0生产计划与控制

1896 1920 1987 2006
Production Planning and Control

Professor JIANG Zhibin
Dr. GENG Na
Department of Industrial Engineering & Logistics
Management
Shanghai Jiao Tong University
Content of this course
Forecast of future demand
Aggregate plan
Master production schedule (MPS)
Schedule of production quantities by
product and time period
Material Requirement Planning (MRP)
Generate production orders and
purchase order

Operations Scheduling
To meet quantities and time
requirements for MRP

Inventory Control
Subject to known and
unknown demand

Push and Pull Production
Control Systems
JIT
Project Scheduling
unknown demand

Scheduling in Supply
Chain Management
Transportation problem
Textbook

Grading Policy:
In-class quiz & Homework 20%
Mid-term Exam 30%
Final Exam 50%
Instructors:
Prof. Zhibin Jiang( )
zbjiang@sjtu.edu.cn
Dr. Na Geng
gengna@sjtu.edu.cn
Chapter 1
Production System and Operations Strategy
Contents
1. Basic Conception of Production & Operations
Management
2. Production System and its Classification
3. Organization of Production System
4. Operations Management &Operations Strategy
5. The Classical View of Operations Strategy
6. Matching Process and Product Life Cycles
7. Capacity Growth Planning
Basic task of enterprise: to gain high efficiency
Basic Conception of Production and Operations
Management
Production
activities
With less
inputs
More product
outputs
Economic and
social benefits
The Production Activities of
Enterprises
1. Decision activities;
2. Technology activities;
3. Supply activities;
4. Manufacturing activities;
5. Sales activities;
6. Financial activities;
Production
activities

The Task of Production & Operations
Management
The task of POM is:
Organizing the production activities reasonably
Making good use of the resources efficiently
The objective is to:
Complete the product production
Achieve the business objective
Reduce cost, improve quality and reduce production time
Promote production systems flexibility.

Chapter 1
Contents
Management
1.2 Production System
System view
An enterprise is a large system.
A production system is a subsystem of the whole
enterprise system, and its main function is to
produce products
A production system uses operations resources to
transforms inputs into some desired outputs.
Production System
Input
Materials, Capital
Demand,
Information,
Energy
Transform
Manufacturing
Assembling
Storing,
Service
Transportation

Output
End products
services

Decision
Decision
Feedback Feedback

Decision
Feedback
Resource: People, Plants, Parts, Process, PPC
Management

Supplier

Customer
Example: inputs-transformation -outputs of
Production system
System

Primary
inputs

Resources

Primary
transforma
-tion

Desired
outputs

automobile
plant

steel, power
parts, other
materials

facilities,
and
workers

Manufactur
-ing and
assembling

cars with
high
quality

wholesale
center

goods

storage
tools

distribution

instant
delivery

Elements of Production System
Technology, e.g. process technology and organization of
equipment
Facility, e.g. scale and layout of facilities
Capacity, e.g. the capacity and flexibility of production
Integratione.g. inner-integrationouter-integration
Structure
Elements
Personnel organization
Production planning
Production control
Production inventory
Quality management
Non-
structure
elements
Classification of Production System
Continuous flow production system
Discrete production system
Classified based on
production process
continuity:
Make to stock
Produce to order
Engineer to order
Classified based on
customization:
Manufacturing industry
Service industry
Classified based on
industry:
Continuous Flow Production System
Continuous flow production
the materials, either dry bulk or fluids that are being
processed are continuously in motion, undergoing chemical
reactions or subject to mechanical or heat treatment.
Continuous usually means operating 24 hours per day, seven
days per week with infrequent maintenance.
Example???

Some common continuous processes:
Oil refining
Natural gas processing
Sanitary waste water treatment
Chemicals
Metal smelting
Etc.
Characteristics:
1. Relatively steady production process;
2. High standardization of products;
4. Few product sorts;
5. Highly automated, highly specialized and expensive facility;
6. The facility failure can bring in great loss, so the equipment
maintenance is very important.
Discrete Production System
Discrete production
A production process in which its output is individually
countable, or identifiable by serial numbers, and is
measurable in distinct units.
Each process can be individually started or stopped and can
be run at a varying production rates.
Example???

Examples of discrete production:
Automobiles
Mobile phone
Home appliances
Computers
Etc.

Characteristics:
Discrete products
Not steady production
Complex products manufacturing
Short product life cycle
Quickly changed products
Complex production and operations management
Most methods in IE aim at this type of production.
Difficulties
Different type of products needs different production process,
needs different layout of equipment;
The production proportion of different type of products (product-
mix) often changes;
The requirement of production capacity is dynamic;
Annual plan for production is usually inaccurate, however, the
short plans such as monthly and weekly plans are the keys.
Difference between Continuous Flow and
Discrete Production
Primary characteristic
Continuous flow Discrete
Customer number
Product sorts
Product standardization
Customization
Demand
few
few
high
low
steady
many
many
low
high
change
Difference in the Facility and Manufacturing
Aspects
Primary characteristic
Continuous flow Discrete
Key production elements
Automatisation level
logistic
facility failure
equipment maintenance
the types of material
WIP (Work in Process)
product inventory
output flexibility
product structure flexibility
operation information
collecting
High dense Capital
high
continuous
great effect
high requirement
fewer
low
high
low
low
easy
high dense workforce
comparatively low
low discontinuous
effect
low
many
high
low
high
high
difficulty
Discrete production systems
Mass production
Multi-products and low volume production
One of a kind production
Mass customization
Mass Production
Only one or a few types of products are produced repeatedly in a
long period. e.g. watches, bicycles, TV, refrigerator, etc.
High productivity
Skilled workers
Simple production planning and scheduling
Easily guaranteed products quality
Low production cost.
Character
-istics
Management emphasis: facility maintenance; worker management;
and quality management

Job Shop
Flow-shop mass production
On October 7, 1913, Henry
Ford installs the first moving
assembly line for the mass production
of an entire automobile in Highland
Park. His innovation reduced the time
it took to build a car from more than
12 hours to two hours and 30 minutes.
First assembly line in Highland
Park, MIch.
Multi-product and Low Volume Production
Many kinds of products
Low volume, low production stability
The high productivity specific equipments can not be used
There are many products in production at the same time so
that the management is very complex.
Characteristic
Optimize product mix
Use the Group Technology (GT) production cell, and
production line,
Improve the connection between processes.
Arrange reasonably the product kind, quantity and
interval to produce products in batches and in turns.
Control and balance WIP, logistic and product inventory
Management
Emphasis
One of a Kind Production
Start to organize production after receiving order. e.g.
shipbuilding, airplane making, generating electricity equipment
Characteristics
many kinds of products
general equipment are required
workers are also required to be generalists.
Management emphasis: effective cooperation
to decide reasonable delivery date,
to improve the flexibility of manufacture system
to improve the generalization of parts.
Mass customization

Mass customization is the customisation and personalisation
of products and services for individual customers at a mass
production price.
Characteristics
Modularity
Commonality
Management emphasis
Improve the standardization of components
Reduce the response time to the customers
Classified based on customization
Make to Stock
Characteristic:
Customer ordered products are satisfied by stocked product;
Produce to replenish the stock while it drops to given level
Production based on the forecasting
Mass production, in big batch
High product inventory
High productivity
Simple production organization
Disadvantage:
High risk of excess inventory or out of stock
Produce to Order
produce based on the order
different product has different specification,
quantity, quality and delivery time
nearly zero inventory
multi-product and low volume production
complex production organization and management
Characte-
ristics
assembly to order;
make to order; and
engineer to order.
the level of
customization
design purchase manufacture assembly delivery
Delivery Engineer to OrderETO Order
MTS
Order
Assembly to OrderATO
Order
Make to OrderMTO
Order
Production to Order
Product
type
Many
Large
Few
Small
Production batch
ATO

MTS

ETO

MTO

Product Type and Production Batch
Classified based on industry -Service Industry
Characteristics
Product is service, and it can not be stored.
Labor denseness, and low labor productivity.
The standard of quality: customer satisfaction degree.
Some service is included in the manufacture industry,
e.g. sales, field support (service-after-sales) and
accounting.
Types of Service Industry
Degree to which customers are touched
High-touch typehotel, clinic, restaurant, and shop
low-touch typewarehouse management, wholesale
Customer to serve
Production oriented service
Consumption oriented service

Service Industry
The End!
Chapter 1
Contents
Management
1.3. The Organization of Production System
A production system is usually divided into
basic units, e.g. workshop, manufacturing cell,
and working team, and all these units can be
organized on one of the following three basic
principles:
Process focused
Product focused
Group technology (GT)
Principle of Process focused
A B C
A B C
part1
part2
part2
part3
part1
part3
(a)
Process
focused
Principle of process focused : organize the production units
according to a production process kind (Turning, milling,
grinding, drilling, etc).
Within one production unit, the same type of workers
complete the same processes using the same type equipment.
e.g. machining workshop lathe section, and drill section,

Principle of Process focused
high equipment utilization rate
an employee do some fixed operations
units have high flexibility to produce
various of products
Advantages
long conveying routes
great amount WIP
WIP needs long waiting time
frequent cooperation and interdependency
among production units
Disadvantages
Principle of Product focused
A B A
A B C
B C A
part1 part1
part2 part2
part3 part3
(b)
Product
focused
Principle of product focused: the production unit is organized
according to a product kind ( or a part or a component kind)
Principle of Product focused
Within one production unit, one products production( all
processes) is completed by different type workers using
different type equipments
It can also be called as closed-production-unit. e.g.
automobile assembling line, TV assembling line.
A B A
A B C
B C A
part1 part1
part2 part2
part3 part3
(b)
Product
focused
Principle of product focused
Advantages:
The material handling distance in production
process can be reduced greatly
The WIP inventory can be reduced greatly.
It can use productive and automatic organization
modeproduction line.
The cooperation among production departments can
be reduced, and hence management can be simplified.
Principle of product focused
Disadvantages
Low equipment utilization rate;
The whole production unit can be influenced once
an equipment breakdown;
Low production flexibility
Group Technology (Manufacturing cell, flexible
manufacturing system)- allocate different equipment
into a production cell to work on a product family that
have similar shapes and process requirements.
Within one cell, the production of a product family
can be completed by different type workers using
different type equipment.

Group Technology
Warehouse
Warehouse Warehouse
Warehouse
Customer
Customer
Customer
Process Specialization vs. GT Manufacturing Cell
Process focused
GT Cell
Discontinuous Flow Line
Advantages
better human relations
improved operator expertise
less work-in-process (WIP) inventory and material
handling
Faster production setup
GT production cell overcomes some disadvantages of
process focused production system.
Group Technology
Chapter 1
Contents
Management
1.4 Operations Management &Operation
Strategy
Operations Management
To design, operate, assess and improve production
system.
The core of Operations Management is the
management of production systems-production
planning & control (PPC)

1.4. Operation Management &Operation
Strategy
Business Strategy
Marketplace
Market Strategy Finance Stragegy Operation Strategy
People
Planning &Control
Operations Management
Processes Parts Plants
Production System
Input
Output
Strategy
Business Strategy is concerned with setting broad
policies and plans for using production resources of a
firm to best support its long-term competitive goals.
It defines
The market where the enterprise compete;
The level of investment;
The measures of allocating resources to and
integrating separate business;
Strategy
Operation Strategy
The means by which a firm configures its resources to achieve
its competitive goals;
For manufacturing firms, key operational decisions include
Where to locate new manufacturing facilities;
How large these facilities should be;
What processes to use for manufacturing and moving
goods through the system; and
What workers to employ.
Strategy
Service firms also require an operation strategy
Disney theme parks continuing record of
success is partly due to its careful attentions to
detail in every phases of its operations.
Establishing monthly sales goals; optimizing
bus routes for 250K daily guest trips;
implementing meter-based maintenance
scheduling for ride vehicles; and building
queuing and simulation models for call centers,
theme park rides, and resort front desks.
Strategy
American companies lesson: OM is important
Years ago, the American big three (General Motor, Ford
Motor Co., and Chrysler Group) placed too much emphasis
on marketing and finance and too little on operations (making
and delivering the products) ;
The enormous success of Japanese auto makers motivated the
American big three to close their inefficient plants and
change the way things are done;
Today, the best American auto makers compete their Japanese
counterpart s by quality and efficiency.

Strategy-Framework
Two Traditional Strategic Dimensions-
Lower Cost
New entrants to a market operate on low cost strategy;
Examples:
Korean automakers: Hyundai and Daewoo;
Discount outlets: Costco;
Retailers: Wal-Mart;
May be successful over short term, but risky in long run-
low cost.
Strategy
Two Traditional Strategic Dimensions
Product Differentiation-Differentiate a companys
products from its competitors.
Providing uniqueness to buyers;
Product differentiation within a firm to capture
different market segments
Henry Ford & GM
Strategy
Other Strategic Dimensions
Quality;
Delivery speed;
Delivery reliability;
Flexibility;
Means different things in different contexts;
Product Functions
Reliability relevance
Delivery speed
Product Design
Strategy
Both Geo Prism and Ferrari are quality products;
Consumers buying these products at greatly different prices
are both looking for good quality but for fundamentally
different objectives.
Ferrari F50 Geo Prism
Biao Car
Transportation Tool
1.4. Operation Management
&Operation Strategy
Quality;
Delivery speed;
Flexibility;
An important competitive weapon in some context;
It is competitive basis for some firms: UPS, Federal Express;
Mail-order and web-based retailers remain competitive by reliably
and quick delivery;
Strategy
Quality;
Delivery speed;
Flexibility;
Being able to deliver products or service as promised;
How do online brokerages retain customers?-Executing trade
reliably and quickly
How do contact manufacturers keep continuously successful?
putting customers first and maintaining a good record of
delivering high-quality products in reliable fashion.
Strategy
Quality;
Delivery speed;
Flexibility;
Offering a wide range of products;
Being able to adjust to unexpected changes in demands of the
product mix offered;
Successful examples:
Flexible assembly lines enable Toyoda adjust product mix in
real time basis;
Dell Computer-Mass customization.
In-class Quiz
1. Describe the classification of production systems.
2. Whats the differences among MTS, ATO, MTO,
ETO?
3. List the six dimensions of operations strategy.
The End!
1896 1920 1987 2006
Chapter 1 Production System and
Operations Strategy
Department of Industrial Engineering &
Logistics Management
Contents
Management
1.5. The Classic View of Operation Strategy
The traditional view of manufacturing strategy was put
forward by Wickham Skinner of the Harvard Business School.
Classical operations strategy thinking relates to the following
issues:
1. Time horizon
2. Focus
3. Evaluation
4. Consistency
1.5. The Classic View of Operation Strategy -
Decision Time Horizons
Decision Time Horizons refer to the length of time required for
the strategy to have impact on.
Fig1-1 Decision Time Horizons of Manufacturing Strategy
Short-term decision horizon
Impacts of Short-term OS are measured in terms of
days or even hours;
Include purchasing, production & personnel
scheduling; policies for control of quality and
maintenance function, short-term inventory control,
production scheduling, and so on.
Medium-term decision horizon
Impacts of medium-range OS are measured in
terms of weeks and months;
Include demand and requirement forecasting,
employment-planning (size & mix), decisions
on the distribution of goods, setting up targets
for inventory and service level ;
Long-term decision horizon
Strategy is usually associated with long-term decision ;
Include choosing the timing, the location, and the scale of new
manufacturing facilities; addressing groundwork for building
proper channel for sales and distribution; and setting up service
objects;
Require information about the forecast for new and exiting
products, the changing patterns of marketplaces , and changes in
costs of availability of resources.
Effects of time horizons on strategy
Impact on decision;
The uncertainties on decisions;
Penalty for wrong decisions.
Example: The Gap
(1) Short time horizons involve many
decisions, each of whose impact may be
small, but cumulatively can make difference.
Manager at the Gap store restock shelves
and reorder every days. Small errors in sales
data and personal judgment required for
reordering mix of items may result in out-of-
stock or wasted shelf space.
The Classic View of Operation Strategy - Decision
Time Horizons
(2) Buyers in San Francisco headquarters of the Gap decide on
what lines of clothes to stock for the coming seasonsMedium-
range decision.
Less information available than store mangers for the decision;
The decisions may have greater impact;
They have to judge fashion trends and the color preference.;
Penalty: A line that does not sell must be on sale at lower price.
(3) Top manager must make long decisions, possibly (a) the
number, location, and size of distribution centers; (b) terms and
conditions of long-term contracts with suppliers;(c) arrangements
for firm-wide supply chain logistics; (d) selection of management
personnel.
More uncertainties: Demographic change makes decision on
location and sizing for DCs wrong; a long-term contract with a
oversea plant may have opposite results, e .g. quotas, tariffs.
1.5. The Classic View of Operation Strategy - Focus
Notion of focus in manufacturing strategy was first put
forward by Skinner. Five key characteristics:
(1) Process technologies
(2) Market demand
(3) Product volumes
(4) Quality levels
(5) Manufacturing tasks
Process Technologies
A natural means of focusing operation of one plant;
New unproven process technologies should be
limited to one per factory;
The number of different mature process
technologies should be kept to the level that the
plant manager can oversee efficiently.
Market demands
Marketplace usually determines the focus of a product or line of
products;
Indications of market on plant focus are
Price-It has been always a key product differentiation factor;
Lead time-It should be shortened for quick production and
distribution of product fight against competitors.
Reliability-Its specifications differ by market segment, even
for identical products.
Product volumes-
The production volumes within a single plant should be
similar for the production systems to be able to operate
smoothly, neither under nor- over utilized.
Quality level
Keep similar quality level for products produced in a
single plant in order to establish a consistent quality
control standards;
Quality control standard are results of several factors:
the statistical control techniques, the monitoring
procedures, and workers training, procedure, and
attitudes.
Manufacturing tasks
Productivity of a plant producing a wide line of
different products are affected by frequent setup and
reconstructing of production lines.
The number of distinct manufacturing tasks at one
location should be limited so that workers may
concentrate on perfecting exiting processes.
Evaluation
Dimensions for evaluating production/operations strategy
Cost-if pricing is a key to market differentiation and
competitiveness, the cost of product delivered to the customer
becomes a major means of strategy evaluation.
Quality-when product quality is a major determinant of product
success in marketplace, or high reliability is required to meet
product specifications, quality should be a dimension for strategy
evaluation;
Evaluation
Dimensions for evaluating production/operations strategy
(Continued)
Profitability- Strategies that achieve short-term profitability
may not necessarily be in the best interests of the firm in the
long term
Customer satisfaction-Maintaining a satisfied and loyal
customer base is important for the success of a firm.
Consistency
Strategy is the composite of all company policies that
affect manufacturing;
Each may aim at optimizing a different objectives,
resulting in a complex situation in which management
and labors assure adversarial positions.
Causes for the common inconsistencies
Professionalism
Product proliferation
Changes in manufacturing tasks
Consistency
Causes for the common inconsistencies
Professionalism-inconsistent goals of different
professionals.
Product proliferation-a firm that produces a smaller
number of products tends to be more profitable,
however become less profitable as more products are
produced within a single plant.
Changes in manufacturing tasks-the objectives that
made sense when the plant was constructed no longer
make sense after the plants function has changed.
The manufacturing task was never made explicit.
1.5. The Classic View of Operation Strategy - Review
Fig.1-2 The Elements of Production and Operations Strategy
The End!
1896 1920 1987 2006
Chapter 1 Production System and
Operations Strategy
Logistics Management
Contents
Management
1.6. Matching Process and Product Life Cycles
Product Life Cycle-the time duration from the time
when a product is put into market to time when it
disappear in the market;
A typical product life cycle consist of four major
segments:
Start up
Rapid growth
Maturation; and
Stabilization or decline;
The Product life cycles-in different phases, different
competition strategy may be adopted
Fig.1-4The Product Life-Cycle Curve
1.6. Matching Process and Product Life Cycles (2)
Start up-the competition is generally not a problem;
concern on improvement of the production and marketing
functions;
Rapid growth-begin to see competition; establish the
product as firmly as possible in the marketplaces by
alternative pricing patterns, improvement and
standardization in manufacturing;
Maturation-maintain and improve the brand loyalty;
increase market share through competitive pricing; save
cost by improving production control and product
distribution;
The final phase-the shape depends on the nature of the
product: indefinite continuing growth (household goods,
processed foods, and auto)/same strategy as Phase 3;
natural decline/squeezing out the most from the product
or product line.
The Process Life Cycle
Early stage- job shop; a varied mix of relatively low-
volume orders; being responsive to changes in the
product design; little control over suppliers;
Middle stage- some automation; batch production and
some transfer lines(assembly lines); more control over
suppliers; unit cost decline;
Last stage-automated process and standardized
production process; few manufacturing innovation;
continuous flow operation.
1.6. Matching Process and Product Life Cycle
The Product-Process Matrix (PPM) was developed by
Hayes & Wheel Wright in 1979;
Link product and process life cycles in two dimensions
Process structure
Product life-cycle stage
The Four Stages in the evolution of the
manufacturing process in term of process structure
J ob Shop;
Batch Production;
Assembly Line; and
Continuous Flow
The Four Phases in product Life Cycles in term of production
volume and the number of product types
low volume , one-of-a kind;
low volume, multiple products;
few majors products, higher volume; and
high volume, high standardization, commodity products.
1.6. Matching Process and Product Life Cycle
I
Low volume,
low
standardization,
one of a kind
II
Multiple
products, low
volume
III
Few major
products, higher
volume
IV
High volume, high
standardization,
commodity
products
I
Jumbled flow
(job shop)
Commercial
printer
Void
II
Disconnected line
flow
(batch)
Heavy
equipment
III
Connected line
flow
(assembly line)
Auto assembly
IV
Continuous flow
Void Sugar refinery
P
r
o
d
u
c
t

l
i
f
e
-
c
y
c
l
e

s
t
a
g
e
Process structure
1To match appropriate industry in its mature phase with the
appropriate process.
1.6 Matching Process and Product Life Cycles
2To identify the proper match of the production
processes with the phases of the product life cycle.
In the start-up phase of product
development, the company would
typically be positioned on the upper
left-hand corner of the matrix;
As the market for the product
matures, the firm would move down
the diagonal to achieve economies of
scale;
Finally, the firm would settle at the
position that would be appreciate
based on the characteristics of the
product.
Contents
Management
1.7 Capacity Growth Planning -a long-term strategy
problem
Capacity of a plant is the number of units that it may produce in
a given time;
Factors to be considered for capacity strategy:
Predicted patterns of demand;
Costs of constructing and operating new facilities;
New process technology;
Competitors strategy
problem
Issues needed to be considered for constructing a new
plant
When to construct? Two affecting factors-construction
lead times and changing patterns;
Where to build? Near suppliers or market outlets?
Where labor cost or tax lower?
What size? Underutilized or shortage of capacity.
1.7 Capacity Growth Planning -a long-term
strategy problem
Economies of Scale and Economies Scope
Economies of Scale
Economies of Scale -The increase inefficiency of
productionas the number of goodsbeing produced
increases.
Economies of scale lowers the average cost per unit
through increased production since fixed costs are
shared over an increased number of goods
1.7 Capacity Growth Planning -a long-term
strategy problem
Economies of Scope:
Introduced by Panzer and Willig, is defined as the cost
savings obtained from combining the production of
two or more product lines at s single location because
of sharing some equipment and personnel.
Economies of Scope can support investment in new
technology-FMS and CIMS result in efficiencies
brought about by variety, rather than volume.
Trade off between combining product lines at a
location and lack of focus.
problem
Make or Buy: A Classical Capacity Expanding Problem.
Buy a part from supplier at cost of c1 $ /unit; or
produce by itself at lower cost of c2 $/unit, but with
investment $K, c
2
<c
1
Total Cost for buy x units parts: c

1
x;
Total Cost for producing x units parts: K+c
2
x
The break-even quantity solves K+c
2
x=c1x
Giving x=K/(c
1
-c
2
)
1.7. Capacity Growth Planning -a long-term strategy
problem
Fig. 1-11 Break-Even Curves
problem
Dynamic Capacity Expanding Policy
The dynamics of changing demand pattern determine
when a firm should invest in new capacity. The two
objectives in capacity planning:
Maximizing market share;
Maximizing capacity utilization.
problem
Discussion
Maximization of capacity utilization may incur
shortages when demand is higher than the anticipated;
Increase productive capacity by producing to inventory
and letting inventory to absorb demand fluctuation,
resulting in obsolete inventories.
Have excess capacity to cope with sudden increase in
demand.
1.7. Capacity Growth Planning -a long-term
strategy problem
Fig.1-12 Capacity Planning Strategies
x: Time interval between
introduction of two
successive new plants
problem
Issues in Plant Location-Factors relevant to location
decisions:
Sizes of the plant- required acreage, space for building
structure;
Product lines to be produced and process technology to be
used
Labor force requirements
Transportation need
Utilities requirement
Environment issues
Interaction with other plants
International considerations
Tax treatment
In-class Quiz
Q1. Describe the elements of production and
operation strategy
Q2. Describe the four stages of product life cycle
Q3. What is the break-even point in capacity
expansion?
Homework for Chapter 1
P31 Q25
P48, Q41
The End!
1896 1920 1987 2006
Dr. GENG Na
Department of Industrial Engineering & Logistics Management
Chapter 1
Review
Management
Classification of Production System
Continuous flow production system
Discrete production system
Classified based on
production process
continuity:
Make to stock
Produce to order
Engineer to order
Classified based on
customization:
Manufacturing industry
Service industry
Classified based on
industry:
The Organization of Production System
A production system is usually divided into
basic units, e.g. workshop, manufacturing cell,
and working team, and all these units can be
organized on one of the following three basic
principles:
Process focused
Product focused
Group technology (GT)
Operation Management &Operation Strategy
Quality;
Delivery speed;
Flexibility;
Two Traditional Strategic Dimensions
Lower Cost
Product Differentiation
The Classic View of Operation Strategy
The traditional view of manufacturing strategy was put
forward by Wickham Skinner of the Harvard Business School.
Classical operations strategy thinking relates to the following
issues:
1. Time horizon
2. Focus
3. Evaluation
4. Consistency
Matching Process and Product Life Cycle
I
Low volume,
low
standardization,
one of a kind
II
Multiple
products, low
volume
III
Few major
products, higher
volume
IV
High volume, high
standardization,
commodity
products
I
Jumbled flow
(job shop)
Commercial
printer
Void
II
Disconnected line
flow
(batch)
Heavy
equipment
III
Connected line
flow
(assembly line)
Auto assembly
IV
Continuous flow
Void Sugar refinery
P
r
o
c
e
s
s

l
i
f
e
-
c
y
c
l
e

s
t
a
g
e
Process structure
To match appropriate industry in its mature phase with the appropriate
process.
Capacity Growth Planning -a long-term strategy
problem
Fig. 1-11 Break-Even Curves
Chapter 2 Forecasting
Contents
1. Introduction
2. The Time Horizon in Forecasting
3. Classification of Forecasts
4. Evaluating Forecast
5. Notation Conventions
6. Methods for Forecasting Stationary Series
7. Trend-Based Methods
8. Methods for Seasonal Series
What is forecasting?
Forecasting is the process of predicting the future
What do you want to forecast?
Career
Marriage
House price

