Supply and Value Chain
Support Through Scheduling
and Simulation: Applications
to the Semiconductor Industry
Dr. James R. Burns, Professor
College of Business Administration
Texas Tech University
Dr. Onur Ulgen, Professor
Department of Industrial and Systems Engineering
University of Michigan, Dearborn
Dearborn, Michigan 48128
Introduction
• Simulation Tools for Supply Chain
Inventory Analysis are presented
• Reductions in inventory result in
• Reductions in cost
• Reductions in cycle time
• Improvements in quality
• Improvements in workflow
2
Simulation Models
• Through use of IT to produce enterprise-wide
visibility, simulation models show
• Significant reductions in uncertainty are possible
• This leads to reductions in between supplier inventory
• Which leads to reductions in cycle (lead) time
• The models show reductions in information
delays through IT investments lead to
significantly improved performance
3
What are stocks and flows??
• A way to characterize systems as stocks and flows
between stocks
• Stocks are variables that accumulate the affects of
other variables
• Rates are variables the control the flows of
material into and out of stocks
• Auxiliaries are variables that modify information
as it is passed from stocks to rates
4
Stock and Flow Notation--
Quantities
• STOCK Stock
• RATE Rate
• Auxiliary i1
o1
i2 Auxiliary o2
o3
i3
5
Stock and Flow Notation--
Quantities
• Input/Parameter/Lookup
i1
i2 Auxiliary
i3
• Have no edges directed toward them
• Output
• Have no edges directed away from them o1
o2
o3 6
Inputs and Outputs
a
• Inputs
• Parameters Input/Parameter/Lookup b
• Lookups
• Inputs are controllable quantities c
• Parameters are environmentally defined quantities over
which the identified manager cannot exercise any
control
• Lookups are TABLES used to modify information as it
is passed along
• Outputs
• Have no edges directed away from them 7
Stock and Flow Notation--edges
• Information
a b
• Flow
x
8
Basic Model Structure
production
time
order rate
orders in
process
prod-trans rate
actual
inventory
sales
9
A Two-player Supply Chain
Model
• First player (the supplier) provides product
to the second player (the firm)
• Second player provides information back to
the first
• Each player received orders from its
“customer” and replenishes inventory
according to its ordering policy
10
Inventory Ordering Policy
• Assume continuous replenishment with
constant demand, fixed order quantity
• Using the Wilson EOQ model, the optimal
order quantity can be calculated to be 2000
widgets
• With annual demand of 6 million, 3000
orders go out every year
• That is an order every 2.9 hours
11
We present first The Two-player
Supply Chain Model…
• Without information visibility
• With discrete ordering policy of ordering
2000 widgets once every 2.9 hours
12
OIT unit cost
production time
Supplier order rate OIT Holding Cost
per mo
orders in
transit Supplier monthly Holding
Holding Cos t Cost
prod-trans rate
actual AI Holding Cost
inventory per mo
TOTAL
INVENTORY sales AI unit cost
TOTAL HOLDING
COST
---------------------------------------------------------------------------------------------------------------
production OIT unit
time 0 cost 0
Firm
order rate
0 Firm's OIT Holding
Cost per mo
orders in
customer transit 0
purc hases
0 Firm's Holding Firm's monthly
prod-trans rate 0 Cost holding cost
actual
inventory 0
Firm's AI Holding
Invent rate Cost per mo
sales rate
sales
0 AI unit cost
0
ACCUM 13
INVENTORY ACCUM SALES
The second Two-Player Model
Assumes ...
• Instantaneous information about end-customer
purchases all the way up and down the supply
chain
• orders cost virtually nothing, as opposed to $100
in the earlier model
• an implied order goes out every time a purchase is
seen at the customer end
• Otherwise, the two models are identical,
structurally
14
OIT un it cost
prod uction time
Supplier order rate OIT Holding Cos t
per mo
orders in
transit Supplier mon thly Holding
Holding Cost Cost
prod -trans rate
actual AI Holding Cost
inventory per mo
TOTAL
sales AI unit cost
INVENTORY
TOTAL HOLDING
COST
-------------------------------------------------------------------------------------------------------
prod uction OIT un it
time 0 cost 0
Firm
order rate
0 Firm's OIT Holding
customer
pu rc hases Cost per mo
0
orders in
transit 0
Firm's Holding Firm's mon thly
prod -trans rate 0 Cost ho lding cost
actual
inventory 0
Firm's AI Holding
Invent rate
sales rate Cost per mo
sales
0
ACCUM
INVENTORY 15
AI unit cost
ACCUM SALES 0
Comparing the two models
• Instantaneous ordering model exhibits greater
sales (less missed sales)
• Instantaneous ordering models exhibits
significantly lower total holding cost--$5,000,000
vs. $13,000,000.
• Results here are approriate for a supplier making
product that costs the firm $1000 each and for
which there is annual demand of 6,000,000 units a
year
16
Why the differences with respect
to inventory?
• In some cases, the discrete ordering policy
“misses” its threshold and does not order more
inventory
• This results in missed sales (there are some time steps
in which no ordering takes place at all)
• Beginning at month four, every other time step is
missed, roughly, so for the last eight months, onl half of
the monthly demand of 500,000 units is met.
