the multi-tier, multi-channel supply chains. Several key findings have emerged:
(1) The bullwhip effect is strongly evident in the supply chain causing inventory
levels to vary greatly throughout all tiers of the supply chain;
(2) Because the ordering process determines recommended procurement and
overhaul actions by subtracting two large numbers, the ordering process and the
resulting supply chain performance are extremely sensitive to noise and
inaccurate data;
(3) Because the requirements determination process does not include the possibility
of production constraints, supply chain performance deteriorates rapidly in the
face of such constraints.
On-going research activities include using the system dynamics model for the following
initiatives:
1. Evaluating alternative, more stable, formulations for the ordering process;
2. Establishing critical push-pull inventories in the multi-tier, multi-channel supply
chain to enable more responsive performance by the extended supply chain; and
3. Investigating the impacts of collaborative planning and forecasting and
information sharing within the supply chain.
28
References
Abramsom, Mark A., and Roland S. Harris Ill, ed. The Procurement Revolution. New
York: Rowman and Littlefield Publishers, Inc., 2003.
Angerhofer, Bernard J. and Marios C. Angelides. “System Dynamics Modeling in Supply
Chain Management: Research Review.” Paper presented at the 2000 Winter Simulation
Conference, Orlando, FL, December 10-13, 2000.
Avery, Susan. “Boeing Executive Steven Schaffer is Named Supply Chain Manager of
the Year for the 787 Dreamliner Project.” Purchasing, Oct. 2007.
http://www.purchasing.com/article/CA6489135.html? q=DistributorHtechnology+can+rem
ove+costs Hrom+he+supply+chain+&q=boeing+suppliertcollaboration (accessed
March 26, 2008).
Folkeson, J ohn R. and Marygail K. Brauner. “Improving the Army's Management of
Reparable Spare Parts.” Santa Monica: RAND Corporation, 2005.
Forrester, J ay W. Industrial Dynamics. Cambridge: MIT Press, 1961.
Forrester, J ay W. “Industrial Dynamics: A Major Breakthrough for Decision Makers.”
Harvard Business Review (J uly-August 1958)
Gansler, J acques S., and Robert E. Luby J r., ed. Transforming Government Supply
Chain Management. New York: Rowman and Littlefield Publishers, Inc., 2004.
GAO. ‘The Army Should Improve Its Requirements Determination System.”
PLRD-82-19, December 1, 1981.
GAO. “Army Inventory: Army Annually Spends Millions to Keep Retention-Level
Stocks.” NSIAD-90-236, September 11, 1990.
GAO. “GAO: Inventory Management: DOD Can Build on Progress by Using Best
Practices for Reparable Parts.” GAO-NSIAD-98-97 Inventory Management,
February 1998.
GAO. “Defense inventory: Opportunities E xist to Improve the Management of DOD's
Acquisition Lead Times for Spare Parts.” GAO-07-281, March 2, 2007.
GAO. “DOD's High-Risk Areas: Efforts to Improve Supply Chain Can Be Enhanced by
Linkage to Outcomes, Progress in Transforming Business Operations, and
Reexamination of Logistics Governance and Strategy.” GAO-07-1064T, J uly 10, 2007.
Huang, Lizhen and Qifan Wang. ‘The Bullwhip Effect in the Closed Loop Supply Chain.”
Paper presented at the 2007 International Conference of the System Dynamics Society
and 50th Anniversary Celebration, Boston, MA, J uly 29 - August 2, 2007.
29
Killingsworth, William R., Regina K. Chavez, and Nelson T. Martin. ‘The Dynamics of
the Government Supply Process for High-Value Spare Parts.” Paper submitted to the
26th International Conference of the System Dynamics Society, Athens, Greece, J uly
20-24, 2008.
Lee, Hau L., V. Padmanabhan, and Seungjin Whang. “The Bullwhip Effect in Supply
Chains.” Sloan Management Review 3837 (Spring 1997): 93-102.
Minnich, Dennis and Frank Maier. “Supply Chain Responsiveness and Efficiency -
Complementing or Contradicting Each Other?” Paper presented at the 2006
International Conference of the System Dynamics Society, Nijmegen, Netherlands, J uly
23-27, 2006.
Minnich, Dennis and Frank Maier. “Responsiveness and Efficiency of Pull-Based and
Push-Based Planning Systems in the High-Tech Electronics Industry.” Paper presented
at the 2007 International Conference of the System Dynamics Society and 50th
Anniversary Celebration, Boston, MA, J uly 29 - August 2, 2007.
Rosenman, B. and Hoekstra, D. “A Management System for High-Value Army Aviation
Components.” IRO Report, October 1964.
Rosenman, Bernard B. “CCSS Supply Management.” IRO Report No. 280, November
1980.
Rosenman, Bernard B. “Supply Control Study Instability.” IRO Report No. 285,
June 1981.
Schroeter, Marcus and Thomas Spengler. “A System Dynamics Model for Strategic
Management of Spare Parts in Closed-Loop Supply Chains.” Paper presented at the
23rd International Conference of the System Dynamics Society, Boston, MA, J uly 17-
21, 2005.
Simchi-Levi, David, Philip Kaminsky, and Edith Simchi-Levi. Designing and Managing
the Supply Chain: Concepts, Strategies and Case Studies. New York: McGraw-Hill
Irwin, 2008.
Sterman, J ohn D. "Business Dynamics: Systems Thinking and Modeling for a Complex
World.” McGraw-Hill Companies, Inc., 2000.
Teresko, J ohn. “The Boeing 787: A Matter of Materials.” Industry Week, Dec 2007.
Thorne, Steve C. “Rightsizing DOD Inventory: A Critical Look at Excesses, Incentives
and Cultural Change.” Master's Thesis, Naval Postgraduate School, 1999.
30