Why forecast?
How forecast?
2.1. Introduction
Why forecast in a business factory?
All business planning is based on a forecast
What a factory should forecast?
Factors affecting the future success of a firm
Sales of existing products
Customer demand patterns for new products
Needs and availabilities of raw materials
Capacity requirements
International policies
2.1. Introduction
Examples of benefiting from good forecasting and paying
price from poor one:
2.1. Introduction
Detroit make slow respond to
customer tastes in automobiles from
heavy gas guzzlers to smaller and
more fuel efficient ones during 1960s,
such that it suffered much when
OPEC oil embargoing in late 1970
speed up the trend of shifting to
smaller cars.
2.1. Introduction
Compaq Computer became a
market leader in the early 1980s
by properly predicating
consumer demand for portable
version of the IBM PC.
2.1. Introduction
Ford Motors early success and later demise
It predicated that customer would want a simpler, less
expensive, and easier to be maintained car, and developed its
Model T car that dominated the market;
However, later, did not see that customer tired of the open
Model T design, and failed to forecast the customers desire for
other designs that almost caused the end of a firm that has
monopolized the industry only a few years ago.
Contents
1. Introduction
Fig.2-1 Forecast Horizons in Operation Planning
2.2. The Time Horizon in Forecasting
The Short-term forecasting is required for day-to-day planning;
Measured usually in day or weeks;
Required for inventory management, production plan, and
resource requirement planning, and shift scheduling
The Intermediate term is measured in weeks or months;
Typical intermediate term forecasting problems include sales
pattern of product families, requirements and availabilities of
workers, and resource requirements.
The long term is measured in months or years;
It is one part of the overall firms manufacturing strategy;
Problems for long term forecasting include long term capacity
planning; long term sales patterns, and growth trend.
One of example is long term planning of capacity.
Contents
1. Introduction
Subjective-based
on human
judgment
Objective-derived
from analysis of
data
Sales force composites;
Customer surveys;
Jury of executive opinion;
The Delphi method.
Causal Models
Time Series Methods
2.3. Classification of Forecasts
Subjective-
based on
human
judgment
Customer surveys;
The Delphi method.
Sale force is in a good position to see changes in their
preferences;
Numbers of the sale force submit sales estimates of the
products they sell in the next year;
Sales manager aggregate individual estimates.
2.3. Classification of Forecasts - Subjective
Subjective-
based on
human
judgment
Customer surveys;
The Delphi method.
It can signal the future trends and shifting preference patterns;
Survey and sampling plans must ensure statistically unbiased
resulting data and representative of the customer base.
Subjective-
based on
human
judgment
Customer surveys;
The Delphi method.
Experts opinion is the only source of information for forecasting,
when no past historic data, as with new products;
The approach is to combine the opinions of experts to derive a
forecast;
Subjective-
based on
human
judgment
Customer surveys;
The Delphi method.
Two ways for the combination.
One is to have individual responsible for preparing for the
forecast interview the executive directly and develop a forecast
from the result of the interview;
Another is to require the executive to meet as a group and
come to consensus.
Subjective-
based on
human
judgment
Customer surveys;
The Delphi method.
Named for the Delphic oracle of ancient Greece who had power
to predicate the future;
Based on collecting the opinion of experts like jury of executive
opinion;
Subjective-
based on
human
judgment
Customer surveys;
The Delphi method.
However, in different manner in which individual opinions are
combined in order to overcome some of the inherent shortcoming
of group dynamics. (The personalities of some group members
overshadow others)
Subjective-
based on
human
judgment
Customer surveys;
The Delphi method.
Procedures for the Delphi method:
Step 1. A group of experts express their opinions;
Step 2. The opinions are then compiled and a summary of results are
returned to the expert, with special attention to those significant
different opinions;
Subjective-
based on
human
judgment
Customer surveys;
The Delphi method.
Procedures for the Delphi method:
Step 3. The experts are asked if they wish to reconsider their original
opinions in light of the group response;
Step 4. The process is repeated until an overall group consensus is
reached.
The End!
Contents
1. Introduction
Subjective-based
on human
judgment
Objective-derived
from analysis of
data
Customer surveys;
The Delphi method.
Causal Models -the forecast for
a phenomenon is some function
of some variables
Time Series Methods -forecast
of future values of some
economic or physical
phenomenon is derived from a
collection of their past
observations
Causal Model
Let
Y-the phenomenon needed to be forecasted; (numbers of house sales)
X
1
, X
2
, , X
n
(interest rate of mortgage) are variables supposed to
be related to Y
Then, the general casual model is as follows:
Y=f(X
1
, X
2
, , X
n
).
Econometric models are lineal casual models:
Y=o
0
+ o
1
X
1
+ o
2
X
2
++ o
n
X
n,
,
where o
i
(i=1~n) are constants.
2.3. Classification of Forecasts - Objective
Causal Model
The method of least squares is most commonly used for finding
estimators of these constants.
Assume we have the past data (x
i
, y
i
), i=1~n; and the causal
model is simplified as Y=a+bX. Define
2
1
( , ) [ ( )]
n
i i
i
g a b y a bx
=
= +
as the sum of the squares of the distances from line a+bX to

data points y
i
.
Causal Model
2
1
( , ) [ ( )]
n
i i
i
g a b y a bx
=
= +
We may choose a and b to minimize

by letting
0
g
a
c
=
c
( )
1
2 0
=
+ = (

n
i i
i
y a bx
( ) ( )
1
0
=
+ = (

n
i i i
i
y a bx x 0
g
b
c
=
c
1 1
2
1 1
= =
= =

n n
i i i
i i
n n
i i
i i
x y y x
b
x x x
| |
1
1
=
= =
n
i i
i
a y bx y bx
n
Causal Model
2
1
( , ) [ ( )]
n
i i
i
g a b y a bx
=
= +
0
g
a
c
=
c
a y bx =
1 1 1 1
2 2
1 1 1 1
n n n n
i i i i i i
xy
i i i i
n n n n
xx
i i i i
i i i i
x y y x n x y ny x
S
b
S
x x x n x nx x
= = = =
= = = =

= = =

0
g
b
c
=
c
2
2
; ( )
n n n n n
xy i i i i xx i i
i i i i i
S n x y x y S n x x = =

1 1
; ;
n n
i i
i i
x x y y
n n
= =

Time Series Methods
The idea is that information can be inferred from the pattern of
past observations and can be used to forecast future values of the
series.
Try to isolate the following patterns that arise most often.
Trend-the tendency of a time series, usually a stable growth
or decline, either linear (a line) or nonlinear (described as
nonlinear function, e. g. a quadratic or exponential curve)
Seasonality-Variation of a series related to seasonal changes
and repeated every season.
Cycles-Cyclic variation similar to seasonality, except that the
length and the magnitude may change, usually associated with
economic variation.
Randomness-No recognizable pattern to the data.
Fig. 2-2 Time Series Patterns
Contents
1. Introduction
The forecast error e
t
in period t is the difference between the
forecast value for that period and the actual demand for that
period.
t t t
e F D =
The three measures for evaluating forecasting accuracy
during n period
2
1
1
n
i
i
MSE e
n
=
=

MAD: The mean absolute deviation, preferred method;
MSE: The mean squared error;
MAPE: The mean absolute percentage error (MAPE)
1
1
| |
n
i
i
MAD e
n
=
=

1
1
[ | / |] 100
n
i i
i
MAPE e D
n
=
=
2.4. Evaluating Forecast

Example 2.1
The production yields forecasts (in percent) of two managers in Artel:
Week P1 O1 |E1| |E1/O1| P2 O2 |E2| |E2/O2|
1 92 88 4 0.0455 96 91 5 0.0549
2 87 88 1 0.0114 89 89 0 0.0000
3 95 97 2 0.0206 92 90 2 0.0222
4 90 83 7 0.0843 93 90 3 0.0333
5 88 91 3 0.0330 90 86 4 0.0465
6 93 93 0 0.0000 85 89 4 0.0449
Example 2.1
MAD
1
=2.83 MAD
2
=3.00 Plant 1 better
MSE
1
=13.17 MSE
2
MAPE
1
=0.0325 MAPE
2
Both MAD
1
and MAPE
1
are less than
MAD
2
and MAPE
2
, however MSE
1
is larger
than MSE
2
?
MSE is more sensitive to one large error than is the MAD.
Plant 1 has got a large error 7.
Fig2-3 Forecast Errors Over Time
A desirable forecast should be unbiased. Mathematically,
E(e
i
)=0.
i
over time should fluctuate randomly above
and below zero.
Contents
1. Introduction
Define D
1
, D
2
, , D
t
, , as the observed values of demand
during periods 1, 2, , t, .
Assume if we are predicating demand for period t, we have
already observed D
t-1
, , but have no D
t
.
Define F
t
as the forecast for period t that is made at the end of
period t-1;
1
2 3 t-1 t
D
1
D
2
D
3
D
t-1
F
t
2.5. Notation Convention
The above is one-step-ahead forecast-they are made for
predicating the demand only in the next period.
A time series forecast is actually obtained by weighting the
past data
1 1 1 2
0
, , o o o

=
=
m
t t i t i t t
i
F D for some set of weights
1
2 3 t-1 t
D
1
D
2
D
3
D
t-1
F
t
2.5. Notation Convention
Contents
1. Introduction
Stationary time series: each observation can be represented by a
constant plus a random fluctuation.
t t
D c = +
where
= an unknown constant corresponding to mean of the series;
c= the random error with mean zero and variation o
2
.
Methods:
Moving average;
Exponential Smoothing
2.6. Methods of Forecasting Stationary Series
Moving Average -A moving average of order N is simply
the arithmetic average of the most recent N observations (one-
step-ahead), denoted as MA(N).
1 2
1
1 1
( )
N
t t i t t t N
i
F D D D D
N N

=
= = + + +

1
2 3 t-1 t
D
1
D
2
D
t-N
D
t-1
F
t
D
t-2
N past data
Example 2.2 Quarterly data for the failures of certain aircraft engines
at a local military base during the last two yrs are 200, 250, 175, 186,
225, 285, 305, 190. Determine the one-step-ahead forecasts for
period 4 through 8 using three-period moving averages, and one-
step-ahead forecasts for periods 7 and 8 using six-period moving
averages.
MA(3)
4
=(1/3)(175+250+200)=208
MA(3)
5
=(1/3)(186+175+250)=204
MA(6)
7
=(1/6)(285+225+186+175+250+200)=220
200, 250, 175, 186, 225, 285, 305, 190.
200, 250, 175, 186, 225, 285, 305, 190.
200, 250, 175, 186, 225, 285, 305, 190.
Quarter Demand MA(3) MA(6)
1 200
2 250
3 175
4 186 208
5 225 204
6 285 195
7 305 232 220
8 190 272 238
Quarter Demand
MA(3)
One-step-
Ahead
MA(3)
Two-step-
Ahead
MA(3)
3-step-
Ahead
1 200
2 250
3 175
4 186 208
5 225 204 208
6 285 195 204 208
7 305 232 195 204
8 190 272 232 195
Example 2.2
Question: Why the forecasts for different periods are the
same?
1
2 3 t-1 t
D
1
D
2
D
t-N
D
t-1
D
t-2
F
t
1SA
F
t+1
2SA
F
t+2
3SA
=
=
1 2
1
1 1
( )
N
t t i t t t N
i
F D D D D
N N

=
= = + + +

1 2 1 1
1
[ ]
+ +
+ + + = + +
t t t t N t t N t N
F
N
D D D D D D
Calculate F
t+1
based on F
t
-simplify the calculation
1 1 1 2 1
1
1 1
[ ]
+ + +
=
= = + + + +
N
t t i t t t t N
i
F D D D D D
N N

1
1
( )
+
= +
t t t t N
F F D D
N
Forecasting by MA for series with trend
Fig.2-4 Moving-Average Forecasts Lag Behind a Trend
Implication: the use of
simple moving averages
is not proper forecasting
method when there is a
trend in the series.
Forecasts lag
behind the
demand
Test: The demand of some product for the past six week is
known. Please forecast the demand in 7
th
week using MA. Note
that the right N may make the forecast more accurate.
Week Demand
1 92
2 87
3 95
4 90
5 88
6 93
In-Class Quiz
Answer: First, the forecasting errors are compared. When N=3,
MAD is smallest.
Week Demand MA(2) MA(3) MA(4)
1 92
2 87
3 95
4 90
5 88 93 91 91
6 93 89 91 90
MAD 4.5 2.5 3
Using MA(3) to predict the demand in period 7, the answer is 90.
The End!
Dr. GENG Na
Contents
1. Introduction
Exponential Smoothing-the current forecast is weighted
average of the last forecast and the last value of demand.
1 1
( ) (1 )(Last forecast)
(1 )
t t t
new forecast current observation of demand
F D F
o o
o o

= +
= +
where 0<os1 is the smoothing constant, which determines
the relative weight placed on the last observation of demand,
while 1- o is weight placed on the last forecast.
1 1
(1 ) o o

= +
t t t
F D F
1 1 1
( ) o

=
t t t
F F D
1 1
o

=
t t
F e
1 1
(1 )
t t t
F D F o o

= +
1 2 2
(1 )
t t t
F D F o o

= +
2
1 2 2
(1 ) (1 ) o o o o

= + +
t t t t
F D D F
2
1 2 2 1
0
(1 ) (1 ) ... (1 )
i
t t t t t i
i
F D D F D o o o o o o

=
= + + = =
Fig.2-5 Weights in Exponential Smoothing

The older of a past data, the
smaller of its contribution to
the forecast for a future
period.
(1 )
0.1
i
o o
o

=
Example 2.3
Consider Example 2.2, in which the observed number of failures
over a two yrs period are 200, 250, 175, 186, 225, 285, 305, 190. We
will now forecast using exponential smoothing. We assume that the
forecast for period 1 was 200, and suppose that o=0.1
F
2
= ES(0.1)
2
= oD
1
+(1- o)F
1
=0.1200+(1-0.1) 200=200
F
3
= ES(0.1)
3
= oD
2
+(1- o)F
2
=0.1250+(1-0.1) 200=205
Assume that the initial forecast is equal to the initial value of
demand to get the method start. Drawback???
Fig.2-6 Exponential Smoothing for Different Values of Alpha
Smaller o turns out a
stable forecast, while
larger o results in better
track of series
Comparing of ES and MA
Similarities
Both methods are based on assumption that underlying demand
is stationary;
Both methods depend on a single parameters;
Both methods will lag behind a trend if one exits.
Differences
MA is better than EA in that it needs only past N data, while
EA needs all the past data.
Contents
1. Introduction
2.7. Trend-Based Methods
Holts Method
Regression
analysis
A method that fits a
straight line to a set of
data
Double exponential
smoothing
Regression Analysis
Let (x
1
, y
1
), (x
2
, y
2
), , (x
n
, y
n
) are n paired data points for the
two variables X and Y; and
Assume that y
i
is the observed value of Y when x
i
is the
observed value of X.
It is believed that there is a relationship between X and Y as
follows
Y a bX = +
Represents the predicated value of Y;
a and b are chosen to minimize the sum
of squared distance between regression line
and the data point
Fig 2-7 An Example of a Regression Line
Y a bX = +
(i, D
i
)
Regression Analysis
2
1
( , ) [ ( )]
n
i i
i
g a b y a bx
=
= +
0
g
a
c
=
c
a y bx =
=
xy
xx
S
b
S
0
g
b
c
=
c
( )
1 1 1 1 1
1
2
= = = = =
+
= =

n n n n n
i i
i
xy i i i i
i i i i
S n x y x
n n
n iD D y
( )
1 / 2 = + a D b n
( )( ) ( )
2 2
1 1
2
2 2
1 2 1 1
6
( )
4
= =
=
+ + +
=

n n
xx i i
i i
n n n n n
S n x x
Example 2.4
Consider Example 2.2 & 2.3, in which the observed number of
failures over a two yrs period are 200, 250, 175, 186, 225, 285, 305,
190. Suppose that we use the first five periods as a baseline in order
to estimate our regression parameters. Then
( )
1 1
1
2
5[200 250*2 175*3 186*4 225*5]
[5*6]/ 2*[200 250 175 186 225]
70
= =
+
=
= + + + +
+ + + +
=

n n
xy i i
i i
n n
S n iD D
( )( ) ( )
2
2 2
1 2 1 1
50
6 4
+ + +
= =
xx
n n n n n
S
7/ 5 = =
xy
xx
S
b
S
( )
1 / 2 211.4 = + = a D b n
211.4 1.4 =
t
F t
8
200 = F
Holts method
A type of double exponential smoothing designed to track
time series with linear trend. The method requires the
specification of two constant o and |, and use two smoothing
equations:
The value of series (the intercept)
The value of trend (the slope)
The -step-ahead forecast made in period t, is given by
( )( )
1 1
1
t t t t
S D S G o o

= + +
( ) ( )
1 1
1
t t t t
G S S G | |

= +
, t t t t
F S G
t
t
+
= +
Example 2.5
The original series was 200, 250, 175, 186, 225, 285, 305, 190.
Assume that both and are equal to 0.1. In order to get the
method started, we need estimates of both the intercept and the
slope at time zero. Suppose that these are S0=200 and G0=10. Then
we obtain
S1=0.1*200+0.9*(200+10)=209
G1=0.1*(209-200)+0.9*10=9.9
S2=0.1*250+0.9*(209+9.9)=222
G2=0.1*(222-209)+0.9*9.9=10.2
F2,5=222+3*10.2=252.6
( )( )
1 1
1
t t t t
S D S G o o

= + +
( ) ( )
1 1
1
t t t t
G S S G | |

= +
, t t t t
F S G
t
t
+
= +
Comparison of Holts method and ES/MA
Holts method is explicitly designed to track the trend in the data,
whereas simple exponential smoothing (ES) and moving averages
(MA) are not.
Comparison of Holts method and regression analysis
Both are designed to handle series that exhibit trend;
With Holts method, it is far easier to update forecasts as new
observations become available.
The original series was 200, 250, 175, 186, 225, 285,
305, 190. At the end of quarter 5, please forecast the
demand for quarter 6~8 by using MA(3), EA(0.1),
Regression analysis, and Holt(0.1,0.1).
MA
6
(3)=MA
7
(3)=MA
8
(3)=(175+186+225)/3=195
Suppose F
5
=203, EA
6
(0.1)=0.1*225+0.9*203=205
The original series was 200, 250, 175, 186, 225, 285,
305, 190. At the end of quarter 5, please forecast the
demand for quarter 6~8 by using MA(3), EA(0.1),
Regression analysis, and Holt(0.1,0.1).
RA: F
t
=211.4-1.4t, F
6
=203, F
7
=202, F
8
=200
Holt (0.1,0.1): Suppose S0=200and G0=10,
S5=238and G5=9
F
6
=248, F
7
=257, F
8
=266
( )( )
1 1
1
t t t t
S D S G o o

= + +
( ) ( )
1 1
1
t t t t
G S S G | |

= +
, t t t t
F S G
t
t
+
= +
Contents
1. Introduction
A seasonal series is one that has a pattern that repeats every N
periods (at least 3).
Fig. 2-8 A Seasonal Demand Series
Length of season-the
number of periods before
the pattern begins to
repeat
A season
2.8. Methods for Seasonal Series
How to represent seasonality?
Seasonal factor-A set of multipliers c
t
, 1s t sN;
c
t
represents the average amount that the demand in t
th
period of
the season is above or below the average.
For example: if c
3
=1.25 and c
5
=0.6, then the demand in the 3
rd
period is 25 percent above the average demand; while demand in
the 5
th
period is 40 percent below the average demand.
2.8. Methods for Seasonal Series
c
t
=N
2.8. Methods of Forecasting Stationary Seasonal Series
Seasonal Factors for Stationary Seasonal Series (No trend)
Compute the sample mean of all data (A minimum of two
seasons of data is required): m
Divide each observation by sample mean the seasonal
factors of observed data for each period in each season:
SF
i,j
=D
ij
/m (D
ij
-the observed dada for period j in season i, total H
seasons)
Average the factors for the same periods within each season
the seasonal factor:
Multiplying the sample mean
by a seasonal factor the
forecast of demand in the
corresponding period of the season.
1
1
H
j ij
i
SF SF
H
=
=

Wk1 Wk2 Wk3 Wk4
Monday 16.2 17.3 14.6 16.1
Tuesday 12.2 11.5 13.1 11.8
Wednesday 14.2 15 13 12.9
Thursday 17.3 17.6 16.9 16.6
Friday 22.5 23.5 21.9 24.3
Compute the sample
mean of all demands:
m= 16.425
Wk1 Wk2 Wk3 Wk4
Mon 16.2 17.3 14.6 16.1
Tues 12.2 11.5 13.1 11.8
Wedn 14.2 15 13 12.9
Thurs 17.3 17.6 16.9 16.6
Fri 22.5 23.5 21.9 24.3
Divide each observation by sample mean m= 16.425
Wk1 Wk2 Wk3 Wk4
Mon 0.99 1.05 0.89 0.98
Tues 0.74 0.70 0.80 0.72
Wedn 0.86 0.91 0.79 0.79
Thurs 1.05 1.07 1.03 1.01
Fri 1.37 1.43 1.33 1.48
Average the factors for the same periods within each
season
Wk1 Wk2 Wk3 Wk4 SF
Mon 0.99 1.05 0.89 0.98
Tues 0.74 0.70 0.80 0.72
Wedn 0.86 0.91 0.79 0.79
Thurs 1.05 1.07 1.03 1.01
Fri 1.37 1.43 1.33 1.48
Wk1 Wk2 Wk3 Wk4 SF
Mon 0.99 1.05 0.89 0.98 0.98
Tues 0.74 0.70 0.80 0.72 0.74
Wedn 0.86 0.91 0.79 0.79 0.84
Thurs 1.05 1.07 1.03 1.01 1.04
Fri 1.37 1.43 1.33 1.48 1.40
SF
Mon 0.98
Tues 0.74
Wedn 0.84
Thurs 1.04
Fri 1.40
SF Forecast
Mon 0.98 16.05
Tues 0.74 12.15
Wedn 0.84 13.775
Thurs 1.04 17.1
Fri 1.40 23.05
Multiplying the sample mean by a seasonal factor
the forecast of demand in the corresponding period of the
season
In-class Quiz
Q1. Describe the difference between exponential
smoothing and double exponential smoothing.
Q2. What is the seasonal factor? What kind of
property this factor has?
Q3. Describe regression analysis.
The End!
Dr. GENG Na
Contents
1. Introduction
2.8. Methods of Forecasting Trend Seasonal Series
Seasonal Decomposition Using Moving Averages
A more complex time series: Trend + Seasonality
Winters Method (triple exponential smoothing)
Deseaonalized- Get
seasonality away;
Make forecast on
deseaonalized data;
Get seasonality
back
2.8. Seasonal Decomposition Using Moving Averages
Example 2.7
Suppose that original demand history of a certain item for the past
eight quarters is given by 10, 20, 26, 17, 12, 23, 30, 22. The graph
of this demand is shown in the following figure.
Procedures:
Draw the demand curves and estimate the season length N;
Computer the moving average MA(N);
Centralize the moving averages;
Get the centralized MA values back on period;
Calculate seasonal factors, and make sure of c
t
=N.
Divide each observation by the appropriate seasonal factor to
obtain the deseasonalizeddemand
Forecast is made based on deseasonalizeddemand.
Final forecast is obtained by multiplying the forecast (with no
seasonality) with seasonal factors.
Fig. 2-9 Demand History for Example 2.7
N=4
Procedures:
t
=N.
Period Demand MA(4)
1 10
2 20
3 26
4 17
5 12 18.25
6 23 18.75
7 30 19.5
8 22 20.5
21.75
Procedures:
t
=N.
Period Demand MA(4) CMA
1 10
2 20
18.25
3 26
18.75
4 17
19.5
5 12 18.25
20.5
6 23 18.75
21.75
7 30 19.5
8 22 20.5
21.75
Since 18.25 is calculated by
moving demand of periods 1, 2,
3, and 4. The center of these
periods is 2.5
Procedures:
t
=N.
Period Demand MA(4) CMA BCMA
1 10
2 20
18.25
3 26 18.5
18.75
4 17 19.125
19.5
5 12 18.25 20
20.5
6 23 18.75 21.125
21.75
7 30 19.5
8 22 20.5
21.75
1 10
2 20
18.25
3 26 18.5
18.75
4 17 19.125
19.5
5 12 18.25 20
20.5
6 23 18.75 21.125
21.75
7 30 19.5
8 22 20.5
21.75
1 10 18.8125
2 20 18.8125
18.25
3 26 18.5
18.75
4 17 19.125
19.5
5 12 18.25 20
20.5
6 23 18.75 21.125
21.75
7 30 19.5 20.5625
8 22 20.5 20.5625
21.75
Procedures:
t
=N.
PeriodDemand BCMA Ratio
1 10 18.813 0.532
2 20 18.813 1.063
3 26 18.5 1.405
4 17 19.125 0.889
5 12 20 0.600
6 23 21.125 1.089
7 30 20.563 1.459
8 22 20.563 1.070
PeriodDemand BCMA Ratio
1 10 18.813 0.532
2 20 18.813 1.063
3 26 18.5 1.405
4 17 19.125 0.889
5 12 20 0.600
6 23 21.125 1.089
7 30 20.563 1.459
8 22 20.563 1.070
Period Factor
1 0.566
2 1.076
3 1.432
4 0.98
c
t
=4.053??
Normalized by
multiply each factor
by 4/4.053=0.9869
Period Factor
1 0.558
2 1.062
3 1.413
4 0.967
Procedures:
t
=N.
Period DemandFactor
Deseasonalized
Demand
1 10 0.558 17.91
2 20 1.062 18.84
3 26 1.413 18.40
4 17 0.967 17.59
5 12 0.558 21.49
6 23 1.062 21.66
7 30 1.413 21.23
8 22 0.967 22.76
Procedures:
t
=N.
Forecast made on Deseasonalized demand by MA
MA(6) for period 9=20.52 (without seasonality);
Forecast for period: 20.520.558=11.45 (with
seasonality)
Question:
How to forecast demand for period 10, 11, 12?
F
10
= MA(6) of 2-step-head=F
9
=20.52
F
10
SF
2
=20.521.061=21.77
Period Factor
1 0.558
2 1.062
3 1.413
4 0.967
Forecast made on Deseasonalized demand by Regression
analysis
F
t
=16.8+0.7092t
Period
Deseasonalized
Forecast
Factor
Seasonalized
Forecast
9 23.18 0.57 13.12
10 23.89 1.08 25.71
11 24.60 1.43 35.23
12 25.31 0.98 24.79
Period Demand MA(4) CMA BCMA Ratio SF
Deseasonalized
Demand
1 10 18.8125 0.532 0.558 17.91
2 20 18.8125 1.063 1.062 18.84
18.25
3 26 18.5 1.405 1.413 18.40
18.75
4 17 19.125 0.889 0.967 17.59
19.5
5 12 18.25 20 0.600 0.558 21.49
20.5
6 23 18.75 21.125 1.089 1.062 21.66
21.75
7 30 19.5 20.5625 1.459 1.413 21.23
8 22 20.5 20.5625 1.070 0.967 22.76
21.75
Summation
2.8. Winters Method for Seasonal Problems
Winters Method
The moving-average method requires that all seasonal factors
be recalculated from scratch as new data become available.
Winters method is a type of triple exponential smoothing.
Easy to update as new data become available.
Assumptions:
Model of the form:
The length of the season is exactly N periods and the seasonal
factors are the same each season and have the property that
( )
t t t t
D G c c = + +
t
c N =

The series. The current level of deseasonalizedseries S
t
The trend.
The seasonal factors
The forecast made in period t for any future period t+ is given
by (assume that <= N)
If N < <= 2N, the seasonal factor would be ct+-2N
If 2N < <= 3N, the seasonal factor would be ct+-3N
and so on
( ) ( )( )
1 1
1
t t t N t t
S D c S G o o

= + +
( ) ( )
1 1
1
t t t t
G S S G | |

= +
( ) ( )
1
t t t t N
c D S c

= +
( )
, t t t t t N
F S G c
t t
t
+ +
= +
The End!
Dr. GENG Na
Contents
1. Introduction
Deseaonalized- Get
seasonality away;
Make forecast on
deseaonalized data;
Get seasonality
back
Procedures:
t
=N.
obtain the deseasonalized demand
Forecast is made based on deseasonalized demand.
Period Demand MA(4) CMA BCMA Ratio SF
Deseasonalized
Demand
1 10 18.8125 0.532 0.558 17.91
2 20 18.8125 1.063 1.062 18.84
18.25
3 26 18.5 1.405 1.413 18.40
18.75
4 17 19.125 0.889 0.967 17.59
19.5
5 12 18.25 20 0.600 0.558 21.49
20.5
6 23 18.75 21.125 1.089 1.062 21.66
21.75
7 30 19.5 20.5625 1.459 1.413 21.23
8 22 20.5 20.5625 1.070 0.967 22.76
21.75
Summation
Winters Method
The moving-average method requires that all seasonal factors
be recalculated from scratch as new data become available.
Winters method is a type of triple exponential smoothing.
Easy to update as new data become available.
Assumptions:
Model of the form:
The length of the season is exactly N periods and the seasonal
factors are the same each season and have the property that
( )
t t t t
D G c c = + +
t
c N =