• Instead of selling 6,000,000 units, only 4,000,000 were
sold
17
Why the differences with respect
to holding cost?
• Overall, the inventory in the pipeline in the
instantaneous ordering model is significantly less.
• Discrete pipeline approach to upstream
information dissemination results in larger
inventories
• Discrete pipeline scenario starts with much higher
initial inventories--500,000 versus only 100 for the
enterprise visibility approach.
• The high initial inventories are needed to compensate
for the missed sales and does so until about month four
18
Cycle times and Little’s Law
• According to Little’s Law
• Cycle time = inventory / throughput
• Inventory was reduced by 58%
• Cycle time would be similarly reduced
19
Reduced inventory leads to...
• reduced cycle (lead) times
• less rework and scrap due to smaller lot
sizes
20
What about a large order
quantity?
• 500,000 once a month would do it
• results are worse that orders of 2000 a
month
21
ACCUMULATIVE SALES
8M
4M
0
0 1 2 3 4 5 6 7 8 9 10 11 12
Time (Month)
Discrete Pipeline Approach
Enterprise Visibility Approach
22
TOTAL HOLDING COST
2M
1M
0
0 1 2 3 4 5 6 7 8 9 10 11 12
Time (Month)
ENTERPRISE VISIBILITY APPROACH
DISCRETE PIPELINE APPROACH
23
ACCUMULATIVE SALES
8M
6M
4M
2M
0
0 1 2 3 4 5 6 7 8 9 10 11 12
Time
(Month)
Discrete Pipeline 2k run
Discrete Pipeline 500k run
Enterprise Visibility Approach
24
TOTAL HOLDING COST
8M
6M
4M
2M
0
0 1 2 3 4 5 6 7 8 9 10 11 12
Time
(Month)
Enterprise Visibility Approach
Discrete 500k run
Discrete 2k run
25
A Three-Player Supply Chain
• Each player is modeled as a first-order balancing
loop structure
• Customer orders run 30 per time steps, but this
happens randomly in only halfof the time steps.
• This model is looked at in both of two contexts--a
delayed information approach and the enterprise-
wide instantaneous information approach
26
First-order Balancing loop
structure
27
desired inventory adjustment time
order/ship rate
First Supplier
information
delay
actual
demand rate inventory
adjustment time 0 Delay time
desired inventory 0
order/ship rate 0
Customer orders
Second
information
Supplier delay 0
actual
demand rate 0 inventory 0
desired inventory 1 adjustment time 1
order/ship rate 1
Firm
information
delay 1
actual
demand rate 1 inventory 1
28
Actual Inventories with one-week information delays
20,000
0
-20,000
0 10 20 30 40 50 60 70 80 90 100
Time (Month)
actual inventory at first supplier
actual inventory 0 (at second supplier)
actual inventory 1 (at the firm)
29
Actual Inventories with two-week information delays
40,000
0
-40,000
0 10 20 30 40 50 60 70 80 90 100
Time (Month)
actual inventory at the first supplier
actual inventory 0 (at the second supplier)
actual inventory 1 (at the firm)
30
Actual Inventories with one-month information delays
200,000
0
-200,000
0 10 20 30 40 50 60 70 80 90 100
Time (Month)
actual inventory at the first supplier
actual inventory 0 (at the second supplier)
actual inventory 1 (at the firm)
31
desired inventory adjustment time
First order/ship rate
Supplier Adjustment
actual
demand rate inventory
desired inventory 0 adjustment time 0
Second order/ship rate 0
Supplier
Customer orders
actual
demand rate 0 inventory 0
desired inventory 1 adjustment time 1
order/ship rate 1
Firm
actual
inventory 1
demand rate 1 32
Actual Inventories Without Information Delays
1,000
500
0
0 10 20 30 40 50 60 70 80 90 100
Time (Month)
actual inventory of the first supplier
actual inventory 0 (at the second supplier)
actual inventory 1 (at the firm)
33
The last figure
• exhibits a rapid ascent to the desired
inventory on the part of all three players, to
the desired inventory, with no overshoot--
very well behaved
34
These models were created using
the VENSIM tool
• www.vensim.com
• a product of Ventana Systems, Inc.
35
Translation of these models to
commercial simulations
• These models can be setup to be driven by
flight simulator front ends with sliders and
dials, meters and such
• Users would decide upon
• Amount of work in process
• Ordering policy
• Ordering parameters (quantity, time between
reviews, lead time, safety stock, etc.)
36
Summary
• Continuous dynamic simulations explain
much of the behavior we see in enterprise
systems and supply chains
• They can be useful tools for deciding
• What effect IT will have on the supply chain
• The actual structure of the simulation tools
can be preprogrammed
37
Summary, Continued
• The only thing the user has to do is use the
simulation model to make decisions about
• Ordering policy
• Order quantities
• Order frequency
• Order lead time
• Amount of work in process
• Etc.
38
Questions from the
AUDIENCE???
• Thank you for coming!!!
39
40
41
42