Introduction to Machine Learning Unit-3 for II MECH
Supply chain design and operation
1. Supply Chain Design and
Operation
Presented by Anqi Guo
Date : 2/12/15 Updated 2/20/15 4/12/15 4/17/15
2. Where do I come from?
George
Danzig
Prof.Ye
Prof. Jackson
Ph.D.
Advisor
Took 3 of his
graduate-level
supply chain
courses
Both got Ph.D. of
Operations
Research from
Stanford ; Friends
Solved Nobel
Winning
Operations
Research
Problem
Client: 华为分仓项目,
中国电力, Boeing,
AMEX
Client: GM inventory
optimization award,
GE, Cleveland Clinic
3. Content
• Operations Research (OR) Intro &Tools
• OR’s application : Supply Chain(SC) Design
• Supply Chain Design Principles &Tools
• Case Study 1 :Velocity Mfg Inc
• Case Study 2: Bike Rental System
• Case Study 3: New Jersey Pharmacy
• Case Study 4 : Amazon(Brief)
5. “Operations Research is
the application of
advanced analytical
methods to help make
better decisions.
-Wikipedia
To Optimize!
6. OR was created and widely used inWW II
Aircraft from black to white, 20% closer, 30%
more sinkingGerman U-boat
1% to 7% of sunken German U-boat by changing
depth charge from 100 ft to 25 ft
2 tubs for washing, 2 tubs for rinsing
3 tubs for washing, 1 tubs for rinsing
Saved a lot of time
Patrick Blackett:
British physicist
7. Cornell professor reduced 85% of number
of students facing three exams in a day
8,000 Students 4,000 Courses
# of possible exam
schedules? More than
50,000,000,000,000,000
,000
11 days 66 time slots
8. Demo 1: Assignment Problem
Simplify the
problem
Model Run Code in AMPL Result
Assume 4 courses(i:
1,2,3,4), 4 time
slot(j:A,B,C,D);
Professor will rate
time slot for each
course from 1(most
favorable) to
4(least favorable);
Each course will get
assigned to a
course;
Minimize sum of
total ranks picked
6, (1C) (2B), (3D),(4A)
See Notes for professors’ ranks
for each time slot
16. ORTools & Application
Linear
Programming
• Transportation
• Assignment
• Integer
Programming
• Goal
Programming
Probabilistic
Technique
• Decision
Analysis
• GameTheory
• Markov
Analysis
• Queuing
Theory
• Simulation
• Forecasting
Inventory
Technique
• Inventory
Models
• Newsvendor
problem
• Economic
Order
Quantity
Network
Technique
• Network
Models
(CPM/PERT)
Non-linear
Programming
• Dynamic
Programming
Fleet Assignment at Delta Airlines : Savings of $220,000 per day in 1993 (with
60,000 variables and 40,000 constraints)
Gasoline Blending atTexaco : Savings of $30 million annually
KeyCorp Service Excellence Management System:Customer processing times
were reduced by 53% , resulted in savings of $98 million in 5 years
17. OR’S application : Supply Chain
Management(SCM)
Global SupplyChain of Maersk
18. Three Basic Measures of Process Analysis
Immigration Department MBA program Auto company
Applications Student Car
Approved or rejected cases Graduating class Sales per year
Processing time 2 years 60 days
Pending cases Total campus population Inventory
Flow Unit
Flow rate/
throughput
FlowTime
Inventory
Flow rate/throughput : number of flow units going through the process per unit of time
FlowTime : time it takes a flow unit to go from the beginning to the end of the process
Inventory: the number of flow units in the process at a given moment in time
Flow Unit : Customer or Sandwich
19. Why is SCM important?
10.87
6.68
5.7
3.25
7.99
0
5
10
15
20
25
8-Dec 9-Dec 10-Dec 11-Dec 12-Dec 13-Dec 14-Dec
Annual InventoryTurnover Rate for China-US Top E-commerce Companies
JD
VIPS
WALMART
JUMEI
DANG
AMAZON
A higher than industry average turnover demonstrates that your business is
competitive, while a smaller than industry average turnover shows that there is room for
improvement.
Global Industry Average is 7.
20. Difficulty of SCM
1. Demand Uncertainty
2. Poor physical characteristics
3. Poor information infrastructures
4. Business process not accurately evolve
5. No Decision support system
22. Current Supply Chain Models
1. MRP Model
Bill of materials(BOM) , Master production schedule
Enterprise Resource Planning(ERP)
2. Mathematical Programming-Based Model
Newsvendor Problem, Assignment Problem
Max flow/Min cut problem, Dynamic Programming, Integer Programming
3. Inventory Model
Safety stock, lead time, etc.
4. Advanced Planning and Scheduling System
28. Guideline for Collaborative Supply Chain Design
1. Know the Customer
2. Construct a lean SC that eliminates waste, variability,
and uncertainty
3. Build tightly coupled information infrastructure
4. Build tightly coupled business process
5. Construct tightly coupled decision support system
1
2
3
4
5
29. Case Study 1 : Revitalization ofVelocity
Mfg Inc.