The series. The current level of deseasonalized series S
t
The trend.
The seasonal factors
The forecast made in period t for any future period t+ is given
by (assume that <= N)
If N < <= 2N, the seasonal factor would be ct+-2N
If 2N < <= 3N, the seasonal factor would be ct+-3N
and so on
( ) ( )( )
1 1
1
t t t N t t
S D c S G o o

= + +
( ) ( )
1 1
1
t t t t
G S S G | |

= +
( ) ( )
1
t t t t N
c D S c

= +
( )
, t t t t t N
F S G c
t t
t
+ +
= +
Initialization Procedure
Assume that exactly two seasons of data are available; that is, 2N
data points. Suppose that the current period is t = 0, so that the past
observations are labeled D-2N+1, D-2N+2 , , D0
Step 1: Calculate the sample means for the two seasons
Step 2: Define the initial slope estimate
Step 3: Set the initial value of the series
0
1 2
2 1 1
1 1
,
N
j j
j N j N
V D V D
N N
= + = +
= =

( )
0 2 1
G V V N =
( )
0 2 0
1 2 S V G N = + (

Initialization Procedure
Step 4(a): Calculate the initial seasonal factors
where i=1 for the first season, i=2 for the second season, j is the
period of the season. That is, j=1 for t=-2N+1 and t=-N+1; j=2
for t=-2N+2 and t=-N+2, and so on.
Step 4(b): Average the seasonal factors
Step 4(c): Normalize the seasonal factors
( )
0
for 2 1 0
1 2
t
t
i
D
c N t
V N j G
= + s s
+ (

2 1 1 0
1 0
, ... ,
2 2
N N N
N
c c c c
c c
+ +
+
+ +
= =
0
1
for 1 0
j j i
i N
c c c N N j
= +
(
= + s s
(

Example 2.8
We have two seasons of data: 10, 20, 26, 17, 12, 23, 30, 22.
Step 1:
V1=(10+20+26+17)/4=18.25; V2=(12+23+30+22)/4=21.75
Step 2:
G0=(21.75-18.25)/4=0.875
Step 3:
S0=21.75+0.875*1.5=23.06
0
1 2
2 1 1
1 1
,
N
j j
j N j N
V D V D
N N
= + = +
= =

( )
0 2 1
G V V N =
( )
0 2 0
1 2 S V G N = + (

Example 2.8
Step 4(a):
c-7=10/[18.25-(5/2-1)*0.875]=0.5904
c-6=20/[18.25-(5/2-2)*0.875]=1.123
c-5=1.391, c-4=0.869, c-3=0.5872
c-2=1.079, c-1=1.352, c0=0.9539
Step 4(b): c-3=0.5888, c-2=1.1010, c-1=1.3720, c0=0.9115
Step 4(c): c-3=0.5900, c-2=1.1100, c-1=1.3800, c0=0.9200
( )
0
for 2 1 0
1 2
t
t
i
D
c N t
V N j G
= + s s
+ (

Example
Step 2: G0=(21.75-18.25)/4=0.875
Step 3: S0=21.75+0.875*1.5=23.06
Step 4(c): c-3=0.5900, c-2=1.1100, c-1=1.3800, c0=0.9200
Suppose that we wish to forecast the following years demand at
time t=0. The forecasting equation is
which results in
F0,1=(S0+G0)c-3=(23.06+0.875)*0.59=14.12
F0,2=(S0+2G0)c-2=(23.06+2*0.875)*1.11=27.54
F0,3=35.44, F0,4=24.38
( )
, t t t t t N
F S G c
t t
t
+ +
= +
Example 2.8
Step 2: G0=(21.75-18.25)/4=0.875
Step 3: S0=21.75+0.875*1.5=23.06
Step 4(c): c-3=0.5900, c-2=1.1100, c-1=1.3800, c0=0.9200
Now, suppose that at time t=1 we observe a demand of D1=16. We
now need to update our equations. Assume that =0.2, =0.1, and
=0.1. Then
S1=(D1/c-3)+(1-)(S0+G0)=0.2*(16/0.59)+0.8*(23.06+0.875)=24.57
G1=(S1-S0)+(1-)G0=0.1*(24.57-23.06)+0.9*0.875=0.9385
c1=(D1/S1)+(1-)c-3=0.1*(16/24.57)+0.9*0.59=0.5961
Example 2.8
S1=24.57, G1=0.9385, c1=0.5961
Recall c-3=0.59, c-2=1.11, c-1=1.38, c0=0.92
Renormc-3=0.59, c1=0.5961, they are close enough.
Forecasting from period 1, we obtain
F1,2=(S1+G1)c-2=(24.57+0.9385)*1.11=28.3144
F1,3=(S1+2G1)c-1=(24.57+2*0.9385)*1.38=36.4969
Review
1. Introduction
Subjective-based
on human
judgment
Objective-derived
from analysis of
data
Customer surveys;
The Delphi method.
Causal Models
Time Series Methods
t
in period t is the difference between the
forecast value for that period and the actual demand for that
period.
t t t
e F D =
The three measures for evaluating forecasting accuracy
during n period
2
1
1
n
i
i
MSE e
n
=
=

MAD: The mean absolute deviation, preferred method;
MSE: The mean squared error;
MAPE: The mean absolute percentage error (MAPE)
1
1
| |
n
i
i
MAD e
n
=
=

1
1
[ | / |] 100
n
i i
i
MAPE e D
n
=
=

Stationary time series: each observation can be represented by a
constant plus a random fluctuation.
t t
D c = +
where
= an unknown constant corresponding to mean of the series;
c= the random error with mean zero and variation o
2
.
Methods:
Moving average;
Holts Method
Regression
analysis
A method that fits a
straight line to a set of
data
Double exponential
smoothing
In-class discussion
Demand v.s. sales???
How to forecast the demand in presence of lost
sales?
Homework for Chapter 2:
Q13, Q24, Q34, Q35
The End!
Dr. GENG Na
Production and Operation Managements
Dr Na GENG
Management
Chapter 3 Aggregate Planning
Content of this course
Aggregate plan
purchase order
Inventory Control
unknown demand
Push and Pull Production
Control Systems
J IT
Project Scheduling
unknown demand
Scheduling in Supply
Chain Management
Transportation problem
Contents
Introduction
Aggregate Units of Production;
Costs in Aggregate Planning;
A Prototype Problem;
Solution of Aggregate Planning Problem by LP
Introduction
Aggregate planning (AP), also called macro
planning, decides how many employees the firm
should retain and, for a manufacturing firm, the
quantity and the mix of products to be produced.
Macro planning strategies are a fundamental part of
the firms overall business strategy.
Introduction
AP begins with the forecast of demand.
Assumption for AP: demand is deterministic, or
known in advance.
The assumption is made to simplify the analysis and
allow us to focus on the systematic or predictable
changes in the demand pattern.
Introduction
Aggregate planning involves several basic competing objectives.
Make frequent and large changes in size of labor force: a
chase strategyto react quickly to anticipated changes in
demand, cost effective, but a poor long-term strategy;
Retain a stable workforce: results in larger buildups of
inventory during period of low demand ;
Develop a production plan for a firm to maximize profit over
the planning horizon subject to constraints on capacity.
AP methodology is designed to translate demand forecast into
blueprint for planning staffing and production level for a firm
over a predetermined planning horizon;
Contents
Introduction
Aggregate Units of Production
AP is predicated on the existence of an aggregate unit of
production.
Describe aggregate units in the following situations:
In terms of average item-when the items produced are
similar, e.g. cars, computers;
In terms of weights (tons of steel), volume(gallons of
gasoline), amount of work required (worker-years of
programming time), and dollar value (value of inventory in
dollars)-when many kinds of items are produced;
Appropriate aggregating schema are determined by context
of the particular planning problem and the level of the
aggregation to be required.
Example 3.1: Decide on aggregating schema for the manager of a
plant that produces six models of washing machines to determine
the workforce and production levels.
Model Number
Number of Worker-
Hours Required to
Produce
Selling Price
Sales
Rate(%)
A5532 4.2 285 32
K4242 4.9 345 21
L9898 5.1 395 17
L3800 5.2 425 14
M2624 5.4 525 10
M3880 5.8 725 6
One possibility is to define aggregate unit as one dollar of output
Unfortunately, it is impossible since the selling prices of the
various models are not consistent with the number of worker-
hours required to produce them.
Given amount of dollars of output, we could not know the total
WH required.
Model Number
Number of
Worker-
Hours
Selling Price Sales Rate(%)
A5532 4.2 285 32
K4242 4.9 345 21
L9898 5.1 395 17
L3800 5.2 425 14
M2624 5.4 525 10
M3880 5.8 725 6
4.2/285=0.0147WH/$
5.4/525=0.008WH/$
Fact: the percentages of the total number of sales for these six
models have been fairly constant
Define the aggregate unit of production as a fictitious washing
machinerequiring .324.2+.21 4.9+.17 5.1+.14 5.2+.10
5.4+.06 5.8=4.856 hrsof labor time.
Problem: Given number of washing machines to be produced,
say 10,000, could we know the total WH required?
Model Number
Number of Worker-
Hours Required to
Produce
Selling Price Sales Rate(%)
A5532 4.2 285 32
K4242 4.9 345 21
L9898 5.1 395 17
L3800 5.2 425 14
M2624 5.4 525 10
M3880 5.8 725 6
Defining an aggregate unit of production at higher level of the
firm is more difficult;
When the firm produces a large of products, a natural aggregate
unit is sales dollars;
Aggregate planning is closely related to hierarchical production
planning (HPP). HPP considers workforce sizes and production
rates at different levels of the firm. The recommended
hierarchy is as follows:
Items-correspond to individual models of washing
machines;
Families-a group of items, e.g. all washing machines;
Types-groups of families, e.g. large house appliances.
Contents
Introduction
Costs in Aggregate Planning
(1) Smoothing cost: Occurs as result of changing the production
level from one period to the next.
Cost for changing size of workforce-advertise positions;
interview prospective employees, and training new hires;
Assumed to be linear;
Fig 3-2 Cost of
Changing the Size of
the Workforce
(2) Holding costs: Occurs as a result of having capital tied up in
inventory.
Always assumed to be linear in the number of units being held at a particular point
in time;
For aggregate planning, it is expressed in terms of dollars per unit held per planning
period; (e.g. 100 $/month for one item)
Charged against the inventory remaining on hand at the end of the planning period;
Negative--Back-orders
Slope = C
P
Slope = C
i
$

C
o
s
t
Positive--Inventory
Fig.3-3 Holding and Back-Order Costs
At the beginning: 50
Week 1: In 100
Week 2: Out 130
Week 3 In 30
Week 4 Out 30
Charge only 20
(3) Shortage costs-
Shortage occurs when demands are higher than anticipated;
For aggregate planning, it is assumed that excess demand is
backlogged and filled in a future period;
In a highly competitive situation, the excess demand may be lost---
lost sales;
Generally considered to be linear.
Negative--Back-orders
Slope = C
P
Slope = C
i
$

C
o
s
t
Positive--Inventory
Fig.3-3 Holding and Back-Order Costs
(4) Regular time costs: Involve the cost of producing one unit of
output during regular working hours;
Actual payroll cost of regular employees working on regular time;
Direct and indirect costs of materials;
Other manufacturing expense;
(5) Overtime and subcontracting costs: Costs of production units
not produced on regular time;
Overtime-production by regular-time employees beyond work day;
Subtracting-the production of items by an outside supplier;
Assumed to be linear;
(6) Idle time costs: Underutilization of workforce;
In most contexts, the idle time cost is zero;
Idle time may have other consequences for the firm, e.g. if the aggregate units
are input to another process, idle time on the line could result in higher costs to
subsequent processes.
Contents
Introduction
A Prototype Problem
Example 3.2
Densepackis to plan workforce and production level for
six-month period J an. to J une. The firm produces a line
of disk drives for mainframe computers. Forecast
demand over the next six months for a particular line of
drives in a plant are 1,280, 640, 900, 1,200, 2,000 and
1,400. There are currently (end of Dec.) 300 workers
employed in the plant. Ending inventory in Dec. is
expected to be 500 units, and the firm would like to have
600 units on hand at the end of J une.
A Prototype Problem
How to incorporate the starting and ending inventory constraints
into formulation?-the simplest way is to modify the values of the
predicated demand;
The net demand in period 1 =(the predicated demand)
(initial inventory);
The net demand in Period T=thepredicated demand + the
minimum ending inventory
Period 1 Period 2 Period 3 Period 4 Period 5
Demand 30 50 70 60 80
Initial
Inventory=10
MinimumEnding
Inventory=20
Actual Demand 20 50 70 60 100
A Prototype Problem
How to handle minimum buffer inventories required?
-By modifying the predicted demand.
If there is a minimum buffer inventory required in every period,
this amount should be added only to the first periods demand.
Demand 30 50 70 60 80
MinimumInv. 10
10 10 10
10
Problem: Why should we require the minimum buffer
inventory?
A Prototype Problem
How to handle minimum buffer inventories?
-By modifying the predicted demand. (Cont.)
If there is a minimum buffer inventory required in only one period,
this amount should be added to that periods demand, and subtracted
from the next periods demand;
Demand 30 50 70 60 80
MinimumInv.
10
A Prototype Problem
Example 3.2
A Prototype Problem
Month Predicated
Demand
Net
Predicated
Demand
Net Cumulative
Demand
Jan. 1,280 780(1280-500) 780
Feb. 640 640 1,420
March 900 900 2,320
April 1,200 1,200 3,520
May 2,000 2,000 5,520
June 1,400 2,000(1400+600) 7,520
A Prototype Problem
If the shortage is not permitted, the cumulative production must
be at least as great as cumulative net demand each period-Feasible
AP
Fig. 3-4 A Feasible Aggregate Plan for
Densepack
AP: Any curve above the
accumulated net demand
A Prototype Problem
How to make cost trade-offs of various production
plans?
Only consider three costs:
C
H
=Cost of hiring one worker=$500;
C
F
=Cost of firing one worker=$1,000;
C
I
=Cost of holding one unit of inventory for one
month=$80
A Prototype Problem
Translate aggregate production in units to
workforce levels:
Use a day as an indivisible unit of measure (since not all
month have equal number of working days) and define:
K=Number of aggregate units produced by one worker in
one day.
A known fact: over 22 working days, with the workforce
constant at 76 workers, the firm produced 245 disk drives.
Average production rate=245/22=11.1364 disk drives per
day;
One worker produced an average of 11.1364/76=0.14653
drive in one day. K=0.14653.
A Prototype Problem
Two alternative plans for managing workforce:
Plan 1 is to change workforce each month in order to
produce enough units to most closely match the demand
pattern-zero inventory plan;
Plan 2 is to maintain the minimum constant workforce
necessary to satisfy the net demand-constant workforce
plan;
In-class Quiz
Describe the chase strategy and constant workforce
strategy
Describe the feasible solution of AP
The End!
Dr. GENG Na
Dr Na GENG
Management
Contents
Introduction
A Prototype Problem
Example 3.2
A Prototype Problem
Month Predicated
Demand
Net
Predicated
Demand
Net Cumulative
Demand
Jan. 1,280 780(1280-500) 780
Feb. 640 640 1,420
March 900 900 2,320
April 1,200 1,200 3,520
May 2,000 2,000 5,520
June 1,400 2,000(1400+600) 7,520
A Prototype Problem
If the shortage is not permitted, the cumulative production must
be at least as great as cumulative net demand each period-Feasible
AP
Fig. 3-4 A Feasible Aggregate Plan for
Densepack
AP: Any curve above the
accumulated net demand
A Prototype Problem
How to make cost trade-offs of various production
plans?
Only consider three costs:
C
H
=Cost of hiring one worker=$500;
C
F
=Cost of firing one worker=$1,000;
C
I
=Cost of holding one unit of inventory for one
month=$80
A Prototype Problem
Translate aggregate production in units to
workforce levels:
Use a day as an indivisible unit of measure (since not all
month have equal number of working days) and define:
K=Number of aggregate units produced by one worker in
one day.
A known fact: over 22 working days, with the workforce
constant at 76 workers, the firm produced 245 disk drives.
Average production rate=245/22=11.1364 disk drives per
day;
One worker produced an average of 11.1364/76=0.14653
drive in one day. K=0.14653.
A Prototype Problem
Two alternative plans for managing workforce:
Plan 1 is to change workforce each month in order to
produce enough units to most closely match the demand
pattern-zero inventory plan;
Plan 2 is to maintain the minimum constant workforce
necessary to satisfy the net demand-constant workforce
plan;
A Prototype Problem
P1: Zero Inventory Plan (Chase Strategy)
minimize inv. level.
A B C
Month No. of
Working
Days
No. of Units Produced
per Worker (BK)
Jan. 20 2.931
Feb. 24 3.517
March 18 2.638
April 26 3.810
May 22 3.224
June 15 2.198
Table 3-1 Initial Calculation for Zero Inv. Plan for Denspack
A Prototype Problem
A B C D
Month No. of
Working
Days
per Worker (BK)
Forecast Net
Demand
Jan. 20 2.931 780
Feb. 24 3.517 640
March 18 2.638 900
April 26 3.810 1,200
May 22 3.224 2,000
June 15 2.198 2,000
A Prototype Problem
A B C D E
Month No. of
Working
Days
per Worker (BK)
Forecast Net
Demand
Minimum No. of
Worker required
(D/C rounded up)
Jan. 20 2.931 780 267
Feb. 24 3.517 640 182
March 18 2.638 900 342
April 26 3.810 1,200 315
May 22 3.224 2,000 621
June 15 2.198 2,000 910
A Prototype Problem
The number of workers employed at the end of Dec. is 300;
Hiring and firing workers each month to match forecast
demand.
A B C D E F G H I
Month No. of
Workers
No.
Hired
No.
Fired
Jan. 267 33
Feb. 182 85
March 342 160
April 315 27
May 621 306
June 910 289
Total 755 145
Table 3-2 Zero Inv. Aggregate Plan for Densepack
A Prototype Problem
demand.
A B C D E F G H I
Month No. of
Workers
No.
Hired
No.
Fired
No. of
Units per
Worker
No. of
Units
Produced
(BE)
Cumulative
Production
Jan. 267 33 2.931 783 783
Feb. 182 85 3.517 640 1,423
Marc
h
342 160 2.638 902 2,325
April 315 27 3.810 1,200 3,525
May 621 306 3.224 2,002 5,527
June 910 289 2.198 2,000 7,527
Total 755 145
A Prototype Problem
demand.
A B C D E F G H I
Month No. of
Workers
No.
Hired
No.
Fired
No. of
Units per
Worker
No. of
Units
Produced
(BE)
Cumulative
Production
Cumulative
Demand
Ending
Inv.
(G-H)
Jan. 267 33 2.931 783 783 780 3
Feb. 182 85 3.517 640 1,423 1,420 3
Marc
h
342 160 2.638 902 2,325 2,320 5
April 315 27 3.810 1,200 3,525 3,520 5
May 621 306 3.224 2,002 5,527 5,520 7
June 910 289 2.198 2,000 7,527 7,520 7
Total 755 145 30
A Prototype Problem
The total cost of this production plan is obtained by multiplying
the totals at the bottom of Table 3-2 by corresponding unit cost:
755500+145 1000+30 80=$524,900;
In addition, the cost of holding for the ending inventory of 600
units, which was considered as the demand for J une, should be
included in holding cost: 600 80=$48,000
The total cost= $524,900+$48,000=$572,900.
Note: the initial inventory of 500 units does not enter into the
calculation because it will be netted out during the month J anuary.
A Prototype Problem
It is possible to achieve zero inventory at the
end of each planning period ?
No! Since it is impossible to have a fractional
number of workers.
It is possible that ending inventory in one or
more period could build up to a point where the
size of the workforce may be reduced by one or
more workers.
A Prototype Problem
P2 Evaluation of the Constant Workforce Plan-to eliminate
completely the need for hiring and firing during the planning
horizon.
In order not incur the shortage in any period, compute the
minimum workforce required for every month in the planning
horizon.
For J anuary, the net cumulative demand is 780 and units
produced per worker is 2.931, thus the minimum workforce is
267(780/2.931) inJ an;
Units produced per worker in J an. and Feb. combined =
2.931+3.517=6.448, andthecumulativedemandis1,420, thenthe
minimumworkforceis221(1420/6.448) tocover bothJ an. andFeb.
Gooncomputinginthesameway.
A Prototype Problem
A B C D
Month Cumulative
Net Demand
Cumulative No.
of Units
Produced per
Worker
Ratio B/C
(Rounded up)
Jan. 780 2.931 267
Feb. 1,420 6.448 221
March 2,320 9.086 256
April 3,520 12.896 273
May 5,520 16.120 343
June 7,520 18.318 411
Table 3-3 Computation of the Minimum Workforce Required by Denspack
The minimum number of workers required for entire six-
month planning horizon is 411, requiring hiring 111 new
workers at the beginning of J an.
A Prototype Problem
A B C D E F
Month No. of Units
Produced per
Worker
Monthly
Production
(B411)
Cumulative
Production
Cumulative
Net
Demand
Ending
Inventory
(D-E)
J an. 2.931 1,205
1,205 780
425
Feb. 3.517 1,445
2,650 1,420
1,230
March 2.638 1,084
3,734 2,320
1,414
April 3.810 1,566
5,300 3,520
1,780
May 3.224 1,325
6,625 5,520
1,105
J une 2.198 903
7,528 7,520
8
Total 5,962+600
Table 3-4 Inventory Level for Constant Workforce Schedule
The total cost is (5,962+600)80+111 500=580,460>569,540 for P1;
P2 is preferred because it has no large difference from P1 in cost, but keeps
workforce stable.
A Prototype Problem
Mixed Strategy and Additional Constraints
The zero inventory plan and the constant workforce strategies
are to target one objective;
Combining the two plans may results in dramatically lower
costs;
Figure 3-4 shows the constant
workforce strategy (a straight
line-a fixed production
rate).
Fig. 3-4 A Feasible Aggregate Plan
the constant
workforce strategy
The zero inventory plan
A Prototype Problem
Mixed Strategy and Additional Constraints
Suppose that we may use two production rates (2 straight lines):
Make net inventory at the end
of April to be zero (P1) by
producing enough in each of
the four months J an. through
April to meet the cumulative
net demand each month:
produce 3,520/4=880 units in
each of the first four months;
The May and J une production
is then set to 2,000, exactly
matching the net demand in
these months.
The two lines are above the cumulative
net demand, the plan is feasible
A Prototype Problem
Month Cumulative Net
Demand
Cumulative Production
J an. 780 880
Feb. 1,420 1,760
March 2,320 2,460
April 3,520 3,520
May 5,520 5,520
J une 7,520 7,520
A Prototype Problem
The graphical solution for additional constraints.
Capacity limitation:
the production capacity
of the plan is only
1,800 units per month
One feasible solution:
produce 980 in each of
the first four months
and 1,800 in each of
the last two months.
(Slopes of the two
segmental lines 1800)
a constraint on the
slope of the straight
line.
Lower than
the cap. limit
Just
approach the
cap. limit
Contents
Introduction
Aggregate Planning by Linear Programming
Linear Programming (LP) is used to determine values of n
nonnegative variables to maximize or minimize a linear function
of these variables that is m linear constraints of these variables.
Cost Parameters
C
H
=Cost of hiring one worker;
C
F
=Cost of firing one worker;
C
I
=Cost of holding one unit of stock for one period;
C
R
=Cost of producing one unit product on regular time;
C
O
=Incremental cost of one unit on overtime;
C
U
=Idle cost per unit of production;
C
S
=Cost of subcontract one unit of production;
n
t
=Number production days in period t;
K=Number of aggregate units produced by one worker in one day;
I
0
=Initial inventory on hand at the start of the planning horizon;
W
0
=Initial workforce at the start of the planning horizon;
D
t
=Forecast of demand in period t;
Problem Variables:
W
t
=Workforce level in period t;
P
t
=Production level in period t;
I
t
=Inventory level in period t;
H
t
=Number of workers hired in period t;
F
t
=Number of workers fired in period t;
O
t
=Overtime production in units;
U
t
=Worker idle time in units (underutilized time);
S
t
=Number of units subcontracted from outside;
If P
t
> Kn
t
W
t
: the number of units produced on overtime :
O
t
=P
t
-Kn
t
W
t
;
If P
t
< Kn
t
W
t
: the idle time is measured in units of production
rather than in time, U
t
= Kn
t
W
t
- P
t
;
Constraints-Three sets of constraints to ensure conservation of
labor and that of units
1. Conservation of workforce constraints
W
t
=W
t-1
+H
t
-F
t
; for 1t T
2. Conservation of units constraints
I
t
=I
t-1
+P
t
+S
t
-D
t
; for 1t T
4. Others
Non negative constraints;
Given I
0
, I
T
, and W
0
.
3. Conservation of relating production level to
workforce levels
P
t
=Kn
t
W
t
+O
t
-U
t
; for 1t T
3T
constraints
Objective function-tochoose variables W
t
, P
t
, I
t
, H
t
, F
t
,O
t
, U
t
and S
t
(total 8T) to
1
( )
T
H I F F I t R t O t U t S t
t
Min c H c H c I c P c O c U c S
Subject to
the above 3T constraints,
nonnegative constraint: W
t
, P
t
, I
t
, H
t
, F
t
O
t
, U
t
and S
t
0, and
I
0
, I
T
, and W
0
.
Rounding the Variables
Some variables such as I
t
, W
t
, F
t
, H
t
should be integers;
May calculate by integer linear programming, however, may be
too complex;
Results of LP should be rounded up- by Conservative approach
Round W
t
to the next larger integer, and then calculate H
t
, F
t
,
and P
t
;
Always feasible solution, but rarely optimized;
Additional constraints
O
t
U
t
=0-either one is zero in case that there are both overtime an
idle production in the same period ;
H
t
F
t
=0- either in case of hiring and firing workers in the same
period
Both the two constraints are linear;
Extensions
Account for uncertainty in demand by minimum buffer inventory
B
t
: I
t
B
t
, for 1tT, where B
t
should be specified in advance;
Capacity constraints on amount of production: P
t
C
t
;
In some cases, it may be desirable to allow demand exceed the
capacity. To treat the backlogging of excess demand, the inventory
needs to be expressed in terms of two different non-negative
variables I
t
+
and I
t
-
, such that I
t
= I
t
+
- I
t
-
, and holding cost is
charged against I
t
+
, while the penalty cost for back orders against
I
t
-
.
Solving AP Problems by LP: An Example (3.2)
Since no subcontracting, overtime, or idle time allowed, and the
cost coefficients are constant with respect to time, the objective
function is simplified as
6 6 6
1 1 1
(500 1000 80 )
t t t
t t t
Min H F I

W
1
-W
0
-H
1
+F
1
=0;
W
2
-W
1
-H
2
+F
2
=0;
W
3
-W
2
-H
3
+F
3
=0;
W
4
-W
3
-H
4
+F
4
=0;
W
5
-W
4
-H
5
+F
5
=0;
W
6
-W
5
-H
6
+F
6
=0;
P
1
-I
1
+I
0
=1,280;
P
2
-I
2
+I
1
=640;
P
3
-I
3
+I
2
=900;
P
4
-I
4
+I
3
=1,200;
P
5
-I
5
+I
3
=2,000;
P
6
-I
6
+I
5
=1,400;
P
1
-2.931W
1
=0;
P
2
-3.517W
2
=0;
P
3
-2.638W
3
=0;
P
4
-3.810W
4
=0;
P
5
-3.224W
5
=0;
P
6
-2.198W
6
=0;
W
i
, P
i
, I
i
, F
i
, H
i
(i=1-6) 0;
W
0
=300, I
0
=500, I
6
=600
Solving AP Problems by LP: An Example (3.2)
Solved by LINGO system;
The value of objective function is $379,320.90, considerably less than that
obtained by either P
1
and P
2
, since this value is obtained by fractional values of
variables .
Rounded up result: the total cost=465 500+1,000 27+(900+600)
80=$379,500
A B C D E F G H I
Month No. of
Worker
s
No.
Hired
No.
Fired
No. of
Units
per
Worker
No. of
Units
Produced
(BE)
Cumulativ
e
Production
Cumulativ
e Demand
Endin
g Inv.
(G-H)
Jan. 273 27 2.931 800 800 780 20
Feb. 273 3.517 960 1,760 1,420 340
March 273 2.638 720 2,480 2,320 160
April 273 3.810 1,040 3,520 3,520 0
May 738 465 3.224 2,379 5,899 5,520 379
June 738 2.198 1,622 7,521 7,520 1
Total 465 27 900
Summary
AP: translate forecasted demand into the blueprint of
workforce level and production level
Cost considered in AP: Smoothing cost, holding
cost, backlog cost, regular time cost, overtime and
subcontract cost, underutilization cost
Solutions: Chase-up strategy, constant workforce
level, and LP
P140 Q14
P141 Q15, 16
The End!
Prof. JIANG Zhibin
Department of Industrial Engineering & Management
Inventory Control
Subject to Known Demand
Contents
Introduction
Types of Inventories
Motivation for Holding Inventories;
Characteristics of Inventory System;
Relevant Costs;
The EOQ Model;
EOQ Model with Finite Production Rate
Quantity Discount Models
Resource-constrained multiple product system
EOQ models for production planning
Power-of-two policies
Introduction
Inventory is the stock of any item or resource used in an
organization.
Generally , inventory is being acquired or produced to meet the
needof customers;
An inventory system is the set of policies and controls that
monitors levels of inventory or determines what levels should be
maintained.
Independent demandsystem
Dependent demandsystem
Introduction
The fundamental problem of inventory management :
When to place order for replenish the stock ?
How much to order?
The complexity of the resulting model depends on the
assumptions about the various parameters of the system
The major distinction is between models for known
demand and random demand.
Introduction
Thecurrent investment ininventoriesinUSA isenormous;
It amountedupto$1.57trillioninthethirdquarter of 2007;
It accountsfor 20-25%of thetotal annual GNP (general net
product);
There exists enormous potential for improving the
efficiency of economy by scientifically controlling
inventories;
Introduction
Fig. 4-1 Breakdown of the Total Investment
in Inventories in the U.S. Economy (2007)
Inventory model
discussed here are most
applicable to
manufacturing,
wholesale, and retail
sectors, composing 82
% of the total .
Contents
Introduction
Relevant Costs;
The EOQ Model;
A natural classification is by value added from manufacturing:
Raw materials-Resources required in the production or
processing activity of the firm.
Components-Raw materials or subassemblies that will later be
included into a final product.
Work-in-process (WIP)-the inventory either waiting in the
system for processing or being processed.
Finished good-also known as end items-the final products.
Contents
Introduction
Relevant Costs;
The EOQ Model;
Why Hold Inventories
For economies of scale-It may be economical to
produce a relatively large number of items in each
production run and store them for future use.
Coping with Uncertainties
Uncertainty in demand;
Uncertainty in lead time
Uncertainty in supply
For speculation-
Purchase large quantities at current low prices and
store them for future use.
Cope with considerable fluctuation in price of
costly commodities required in large quantities;
Cope with labor strike;
For Transportation
Pipeline inventories is the inventory moving from
point to point, e.g., materials moving from suppliers
to a plant, from one operation to the next in a plant.
It exists for purpose of transportation or materials
handling in a plant;
Smoothing-Producing and storing inventory in anticipation of peak
demand helps to alleviate the disruptions caused by changing
production rates and workforce level.
Logistics-To cope with constraints in purchasing, production, or
distribution of items that may cause a system maintain inventory;
Purchase an item in minimum quantities;
Logistics of manufacture-zero inventory is impossible in order
to keep continuity in manufacturing process;
Control costs -More inventory need less and simpler control
Why we should control the inventory
Quality
Working efficiency
delay
Production plan
Machine
down
Shut down
WIP
lost
Inventory Level
(a) Higher inventory level
quality
Working efficiency
delay
Production plan
Machine
down
Shut down
WIP
lost
(b) Lower inventory level

Contents
Introduction
Relevant Costs;
The EOQ Model;
Characteristics of Inventory Systems
Demand: patterns and characteristics
Constant versus variable;
Known versus random
Lead Time
Ordered from the outside;
Produced internally
Characteristics of Inventory Systems
Review patterns
Continuous-supermarket;
Periodic-regular stock-taking for a grocery store;
Excess demand-demand that cannot be filled immediately from
stock; backordered or lost.
Changing inventory
Limited shelf life- perishability;
Become obsolete- obsolescence;
Contents
Introduction
Relevant Costs;
The EOQ Model;
Relevant Costs- Holding cost
Holding cost (carrying or inventory cost)-the sum of
costs that are proportional to the amount of inventory
physically on hand at any point in time.
Some items of holding costs:
Cost of providing the physical space to store the items;
Taxes and insurance;
Breakage, spoilage, deterioration, and obsolescence;
Opportunity cost of alternative investment;
Opportunity cost, also known as cost of capital, turns out to be
the most significant one
If we have 10,000RMB on hand, and save it in bank for one
month at interest rate 0.25/month, then we may earn
25RMB/month from bank;
If we use this amount money to buy some goods and store
them in warehouse, then we lose 25RMB/month.
Holding cost considers the aggregated interest rate comprised
of the four components
For example, 28% cost of capital, 2% taxes and insurance, 6%
cost of storage, 1% breakage and spoilage, mean 37% total
interest charge.
We would assess a charge of 37 cents for every dollar that we
have invested during a one-year period.
Holding cost h (in terms of dollars per unit per year)
h=Ic
I: the annual interest rate
c: the dollar value of one unit of inventory
For example, an item valued at $180 would have an annual
holding cost of h=0.37*180=66.60.
Inventory cost fluctuates with time-inventory as a function of time
The holding cost incurred at any point in time is proportional to
the inventory level at that point in time.
Fig.4-2 Inventory as a Function of Time
2
1
2 1
( ( ) )/( )
t
t
I I t dt t t =
}
Relevant Costs- Order cost
It depends on the amount of inventory that is ordered or produced.
Two components
The fixed cost K: independent of size of order as long as it is
not zero; which includes only those costs that are relevant to
the current ordering decision, maybe the costs of order
generation and receiving, and handling cost; not includes the
overhead costs
The variable cost c: incurred on per-unit basis;
0 0;
( )
0
if x
C x
K cx if x
=
=

+ >
Relevant Costs- Order cost

Fig.4-3 Order Cost Function
Relevant Costs- Penalty Cost
Also know as shortage cost or stock-out cost-is the cost of not
havingsufficient stockonhandtosatisfyademandwhenit occurs.
Twointerprets:
In back-order case: include whatever bookkeeping and/or delay
costsmaybeinvolved;
Inlost-salecase: includeof loss-of-goodwill cost, ameasureof
customer satisfaction.
Relevant Costs- Penalty Cost
Twoapproaches:
Penalty cost, p, is charged per-unit basis. Each time a demand
occurs that cannot be satisfied immediately, a cost p is incurred
independent of howlongit takestoeventuallyfill thedemand.
Chargethepenaltycost onaper-unit-timebasis.
The EOQ Model-Basic Model
The objective is to choose Q to minimize the average cost per
unit time (usually, a year)
Fig 4-4 Inventory Levels for the EOQ Model
Average holding level is Q/2
Considerations:
On-hand inventory
level at the time
zero is zero;
An order must be
placed at time zero;
Q is the size of the
order (lot size);
Next order is
placed just when
the inventory level
drops to zero;
The order cycle
T=Q/
Express the average annual cost as a function of the lot
size
Order cost in each order cycle: C(Q)=K+cQ;
The average holding cost during one order cycle is hQ/2;
The average annual cost (suppose there are n cycles in a year) :
( )
( )
2 / 2 2
K cQ n hQ K cQ hQ K hQ
G Q c
nT Q Q
+ +
= + = + = + +
2 3
2
'( ) ; ''( ) 0 0;
2
K h K
G Q G Q for Q
Q Q

= + = > >
The annual total ordering cost
Average holding cost for one cycle = that for one year.
Since Q>0G(Q) is convex function of Q;
G(Q) is minimized at Q*--economic order quantity, EOQ
*
2
'( ) 0
K
G Q Q
h
= =
Fig.4-5 The Average Annual
Cost Function G(Q)
Notes:
Q* is the value of Q where the two
curves intersect;
The unit order cost, c, does not
appear explicitly in the expression of
Q*, since c is independent from Q
in the C(Q);
Because c is constant, it is
ignored while computing average
cost.
Example 4.1
Pencils are sold at a fairly steady rate of 60 per week;
Pencils cost 2 cents each and sell for 15 cents each;
Cost $12 to initiate an order, and holding costs are based on annual interest
rate of 25%.
Determine the optimal number of pencils for the book store to purchase
each time and the time between placement of orders
Solutions
Annual demand rate =60 52=3,120;
The holding cost is the product of the variable cost of the pencil and the
annual interest-h=0.02 0.25=0.05
*
2 2 12 3,120
3,870
0.05
K
Q
h

= = =
3,870
1.24
3,120
Q
T yr
= = =
Back
The End!
Prof. JIANG Zhibin
Dr. GENG Na
Management
Inventory Control
Contents
Introduction
Relevant Costs;
The EOQ Model;
Express the average annual cost as a function of the lot
size
Order cost in each order cycle: C(Q)=K+cQ;
The average holding cost during one order cycle is hQ/2;
The average annual cost (suppose there are n cycles in a year) :
( )
( )
2 / 2 2
K cQ n hQ K cQ hQ K hQ
G Q c
nT Q Q
+ +
= + = + = + +
2 3
2
'( ) ; ''( ) 0 0;
2
K h K
G Q G Q for Q
Q Q

= + = > >
*
2
'( ) 0
K
G Q Q
h
= =
Example 4.1
Pencils are sold at a fairly steady rate of 60 per week;
Pencils cost 2 cents each and sell for 15 cents each;
Cost $12 to initiate an order, and holding costs are based on annual interest
rate of 25%.
Determine the optimal number of pencils for the book store to purchase
each time and the time between placement of orders
Solutions
Annual demand rate =60 52=3,120;
The holding cost is the product of the variable cost of the pencil and the
annual interest-h=0.02 0.25=0.05
*
2 2 12 3,120
3, 870
0.05
K
Q
h

= = =
3, 870
1.24
3,120
Q
T yr
= = =
Since there exits lead time t (4 moths for Example 4.1), order
should be placed some time ahead of the end of a cycle;
Reorder point R -determines when to place order in terms of
inventory on hand, rather than time.
The EOQ Model-Considering Lead Time
Fig. 4-6 Reorder Point Calculation for Example 4.1
4
( ) 3,120( / ) 1, 040
12
R yr units yr t = = =
Determine the reorder point
when the lead time exceeds a
cycle.
The EOQ Model-Considering Lead Time
Fig 4-7 Reorder Point Calculation for
Lead Times Exceeding One Cycle
Example:
EOQ=25;
=500/yr;
t=6 wks;
Computing R for placing order
2.31 cycles ahead is the same as
that 0.31 cycle ahead.
T=25/500=2.6 wks;
t/T=2.31---2.31
cycles are included
in LT.
Action: place every
order 2.31 cycles in
advance.
How sensitive is the annual cost function to errors
in the calculation of Q?
The EOQ Model- Sensitivity
Considering Example
4.1. Suppose that the
bookstore orders pencils
in batches of 1,000, rather
than 3,870 as the optimal
solution indicates. What
additional cost is it
incurring by using a
suboptimal solution?
By substituting Q=1,000,
we can find the average
annual cost for this lot size.
( ) / / 2
(12)(3,120) /1, 000 (0.005)(1, 000) / 2
$39.94
G Q K Q hQ = +
= +
=
Which is considerably larger
than the optimal cost of $19.35.
Lets obtain a universal solution to the sensitivity problem.
Let G* be the average annual holding and setup cost at the
optimal solution. Then
* / * */ 2
2

2
2 /
2
2
2
G K Q hQ
K h K
h
K h
K h
K h
= +
= +
=
=
( ) / / 2
*
2
1 2

2 2 2
*

2 2 *
1 *

2 *
It follows that for any Q,
G Q K Q hQ
G
K h
K Q h
Q h K
Q Q
Q Q
Q Q
Q Q
+
=
= +
= +
(
= +
(

To see how one would use this result, consider using a
suboptimal lot size in Example 4.1.
The optimal solution was Q*=3,870, and we wished to
evaluate the cost error of using Q=1,000. Forming the
ratio Q*/Q gives 3.87. Hence,
G(Q)/G*=(0.5)(3.87+1/3.87)=(0.5)(4.128)=2.06.
This says that the average annual holding and setup cost
with Q=1,000 is 2.06 times the optimal average
holding and setup cost.
In general, the cost function G(Q) is relative
insensitive to errors in Q. For example,
if Q is twice as large as Q*, then G/G(Q*)1.25 ,
meaning that an error of 100% in Q will generate an
error of 25% in annual average cost.
And suppose that the order quantity differed from the
optimal by Q units. A value of Q=Q*+ Q would
result in a lower average annual cost than a value of
Q=Q*- Q. ---Not symmetric.
( * ) ( * ) 1 * * 1 * *
* * 2 * * 2 * *
1 * * 1 * *

2 * * 2 * *

( ( +A A +A A
= + +
( (
+A A

( ( +A A
= +
( (
+A A

G Q Q G Q Q Q Q Q Q Q Q
G G Q Q Q Q Q Q
Q Q Q Q Q Q
Q Q Q Q Q Q
( )( )
( ) ( )
( )
2 2
2
1 2 * 1 2

2 * * 2 *
*

*
*
0
*
* / *
(
( A A
= +
(
(
+A A

A A
= +
A
A A
= + s
A
Q Q Q
Q Q Q Q Q
Q Q Q
Q
Q Q
Q Q
Q
Q Q Q
Contents
Introduction
Relevant Costs;
The EOQ Model;
The simple EOQ model is based on assumption that
the items are obtained from an outside supplier, and
thus entire lot is delivered at the same time;
EOQ model is also effective when units are internally
produced, based on assumption that production rate is
infinite , such that the entire lots are delivered at the
same time.
If the production rate is finite and comparable to the
rate of demand, the simple EOQ model will be
ineffective.
The EOQ Model for Finite Production Rate
Assumption:
Items are produced at a rate P during a production run;
P> for feasibility (comparable to the rate of demand );
Let
Q is the lot size of each production run;
T is the cycle length, the time between successive startups.
T=T
1
+T
2
, where T
1
is production time, while T
2
is the
downtime (no production);
Note that the maximum level of
on-hand inventory during a cycle
is no longer Q.
Fig. 4-8 Inventory Levels for Finite
Production Rate Model
The number of units
consumed in each cycle is T;
The number of units produced
at rate P in a production time
T
1
is Q=T
1
P;
The number of produced units
during the production time
should satisfies the demand
for the cycle time:
T=T
1
P =QT
1
=Q/P;
The maximum level of
inventory on hand is H=T
1
(P-
)=Q(1- /P);
Since average inventory level
is H/2, thus the annual average
inventory cost follows:
Fig. 4-8 Inventory Levels for Finite
Production Rate Model
( ) (1 / )
2 2
K hH K Q
G Q h p
T Q
= + = +
2
' (1 / ), *
'
k
Let h h P Then Q
h
= =
Solutions:
h=h(1- / P)=0.6(1- 2,500 / 10,000)=0.45;
Q
*
=(2K /h)
1/2
=745;
T=Q
*
/ =745/2,500=0.298 yr;
The production time (uptime) T
1
=Q*/P=0.0745 yr;
The downtime is T
2
=T-T
1
=0.2235 yr;
G(Q
*
)=K / Q
*
+hQ
*
/2=335.41;
H=Q
*
(1- /P)=559 units;
Determine:
Optimized size of a production run Q;
The length of each production run T;
The average annual cost of holding cost and
setup;
Maximum level of inventory on hand.
Given
P=10,000 units/yr;
K=$50;
=2,500 units/yr;
h=$20.3=0.6
Example 4.3
In-class quiz
Determine:
EOQ;
The length of each order cycle T;
The average annual cost of holding cost and
setup;
Maximum level of inventory on hand.
Given
c=$2.4/ unit;
K=$45;
=280 units/yr;
I=0.2
Solutions:
h=Ic=0.48;
Q
*
=(2K /h)
1/2
=229;
T=Q
*
/ =229/280= .8179 yrs. (= 9.81 months);
G(Q
*
)=K / Q
*
+hQ
*
/2= $109.98;
H=Q
*
=229 units;
The End!
Prof. JIANG Zhibin
Dr. GENG Na
Management
Inventory Control
Contents
Introduction
Relevant Costs;
The EOQ Model;
The suppliers may charge less per unit for
larger orders to encourage the customer to buy
their products in larger batches.
Two popular ways of discounts:
All-units: discount is applied to all of the
units in an order;
Incremental: only applied to additional units
beyond the breakpoints;
Quantity Discount Model
Quantity Discount Model
Example 4.4 Weighty Trash Bag Companys pricing schedule
for its large trash can liners:
For orders of less than 500 bags, charges 30 cents per bag;
for orders of 500 or more but fewer than 1,000 bags, charges
29 cents per bag; and
for orders of 1,000 or more, charges 28 cents per bag.
The breakpoints are 500 and 1,000. The discount schedule is
all-units;
The order cost function C(Q) is defined as
Fig 4-9 All-Units Discount Order Cost
Function
Quantity Discount Model- Optimal Policy for All-
Units Discount Schedule
0.30 0 500,
( ) 0.29 500 1, 000,
0.28 1, 000
Q for Q
C Q Q for Q
Q for Q
s <
= s <
Example 4.4: If Weighty uses trash bags at a fairly constant rate

of 600 per yr, how to place order?
Suppose that fixed cost of placing an order is $8, and holding
costs are based on 20% annual interest rate.
First compute EOQ values corresponding to each of the unit cost.
0 0
0
2 2 8 600
400, 0 500
0.2 0.3
K
Q C Q
IC

= = = s <
1 1
1
2 2 8 600
406, 500 1000
0.2 0.29
K
Q C for Q
IC

= = = s <
2 2
2
2 2 8 600
414, 1000
0.2 0.28
K
Q C for Q
IC

= = = >

Each curve is valid only for certain values of Q, thus the
average annual cost function is given by discontinuous curves.
0
1
2
( ) / / 2
0,1, 2
( ) 0 500,
( ) ( ) 500 1, 000,
( ) 1, 000
j
j j C
G Q c K Q I Q
for j and
G Q for Q
G Q G Q for Q
G Q for Q
= + +
=
s <
= s <
Fig. 4-12 All-Units Discount Average

Annual Cost Function
400 0 500,
406 500 1, 000,
414 1, 000
s <
= s <
for Q
EOQ for Q
for Q
Fig. 4-12 All-Units Discount Average
Annual Cost Function
An EOQ value is realizable, if it falls within the interval of EOQ
that corresponds to the unit cost that has been used to compute it.
(Q
0
for the example);
The goal is to find the minimum of this discontinuous curve,
which corresponds to the EOQ.
Generally, the optimal solution will be either the largest
realizable EOQ or one of the breakpoints that exceeds it.
The three candidates are 400, 500, and 1,000.
0
(400) (400) 600 0.3 600 8/ 400 0.2 0.3 400/ 2 $204.00 G G = = + + =
1
(500) (500) 600 0.29 600 8 / 500 0.2 0.29 500 / 2 $198.10 = = + + = G G
2
(1, 000) (1, 000) 600 0.28 600 8/1, 000 0.2 0.28 1, 000/ 2 $200.80 G G = = + + =
Conclusion: the optimal solution is to place a standing order for
500 units with Weighty at an annual cost of $198.10.
Quantity Discount Model- Optimal Policy for All-Units
Discount Schedule
Summary of the Solution Technique for All-Units Discounts
Determine the largest realizable EOQ value. The most
efficient way to do this is to compute the EOQ for the lowest
price first, and continue with the next higher price. Stop when
the first EOQ is realizable (that is, within the correct interval)
Compare the value of the average annual cost at the largest
realizable EOQ and at all the price breakpoints that are
greater than the largest realizable EOQ. The optimal Q is the
point at which the average annual cost is a minimum.
Quantity Discount Model- Incremental Quantity
Discounts
0.30 0 500,
( ) 150 0.29( 500) 5 0.29 500 1, 000,
295 0.28( 1000) 15 0.28 1, 000
Q for Q
C Q Q Q for Q
Q Q for Q
s <
= + = + s <
+ = + s
The order cost function follows

The average annual cost function follows
G(Q)=C(Q)/Q+K
/Q+I[C(Q)/Q]Q/2
G(Q) has been divided into three
segments G
1
(Q), G
2
(Q), and G
3
(Q),
each of which is obtained by using
one of the three segments of C(Q) .
Fig.4-10Incremental Discount Order Cost
Function
Discounts
( ) ( )( ) ( )( ) ( )( )
( )( ) ( )( ) ( )( ) ( )( )
1
600 0.29 5 / 8 600 / 0.20 0.29 5 / / 2
0.29 600 13 600 / 0.20 0.29 / 2 0.20 5 / 2
G Q Q Q Q Q
Q Q
= + + + +
= + + +
( )
( )( )( )
( )( )
2
2 23 600
702
0.20 0.28
Q = =
( )
( )( )( )
( )( )
0
2 8 600
400
0.20 0.30
Q = =
The optimal solution occurs at the minimum of one of the
average annual cost curves.
( )
( )( )( )
( )( )
1
2 13 600
519
0.20 0.29
Q = =
( ) ( )( ) ( )( ) ( )( )
( )( ) ( )( ) ( )( ) ( )( )
2
600 0.28 15 / 8 600 / 0.20 0.28 15 / / 2
0.28 600 23 600 / 0.20 0.28 / 2 0.20 15 / 2
G Q Q Q Q Q
Q Q
= + + + +
= + + +
( ) ( )( ) ( )( ) ( )( )
0
600 0.30 8 600 / 0.20 0.30 / 2 G Q Q Q = + +
Discounts
( )
( )( )( )
( )( )
2
2 23 600
702
0.20 0.29
Q = =
( )
( )( ) ( )
( ) ( )
( )
0
2 8 600
400
0.20 0.29
Q realizable = =
( )
( )( )( )
( )( )
( )
1
2 13 600
519
0.20 0.29
Q realizable = =
Fig 4-12 Average Annual Cost Function
for Incremental Discount Schedule
( )
( )
0
0
$204.00 G Q =
( )
( )
1
1
$204.58 G Q =
The optimal solution occurs at the
minimum of one of the three average annual
cost curves.
Quantity Discount Model- Incremental Quantity Discounts
Procedures:
(1) Determine an algebraic expression for C(Q) corresponding to
each price interval. Use that to determine an algebraic
expression for C(Q)/Q;
(2) Substitute the expressions derived for C(Q)/Q into defining
equation for G(Q). Compute the minimum value of Q
corresponding to each price interval separately.
(3) Determine which minima computed in (2) are realizable (that is
fall into the correct interval). Compare the values of the average
annual costs at the realizable EOQ values and pick the lowest.
Other Discount Schedules
A carload consists of M
units.
The supplier charges a
constant c per unit up until
you have paid for the cost of
a full carload, at which point
there is no charge for the
remaining units in that
carload.
Once the first carload is full,
you again pay c per unit
until the second carload is
full, and so forth.
In-class Quiz
Buy tissue either one roll or one pack (12 rolls):
c1=45 cents/roll, c2=$5/1pack, I=0.25
Fixed cost K=$1, =4 rolls/year
Determine the optimal quantity to buy.
Optimal number of single rolls to purchase
If you buy in single packs, the average annual cost is
(2)(1)(4)
1 8
(.25)(.45)
= = Q
1 1
2 2.75 + = c K Ic
In-class Quiz
Buy tissue either one roll or one pack (12 rolls):
c1=45 cents/roll, c2=$5/12rolls, I=0.25
Fixed cost K=$1, =4rolls/year
Determine the optimal quantity to buy.
If you buy in 12 rolls, the average annual cost is
2
2
(.25)(.42)(12) (1)(4)
(4)(.42) $2.64
2 2 12
+ + = + + =
Ic Q K
c
Q
The End!
Prof. JIANG Zhibin
Dr. GENG Na
Management
Inventory Control
Contents
Introduction
Relevant Costs;
The EOQ Model;
When the EOQ model is used in companies stocking many
different items, although we can obtain the optimal order
quantities separately for each of different items, there could
exist constraints (such as stock space, financial budget) that
would make the resulting solution infeasible.
Example 4.5: three items are produced in a small fabrication shop.
I=0.25
The shop never have more than $30,000 invested in the inventory
of these items at one time
What lot sizes should the shop be producing so that they do not
exceed the budget?
1 2 3
Demand rate 1,850 1,150 800
Variable cost 50 350 85
Setup cost 100 150 50
First check whether the budget is feasible when using EOQ
values of these three items
If the EOQ value is used, the maximum investment in
inventory would be 72*50+63*350+61*85=$35835>$30000
We therefore need to reduce these lot sizes. How to do?
1
2
3
2*100*1, 850
172,
0.25*50
2*150*1,150
63,
0.25*350
2*50*800
61.
0.25*85
EOQ
EOQ
EOQ
= =
= =
= =
We multiply each EOQ value by the ratio 30,000/35,835 =
0.8372. In order to guarantee that we do not exceed the
$30,000 budget, we round each value down, we get the
optimal values as follows:
Q
1
=172*0.8372144
Q
2
=63*0.837252
Q
3
=61*0.837251,
Lastly, we can increase the lot sizes of some items if there is
remaining budget.
Suppose that n items have unit costs of c
1
, c
2
,, c
n
,
respectively, and the total budget available for them is C.
Then the budget constraint can be written
c
1
Q
1
+c
2
Q
2
+, c
n
Q
n
<=C
In general, budget-constrained problems can be solved as
follow:
2
Let
i i
i
i
K
EOQ
h
=
1
Case 1: If , then the optimal solution is ;
n
i i i i
i
c EOQ C Q EOQ
=
s =
Note that ( ) 1 , the condition is equivalent to

the requirement that the same interest rate be used, which is
reasonable in most circumstances.
i i i i i i
c h c I c I = =
1 1 2 2
1
*
1
Case 2: If , then if ,
we can prove that the optimal solution is * ;
where .
n
i i n n
i
i i
n
i i
i
c EOQ C c h c h c h
Q m EOQ
C
m
c EOQ
=
=
> = = =
=
=
Contents
Introduction
Relevant Costs;
The EOQ Model;
We consider an extension of the EOQ model with a finite
production rate, discussed in sec. 4.6, to the problem of
production n products on a single machine
The goal is to determine the optimal procedure for producing
n products on the machine to minimize the cost of holding
and setup, and to guarantee that no stock-outs occur during
the production cycle.
1
Assumption 1 is required, which is stronger than
for each . Why??
n
i i
i
i i
P
P i
=
s
<

Rotation cycle policy: in each cycle there is exactly one
setup for each product, and products are produced in the
same sequence in each production cycle.
The optimal solution is to sequentially produce lot sizes
for each product optimized respectively.
But we have only single production facility, it is likely that
some of the lot sizes Q
j
will not be large enough to meet
the demand between production runs for product j, thus
resulting in stock-out.
'
2
From sec. 4.6, a lot size for product j, , where (1 ).
j j
j j j j j
j
K
Q h h P
h
'
= =
We assume that exactly one lot of each product is produced
between time T (cycle time)
In order that the lot for product j will be large enough to
meet the demand occuring during time T, it follows that the lot
size must be .
j j
Q T =
Then the average annual cost associated with product j:
( ) 2
j j j j j j
G Q K Q h Q
'
= +
1 1
The average annual cost for all products: ( ) ( 2);
n n
j j j j j j
j j
G Q K Q h Q
= =
'
= +