In 2011, a second tier supplier in the U.S. hydraulic hose
Note: Course project
30. Case Study 1 : Revitalization ofVelocity
Mfg Inc.
31. Case Study 1 : Revitalization ofVelocity
Mfg Inc.
Current Problems
Our Solutions
Projected Results
32. Summary of Problems
• Sales is declining (2010: 89% of 2008)
• Lousy quality of products
• Material cost is high
• RONA is low
• Long delivery LeadTime
• Net Income / Assets is unacceptably low (2.7%)
• Fill Rate is 67%, but should be about 98%
• WIP can be reduced
• Dividends are unsustainable given current cash flow
37. Acquire production
info from white
boardStart of Shift
Are there
still input
materials
to process?
Choose next
unit to work on
Setup Tester for
Current End
Fitting
Run Test
Good
Unit?
No
Yes
Place into
Scrap Pile and
report to plan
controller
Place into WIP1
and update WIP1
Produce Blanks
Assemble Blanks and
Elbows
Lathe Operators
Assemblers
Oven and Test
Operator
Are there
still input
materials
to
process?
Choose next
unit to work on
Lathe Operators
Assemblers
Oven 1
running
?
Both Oven
in
operation?
Yes
Put end fitting into
oven 1 until it’s filled
Run Oven
Yes
No
No
Put end fitting into oven
2 until it’s filled
Lathe 2
operating
?
Yes
Put blanks into lathe1
until it’s filled
Put blanks into lathe2
until it’s filled
No
End of Shift
Run Lathe
Sector 1 Shop Floor Operation Plan
Update White
Board
3
39. Case Study 2 : Bike Rental System
• 6 Bike Stations
• 18 Bikes
• Simulated effect of changing incentives for people to
reposition bike on customer satisfaction rate
5
41. Case Study 3: New Jersey Pharmacy
Provide
Antivirals
to
patients
Set
Patient
Travel
Range <
25 Miles
Ensure Short
WaitingTimes
at Pharmacies
0
5000
10000
15000
20000
25000
30000
1
14
27
40
53
66
79
92
105
118
131
144
157
170
183
196
209
222
235
248
261
274
287
300
313
326
339
352
Demand
Days
Actual Historical Annual Demand
(Pediatric + Adult)
Phase 1: Initiation
Phase 2: Peak
Phase 3: Resolution
42. Reduce the Complexity of the CDC Supply Chain
Determining Which Specific Pharmacies to Use (3 out of 3)
Divided States into 45
Hexagon Shaped regions
Furthest distance <25
miles
Unify many inventories
into one inventory
(governmental policy)
Link the hexagons to
pharmacy locations and
regional population
Determine whether
regional capacity fits the
peak day demand
Discovered certain high
population hexagons need
additional capacity on
peak days!
(Divide into three phases)
OR even improved the most mundane processes. After observing a long line of soldiers washing their mess kits after a meal, a scientist saw the troops used two tubs to wash their kits and two to rinse them, even though washing took three times longer than rinsing. By providing three tubs for washing and just one to rinse, the bottleneck was eliminated and the line disappeared, saving much valuable time. - See more at: http://www.navyhistory.org/2013/06/book-review-blacketts-war-men-who-defeated-nazi-u-boats-brought-science-art-warfare/#sthash.srW09vV0.dpuf
http://en.wikipedia.org/wiki/Operations_research
Blackett’s Circus: group of mathematician, biologist, army officer, etc. Later become first official Operations Research Group
Contribution:
Change color of aircraft to White to increase 30% of sinking a German U-boat
Change Depth Charge from 100ft to 25ft, sunk 7% instead of 1% of German U-boat
Reduce bullets to shoot down a aircraft from 20,000 to 4,000
Save soldiers’ washing time by remove bottleneck
News : http://www.orie.cornell.edu/news/index.cfm?news_id=62099&news_back=news_archive%26y%3D2005
Note : 66 is by estimate
AMPL: Modeling Language for Mathematical Programming, also accessible in Python
param: E : c :=
1 A 4
1 B 3
1 C 2
1 D 1
2 A 3
2 B 1
2 C 2
2 D 4
3 A 1
3 B 3
3 C 4
3 D 2
4 A 1
4 B 2
4 C 3
4 D 4;
Intro to Supply Chain Management: http://www.youtube.com/watch?v=Mi1QBxVjZAw
Intro to Operation Management by Upenn : https://class.coursera.org/whartonoperations-004/lecture/2
Data Source : http://www.gurufocus.com/term/InventoryTurnover/EBAY/Inventory%2BTurnover/eBay%2BInc
the Inventory turnover is a measure of the number of times inventory is sold or used in a time period such as a year. The equation for inventory turnover equals the Cost of goods sold divided by the average inventory.