1
Substituting ;
The average annual cost associated with n products in terms of the cycle time T:
( ) ( 2)
j j
n
j j j
j
T Q
G T K T h T
=
=
'
= +
( )
The necessary condition for an optimal T is 0
dG T
dT
=
1 *
1
2
The optimal cycle time is
n
j
j
n
j j
j
K
T
h
=
=
=
'
1
If set time is a factor, let be the setup time for product j, the following constraint
should be satisfied ( / )
j
n
j j j
j
s
s Q P T
=
+ s
1
min
1
Using the fact , and rearranging terms,
1 ( / )
n
j
j
j j
n
j j
j
s
Q T T T
P
=
=
= > =
* *
* min
min
min
if
Therefore, the cycle time is max{ , }
else
T T T
T T T
T
>
= =


Example 4.7 Bali produces several styles of mens and
womens shoes at a single facility near Bergamo, Italy.
ROTATION policy used to schedule a single cut machine
Setup costs $110 per hour, holding costs 0.22. the relevant data for
this problem is in following table.
Style Annual demand
(units/year)
Production
rate(units/year)
Setup time
(hours)
Variable cost
($/unit)
W pump 4,520 35,800 3.2 40
W loafer 6,600 62,600 2.5 26
W boot 2,340 41,000 4.4 52
W sandal 2,600 71,000 1.8 18
M wingtip 8,800 46,800 5.1 38
M loafer 6,200 71,200 3.1 28
M oxford 5,200 56,000 4.4 31
Next we compute the value of optimal cycle time.
We first obtain setup costs and modified holding costs, and
then get T
*
=0.1529 and optimal lot sizes for each production
run in following.
The first step is to verify that the problem is feasible.
/ 0.69355 1
i i
P = s
Setup cost K
j
Modified holding costs h
j
Optimal lot sizes for each production run

352 7.69 691
275 5.12 1,009
484 10.79 358
198 3.81 398
561 6.79 1,346
341 5.62 948
484 6.19 795
T
min
=0.04. thus T* =0.1529> T
min
is feasible and, hence,
optimal.
The total average annual cost of holding and setups at an
optimal policy is G(T*)=35,244.44.
Note that
the optimal rotation policy that minimizes total holding
and setup costs results in the cutting operation remaining
idle time. The total uptime each rotation cycle is 0.106
year.
the relatively simple solution was the result of two
assumptions: One was that the setup costs were not
sequence dependent. A second assumption was that the
plant used a rotation cycle policy.
Contents
Introduction
Relevant Costs;
The EOQ Model;
Even when demand is assumed known, there exist several
extensions of these basic models whose optimal solutions may
be difficult or impossible to find. In those cases, effective
approximations are very valuable.
An approach has proven to be successful in a variety of
deterministic environments, based on choosing the best
replenishment interval from a set of possible intervals
proportional to powers of two.
The optimal solution given by EOQ model:
* 2 / , * */ 2 / Q K h T Q K h = = =
Is it realizable or convenient?
No. People like to place the order at the beginning of a day or
a week.
To find the optimal solution under the constraint that the order
interval must be a multiple of T
L
, one would simply compare
the costs at the two closest multiples of T
L
to T* (kT
L
T*
(k+1)T
L
) and pick the order interval with the lower cost.
Suppose we add the additional restriction that the order
interval must be a power of two time T
L
. That is, the order
interval must be of the form 2
k
T
L
for some integer k0.
Under such a restriction (known as a power-of-two policy),
what is the worst cost error we will incur relative to that of
the optimal reorder interval T*?
If k=12, 2
k
=4096, and 2
k+1
=8192, T*=6000. Even with T
L
=1,
the order interval to be either 4096, or 8192, quite far from
T*. How large is the cost error?
The cost error in every case is proven to be bounded by
slightly more than 6%.
( )
( )
/ / 2
From , / / 2
G Q K Q hQ
Q T G T K T h T

= +
= = +
( )
( ) ( )
In the case of a power-of-two policy, we wish to find the value of k that minimizes 2 .
Since G is a continuous convex function of T, it follows that 2 is a discrete convex
function of . That mea
k
L
k
L
G T
T G T
k
( ) ( ) { }
1
ns that the optimal value of , say *, satisfies
* min : 2 2
k k
L L
k k
k k G T G T
+
= >
1
1
1
substituting for G(T)
2 2
2 2 2 2

/ 2
2
2
* 2 /
1 1
2 or 2 *
2
k k
L L
k k
L L
k
L
k
L
k k
L
L
h T h T K K
T T
K
h T
T
T K h
K K
T T
T h h

+
+
>
s
=
> > =
We assume that * 2 , which means that 2 * 2 .
Therefore, we seek the smallest value of that satisfies
1
* 2 2 *
2
which means that as long as * , the optimal power-of-two
solution will always li
k k
L L
k
L
L
T T T T
k
T T T
T T
> >
s s
>
( )
( )
The worst-case cost error of the power-of-two policy is gi
e between 0.707 * and 1.41 *
ven by
1 * 1 1
2 1.0607
* 2 * 2
.
2
G Q
Q Q
G Q
T
Q
T
Q
| |
| |
= + = + =
| |
\ .
\ .
Summary
Economic motives for holding inventories, including
economies of scale, uncertainties, speculation, and logistics
Physical characteristics of inventory systems, including
demand, lead time, review time, back-order/lost-sales
assumptions
Three significant classes of costs: holding cost, setup cost (or
fixed order cost), and penalty cost
EOQ model
Two types of quantity discounts: all units and incremental
Limitation of budgets or space
Rotation cycle policy
Homework
Page 188, Q12,15;
Page 197, Q22, 23;
Page 206, Q30
The End!
Prof. JIANG Zhibin
Inventory Control
Subject to Unknown Demand
Inventory Control Subject to Unknown
Demand
Contents
Introduction
The newsboy model
Lot Size-Reorder Point System;
Service Level in (Q, R) System;
Additional Discussion of Periodic-review Systems
Multiproduct Systems
Introduction
Sources of Uncertainties
In consumer preference and trends in the market;
In the availability and cost of labor and resources;
In vendor resupply times;
In weather and its effects on operations logistics;
Of financial variables such as stock prices and interest rates;
Of demand for products and services.
Introduction
Uncertainty of a quantity means that we cannot predicate its
value in advance.
A department store cannot exactly predicate the sales of a
particular item on any given day;
An airline cannot exactly predicate the number of people
that will choose to fly on any given flight.
How can these firms choose the number of items to keep in inventory
or the number of flights to schedule on any given route?
Based on the past experience for planning;
Probability distribution is estimated based on historical data;
Minimize expected cost or maximize the expected profit when
uncertainty is present.
Introduction
Some Examples
In the economic recession of the early 1990s, some business that
relied on direct consumer spending, suffered severe losses.
Sears and Macys department stores, long standing successes
in American retail market made poor earning in 1991.
Several retailers enjoyed dramatic successes.
Both The Gap and Limited in the fashion business did very
well.
Wal-Mart Stores continues its ascendancy and surpassed Sear
as the largest retailer in the United State.
Intelligent inventory management in the face of uncertainty
certainly played a key role in the success of these firms.
Introduction
As almost all inventory management refers to some
level of uncertainty, what is the value of the
deterministic inventory control model?
Provide a basis for understanding the fundamental trade-offs
encountered in inventory management;
May be good approximations depending on the degree of
uncertainty in the demand.
Introduction
Let D be the demand for an item over a given period of
time. We express it as the sum of two parts D
Det
and D
ran
:
D=D
Det
+D
Ram
In many cases D~D
Det
even D
Ram
=0:
When the variance of the random component, D
Ram
is
small relative to the magnitude of D
Det
When the predictable variation is more important than

random variation;
When the problem is too complex to include an explicit
representation of randomness in the model.
Introduction
In many situations the random component of the
demand is too important to ignore.
As long as the expected demand per unit times is
relatively constant, and the problem structure is not
too complex, explicit treatment of demand
uncertainty is desirable.
Introduction
Two basic inventory control models subject to uncertainty:
Periodic review-the inventory level is known at discrete points
in time only;
For one planning period-the objective is to balance the costs of overage
and underage; useful for determining run sizes for items with short useful
lifetimes (Fashions, foods, newspaper)-newsboy model.
For multiple planning period-Complex, topics of research, and rarely
implemented.
Continuous review-the inventory level is known at all times.
Extensions of the EOQ model to incorporate uncertainty, service level
approaches are frequently implemented
Easy to compute and implement
Accurately describe most systems in which there is ongoing replenishment
of inventory items under uncertainty
Demand
Contents
Introduction
The newsboy model
The newsboy model
Example 5.1-Mac wishes to determine the number of copies of the
Computer J ournal he should purchased each Sunday. The demand
during any week is a random variable that is approximately
normally distributed, with mean 11.73 and standard deviation 4.74.
Each copy is purchased for 25 cents and sold for 75 cents, and he
is paid for 10 cents for each unsold copy by his supplier.
Discussion:
One obvious solution is to buy enough copies to meet the
demand, which is 12 copies.
Wrong: If he purchase a copy that does not sell, his out-of-
pocket expense is only 25-10=15 cents. However, if he is unable
to meet the demand of a customer, he loses 75-25=50cents.
Suggestion: He should buy more than the mean. How many?
The newsboy model
Notation-the newsboy model
A single product is to be ordered at the beginning of a period and can
be used only to satisfy the demand during that period.
Assume that all relevant costs can be determined on the basis of ending
inventory. Define:
c
0
=Cost per unit of positive inventory remaining at the end of the period
(overage cost);
c
u
=Cost per unit of unsatisfied demand, which can be thought as a cost per
unit of negative ending inventory (underage cost).
Assume that the demand D is a continuous nonnegative random variable with
density function f(x) and cumulative distribution function F(x).
The decision variable Q is the number of units to be purchased at the beginning
of the period.
The goal is to determine Q to minimize the expected costs incurred at the end
of the period.
The newsboy model
A general outline for analyzing most stochastic inventory problems
is as follows:
1. Develop an expression for cost incurred as a function of both
the random variable D and the decision variable Q.
2. Determine the expected value of this expression with respect to
the density function or probability function of demand.
3. Determine the value of Q such that the expected cost function is
minimized.
The newsboy model
Development of Cost Function
Define G(Q, D) as the total overage and underage cost incurred at
the end of the period when Q units are ordered at the start of the
period and D is the demand.
Q-D is the demand units left at the end of the period as long as
Q>D;
If Q<D, then Q-D is negative and the number of units remaining
on hand at the end of the period is 0.
The newsboy model
max{Q-D, 0} represents the units left at the end of the period.
max {D-Q, 0} indicates the excess demand over supply, or
unsatisfied demand.
,
max{ , 0}
0 .
Q D if Q D
Q D
if Q D
>
=

s
,
max{ , 0}
0 .
D Q if D Q
D Q
if D Q
>
=

s
G(Q, D)=c
0
max{Q-D, 0}+c
u
max{D-Q, 0}
The expected cost function is defined as:
G(Q)=E(G{Q, D))
0
0 0
0
0
( ) max{ , 0} ( ) max{ , 0} ( )
( ) ( ) ( ) ( )
u
Q
u
Q
G Q c Q x f x dx c x Q f x dx
c Q x f x dx c x Q f x dx

= +
= +
} }
} }
The newsboy model
Determining the Optimal Policy
Determine the value of Q that minimizes the expected cost G(Q).
G(Q) is convex such
that G(Q) has minimal
value
0
0
0
( )
( ) ( )
( ) (1 ( ))
=
=
} }
Q
u
Q
u
dG Q
c f x dx c f x dx
dQ
c F Q c F Q
2
0
2
( )
( ) ( ) 0 0
u
d G Q
c c f Q for all Q
dQ
= + > >
0
0
( )
(0) (1 (0))
0, (since (0) 0)
u
Q
u
dG Q
c F c F
dQ
c F
=
=
= < =
Since the slope is
negative at Q=0, G(Q)
is decreasing at Q=0.
The newsboy model
Optimal solution, Q
*
, such
that
0
( )
( ) (1 ( )) 0
u
dG Q
c F Q c F Q
dQ
= =
Fig5-3 Expected Cost Function
for Newsboy Model
* *
0
*
0
( ) (1 ( )) 0
( ) /( )
u
u u
c F Q c F Q
or
F Q c c c
=
= + The critical
ratio.
The critical ratio is strictly between 0 and 1, meaning that for a
continuous demand, this equation is always solvable.
The newsboy model
Since F(Q*) is defined as the probability that the demand does not
exceed Q*, the critical ratio is the probability of satisfying all the
demand during the period if Q* units are purchased at the beginning
of the period.
Example 5.1- Macs newsstand
Suppose that the demand for the J ournal
is approximately normally distributed
with mean =11.73 and standard
deviation o=4.74. c
0
=25-10=15, and
c
u
=75-25=50 cents. The critical ratio is
c
u
/(c
o
+c
u
)=0.50/(0.15 +0.5)=0.77. Hence,
he ought to purchase enough copies to
satisfy all of the weekly demand with
probability 0.77. The optimal Q* is the
77
th
percentile of the demand distribution.
Fig. 5-4 Determination of the Optimal
Order Quantity for Newsboy Example
Q*=oz+=4.740.
74+11.73=15.24~15
F(Q*)
In-class Quiz
The unit cost of handbag to the store is $28.5
The handbag will sell for $150
Any handbags not sold are purchased by a discount
firm for $20.
In addition, there is a cost of $0.4 for each dollar tied
up in inventory
The sale of handbag follows uniform distribution
(50,250).
How much handbag should be purchased in the
beginning
In-class Quiz
c
o
= 28.50 - 20.00 + (.4)(28.50) = 19.90
c
u
= 150 - 28.50 = 121.50.
Critical Ratio = c
u
/(c
u
+ c
o
) = 121.5/(121.5 + 19.9) =.86
F(Q) = .86
From the picture it follows that
(Q - 50)/200 = .86, Q = (.86)(200) + 50 =222
The End!
Dr. Na GENG
Inventory Control
Demand
Contents
Introduction
The newsboy model
The newsboy model- Optimal Policy for
Discrete Demand
In some cases, accurate representation of the observed
pattern of demand in term of continuous distribution is
difficult or impossible.
In the discrete case, the critical ratio will generally fall
between two values of F(Q).
The optimal solution procedure is to locate the critical
ratio between two values of F(Q) and choose the Q
corresponding to the higher value.
The newsboy model- Optimal Policy for
Discrete Demand
Example 5.2- Macs newsstand
f(4) =3/52 is obtained by dividing frequencies 4 (the numbers
of times 3 that a given weekly demand 4 occur during a year, i.e.
52 weeks) by 52;
The critical ratio is 0.77, which corresponds to a value of F(Q)
between Q=14 and Q=15.
Q f(Q) F(Q) Q f(Q) F(Q) Q f(Q) F(Q)
0 1/52 0.02 8 4/52 0.25 16 1/52 0.81
1 0 0.02 9 6/52 0.37 17 3/52 0.87
2 0 0.02 10 2/52 0.40 18 3/52 0.92
3 0 0.02 11 5/52 0.50 19 3/52 0.98
4 3/52 0.08 12 4/52 0.58 20 0 0.98
5 1/52 0.10 13 1/52 0.60 21 0 0.98
6 2/52 0.13 14 5/52 0.69 22 1/52 1.00
7 2/52 0.17 15 5/52 0.79
The newsboy model- Extension to Include
Starting Inventory
Suppose that the starting inventory is some value u and u>0
.
The optimal policy is simply to modify that for u=0.
The same ideal is that we still want to be at Q* after ordering.
If u<Q*, order Q*-u; If u>Q*, do not order.
Note that Q* should be understood as order-up-to point rather
than the order quantity when u>0.
Example 5.2 (Cont.)-Suppose that Mac has received 6 copies of
the Journal at the beginning of the week from other supplier. The
optimal policy still calls for having 15 copies on hand after
ordering, thus he would order the difference 15-6=9 copies.
The newsboy model- Extension to Multiple
Planning Periods
The ending inventory in any period becomes the starting
inventory in the next period.
Consider the case in which there are infinitely many periods
remaining:
If excess demand is back-ordered, interpret c
u
as the loss-
of-goodwill cost and c
o
as the holding cost.
If excess demand is lost, interpret c
u
as the loss-of-
goodwill cost plus the lost profit and c
o
as the holding cost.
Example 5.3: Suppose that Mac is considering how to
replenish the inventory of a very popular paperback thesaurus
that is ordered monthly. Copies of the thesaurus unsold at the
end of a month are still kept on the shelves for future sales.
Assume that customers who request copies of the thesaurus
when they are out of stock will wait until the following
month. Mac buys the thesaurus for $1.25 and sells it for $
3.75. Mac estimates a loss-of-goodwill cost of 80 cents each
time a demand for a thesaurus must be back-ordered. Monthly
demand for the book is fairly closely approximated by a
normal distribution with mean 20 and standard deviation 10.
Mac uses a 20 percent annual interest rate to determine his
holding cost. How many copies of the thesaurus should be
purchased at the beginning of each month?
Planning Periods
Answer for Example 5.3:
=20 and standard deviation o=10
c
o
=1.25*0.2/12=0.208 holding cost
c
u
=0.80.
The critical ratio is c
u
/(c
o
+c
u
)=0.8/(0.208 +0.8)=0.794
Hence, he ought to purchase enough copies to satisfy all of
the monthly demand with probability 0.794. The optimal Q*
is the 79.4
th
Q*=oz+=100.82+20=28.2~28
Planning Periods
Assume the excess demands are lost, and the cost for lost-of-
sale is 0.8:
=20 and standard deviation o=10
c
o
=1.25*0.2/12=0.208 holding cost
c
u
=0.80+2.5=3.3.
The critical ratio is c
u
/(c
o
+c
u
)=3.3/(0.208 +3.3)=0.94
Hence, he ought to purchase enough copies to satisfy all of
the monthly demand with probability 0.94. The optimal Q* is
the 94
th
Q*=oz+=101.56+20=35.6~36
Planning Periods
Planning Periods
One serious limitation of the multi-period newsboy model :
it did not include a setup cost for placing an order.
In most real systems, there are fixed costs associated with
ordering, and it is not optimal to place orders each period.
The problem of random demand is approached when a fixed
charge for ordering is present in a different way.
Demand
Contents
Introduction
The newsboy model
Lot Size-Reorder Point System
Service Level in (Q, R) System
For random demand, Q and R are regarded as independent
decision variables;
Assumptions
Continuous review-demands are recorded as they occur;
Random and stationary demand-the expected value of demand
over any time interval of fixed length is constant; the expected
demand rate is units/year.
Fixed positive lead time, , for placing an order;
decision variables;
Assumptions
Assume the following costs
Setup cost $K per order;
Holding cost at $h per unit held per year;
Proportional order cost of $c per item;
Stock-out cost $p per unit of unsatisfied demand, or
shortage cost or penalty cost;
1. Demand Description
The response time is the amount of time required to effect a
change in the on-hand inventory level.
The response time is the reorder lead time t
The demand during the lead time is the random variable of
interest.
It is assumed that demand during lead time is continuous
random variable D with probability density function (pdf) f(x)
and cumulative distribution function (cdf) F(x).
( ) E D = is the mean of the demand during the lead time.
( ) Var D o =
is the standard deviation.
2. Two Independent
Decision Variables
Q = the lot size or order
quantity;
R=the reorder level in
units of inventory.
The policy is that when the
level of on-hand inventory
drops to R, an order for Q
units is placed such that it
will arrive in t units of time.
Fig 5-5 Changes in Inventory Over
Time for Continuous-Review (Q, R)
System
3. Derivation of the Expected Cost Function
1) develop an expression for the expected average annual
cost in terms of the decision variables (Q, R)
2) search for the optimal values of (Q, R) to minimize this
cost.
4. Holding cost
Assume that the mean rate of
demand is units per year;
The expected inventory level
varies between s and Q+s, where s
is the safety stock, defined as the
expected level of on-hand inventory
just before an order arrives, s=R-t.
The average inventory is s+Q/2=R- t+Q/2.
The holding cost should not be charged against the inventory
level when it is negative.
5. Setup cost
A cycle is defined as the time
between the arrival of successive
orders of size Q.
Let T represent the expected cycle
length.
T=Q/.
The average setup cost incurred
per unit time is K/T=K /Q.
6. Penalty Cost
Occurs only when the system is subject to shortage.
The number of units of excess demand is simply the amount by
which the demand over the lead time, D, exceeds the reorder level,
R.
The expected number of shortages that occurs in one cycle is
determined by
(max( , 0)) ( ) ( )
R
E D R x R f x dx
=
}
n(R)
As n(R) represents the expected number of stock-outs
incurred in a cycle, the expected number of stock-outs
incurred per unit time is n(R)/T=n(R)/Q.
7. Proportional Ordering Cost Component.
The expected proportional order cost per unit of time is c;
Since this item is independent of variables Q and R, it does not
affect the optimization, and thus may be ignored.
The Cost Function:
G(Q, R)=h(Q/2+R-t)+K/Q+pn(R)/Q.
The objective is to choose Q and R to minimize G(Q, R).
0, 0
G G
Q R
c c
= =
c c
2 [ ( )]
(1)
1 ( ) / (2)
K pn R
Q
h
F R Qh p
+
=
=
The solution procedure requires iterating between (1) and (2)
until the two successive values of Q and R are the same.
When the demand is normally distributed, n(R) is computed by
using the standardized loss function L(z).
( ) ( ) ( )
z
L z t z t dt |
=
}
If lead time demand is normal with mean and standard
deviation o, then
( ) ( ) ( ) ( ) ( ),

o o
o o

= = = =
}
R
R R
L L z where z n R x R f x dx
Calculations of the optimal policy are carried out using Table A-4
at Page 528 of the book.
where |(x) is standardized normal
density
The procedure of computing Q and R: Q
0
=EOQ
1 ( ) /
i i
F R Qh p =
check z and L(z ) value in Table A-4
i i
i i
R z o = + ( )
( )
i i
n R L z o =
1
2 [ ( )] /
i i
Q K pn R h
+
= +
Q
i
, Q
i+1
close?
Q
i
, Q
i+1
close?
Yes
Stop
No, i=i+1
The End!
Dr. Na GENG
Inventory Control
Demand
Contents
Introduction
The newsboy model
0
=EOQ
1 ( ) /
i i
F R Qh p =
i i
i i
R z o = + ( )
( )
i i
n R L z o =
1
2 [ ( )] /
i i
Q K pn R h
+
= +
Q
i
, Q
i+1
close?
Q
i
, Q
i+1
close?
Yes
Stop
No, i=i+1
Example 5.4 Harveys Specialty Shop sells a popular mustard that
purchased from English company. The mustard costs $10 a jar and
requires a six-month lead time for replenishment stock. The
holding cost is computed on basis 20% annual interest rate; the
lost-of-goodwill cost is $25 a jar; and bookkeeping expenses for
placing an order amount to about $50. During the six-month lead
time, average 100 jars are sold, but with substantial variation from
one six-month period to the next. The demand follows normal
distribution and the standard deviation of demand during each six-
month period is 25. How should Harvey control the replenishment
of the mustard?
Solution to Example 5.4
To find the optimal values of R and Q
The mean lead time demand in six-month lead time is 100, the
mean yearly demand is 200, giving =200;
h=100.20=2; K=50; P=25;
0
Initialization: =EOQ= 2 / 2 50 200 0 / 0 2 1 ; Q K h = =
( )
0 0
Step 1: 1 ( ) / 100*2 / 25*200 0.04; F R Q h p = = =
0 0
Step 2: check in Table A-4 z =1.75 and L(z ) =0.0162;
0 0
Step 3: 100 1.75*25 144; R z o = + = + =
( )
0 0
Step 4: ( ) 25*0.0162 0.405; n R L z o = = =
0
1
2 [ ( )] 2 200[50 25 0.40
11
5]
Step 5:
2
0;
K pn R
Q
h
+ +
= = =
0 1
Compare: Q and Q are not close, so return to step 1 and continue.
( )
( )
1
1 1
1 1
1 1
1 1
2
=110;
Step 1: 1 ( ) / 110*2 / 25*200 0.044;
Step 2: check in Table A-4 z =1.70 and L(z ) =0.0183;
Step 3: 100 1.70*25 143;
Step 4: ( ) 25*0.0183 0.4575;
2 [
Step 5:
Q
F R Qh p
R z
n R L z
K
Q
o
o
= = =
= + = + =
= = =
+
=
1
1 2
( )] 2 200[50 25 0.4575]
;
2
Compare: Q and Q are close, sto
1
.
1 1
p
pn R
h
+
= ~
2 2 2
2 1 2 1
Substitue Q =111 into 1 ( ) / 0.044,
z =z =1.70, R =R =143
F R Q h p = =
Results for Example 5.4: The optimal values of (Q, R)=(111,
143), that is, when Harveys inventory of this type mustard hits
143 jars, he should place an order for 111 jars.
Example 5.4 (Cont.): determine the following
(1) Safety stock;
(2) The average annual holding, setup, and penalty costs
associated with the inventory control of the mustard;
(3) The average time between placement of orders;
(4) The proportion of order cycles in which no stock-outs
occur>Among given number of order cycles, how many
order cycles do not have stock-outs?
(5) The proportion of demands that are not met.
Solution to Example 5.4 (Cont.)
1) The safety stock is s=R-=143-100=43 jars;
2) Three costs:
The holding cost is h(Q/2+s)=2(111/2+43)=$197/jar;
The setup cost is K/Q=50200/111=$90.09/jar;
The penalty cost is p n(R)/Q=25 2000.4575/111=$20.61/jar
Hence, the total average cost under optimal inventory control policy is
$307.70/jar.
3) The average time between placement of orders:
T=Q/ =111/200=0.556 yr=6.7months;
3) Compute the probability that no stock-out occurs in the lead time, which
is the same as that the probability that the lead time demand does not
exceeds the reorder point: P(DsR)=F(R)=1-Qh/p =1-0.044=0.956;
4) The proportion of demand that stock out is n(R)/Q=0.4575/111=0.004.
Demand
Contents
Introduction
The newsboy model
Service Levels in (Q, R) System
In reality, it is difficult to determine the exact value of stock-out
cost p. Acommon substitute for a stock-out cost is a service level.
Service level generally refers to the probability that a demand or
a collection of demand is met.
Service level can be applied both to periodic review and
continuous review systems, that is, (Q, R) system.
Two types of service levels for continuous review system: Type 1
and Type 2
Type 1 Service
Specify the probability of not stocking out in the lead time,
denoted as o.
As the value of R can be completely specified by o,
computation of R and Q can be decoupled.
The computation of the optimal (Q, R) values subject to Type
1 service constraint is straightforward:
(1) Determine R to satisfy the equation F(R)= o;
(2) Set Q=EOQ
Discusses on Type 1 Service
1) o is interpreted as the proportion of cycles in which no
stock-out occurs;
2) A Type 1 service objective is suitable when a shortage
occurrence has the same consequence independent of its time
or amount. For example, a production line is stopped whether
1 unit or 100 units are short.
3) Usually, when we say we would like provide 95% service, we
mean that we would like to be able to fill 95% of the demand
when they occur, rather than fill all of the demands in 95% of
the order cycles. not be specified by Type 1 Service.
4) In addition, different items have different cycle lengths, this
measure will not be consistent among different products,
making the proper choice of o difficult.
Type 2 Service
Measures the proportion of demands that are met from stock,
denoted by |.
Since n(R)/Q is the average fraction of demands that stock out
each cycle, then specification of | results in constraint n(R)/Q
=1- |.
This constraint is more complex than that arising from Type 1
service, because it involves both Q and R.
Although EOQ is not optimal in this case, it usually gives pretty
good results.
If EOQ is used to estimate the lot size, then we would find R to
solve n(R)=EOQ(1- |).
Example 5.5
Harvey feels uncomfortable with assumption that the stock-out
cost is $25 and decide to use a service level criterion instead.
Suppose that he chooses to use 98%.
1) Type 1 service: o=0.98, find R to solve F(R)=0.98. From
Table A-4, z=2.05, R= oz+=252.05+100=151.
2) Type 2 service: |=0.98, n(R)=EOQ(1- |), which
corresponds to L(z)= EOQ(1- |)/ o=100(1-0.98)/25=0.08.
From Table A-4, z=1.02, then R= oz+=251.02+100=126.
The same values of o and | gives considerably different values
of R.
To understand more clearly the difference between these two
measure of service, consider the following example.
Order Cycle Demand Stock-outs
1 180 0
2 75 0
3 235 45
4 140 0
5 180 0
6 200 10
7 150 0
8 90 0
9 160 0
10 40 0
Sum 1450 55
o=0.8
|=(1450-55)/1450=0.96
Optimal (Q, R) Polices Subject to Type 2 Constraints
EOQ is only an approximation of the optimal lot size.
A more accurate value of optimal Q can be obtained as follows
/[(1 ( )) ] p qh F R =
Solving for p in Equation (2) gives
Substituting p in Equation (1) results in
2 { ( ) /[(1 ( )) ]} K Qhn R F R
Q
h
+
=
2
( ) 2 ( )
( ) (3)
1 ( ) 1 ( )
n R K n R
Q
F R h F R
= + +

Service level
order quantity,
SOQ (Service
level order
quantity) formula
( ) (1 ) (4) n R Q | = ( ) (1 ) / (4') L z Q | o =
2 [ ( )]
(1)
1 ( ) / (2)
K pn R
Q
h
F R Qh p
+
=
=
The procedure of computing Q and R for type 2
service: Q
0
=EOQ
( )
( ) (1 ) and L ( ) /
i i i i
n R Q z n R | o = =
( )
check z , 1-F in Table A-4
i i
R
i i
R z o = +
2
1
( ) ( ) 2
( )
1 ( ) 1 ( )
i i
i
i i
n R n R K
Q
F R h F R
+
= + +

Q
i
, Q
i+1
close?
Q
i
, Q
i+1
close?
Yes
Stop
No, i=i+1
0 0
Computing procedure for Exam 5.3:
Q =100, 125, 25. R o = =
( )
( )
0 0
0
Step 1: ( ) (1 ) = 1 0.98 *100=2
and L ( ) / 2 / 25 0.08,
n R Q
z n R
|
o
=
= = =
( )
0
Step 2: check z=1.02, 1-F =0.154 ; R
2 2 2
0 0
1
0 0
( ) ( ) 2 2 2
Step 4: ( ) 100 ( ) 114;
1 ( ) 1 ( ) 0.154 0.154
n R n R K
Q
F R h F R
= + + = + + =

1 0 1 0
Compare Q 114, Q =100, and are not close. Q Q =
( )
( )
1
1 1
1
1
2
1 1
2
1 1
Q 114
Step 1: ( ) (1 ) (1 0.98) *114 2.28
and L ( ) / 2.28 / 25 0.0912,
Step 2: check z=0.95, 1-F =0.171, and
Step 3: compute 124;
( ) ( ) 2
Step 4: ( )
1 ( ) 1 ( )
i
n R Q
z n R
R
R z
n R n R K
Q
F R h F R
|
o
o
=
= = =
= = =
= + =
= + + =

2 2
2 1 2 1
2 2.28
100 ( ) 114
0.171 0.171
Compare and are the same, stop. 124 Q Q R R
+ + =
= =
The optimal values of Q and R satisfying a 98 percent fill rate
constraint are (Q, R)=(114, 124).
Consider the cost error resulting from EOQ substituted for the SOQ.
In order to compare these policies, we compute the average annual
holding and setup costs for the policies (Q, R)=(100, 126) and (Q,
R)=(114, 124).
Recall the formulas for average annual holding cost and setup costs
Holding cost = h(Q/2+R-), Setup cost = K/Q
For (Q, R)=(100, 126)
For (Q, R)=(114, 124)
Holding cost = 2*(100/2+126-100=$152
Setup cost = 50*200/100=$100
Holding cost = 2*(100/2+126-100=$152
Setup cost = 50*200/100=$100
Total cost
$252
Holding cost = 2*(114/2+124-100=$162
Setup cost = 50*200/114=$88
Holding cost = 2*(114/2+124-100=$162
Setup cost = 50*200/114=$88
Total cost
$250
Imputed Shortage Cost
The imputed shortage cost is a useful way to determine whether the
value chosen for the service level is appropriate.
Example 5.5 (Continued) Consider again Harveys Specialty Shop.
Using a value of =0.98 (Type 1 service), we obtain the policy
(100,151). The imputed shortage cost is p=100*2/0.02/200=$50
Using a value of =0.98 (Type 2 service) we obtain the policy (114,
124). In this case, the imputed shortage cost is p=114*2/0.171/200
=$6.67
/[(1 ( )) ] p qh F R =
2 [ ( )]
(1)
1 ( ) / (2)
K pn R
Q
h
F R Qh p
+
=
=
Scaling of Lead Time Demand
Assume that demands follow a normal distribution.
Let the periodic demand have mean and standard deviation ,
and let be the lead time in periods.
The mean demand during lead time is = , and the standard
deviation of demand during lead time is
Example 5.6 Weekly demand for a certain type of automotive spark
plug in a local repair shop is normally distributed with mean 34 and
standard deviation 12. Procurement lead time is six weeks. Determine
the lead time demand distribution.
Solution. The demand over the lead time is normally distributed with
mean 34*6=204 and standard deviation 12* 6=29.39
In-class Quiz
Weisss paint store uses a (Q,R) inventory system to control
its stock levels. The monthly demand of a particularly popular
white latex paint follows N(28, 8
2
). The lead time is 14
weeks.
Each can of paint costs the store $6. Fixed costs of
replenishment are $15 per order, and holding costs are based
on 30% annual rate of interest. Although excess demands are
back-ordered, the store owner wants to determine the lot size
by EOQ formula and determines his reorder point so that
there is no stock-out in 90% of the order cycle.
a) Find the resulting (Q,R) values.
b) Suppose that, the owner really wanted to satisfy 90% of his
demands. What fill rate did he actually achieve from the policy
determined in part (a)
In-class Quiz
a) Type 1 service of 90%
EOQ = 75
F(R) = 0.9, z = 1.28,
R = oz + = (15)(1.28) + 98 = 117
(Q, R) = (75,117)
b) Find Type II service level achieved in part (a).
| = .9905 (99.05% service level)
3.5*28 98, 8* 3.5 15,
15, 0.3*6 1.8, 28*12 336 K h Ic
o
= = = =
= = = = = =
( ) ( ) 15*0.0485
1 0.0095
75
n R L z
Q Q
o
| = = = =
P235 Q8
P251, Q13
P251, Q20
The End!
Dr. Na GENG
Inventory Control
Demand
Contents
Introduction
The newsboy model
(s, S) Policies
It is difficult to implement a continuous-review solution in a
periodic-review environment because the inventory level is
likely to overshoot the reorder point R during a period, which
makes it impossible to place an order the instant the inventory
reaches R.
Define two numbers, s and S, to be used as follows: When the
level of on-hand inventory is less than or equal to s, an order for
the difference between the inventory and S is placed.
Additional Discussion of Periodic-review
Systems
(s, S) Policies
If u is the starting inventory in any period, then the (s, S) policy
is
If us, order S-u;
Else, do not order.
Approximation: to compute a (Q,R) policy using the methods
described earlier, and set s=R and S=R+Q.
This approximation will give reasonable results in many
cases, and is probably the most commonly used.
Systems
Service Level in Periodic-Review Systems
Type 1 service
Objective-find the order-up-to point Q so that all of the demand
is satisfied in a given percentage of the periods
F(Q)=o,
where F(Q) is the probability that the demand during the period
does not exceed Q.
Systems
Type 2 service
To find the Q to satisfy the Type 2 service objective |, it is
necessary to obtain an expression for the fraction of demand
that stock out each period.
Define n(Q), the expected number of demands that stock out
at the end of period.
Systems
( ) ( ) ( )
Q
n Q x Q f x dx
=
}
Since the demand per period is , then the proportion of
demand that stock out each period is n(Q)/=1-|, giving n(Q)
=(1-|).
Example 5.9: Mac, the owner of the newsstand described in
Example 5.1, wishes to use a Type 1 service level of 90
percent to control his replenishment of the Computer Journal.
F(R)= o=0.9
Check in Table A-4, z=1.28
Q*=z+=(4.74)(1.28)+11.73=17.818
Systems
Using a Type 2 service of 90 percent, we obtain
n(Q)=(1-) =(0.1)(11.73)=1.173
It follows that L(z)=n(Q)/ =1.173/4.74=0.2475;
From Table A-4, we find z 0.35
Then Q*= z+ =(4.74)(0.35)+11.73=13.4 13
Systems
Demand
Contents
Introduction
The newsboy model
Example 5.10, Performance of 20 Stock Items Selected at
Random
Part Number Price ($) Yearly Demand Dollar Volume ($)
5497J 2.25 260 585.00
3K62 2.85 43 122.55
88450 1.50 21 31.50
P001 0.77 388 298.76
2M993 4.45 612 2723.40
4040 6.10 220 1342.00
W76 3.10 110 341.00
JJ335 1.32 786 1037.52
R077 12.80 14 179.20
70779 24.99 334 8346.66
4J65E 7.75 24 186.00
334Y 0.68 77 52.36
8ST4 0.25 56 14.00
16113 3.89 89 346.21
45000 7.70 675 5197.50
7878 6.22 66 410.52
6193L 0.85 148 125.80
TTR77 0.77 690 531.30
39SS5 1.23 52 63.96
93939 4.05 12 48.60
Random
70779 24.99 334 8346.66
R077 12.80 14 179.20
4J65E 7.75 24 186.00
45000 7.70 675 5197.50
7878 6.22 66 410.52
4040 6.10 220 1342.00
2M993 4.45 612 2723.40
93939 4.05 12 48.60
16113 3.89 89 346.21
W76 3.10 110 341.00
3K62 2.85 43 122.55
5497J 2.25 260 585.00
88450 1.50 21 31.50
JJ335 1.32 786 1037.52
39SS5 1.23 52 63.96
6193L 0.85 148 125.80
P001 0.77 388 298.76
TTR77 0.77 690 531.30
334Y 0.68 77 52.36
8ST4 0.25 56 14.00
ABC Analysis
Borrow a concept from economics: Pareto effect
The economist Vilfredo Pareto (18481923) , studying
the distribution of wealth in the 19
th
century, noted that a
large portion of wealth was owned by a small segment of
the population
Pareto effect in inventory control: a large portion of the total
dollar volume of sales is often accounted for by a small
number of inventory items.
Assume that items are ranked in decreasing order of the
dollar value of annual sales.
A items: the top 20 percent of the
items account for about 80 percent
of the annual dollar volume of
sales.
B items: the next 30 percent of the
items for the next 15 percent of
sales
C items: the remaining 50 percent
for the last 5 percent of dollar
volume.
A items should be watched most closely. Inventory levels for A
items should be monitored continuously. More complicated
forecasting procedures is needed.
For B items inventories could be reviewed periodically, items
could be ordered in groups rather than individually, and somewhat
less sophisticated forecasting methods could be used.
The minimum degree of control would be applied to C items.
For very inexpensive C items with moderate levels of
demand, large lot sizes are recommended.
For expensive C items with very low demand, the best
policy is generally to order these items as they are
demanded.
Random
5497J 2.25 260 585.00
3K62 2.85 43 122.55
88450 1.50 21 31.50
P001 0.77 388 298.76
2M993 4.45 612 2723.40
4040 6.10 220 1342.00
W76 3.10 110 341.00
JJ335 1.32 786 1037.52
R077 12.80 14 179.20
70779 24.99 334 8346.66
4J65E 7.75 24 186.00
334Y 0.68 77 52.36
8ST4 0.25 56 14.00
16113 3.89 89 346.21
45000 7.70 675 5197.50
7878 6.22 66 410.52
6193L 0.85 148 125.80
TTR77 0.77 690 531.30
39SS5 1.23 52 63.96
93939 4.05 12 48.60
Twenty Stock Items Ranked in Decreasing Order of Annual Dollar
Volume
Cumulatvie Dollar
Volume ($)
70779 24.99 334 8346.66 8346.66
A items:
20% of items
account for 80%
of total value
45000 7.70 675 5197.50 13544.16
2M993 4.45 612 2723.40 16267.56
4040 6.10 220 1342.00 17609.56
JJ335 1.32 786 1037.52 18647.08
B items:
30% of items
account for 14.8%
of total value
5497J 2.25 260 585.00 19232.08
TTR77 0.77 690 531.30 19763.38
7878 6.22 66 410.52 20173.90
16113 3.89 89 346.21 20520.11
W76 3.10 110 341.00 20861.11
P001 0.77 388 298.76 21159.87
C items:
50% of items
account for 5.1%
of total value
4J65E 7.75 24 186.00 21345.87
R077 12.80 14 179.20 21525.07
6193L 0.85 148 125.80 21650.87
3K62 2.85 43 122.55 21773.42
39SS5 1.23 52 63.96 21837.38
334Y 0.68 77 52.36 21889.74
93939 4.05 12 48.60 21938.34
88450 1.50 21 31.50 21969.84
8ST4 0.25 56 14.00 21983.84
Exchange Curves
Much of our analysis assumes a single item in isolation and
that the relevant cost parameters K, h, and p ( or just K and h
in the case of service levels) are constants with correct
values that can be determined.
The correct values are those that result in a control system
with characteristics that meet the needs of the firm and the
goals of management.
In a typical multiproduct system, the ratio K/I can be treated
as a policy variable.
Exchange Curves
EOQ:
2Kx
i
Ic
i
, 1in
Replenish per year:
x
i
i
n
=1
Value of average inventory
c
i
i
2
n
=1
K/I
Exchange Curves
Exchange curves can be used to compare various
safety stock and service level strategies.
Exchange Curves
Consider the 20 stock items listed in Table 5-1 and 5-2.
Suppose that Harvey, the owner of Harveys Specialty Shop,
is reconsidering his choices of the setup cost of $50 and
interest charge of 20%. Harvey uses EOQ formula to compute
lot sizes for the 20 items for a range of K/I from 50 to 500.
The resulting exchange curve appears in the figure.
Demand
Review
Introduction
The newsboy model
The newsboy model
Notation-the newsboy model
A single product is to be ordered at the beginning of a period and can
be used only to satisfy the demand during that period.
Assume that all relevant costs can be determined on the basis of ending
inventory. Define:
c
0
=Cost per unit of positive inventory remaining at the end of the period
(overage cost);
c
u
=Cost per unit of unsatisfied demand, which can be thought as a cost per
unit of negative ending inventory (underage cost).
Critical ratio F(Q*)=c
u
/(c
u
+c
o
)
1) For a continuous demand, check z value in Table A-4, Q-=+oz.
2) In the discrete case, the optimal solution procedure is to locate the critical
ratio between two values of F(Q) and choose the Q corresponding to the higher
value.
The newsboy model- Extensions
Suppose that the starting inventory is some value u and u>0
.
If u<Q*, order Q*-u; If u>Q*, do not order.
Consider the case in which there are infinitely many periods
remaining
If excess demand is back-ordered, interpret c
u
as the loss-
of-goodwill cost and c
o
as the holding cost
If excess demand is lost, interpret c
u
as the loss-of-
goodwill cost plus the lost profit and c
o
as the holding
cost.
decision variables;
Assumptions
Continuous review-demands are recorded as they occur;
Random and stationary demand-the expected value of demand
over any time interval of fixed length is constant; the expected
demand rate is units/year.
Fixed positive lead time, , for placing an order;
The Cost Function:
G(Q, R)=h(Q/2+R-t)+K/Q+pn(R)/Q.
The objective is to choose Q and R to minimize G(Q, R).
0, 0
G G
Q R
c c
= =
c c
2 [ ( )]
(1)
1 ( ) / (2)
K pn R
Q
h
F R Qh p
+
=
=
The solution procedure requires iterating between (1) and (2)
until the two successive values of Q and R are the same.
0
=EOQ
1 ( ) /
i i
F R Qh p =
i i
i i
R z o = + ( )
( )
i i
n R L z o =
1
2 [ ( )] /
i i
Q K pn R h
+
= +
Q
i
, Q
i+1
close?
Q
i
, Q
i+1
close?
Yes
Stop
No, i=i+1
In reality, it is difficult to determine the exact value of stock-out
cost p. Acommon substitute for a stock-out cost is a service level.
Two types of service levels for continuous review system: Type 1
and Type 2
Type 1: the probability of not stocking out in the lead time,
denoted as o. F(R)= o, Q=EOQ
Type 2: the proportion of demands that are met from stock,
denoted by |. n(R)/Q =1- |
The procedure of computing Q and R for type 2
service: Q
0
=EOQ
( )
( ) (1 ) and L ( ) /
i i i i
n R Q z n R | o = =
( )
check z , 1-F in Table A-4
i i
R
i i
R z o = +
2
1
( ) ( ) 2
( )
1 ( ) 1 ( )
i i
i
i i
n R n R K
Q
F R h F R
+
= + +

Q
i
, Q
i+1
close?
Q
i
, Q
i+1
close?
Yes
Stop
No, i=i+1
Imputed Shortage Cost
Assume that demands follow a normal distribution.
Let the periodic demand have mean and standard deviation , and
let be the lead time in periods.
The mean demand during lead time is = , and the standard
deviation of demand during lead time is o = :
/[(1 ( )) ] p qh F R =
(s, S) Policies
If u is the starting inventory in any period, then the (s, S) policy
is
If us, order S-u;
Else, do not order.
Type 1 service: F(Q)=o
Type 2 service: n(Q)/=1-|
Systems
Assume that items are ranked in decreasing order of the
dollar value of annual sales.
A items: the top 20 percent of the
items account for about 80 percent
of the annual dollar volume of
sales.
B items: the next 30 percent of the
items for the next 15 percent of
sales
C items: the remaining 50 percent
for the last 5 percent of dollar
volume.
P253, Q23
P259 Q24
The End!
Dr GENG Na
Scheduling in Supply Chain Management
Contents
Introduction
Transportation Problem
Generalizations of the Transportation;
More General Network Formulations
Distribution Resources Planning;
Determining Delivery Routes in Supply Chain
The Role of information in the SCM
Multilevel Distribution Systems
Designing the Supply Chain in a Global Environment
Supply Chain
What is Supply Chain?
Nitrogen
&
phosphor
alga
Plankton
Small
fish
Big fish
Marine food chain
Raw material
supplier
Manufactu-
rer
Distributer Retailer
Users
Introduction
The Definition of Supply Chain Management (SCM)
Hau-Lee, the head of Stanford Supply Chain Forum (1999):
SCM deal with the management of materials, information
and financial flows in a network consisting of suppliers,
manufacturers, distributors, and customers.
Simchi-Levi et al.(1999,p1): SCM is a set of approaches
utilized to efficiently integrate suppliers, manufactures,
warehouses, and stores, so that merchandise is produced and
distributed at the right quantities, to the right locations, and
at right time, in order to minimize system wide costs while
satisfying service level requirements.
Supply Chain Management
The clearest description appeared in a Fortune magazine article
devoted to the subject:
Call it distribution or logistics or supply chain management. By
whatever name, it is the sinuous, gritty, and cumbersome process
by which companies move material, parts, and products to
customers. In industry after industry, fromcars and clothing to
computersandchemicals, executiveshavepluckedthisoncedismal
discipline off the loading dock and placed it near the top of the
corporate agenda. Hard-pressed to knock out competitors on
quality or price, companiesaretryingtogainanedgethroughtheir
abilitytodeliver theright stuff intheright amount at theright time.
(Henkoff, 1994)
Introduction
The Supply Chain as a Strategy Weapon
The design of supply chain reflects a firms strategic
positioning; (low cost /fast response)
Supply chain trade off between cost and response time;
Third party logistic (3PL) is becoming more common
Less expensive to subcontract logistics than to do it in-
house.
Maintain core competency
SCM make it possible to move product quickly and efficiently
so as to gain an edge over their competitors; (Outsourcing)
New technologies will trim costs from the supply chain without
compromising service.
Contents
Introduction
The transportation problem is a mathematical model for
optimally scheduling the flow of goods from production
factoriestodistributioncenters.
The transportation problemcan be viewed as prototype supply
chainproblem
Example 6.1
The Pear Disk Corp. produces 80GB drives in three plants and
supply four distribution centers. The production capacities for the
three factories in the next month are 45,000, 120,000, and 95,000.
The amounts shipped to the four DCs. are 80,000, 78,000, 47,000,
and 55,000. The unit cost for shipping 1,000 units drives from each
plant to each warehouse is given. The goal is to determine a pattern
of shipping that minimize the total transportation costs.
Fig.6-2 Pear Disk Drive Transportation Problem
80
78
47
55
45
120
95
250
420
380
280
1280
990
1440
1520
1550
1420
1660
1730
Solve TP by Greedy Heuristic: sub-optimal
Amarillo Teaneck Chicago Sioux Falls
Warehouse (Sinks)
Factories
(sources)
Sunnyvale
Dublin
Bangkok
250 420 380 280
1,280 990 1,440 1,520
1,550 1,420 1,660 1,730
45
120
95
80 78 47 55
55
45
Min (45, 80)
78
Min (78, 120)
35
Min (80-45, 120-78)
7
Min (120-35-78, 47)
40
Min(47-7, 95)
The total cost=
45250+35,1280+
=$304,900,
Solve TP by LP
Let m (=3) be the number of sources and n (=4) be the number of
sinks;
a
i
= the supply from the source i, 1 im
b
j
=the demand of sink j, 1jn;
c
ij
=the cost of shipping one unit from source i to sink j;
x
ij
=the flow from the source i to sink j for 1 imand 1jn;
1 1
m n
ij ij
i j
Min c x

Subject to
1
1
: , 1 ;
: , 1 ;
: 0, 1 1
n
ij i
j
m
ij j
i
ij
Supply x a for i m
Demand x b for j n
Nonegtive x for i mand j n
Fig 6-3 Solution of Pears Transportation Problem Using Excel Solver
Solution of Example 6.1
Variables x11 x12 x13 x14 x21 x22 x23 x24 x31 x32 x33 x34 Operator Value
Values 0 0 0 45 42 78 0 0 38 0 47 10
Heuristic 304900
Objective
Coeff
250 420 380 280 1280 990 14401520 1550 1420 1660 1730 LP Min 297800
st LHS RHS
Constraint 1 1 1 1 1 = 45 45
Constraint 2 1 1 1 1 = 120 120
Constraint 3 1 1 1 1 = 95 95
Constraint 4 1 1 1 = 80 80
Constraint 5 1 1 1 = 78 78
Constraint 6 1 1 1 = 47 47
Constraint 7 1 1 1 = 55 55
The End!
Prof. JAING Zhibin
Dr GENG Na
Scheduling in Supply Chain
Management
Contents
Introduction
Generalizations of the Transportation
Infeasible Routes
Eliminate routes from some sources to some sinks=Set the costs
of these routes very high;
Traditionally M is used to denote the very high cost. However, in
practice, costs of these routes are assigned numbers much larger
than the costs of other feasible routes, so that optimal solution
will never assign flow to these routes.
Unbalanced Problems
The total amount shipped from the sources is not equal to the total
amount required at the sinks.
Occurs when the demand exceeds the supply or vice versa.
Two measures
either add a dummy row or a dummy column to absorb the
excess supply or demand;
Alter the appropriate set of constraints to either or form.
Suppose that the demand for the drives was higher than
anticipated.
The anticipated demand for the four sinks:
80 78 47 55 (the anticipated total demand is 260)
The actual demands:
90 78 55 55 (the actual total demand is 278)
Treatments to turn into balanced problem in Greedy Heuristic
Add an additional fictitious factory to account for this 18-unit
shortfall: add a dummy row in the transportation tableau and all
entries for this row are assigned an arbitrarily large unit cost.
Demand 278>supply 260
Amarillo Teaneck Chicago Sioux Falls
Warehouse
Factories
Sunnyvale
Dublin
Bangkok
250 420 380 280
1,280 990 1,440 1,520
1,550 1,420 1,660 1,730
45
120
95
90 78 55 55
18 A
M M M M
Example 6.1
Variables x11 x12 x13 x14 x21 x22 x23 x24 x31 x32 x33 x34 x41 x42 x43 x44
Oper
ator
Value
Values 0 0 0 45 42 78 0 0 48 0 47 0 0 0 8 10
Objective
Coeff
250 420 380 280 1280 990 1440 1520
155
0
142
0
16601730
1E+0
6
1E+0
6
1E+0
6
1E+06 Min 1.8E+07
st LHS RHS
Constraint 1 1 1 1 1 = 45 45
Constraint 2 1 1 1 1 = 120 120
Constraint 3 1 1 1 1 = 95 95
Constraint 4 1 1 1 1 = 18 18
Constraint 5 1 1 1 1 = 90 90
Constraint 6 1 1 1 1 = 78 78
Constraint 7 1 1 1 1 = 55 55
Constraint 8 1 1 1 1 = 55 55
Treatment for LP
Unbalanced transportation problem can be formulated as LP
by inequality (demand or supply )constraints.
When the demand exceeds the anticipated, use equality
supply constraint (first three) to make sure of complete
shipment; and use less than or equal demand constraint.
1 1
1
1
..
1 ;
1 ;
0 1 1
n m
ij ij
i j
m
ij i
j
n
ij j
i
ij
Min c x
St
x a for i n
x b for j m
x for i nand j m

1 1
1
1
..
1 ;
1 ;
0 1 1
n m
ij ij
i j
m
ij i
j
n
ij j
i
ij
Min c x
St
x a for i n
x b for j m
x for i nand j m

Unbalanced TPDemand > Supply

Solve TP by LP
Let m be the number of sources and n be the number of sinks;
a
i
= the supply from the source i, 1 im
b
j
=the demand of sink j, 1jn;
c
ij
=the cost of shipping one unit from source i to sink j;
x
ij
=the flow from the source i to sink j for 1 imand 1jn;
Contents
Introduction
Transportation problem is a special network: all nodes are either
supply nodes or demand nodes;
More complex problems, e.g. transshipment problem, where one
or more nodesin the network are transshipment nodes rather a
supply or demand node;
Note a transshipment node can be also a supply or a demand
node (but no node is both a supply and a demand node).
Rule
Supplya negative number attached the node;
Demanda positive number attached to the node
x
ij
the total flow from node i to node j;
If Apply the following rules at each node
1. Total supply>Total demand
2. Total supply<Total demand
3. Total supply=Total demand
Inflow - Outflow Supply or demand
Inflow - Outflow Supply or demand
Inflow - Outflow= Supply or demand
Rules for balancing flow of general network
Example 6.2 In Example 6.1, the company considers to place a
warehouse in Sacramento to be used as a transshipment node and
has also expanded the Chicago facility to also allow for
transshipment. The two nodes are also demand nodes.
Fig 6-5 Pear Disk Drive Problem with Transshipment Nodes
Supply 260<Demand 285
Rule 2
Inflow-Outflow Supply
or demand
LP
min 250x
16
+76x
14
+380x
15
+1440x
25
+1660x
35
+110x
46
+95x
48
+180x
56
+120x
57
+195x
58
Subject to
Node 1: -x
14
-x
15
-x
16
-45;
Node 2: -x
25
-130;
Node 3: -x
35
-95;
Node 4: x
14
-x
46
-x
48
25;
Node 5: x
15
+x
25
+x
35
-x
56
-x
57
-x
58
47;
Node 6: x
16
+x
46
+x
56
80;
Node 7: x
57
78;
Node 8: x
48
+x
58
55;
Fig 6-5 Pear Disk Drive Problem with
Transshipment Nodes
Variables x14 x15 x16 x25 x35 x46 x48 x56 x57 x58
Values 45 0 0 120 95 0 20 80 78 10
Objective Coeff 76 380 250 1440 1660 110 95 180 120 195 Min 361530
st LHS relation RHS
Constraint 1 -1 -1 -1 -45 <= -45
Constraint 2 -1 -120 <= -120
Constraint 3 -1 -95 <= -95
Constraint 4 1 -1 -1 25 <= 25
Constraint 5 1 1 1 -1 -1 -1 47 <= 47
Constraint 6 1 1 1 80 <= 80
Constraint 7 1 78 <= 78
Constraint 8 1 1 30 <= 55
Discussion:
If demands are more than supply, shortfall will occur;
Node 4 (Sacramento): x
14
-x
46
-x
48
=45-0-20=25, the demand is
met;
Node 5 (Chicago): x
15
+x
25
+x
35
-x
56
-x
57
-x
58
=0+120+95-80-78-
10=47, the demand is also satisfied;
Node 8 (Sioux Falls): x
48
+x
58
=20+10=30<55, all 25 units
shortfall is absorbed in Sioux Falls Node 8.
Variables x14 x15 x16 x25 x35 x46 x48 x56 x57 x58
Values 45 0 0 120 95 0 20 80 78 10
Contents
Introduction
Distribution Resources Planning
Distribution Resource Planning
DRP was first coined by Martin;
It uses MRP logic to manage distribution systems;
The primary advantages of DRP over traditional ROP(reorder
point systems) is that it can cope with changing demand
patterns;
DRP does not handle uncertain demand;
Example 6.3-The campus bookstore decides to implement a DRP
approach for reordering blue books because of the wide demand
variation. Sales become high just prior to the exam period and
very low during winter and spring breaks.
DRP-Based Order Profile for Blue Book
On hand 26, Order Lead Time 2 weeks, Safety Stock 0
Weeks 0 1 2 3 4 5 6 7 8
Sales forecast 22 35 60 12 0 19 85 33
In transit 40
On order 26
Project balance 26 44 35 -25
On hand 25
Order Lead Time 2 weeks
Safety Stock 0
Weeks 0 1 2 3 4 5 6 7 8
Sales forecast 22 35 60 12 0 19 85 33
In transit 40
On order 26
Project balance 26 44 35
Planned shipments
Order date
Planned shipments
receipt date
25
25
0 0 0 0
12
12
Place 2 orders before wk 6
Combine 2
orders into 1
19
19
On hand 25
Order Lead Time 2 weeks
Safety Stock 0
Weeks 0 1 2 3 4 5 6 7 8
Sales forecast 22 35 60 12 0 19 85 33
In transit 40
On order 26
Project balance 26 44 35
Planned shipments
Order date
Planned shipments
receipt date
37
37
12 0 0 -19
Which one is better ?
(in-class quiz)
Discussion
How to determine the size of the planned shipment?
Precisely the same as the method adopted for deciding lot
size in MRP as will be addressed in Chapter 7 ;
Positive safety stock may be incorporated in DRP approach-
orders are placed so that anticipated on-hand inventory never
drops below the safety stock.
The End!
Prof. JIANG Zhibin
Dr GENG Na
Scheduling in Supply Chain
Management
Contents
Introduction
Determining Delivery Routes in Supply Chains
Traveling Salesman Problem Description:
A salesman starts at his home base, labeled City 1. He must make
stops at n-1 other cities, visit each city exactly once. The problem
is to optimize sequence in which to visit all the cities to minimize
the total distance traveled.
The complexity in finding the solution:
If the number is small, it is possible to enumerate all the
possible tours, and select the shortest one;
For n cities, there are n! different sequences;
If we could evaluate 1 trillion (10
12
) sequences per second on a
supercomputer, then a 25-city problem, nearly 500,000 years are
required to evaluate all the sequences;
Mathematically NP hard--No Polynomial, meaning that the
time required to solve such problems is an exponential function
of the number of cities rather than a polynomial function.
Finding optimal rout in vehicle scheduling is similar, but more
complex problem;
Assume that there is a central depot with one or more delivery
vehicles and n customer locations, each having a known
requirement. Thequestionis howto assignvehicles to customer
locations to meet customer demand and satisfy whatever
constraintstheremayexitsat minimumcost.
Optimal solutionisimpossibleandonlygoodsolutionispossible
(proposedbyClarkandWright).
Supposethat thereis asingledepot all vehicles depart fromand
returnto.
Identify thedepot aslocation0andthecustomersaslocations 1,
2, , n.
Assume all the costs traveling from depot to each customer
locationandamongcustomer locationsarec
0j
andc
ij
respectively.
Clark and Wrights Saving Method
Only consider situation of c
ij
=c
ji.
Suppose that initially a separate vehicle is assigned to each
customer location. Then the initial solution consists of n
separate routes from the depot to each customer location and
back, so that the initial solution is
0
1
2
n
j
j
c
=
Suppose that link customer i and

j, then save two trips 0-j and 0-i,
however add one trip i-j. The
saved cost by linking i and j is :
0 i
j
s
ij
=c
0i
+c
0j
-c
ij
Clark and Wrights Saving Method
Procedures:
1. Compute s
ij
for all possible pairs of customer locations i and j,
and rank the s
ij
in decreasing order.
2. According to descending order of savings, include link (i, j) in a
route if it does not violate feasibility constraints.
3. If including the current link (i, j) violates the feasibility,
consider other feasible link (i, j) until all the lists is exhausted, and
then stop the current route.
4. If the current route includes all the remaining locations, stop;
otherwise, go to Step 2 to form a new route.
The total number of calculation of is:
! ( 1)
2 2!( 2)! 2
ij
s
n n n n
n
| |
= =
|

\ .
Example 6.4-Whole Grains is a bakery that
supplies five major customers with bread each
morning. Thelocationsof thebakery andthefive
customers and their demand are given as
described in a grid. The bakery has several
delivery trucks, each having a capacity of 300
loaves.
Custom Location Daily Req.
( Loaves)
1
2
3
4
5
(15, 30)
(5, 30)
(10, 20)
(5, 5)
(20, 10)
85
162
26
140
110
Cost Matrix (c
ij
) (=distance)
TO
F 0
R 1
O 2
M 3
4
0 1 2 3 4 5
33.5 30.4 22.4 7.1 22.4
10.0 11.2 26.9 20.6
11.2 25.0 25.0
15.8 14.1
15.8
Saving for all pairs (i, j), i, j=1~5:
s
12
=c
01
+c
02
-c
12
=33.5+30.4-10=53.9
s
13
=44.7, s
14
=13.7, s
15
=35.3, s
23
=41.6,
s
24
=12.5, s
25
=27.8
s
34
=13.7, s
35
=30.7, s
45
=13.7
Rank the customer pairs in the
decreasing order of all the saving
values: (1, 2)(1, 3) (2, 3) (1, 5)
(1, 4), (3, 4), (4, 5) (2, 4)
Customer Daily Req.( Loaves)
1
2
3
4
5
85
162
26
140
110
85+162=247<300
247+26=273<300
110+140=250<300
Practical Issues in VS
Frequency requirements;
Visits to customers may have to occur at a certain frequency,
and that frequency may vary from customer to customer.
Time windows;
This refers to the requirement that visits to customer
locations be made at specific times.
Time-dependent travel time;
When deliveries are made in urban centers, rush-hour
congestion can be an important factor.
Practical Issues in VS (Con)
Multidimensional capacity constraints;
There may be constraints on weight as well as on volume.
Vehicle types.
Large firms may have several vehicle types from which to
choose, which makes the number of feasible alternatives
increases dramatically.
Split deliveries
If one customer has a particular requirement, it could
make sense to have more than one vehicle assigned to
that customer.
Uncertainty
The time required to cross certain portion of a network
could be highly variable.
Contents
Introduction
The Role of Information in the Supply Chain
TheBullwhipEffect
Thehistoryof bullwhipeffect
Executive at Procter & Gamble (P&G) studied the
replenishment patters for one of their best selling products
(diapers).
Orders placed by distributors had much more variation than
salesat retail stores.
Ordersof materialstosuppliershadevengreater variability.
Demand for diapers is pretty steady, however, variation in the
entiresupplychainishigh, rather thanlowasonewouldexpect.
P&Gdescribedthisphenomenonasbullwhip effect.
Beer games demonstrated bullwhip effect.
Participants play roles of retailers, wholesales, and
manufacture of beer;
Each player must make ordering decision based on what is
demanded from the downstream play only;
Even the original demands are fairly stable, wild swings in
orders placed downstream are found.
The Main causes for the bullwhip effect
Demand forecast updating;
Order batching;
Price fluctuation;
Shortage gaming;
Human Behavior
Measures to alleviate the effect
Information sharing;
Channel alignment;
Price stabilization;
Discouragement of shortage gaming
VMI(Vendor-managed inventory )
This is one example of what today is known as Vendor-
managedinventory(VMI).
With VMI programs, it becomes the manufacturers
responsibility to keep the shelves stocked. VMI programs
have been the mainstay of many successful retailers in the
UnitedStates.
Electronic commerce
Electronic commerce is a catch-all term for a wide range
of methods for effecting business transactions without the
traditional paper-based system.
Electronic Data Interchange (EDI)
EDI refers to the electronic transmission of standard
business documents in a predetermined format from one
companys business computer to its trading partners
computer.
Web-based transactions systems
Virtually, everyone has access to the Web.
The Web does not require the complex protocols that
characterize EDI systems.
Web-based systems can be set up and implemented at a
fraction of the cost of EDI.
RFID technology provides faster product flow
Contents
Introduction
Advantages:
Risk pooling
Distribution centers that can be designed to meet local
needs.
Economies of scale in storage and movement of goods.
Faster response time.
Disadvantage
May require more inventory.
May increase total order lead times from plants to
customers.
Could ultimately result in greater costs for storage and
movement of goods.
Could contribute to bullwhip effect.
Contents
Introduction
Designing the supply chain in a global environment
Four driving forces for globalization process:
Global market forces
Technology forces
Global cost forces
Political and macroeconomic forces.
Home works
Q1, Q2, Q3
Q4, Q5
Q9
Q16
The End!
Dr GRANG Na
Management
Chapter 7
Production Planning
Contents
Introduction
Master Production Scheduling (MPS)
Capacity Planning
Improvement in MRP
Overview of Production Planning Activities
Production Planning:
Given specific process planning,
process technologies, and
production conditions, production
planning predetermine varieties,
quantities, quality, and schedules
of products to be produced
according to market demand of
products.
Figure 4.1 Framework of Production Planning Activities
Process Planning
Aggregate Planning
Master Production Scheduling
(MPS)
Material Requirement Planning
(MRP)
Operations Scheduling (Shop
floor Scheduling & Control)
Demand Forecast
Rough-cut
capacity Planning
Long Range
Medium
Range
Short range
Strategic capacity Planning
Capacity Req.
Planning
Order
Management
Time Dimensions:
Long-range planning is done
annually and focus on a
planning horizon greater than
oneyear;
Medium-range planning
usually covers aperiod from6
months to 18 months, with
monthly or sometimes
quarterly timeincrements;
Short-range planning covers a
period fromoneday or less to
six months, with weekly time
increment usually.
Process Planning
Aggregate Planning
(MPS)
(MRP)
Demand Forecast
Rough-cut
capacity Planning
Long Range
Medium
Range
Short range
Capacity Req.
Planning
Order
Management
Process Planning determines
the specific technologies and
procedures required to
produceaproduct aservice.
Strategic capacity planning
determines long-term
capabilities (e. g. size and
scope)
Aggregateplanning concerns
with setting up production
rate by product family or
other categories for
intermediate term (6-18
months).
Process Planning
Aggregate Planning
(MPS)
(MRP)
Demand Forecast
Rough-cut
capacity Planning
Long Range
Medium
Range
Short range
Capacity Req.
Planning
Order
Management
Master production
scheduling generates the
amountsanddatesof specific
itemsrequiredbyorders.
The inputs into MPS are
arrivedordersandAPresults.
Rough-cut capacity planning
is used to verify that the
production and warehouse
facilities, equipment, and
labor are available , and the
key suppliers have allocated
sufficient capacity to provide
materialswhenneeded.
Process Planning
Aggregate Planning
(MPS)
(MRP)
Demand Forecast
Rough-cut
capacity Planning
Long Range
Medium
Range
Short range
Capacity Req.
Planning
Order
Management
Material requirement planning
takes the end product
requirements from MPS and
breaks them down into their
components and subassemblies
to create a material plan
(production orders and
purchase order).
Capacity requirement planning
(CAP) allocate production
resources to each production
order.
Operations scheduling
allocates jobs to specific
machines, production lines or
work centers.
Process Planning
Aggregate Planning
(MPS)
(MRP)
Demand Forecast
Rough-cut
capacity Planning
Long Range
Medium
Range
Short range
Capacity Req.
Planning
Order
Management
Chapter 7 Production Planning
Contents
Introduction
Capacity Planning
Improvement in MRP
Month 1 2
Mattress production 900 950
Week 1 2 3 4 5 6 7 8
Model 327 200 400 200 100
Model 538 100 100 150 100
Model 749 100 200 200
MPS for mattress models
Aggregate production plan
for mattress
Aggregate production plan for mattress specifies the total number
of mattress planned per month, without regard of mattress types;
MPS specifies the exact types of mattress and quantities planned
for production by week.
Mater Production Scheduling
Aggregate planning specifies product groups, rather than
exact items;
As thenext level downintheplanningprocess, MPS is time
phasedplanthat specifieshowmanyandwhenafirmtobuild
eachenditem.
Inthecaseof thefurniturecompany
Its AP may specify the total volume of mattress it plan to
produceover next month, e. g. 900for thenext 1month;
ItsMPSidentifiesperiodbyperiod( usuallyweekly)
whichmattressstylesandhowmanyof thesemattressstylesare
needed, 200Model 200for Wk 1, 100Model 538for bothWks
2and3respectively, and100Model 749for Wk3...
Week 1 2 3 4 5 6 7 8
Model 327 200 400 200 100
Model 538 100 100 150 100
Model 749 100 200 200
Could a MPS be changed?-Flexibility of MPS
When updated?
New order arrives;
Unfinished orders when the products are delayed;
Some work center becomes bottleneck;
There are shortage of raw material
Other reasons.
MPS needs the continuous revisions:
Production lead time;
The cycle of demand forecasting;
The change of orders
MPSneedscontinuousrevisionsinorder tomeet theneeds
of customersbyreducingtheinventory, ontimedelivering,
andimprovingtheproductionefficiency.
Rolling production plan is used to make MPS.
Rolling production plan is to update the production plan
over a predetermined horizon based on the planned
production.
Key problems when making MPS
TheconnectionwithAPP
APP considers the production level and workforce
level;
Production level in APP is broken down into the
demand in every period with respect to each end
product;
Combining with the initial inventory, the broken
productionlevel istheforecasteddemandinMPS.
Thefrozenperiodof MPS:
MPSisthebasisof MRP.
The revision of MPS, especially that has been
implemented, impactsMRP, whichleadstotheincrease
of thecost.
The way to deal with this problemis to set different
time fences during which MPS cant be revised or
easily revised: demand frozen timefenceand planning
frozentimefence.
Thefrozenperiodof MPS:
Demandfrozentimefence: Duringtheperiodfromnow
to demand frozen time fence, MPS cant be revised
without thepermit of managingexecutives.
Planning frozen time fence: it is longer than demand
frozen time fence. During the period from demand
frozen time fence to planning frozen time fence, the
application softwarecant reviseMPS, but the planner
canreviseit.
Figure 4.2 MPS Time Fences
Each company may have its
own time fences and operating
rules
Frozen: absolutely no change
couldbemadeinafirm, or the
most minor changes may be
allowedinanother.
Moderately firm: some
changes to specific products
within a products group so
longaspartsareavailable;
Flexible: almost an variations
in products are allowable,
providing that capacity
remains about the same and
that there are no long lead
timeitemsinvolved.
Alternatives of MPS in different production types
MPS is to determine the quantity of each end product in
everyperiod.
For MTS, endproduct isthefinishedproduct.
ForATO, endproductsarethecomponentsandparts.
For MTO, end products are raw materials and basic
parts.
In-class Quiz
What kinds of constraints should be considered when
making MPS?
Meet the needs of demands
The constraints of capacity
The End!
Dr GENG Na
Management
Chapter 7
Production Planning
Contents
Introduction
Capacity Planning
Improvement in MRP
The procedure for making MPS
1. The preparation of demand information:
For mass production, demand is forecasted based on
historical data.
For one-of-a-kind production, demand is calculated
based on the order or the estimation of demand by
visiting customers.
For batch production, demand is combined from the
orders and forecasting.
1. The preparation of demand information:
The production of end items is estimated
according to APP and plan bills.
The total demand is calculated according to the
production estimation, orders, the demand as
dependent items.
2. The determination of output plan
Scheduled receipt (SR) and projected available balance
(PAB) are calculated according to the total demand, the
order strategy and lot size, safety stock and initial
inventory level.
PAB
t+1
=PAB
t
+SR
t+1
-(Total Demand)
t+1
PAB
0
=0
If PAB0, the demand can be satisfied, and no MPS is
scheduled;
If PAB<0, one or multiple lots of products should be
scheduled as SR.
The output plan is used for directing production, not for
sales.
The part that can be satisfied from inventory is not
considered.
The production lot size and interval are chosen to ensure
the economic efficiency of production.
The large fluctuation of workload should be avoided in
order to keep the stable production.
Some abstract demand needs to be turned into detailed
products.
For mass production, the output is evenly distributed
over every month.
For batch production, the reasonable coordination is the
focus.
For the products with small demands, the concentrated
production is preferred.
The due date should be obeyed.
MPS MPS
Capacity constraints
Forecast
Production Rules
Orders
Product types
Production time
Production quantity
Production lead time
3. The check of enough capacity
The production capacity unit for MPS is workshop.
Allocate the production in the capacity unit, and
evaluate whether the capacity is enough for the
production.
The end products are decomposed into the dependent
items.
The workload is calculated with these items, their due
date, and the process flow.
The principles for making MPS
1. The basic principles for making MPS
First of all to ensure that the time and quantity of output
meet the demands of orders.
The reasonable combination of products should be
considered for multiple products firms.
The pilot runs of new products should be distributed over
the year.
The purchase of raw materials, components, and parts
should consider the output plan.
Pay attention to the connection of output plans in
sequential years.
The principles for making MPS
2. The revision of MPS
Determine the time fence
The implementation of MPS is the responsibility of
manufacturing and marketing departments.
Timely response for revision is required.
The constraints for MPS
1. The output of MPS should meet the requirement of APP.
The total output of every product in MPS should equal
with that determined by APP.
The total output of every product in MPS should be
efficiently allocated in the corresponding periods.
2. The output of MPS should consider the constraint of
capacity.
Allocate the capacity in the order of priority.
Critical resources should be used in critical products.
The constraints for MPS
Product\Week 1 2 3 4 5 6 7 8
End Item A 50 - 50 - 60 - 60 -
End Item B 610 610 610 610 610 610 610 610
End Item C 340 360 380 400 300 310 320 330
End Item D 180 180 180 180 150 150 150 150
End Item E 15 - - - - - 20 -
End Item F 205 205 205 205 225 225 225 225
Table of MPS
The calculation of MPS
Example. The calculation of MPS.
Material No. LA001 Material Name Lamp
Initial Inventory 15 Safety inventory 5
Lead time 1 week Lot size 60
Demand time fence 3rd week Plan time defence 8th week
Week Now 1 2 3 4 5
Forecasting 20 20 20 20 20
Order 25 18 23 16 28
Week 6 7 8 9 10
Forecasting 20 20 20 20 20
Order 15 24 18 20 18
Demand time defnece
Plan time defence
1. The first step is to calculate the gross demand (Gross Requirement, GR)
Suppose the gross requirement of product i in period t is GR
i
(t), then
where, D
i
(t) is the real demand from orders of product i in period t;
F
i
(t) is the forecasted demand of product i in period t;
t
d
is the demand time fence;
t
p
is the plan time fence;

max ,
<
i d
i i i d p
i p
D t t t
GR t D t F t t t t
F t t t

Week Now 1 2 3 4 5 6 7 8 9 10
Forecasting 20 20 20 20 20 20 20 20 20 20
Order 25 18 23 16 28 15 24 18 20 18
GR 25 18 23 20 28 20 24 20 20 20
Demand time defnece Plan time defence
2. Determine the Scheduled Receipts, SR
The scheduled receipts is that has been ordered or produced, and
will be due in period t.
SR
i
(t) is the scheduled receipts of product i in period t.
3. Calculate the Projected On-Hand, POH
If there is no Planned Order Receipts in period t, then the inventory
at the end of period is POH, which is used to determine whether
there is net demand.
POH in first periodP0E
1 = 0E +SR
1 -0R
1
POH in other periods: P0E
t = PAB
t -1 +SR
t -0R
t
Where OH is the initial inventory, PAB
i
(t-1) is the projected
available balance of product i in period t-1. We will explain it
later.
4. Calculate the Net Requirement, NR
Suppose that the net requirement of product i in period t is NR
i
(t)
If safety inventory is not considered,
NR
t = 0R
t -SR
t -PAB
t - 1
If safety inventory SS is considered,
NR
t = SS +0R
t - SR
t -PAB
t - 1 = SS - P0E
t
If NR is negative, there is not net requirement; otherwise, there
exists net requirement.
5. Calculate the Planned Order Receipts, PORC
Suppose that the Planned Order Receipts of product i in period t is
PORC
i
(t).
If there is net demand in period t, it is required that the number of
materials is at least the number of net demand, and the output
time is the same with that of net demand.
PORC should consider the economic lot size, therefore, PORC is
the multiple times of lot size.
6. Calculate the Planned Order Release, POR
Suppose that the Planned Order Release of product i in period t is
POR
i
(t). POR
i
(t)= PORC
i
(t-LT) where LT is the lead time.
7. Calculate the Projected Available Balance, PAB
Suppose that the projected available balance of product i in period t
is PAB
i
(t)
PAB
t = PAB
t -1 +SR
t +P0RC
t -0R
t
= P0E
t +P0RC
t
8. Calculate the Available to Promise, ATP
ATP is the excess product before the next output, which is the
number of output minus the total demand during the period from
now to the period of next output.
Suppose that the Available to Promise of product i in period t is
ATP
i
(t).
AIP
t = P0RC
t +SR
t - 0R
t`-1
]=t
]
Where t is the period of next output.
0 1 2 3 4 5 6 7 8 9 10
GR 25 18 23 20 28 20 24 20 20 20
SR 20 0 0 0 0 0 0 0 0 0
POH 10 -8 29 9 -19 21 -3 37 17 -3
PAB 15 10 52 29 9 41 21 57 37 17 57
NR 0 13 0 0 24 0 8 0 0 8
PORC 0 60 0 0 60 0 60 0 0 60
POR 60 0 0 60 0 60 0 0 60 0
ATP 10 0 0 0 12 0 0 0 0 0
Periodsweek)
1 2 3 4 5 6 7 8 9 10
25 18 23 20 28 20 24 20 20 20
20 0 0 0 0 0 0 0 0 0
10 -8 29 9 -19 21 -3 37 17 -3
15 10 52 29 9 41 21 57 37 17 57
0 13 0 0 24 0 8 0 0 8
0 60 0 0 60 0 60 0 0 60
60 0 0 60 0 60 0 0 60 0
10 0 0 0 12 0 0 0 0 0
The End!
Dr GENG NA
Management
Chapter 7
Production Planning
Contents
Introduction
Capacity Planning
Improvement in MRP
MRP-Overview
MRP create schedules identifying
the specific parts and materials required to produce end items
planned by MPS;
The exact numbers needed; and
The date when orders for these materials should be released
and be received or completed within the production cycle.
MRP-Overview
The main purpose of MRP are to control inventory levels,
assignoperatingprioritiesfor items, andplancapacity toload
theproductionsystem.
Inventory-Order the right part in right quantity at the right
time;
Priorities-Order withright duedate; keeptheduedatevalid;
Capacity-Plan for a complete load, an accurate load, or for an
adequate time to view future load
MRP-Overview
Philosophy MRP Materials should be expedited (hurried)
when their lack would delay the overall production schedule.
The main advantages of MRP
Ability to price more competitively;
Reduced inventory;
Reduced price;
Better customer service;
Better response to market demand;
Ability to change master schedule;
Reduced setup and tear-down costs;
Reduced idle time;
MRP-Principle
Aggregate
product plan
MPS
MRP
Explosion
Calculus
Bill of
Mterials
(BOM)
Inventory
record files
Production
order
Purchase
order
Firm orders
from known
customer
Forecast of
demand from
unknown
customer
Inputs for MRP
Master production specifies the
number of items (end products or
subassemblies or parts for the assembly
of end products) to be produced during
specifictimeperiods;
BOM (Bill of material) identifies the
specific materials used to make each
itemandcorrect quantitiesof each;
The inventory records file supplies
datasuchasthenumber of unitsonhand
andonorder;
MRP explosion calculus
implements the calculation of MRP
MRP-Inputs (BOM)
BOM contains complete description, listing not only the
materials, parts, and components but also the sequence in
which the product is created;
BOM is often called product structure file or product tree
because it show how all the materials, parts, components, and
subassemblies are put together to form a product.
Fig 4-4 Typical Product Structure Diagram
MRP-Inputs (BOM)
Deferent representations of
BOM
Indented part list
Single level part list
Indented Part Lists Single-Level Part Lists
End Item
A(2)
C(1)
D(2)
B(1)
C(2)
E(3)
End Item
A(2), B(1)
A
C(1)
D(2)
B C(2)
E(3)
MRP-Inputs (BOM)
Low-level Coding
Although Cs appear at both Level 1 and 2, they should be
listed at the lowest level they appear, so that computation and
implementation may become easier. Why?
MRP-the Explosion Calculus
Example 7.1: the Harmon Music Company, model 85C trumpet:
Minimize the amount of money tired with inventory, and
production level should be set to match the predicted demand as
closed as possible.
Fig 7-4 Trumpet and Subassemblies
The bell assembly
Valve casting
Valve Valve slide
Fig 7-5 Product Structure Diagram for Harmon Trumpet
Indented BOM
Level 0 Level 1 Level 2
1 Trumpet
1 Bell assembly
1 Valve assembly
3 Slide assemblies
3 Valves
Right? Right?
If 100 trumpets are ordered, the numbers of parts needed are
Trumpet: 100-0=100;
Bell assembly: 100-5=95; Valve assembly: 100-10=90;
Slide assembly:100-5=95; Valves: 100-10=90
Level 0 Level 1 Level 2 On-hand inventory
1 Trumpet 0
1 Bell assembly 5
1 Valve assembly 10
3 Slide assemblies 5
3 Valves 10
Level 0 Level 1 Level 2 On-hand inventory
1 Trumpet 0
1 Bell assembly 5
1 Valve assembly 10
3 Slide assemblies 5
3 Valves 10
Trumpet
Bell
assembly
Valve
assembly
Slide
assemblies
Valves
Gross
Requirement
100 100 100 270 270
On-hand
inventory
0 5 10 5 10
Net
requirement
100 95 90 265 260
The lead time for producing a trumpet is 7 weeks: the company
should begin the production now on trumpet to be shipped in 7
weeks.
Only consider forecasts for demand that are at least 7 weeks into
the future. (if the current week is labeled as week 1, then requires
forecasts for the sales for week 8 and later)
Week 8 9 10 11 12 13 14 15 16 17
Demand 77 42 38 21 26 112 45 14 76 38
The predicated demand for week 8 through 17
Week 8 9 10 11 12 13 14 15 16 17
Scheduled
Receipts 12 0 6 9
Harmon periodically receive returned trumpet which are defective
for some reason or damaged in shipping. After necessary repair, they
are returned into pool of those ready for shipping. The receipts are
12, 6, 9 in weeks 8, 10 and 11 respectively.
These are scheduled receipts : not on hand yet, but will arrive and
could be used to fill the order in due periods.
The on-hand inventory at the end of week 7 is 23;
Net Demand = Demand SR Inventory of last period
The predicated demand for week 8 through 17
Week 7 8 9 10 11 12 13 14 15 16 17
Demand 77 42 38 21 26 112 45 14 76 38
SR 12 6 9
Initial
Inventory
23
Net
Demand
42 42 32 12 26 112 45 14 76 38
Translates MRP for the end product into a production schedule for
the components at the next level of the product structure (bell
assembly and valve casting assembly).
Determine the gross requirement with the BOM and the parts of
upper level;
Gross requirement-on hand inventory-scheduled receipts
=Net requirement;
Net requirement is shifted by LT=Time-phased requirement;
Time-phased requirement is treated by lot-sizing algorithm =
planned order release
MRP for the bell assembly (LT=2 weeks)
Batch method: LFL(lot-for-lot): Order as actually required.
Week 6 7 8 9 10 11 12 13 14 15 16 17
Gross Requirement
(GR)
42 42 32 12 26 112 45 14 76 38
Net Requirement
(NR)
42 42 32 12 26 112 45 14 76 38
Time-phased NR 42 42 32 12 26 112 45 14 76 38
Planned order
release (LFL)
42 42 32 12 26 112 45 14 76 38
Batch method: LFL(lot-for-lot): Order as actually required.
Week 4 5 6 7 8 9 10 11 12 13 14 15 16 17
Gross Requirement
(GR)
42 42 32 12 26 112 45 14 76 38
Net Requirement
(NR)
42 42 32 12 26 112 45 14 76 38
Time-phased NR 42 42 32 12 26 112 45 14 76 38
Planned order
release (LFL)
42 42 32 12 26 112 45 14 76 38
MRP for the valve casting assembly (LT=4 weeks)
MRP for the valves
On-hand inventory at the end of week 3=186;
Receive 96 at the end of week 4 from outside supplier;
LT=3 weeks
Week 2 3 4 5 6 7 8 9 10 11 12 13
Gross Requirement
(GR)
126 126 96 36 78 336 135 42 228 114
Scheduled receipts 96
On-hand inventory 186 60 30
Net Requirement
(NR)
0 0 66 36 78 336 135 42 228 114
Time-phased NR 66 36 78 336 135 42 228 114
Planned order
release (LFL)
66 36 78 336 135 42 228 114
3
The End!
Dr GENG Na
Management
Chapter 7
Production Planning
Contents
Introduction
Capacity Planning
Improvement in MRP
MRP- Alternative Lot-sizing Schemes
Lot for lot (LFL): the number of units scheduled for production
eachperiodisthesameasthenet requirement.
LFL isonlyfor convenienceandeaseof use, rather thanoptimal;
Theproblemof findingthebest or near optimal productionplanis
describedas
Having known the time-varying demand and costs of setup and
holding, what productionquantities will minimizethetotal holding
andsetupcost over planninghorizon?
Neither themethods of Chapter 4 nor thoseof Chapter 5 may be
appropriate.
EOQ Lot Sizing
Three parameters are required: the average demand , the holding
cost h, and setup cost K;
Consider the valve casting assembly: K=$132; the holding cost is
$0.6 per unit per week;
The planned order release resulting from a LFL policies requires
production in each week;
If the holding cost is charged against the inventory each week,
the total holding cost incurred from week 6 through 15 is 0;
Since there is one setup each week, the total setup cost incurred
over 10 weeks planning horizon is 10132=$1,320.
EOQ Lot Sizing (Cont.)
The cost can be reduced largely by producing larger amounts less
often;
As the first cut, EOQ formula is used to determine an
alternative production policy;
The total of the time-phased requirements over week 8 through
17 is 439, the average is 43.9/week ;
Using ==43.9, h=0.6, and K=132,
2 2 132 43.9
139.
0.6
K
Q
h

If we schedule the production in lot size 139, while ensuring that
the net requirements are satisfied, then resulting MRP is as follows
Cost of Using Sizing
During week 8-17, there are 4 setups, the total setup cost is $528;
The cumulative ending inventories are 97+55+=653, the total holding
cost is 6530.6=$391.80;
The total cost is 391.80+653=$919.80<&1,320 for LFL.
Ending Inv.=Beginning Inv.(Ending Inv. in the last period) +Planned deliveries -
Net requirement
Week 4 5 6 7 8 9 10 11 12 13 14 15 16 17
Net requirement (NR) 42 42 32 12 26 112 45 14 76 38
Planned deliveries 139 0 0 0 139 0 139 0 0 139
Planned order release (EOQ) 139 0 0 0 139 0 139 0 0 139
Ending inventory 97 55 23 11 124 12 106 92 16 117
The Silver-Meal Heuristic (Named for Harlan Meal and
Edward Meal)
A forward method that requires determining the average
cost per period as a function of the number of periods the
current order to span, and stopping the computation when
this function first increase.
Define C(T) as the average holding cost and setup cost
per periodif the current order spans the next T periods;
Let (r
1
, r
2
, , r
n
) be the requirements over the n-period horizon;
Consider period 1,
if we produce just enough in period 1 to meet the demand in
period 1, then only the order cost K is incurred. Hence,
C(1)=K/1;
If we produce just enough in period 1 to meet the demand in
both periods 1 and 2, then r
2
is held for one period. Hence,
C(2)=(K+hr
2
)/2;
Similarly, C(3)=(K+hr
2
+2hr
3
)/3;
In general, C(j)=((K+hr
2
+2hr
3
++(j-1) hr
j
)/j;
Once C(j)>C(j-1), stop and set y
1
=r
1
+r
2
++r
j-1
and start the
process at period j.
Hint: Streamline the computation by
C(j+1)=(j/(j+1))(C(j)+hr
j+1
)
Example 7.2
r=(18, 30, 42, 5, 20); h=2; K=80
C(1)=K/1=80;
C(2)= (K+hr
2
)/2
=(80+(2)(30))/2=70;
C(3)=(K+hr
2
+2hr
3
)/3=(80+(2)(30)+(2)(2)(42))/3=102.67;
70<80
102.67>70
Stop at period 2, y
1
=r
1
+r
2
=18+30=48
Starting in period 1
C(1)=K/1=80;
C(2)= (K+hr
4
)/2
=(80+(2)(5))/2=45;
C(3)=(K+hr
4
+2hr
5
)/3=(80+(2)(5)+(2)(2)(20))/3=56.67;
45<80
56.67>45
Stop at period 4, y
3
=r
3
+r
4
=42+5=47
Starting in period 5 y
5
=r
5
=20 y=(48, 0, 47, 0, 20)
Example 7.3
r=(10,40,30); h=1; K=50
y=(50,0,30) from Silver-Meal
y=(10,70,0)
optimal
Notes: Silver-Meal heuristic does not always result in optimal
solution.
Least Unit Cost (LUC)
LUC heuristics is similar to Silver-Meal Heuristic method
except that instead of dividing the cost over j periods by the
periods, j, but by the total number of units demanded from 1
through j period, r
1
+r
2
++r
j
;
C(1)=K/r
1
;
C(2)=(K+hr
2
)/(r
1
+r
2
);
.
.
C(j)= (K+hr
2
+2hr
3
++(j-1)hr
j
)/(r
1
+r
2
++r
j
)
The computation is stopped when C(j)>C(j-1), and the
productionlevel isset equal tor
1
+r
2
++r
j-1
. Theprocessisthen
repeated, startingat periodj andcontinuinguntil theendof the
planninghorizonisreached.
Example 7.4
r=(18,30,42,5,20); h=2; K=80
C(1)=K/r
1
=80/18=4.44;
C(2)=(K+hr
2
)/(r
1
+r
2
) =(80+(2)(30))/(18+30)=2.92;
C(3)=(K+hr
2
+2hr
3
)/(r
1
+r
2
+r
3
)=(80+(2)(30)+(2)(2)(42))/ (18+30+42) =3.42;
2.92<4.44
3.42>2.92
Stop at period 2: y
1
=r
1
+r
2
=18+30=48
Starting in period1
C(1)=K/r
3
=80/42=1.90;
1.92>1.90
Stop at period 3: y
3
=r
3
=42
Starting in period3
C(2)=(K+hr
4
)/(r
3
+r
4
) =(80+(2)(5))/(42+5)=1.92;
Example 7.4
r=(18, 30, 42, 5, 20 ); h=2; K=80
C(1)=K/r
4
=80/5=16;
C(2)=(K+hr
5
)/(r
4
+r
5
) =(80+(2)(20))/(5+20)=4.8;
4.8<16
y
4
=r
4
+r
5
=5+20=25
Starting in period4
y=(48, 0, 42, 25, 0)
Silver-Meal (48,0,47,0,20) $310
LUC (48,0,42,25,0) $340
optimal
Part Period Balancing
More popular in practice;
Set the order horizon equal to the number of periods that mostly
matches the total holding cost with the setup cost over that period.
The exact matching is rare in an integer number.
Example 7.5 r=(18,30,42,5,20); h=2; K=80
Order
Horizon
Total holding
cost
Starting in period1
1 0
2 60
3 228
y
1
=r
1
+r
2
=18+30=48
228 exceeds the setup cost of 80.
80 is closer to 60 than to 228
Order
Horizon
Total holding
cost
1 0
2 10
3 90(5*h+20*2*h)
y
3
=r
3
+r
4
+r
5
=42+5+20=67
Silver-Meal (48,0,47,0,20) $310
LUC (48,0,42,25,0) $340
PPB (48,0,67,0,0) $310
90 is close to 80 than is 10
Example 7.5 r=(18,30,42,5,20); h=2; K=80
MRP-Lot Sizing with Capacity Constraints
Given:
Requirements: ( r
1
, r
2
, , r
n
);
Capacity: (c
1
, c
2
, , c
n
);
The objective is to find optimal production quantities (y
1
, y
2
, , y
n
)
subject to capacity constraints : y
i
c
i
, i=1~n.
r=(52, 87, 23, 56), c=(60, 60, 60, 60), feasible or not
Feasibility condition
1 1
1~
t t
i i
i i
c r for t n

If the above feasibility cannot be satisfied, no solution is available.
Even feasibility is OK, however, requirements in some period
may exceeds corresponding capacities.
Lot-shifting technology for obtaining initial feasible solution,
such that r
i
c
i
, i=1~n.
Method:
Back-shift demand from periods in which demand exceeds
capacity to earlier periods in which there is free capacity; if
there are more than 1 earlier free capability, priority is given to
the closet earlier free capacityWhy?)
Repeat the process for the period in which demand exceeds
capacity until r
i
c
i
, i=1~n, that is, feasible for lot-for-lot.
1 2 3 4 5 6 7 8 9
r 100 79 230 105 3 10 99 126 40
c 120 200 200 400 300 50 120 50 30
200
30
109 50
21
120
40
50
15
18
30
10
28
y 100 109 200 105 28 50 120 50 30
Excess capacity 20 91 0 295 272 0 0 0 0
242
58
192
108
142
158
Extra holding cost=(2)(30)(4)=$240<450
Extra holding cost=(2)(50)(3)=$300< 450
Extra holding cost=(2)(158)
=$316<450
300
0
137
263
Extra holding cost=(2)(50)(1)=$100< 450
Extra holding cost=(2)(120)(2)=$480>450
y= (100, 109, 200, 263, 0, 0, 120, 0, 0 )
r= (100, 79, 230, 105, 3, 10, 99, 126, 40)
Ending inv.=(0, 30, 0, 158, 155, 145, 166, 40, 0)
The total cost=5( 450)+2(694)=2250+1388=3638.
The End!
Dr GEAN Na
Management
Chapter 7
Production Planning
Contents
Introduction
Capacity Planning
Improvement in MRP
Optimal Lot Sizing for Time-Varying Demand
Optimal in this context means the policy that minimizes the total
holding and setup cost over the planning horizon.
The optimal policies can be determined by casting the problem as
a shortest-path problem.
Assume that
Forecasted demands over the next n periods are known and
given by the vector r=(r
1
, r
2
, , r
n
)
Costs are charged against holding at $h per unit per period
and $K per setup. We will assume that the holding cost is
charged against ending inventory each period.
Example: The forecast demand for an electronic assembly
produced at Hi-Tech, a local semiconductor fabrication shop, over
the next four weeks is 52, 87, 23, 56. There is only one setup each
week for production of these assemblies, and there is no back-
ordering of excess demand. Assume that the shop has the capacity
to produce any number of the assemblies in a week.
Let y
1
,,y
4
be the order (or Production) quantity in each of
the four weeks.
52 y
1
52+87+23+56=218
139- y
1
y
2
218- y
1
if y
1
139
0y
2
218- y
1
if y
1
>139
10200
pairs
10200
pairs
Search all the feasible policies is unreasonable.
The Wagner-Whitin algorithm observation:
An optimal policy has the property that each value of y is
exactly the sum of a set of future demands.
y
1
=r
1
, or y
1
=r
1
+r
2
, , or y
1
=r
1
+r
2
++r
n
y
2
=0, y
2
=r
2
, or y
2
=r
2
+r
3
, , or y
2
=r
2
+r
3
++r
n
.
y
n
=0, y
n
=r
n
.
The number of exact requirements policies is much smaller
than the total number of feasible policies.
Example (continued): We continue with the four-period scheduling
problem.
y
1
=52, or y
1
=139, y
1
=162, or y
1
=218
y
2
=0, y
2
=87, or y
2
=110, or y
2
=166
y
3
=0, y
3
=23, or y
3
=78
y
4
=0, or y
4
=56
It is easy to see that every exact requirements policy is completely
determined by specifying in what periods ordering should take
place.
problem.
Define
This problem can be regarded as a one-way network with the
number of nodes equal to exactly one more than the number of
periods.
Every path through the network corresponds to a specific exact
requirements policy.
1 if the production takes place in period
0 else
j
j
i

=

11 22
33 44
55
problem.
The next is to assign a value to each arc in the network.
The value or length of the arc (i, j), called c
ij
, is defined as
the setup and holding cost of ordering in period i to meet the
requirements through j-1.
Finally, dynamic programming is used to determine the minimum-
cost production schedule, or shortest path through the network.
problem.
Path enumeration is used to solve the Hi-Tech problem. Recall that
r=(52, 87, 23, 56). In addition, h=$1, and K=$75.
c
12
=75 (setup cost only)
c
13
=75+87=167
c
14
=75+87+23*2=208
c
15
=75+87+23*2+56*3=376
c
23
=75
c
24
=75+23=98
c
25
=75+23+56*2=210
c
34
=75
c
35
=75+56=131
c
45
=75
1-2-3-4-5: 757575753001-2-4-5: 75+98+75=248
i\j 1 2 3 4 5
1 75 162 208 376
2 75 98 210
3 75 131
4 75
problem.
Path enumeration is used to solve the Hi-Tech problem. Recall that
r=(52, 87, 23, 56). In addition, h=$1, and K=$75.
The optimal path is 1-2-4-5 at cost 248.
The optimal ordering is
y
1
=52, y
2
=110, y
3
=0, y
4
=56
i\j 1 2 3 4 5
1 75 162 208 376
2 75 98 210
3 75 131
4 75
Path Cost($)
1-2-3-4-5 300
1-2-4-5 248
1-2-5 285
1-2-3-5 281
1-3-4-5 312
1-3-5 293
1-4-5 283
1-5 376
problem.
Dynamic programming is a recursive solution technique that can
significantly reduce the number of computations required, although
it too can be quite cumbersome.
Dynamic programming is based on the principle of optimality.
If a problem consists of exactly n stages and there are r<n stages
remaining, the optimal policy for the remaining stages is
independent of policy adopted in the previous stages.
problem.
Define f
k
as the minimum cost starting at node k, assuming that an
order is placed in period k.
The principle of optimality for this problem results in the
following system of equations:
The initial condition is f
n+1
=0
( )
min for 1, 2,...,
k kj j
j k
f c f k n
>
= + =
problem.
Solution:
f
5
=0
( )
4 4
4
min =75 at =5 (the only possible value of )
j j
j
f c f j j
>
= +
i\j 1 2 3 4 5
1 75 162 208 376
2 75 98 210
3 75 131
4 75
( )
34 4
3 3
3
35 5
75 75 150
min =min min min =131 at 5
131 0 131
j j
j
c f
f c f j
c f
>
+ +

= + = = =
` ` `
+ +
) ) )
( )
23 3
2 2 24 4
2
25 5
75 131 206
min =min min 98 75 min 173 =173 at 4
210 0 210
j j
j
c f
f c f c f j
c f
>
+ +

= + + = + = =
` ` `

+ +
) ) )
problem.
Solution:
5
4
3
2
0
75 at =5
131 at =5
173 at =4
f
f j
f j
f j
=
=
=
=
i\j 1 2 3 4 5
1 75 162 208 376
2 75 98 210
3 75 131
4 75
( )
12 2
13 3
1 1
1
14 4
15 5
75 173 248
162 131 293
min =min min min =248 at 2
208 75 283
376 0 376
j j
j
c f
c f
f c f j
c f
c f
>
+
+

+ +

= + = = =
` ` `
+ +

+ +
) )
)
Example (continued): We continue with the four-
period scheduling problem. r=(52, 87, 23, 56).
Solution:
To determine the optimal order policy, we trace the
solution back from the beginning.
In period 1, the optimal value of j is j=2. y
1
=r
1
= 52
2
=r
2
+r
3
=
110
4
=r
4
= 56
y=(52,110,0,56)
5
4
3
2
1
0
75 at =5
131 at =5
173 at =4
248 at 2
f
f j
f j
f j
f j
=
=
=
=
= =
Contents
Introduction
Capacity Planning
Improvement in MRP
Capacity Planning -Framework
Long range:
Strategic capacity
planning;
Medium range:
Rough-cut capacity
planning and
capacity
requirement
planning;
Short-range:
Finite loading and
Input/output
analysis
Process Planning
Aggregate Planning
(MPS)
(MRP)
Demand Forecast
Rough-cut
capacity Planning
Long Range
Medium
Range
Short range
Capacity Req.
Planning
Order
Management
Finite
loading
Input/outpu
t analysis
Capacity Planning Framework
Strategic capacity planning or resources planning
Directly linked to aggregate production planning;
The most highly aggregated and longest range capacity
planning decision;
Typically revolves converting monthly, quarterly, or even
annual data from aggregate production planning into
aggregate resources such as gross labor-hours, floor space,
and machine-hours;
Involves new capital expansion, bricks and mortar, machine
tools, warehouse space, and so on, which requires a time
horizon of months or years.
Capacity Planning - Framework
Rough-cut capacity planning
MPS is the primary information sources;
Several techniques: Capacity planning using overall planning
factors (CPOF); Capacity bill; or Resource profiles;
These techniques provide information for modifying the
resource levels or material plan to ensure execution of MPS;
Capacity requirement planning (CRP)
Time-phased material plan supplied by MRP is the basis for
calculating time-phased capacity requirements;
Data used by CRP techniques include WIP, routing,
scheduled receipts, and planned order;
Information provided by CRP can be used to determine
capacity requirement for all the work centers and labor skills.
Finite loading technique
Also relates to a firms that use time-phased detailed material
plan;
Can be better viewed as a shop floor scheduling technique;
Clarifies the relationship between scheduling and availability;
Is a method for scheduling work center or resource group.
Input/output analysis
Provides a method of monitoring the actual consumption of
capacity during execution of detailed material plans obtained
by MRP;
Can indicate the need to update capacity plans as actual shop
performance deviates from the current plan, and the need to
modify planning factors used in capacity planning techniques.
The End!
Dr.GENG Na
Management
Chapter 7
Production Planning
Contents
Introduction
Capacity Planning
Improvement in MRP
Capacity Planning Techniques (CPOF)
Standard or
historical data of
end product
CPOF(Capacity planning using overall planning factors) is
relatively simple approach to rough-cut capacity planning;
CPOF plans are usually stated in terms of weekly or monthly
time periods and accordingly are revised as the firm changes the
MPS.
Planning factor (e.g.
direct labor time/
end product unit)
MPS
Cap. req. on
overall labor or
machine-hr
Historical data
on shop load
Cap. Req. on
individual work
center
The procedure for COPF
Example Problem data
Estimated CPOF in standard direct labors
Total Cap. Req. in P1: (0.95*33)+(1.85*17) = 62.8
End product 1 2 3 4 5 6 7 8 9 10 11 12 13 Total
A 33 33 33 40 40 40 30 30 30 37 37 37 37 457
B 17 17 17 13 13 13 25 25 25 27 27 27 27 273
End product
A 0.95 hours
B 1.85 hours
Master production schedule (in units)
Period
Direct labor time per end product unit:
Total direct labor in standard hours/unit
1 2 3 4 5 6 7 8 9 10 11 12 13 Total
62.80 62.80 62.80 62.05 62.05 62.05 74.75 74.75 74.75 85.10 85.10 85.10 85.10 939.20
Period
Total required
Work center
Example Problem data
Estimated CPOF in standard direct labors
Cap Req. for WC 100 in P1: 62.8*60.3%=37.87
A 33 33 33 40 40 40 30 30 30 37 37 37 37 457
B 17 17 17 13 13 13 25 25 25 27 27 27 27 273
End product
A 0.95 hours
B 1.85 hours
Period
Direct labor time per end product unit:
Total direct labor in standard hours/unit
Work
center
Historical
percentage
1 2 3 4 5 6 7 8 9 10 11 12 13 Total
100 60.3 37.87 37.87 37.87 37.42 37.42 37.42 45.07 45.07 45.07 51.32 51.32 51.32 51.32 566.34
200 30.4 19.09 19.09 19.09 18.86 18.86 18.86 22.72 22.72 22.72 25.87 25.87 25.87 25.87 285.52
300 9.3 5.84 5.84 5.84 5.77 5.77 5.77 6.95 6.95 6.95 7.91 7.91 7.91 7.91 87.35
62.80 62.80 62.80 62.05 62.05 62.05 74.75 74.75 74.75 85.10 85.10 85.10 85.10 939.20
Period
Total required capacity
Capacity Planning Techniques (Capacity Bills)
Capacity bill is a rough-cut method, providing more direct
linkagebetweenindividual endproduct inMPS andthecapacity
requiredfor eachworkcenters;
considersanyshiftsinproduct mix;
requires moredatathan CPOF: not only BOM and routing
dateare, but alsodirect labor-hour or machine-hour data.
Calculationprocedures:
First calculate bill of capacity of end product, that is total
time/unit for each end product on each resource, based on
BOM, routingdata, direct labor-hour or machine-hour;
Then, apply bill of capacity to MPS to obtain capacity
requirementsof eachresourcesin eachperiod;
Routing and standard time data of an example
0.05=standard setup time per unit+ standard run time per unit=0.025+0.025
End products Lot sizes Operation
Work
center
Standard
setup hours
Standard setup
hours per unit
Standard run time
hours per unit
Total hours
per unit
A 40 1 of 1 100 1.00 0.025 0.025 0.050
B 20 1 of 1 100 1.00 0.050 1.250 1.300
Components
C 40 1 of 2 200 1.00 0.025 0.575 0.600
40 1 of 2 300 1.00 0.025 0.175 0.200
D 60 1 of 1 200 2.00 0.033 0.067 0.100
E 100 1 of 1 200 2.00 0.020 0.080 0.100
F 100 1 of 1 200 2.00 0.020 0.043 0.063
0.025=setup time/lot size=1/40
Routing and standard time data of an example
End products Lot sizes Operation
Work
center
Standard
setup hours
Standard setup
hours per unit
Standard run time
hours per unit
Total hours
per unit
A 40 1 of 1 100 1.00 0.025 0.025 0.050
B 20 1 of 1 100 1.00 0.050 1.250 1.300
Components
C 40 1 of 2 200 1.00 0.025 0.575 0.600
40 1 of 2 300 1.00 0.025 0.175 0.200
D 60 1 of 1 200 2.00 0.033 0.067 0.100
E 100 1 of 1 200 2.00 0.020 0.080 0.100
F 100 1 of 1 200 2.00 0.020 0.043 0.063
A B
Work center Total time/unit Total time/unit
100 0.05 1.3
200 0.7 0.55
300 0.2 0
Total time/unit 0.95 1.85
Bill of capacity
End product
(0.1 * 1D + 0. 1 * 2E + 0.063*4F)
on WC 200=0.55
A 33 33 33 40 40 40 30 30 30 37 37 37 37 457
B 17 17 17 13 13 13 25 25 25 27 27 27 27 273
Period
Work
center
1 2 3 4 5 6 7 8 9 10 11 12 13 Total
100 23.75 23.75 23.75 18.9 18.9 18.9 34 34 34 36.95 36.95 36.95 36.95 377.75
200 32.45 32.45 32.45 35.15 35.15 35.15 34.75 34.75 34.75 40.75 40.75 40.75 40.75 470.05
300 6.6 6.6 6.6 8 8 8 6 6 6 7.4 7.4 7.4 7.4 91.4
Capacity requirement
Period
A B
Work center Total time/unit Total time/unit
100 0.05 1.3
200 0.7 0.55
300 0.2 0
Total time/unit 0.95 1.85
Bill of capacity
End product
0.05*33+1.3*17=23.75
Capacity Planning Techniques (CRP)
CRP is to calculate capacity requirements placed on a work
center or resource group by using the output of MRP, that is, the
time-phased material plan information generated by MRP.
The information include actual lot sizes and lead time for both
open shop orders (scheduled receipts) and orders planned for
future release (planned orders).
By calculating capacity requirements both for open shop orders
(scheduled receipts) and orders planned for future release
(planned orders) in the MRP data base, CRP accounts for the
capacity already stored in the form of finished and WIP
inventories.
1 2 3 4 5 6 7 8 9 10 11 12 13
33 33 33 40 40 40 30 30 30 37 37 37 37
Product A MPS
1 2 3 4 5 6 7 8 9 10 11 12 13
33 33 33 40 40 40 30 30 30 37 37 37 37
40
4 11 18 18 18 18 28 38 8 11 14 17 20
0 0 40 40 40 40 40 40 40 40 40 40
40 40 40 40 40 40 40 40 40 40
Component C
Lot size=40; Lead time=2;
On hand inv.=37
Gross Req.
Scheduled Receipts
Inventory balance
Planned deliveries
Planned order release
The time to fabricate a component C in machine center 300
is 0.2 hrs;
1 2 3 4 5 6 7 8 9 10 11 12 13
40 40 40 40 40 40 40 40 40 40
8 8 8 8 8 8 0 8 8 8 8 0 0
Worker center 300 capacity requirements using CRP
Planned order release
The 8 hrsof capacity is derived from the scheduled receipt
and planned order quantities of 40 units multiplied by 0.2;
Periods
Hrsof capacity
Contents
Introduction
Aggregate Planning
Material Requirement Planning (MPS)
Master Production Planning (MRP)
Capacity Planning
Improvement in MRP
Improvement for MRPClosed-Loop MRP
When MRP system has
information feedback from
its module output, this is
termed closed loop MRP;
The closed-loop is realized
by checking whether
sufficient capacity is
available. If not, the
schedule should be
modified.
The modification may be
made either by changing
MPS or by regulating the
planned order release.
Aggregate
product plan
MPS
MRP
Explosion
Calculus
Bill of
Mterials
(BOM)
Inventory
record files
Production
order
Purchase
order
Firm orders
from known
customer
Forecast of
demand from
unknown
customer
OK
No
Yes
CRP
Improvement for MRPMRP-II
Manufacturing Resource Planning (MRP-II) was first proposed
by Ollie Wright 1977;
As a complex planning computer system, MRP-II was to plan
and monitor all the resources of a manufacturing firm through a
closed-loop system.
The manufacturing resource includes those for manufacturing,
marketing, finance, and engineering.
MRP-II helps a firm cope with complex balancing problem
among production, purchasing, and sales to shorten lead time for
manufacturing and purchasing, reduce production cost, enhance
adaptability, and increase economic efficiency.
Improvement for MRPMRP-II
Aggregate
product plan
MPS
MRP
Explosion
Calculus
Inventory
Management
Bill of
Materials
(BOM)
Production
order
Purchase
order
Firm orders
from known
customer
Forecast of
demand from
unknown
customer
OK
No
Yes
CRP
Purchase
Management
Shop Floor
Management
Account and
Financial
Management
Rough-cut
Capacity
Planning
Additional common functions
for MRP-II
Purchasing management;
Production costing ;
Shop floor management;
Inventory management;
Accounting;
Improvement for MRP-II: ERP
1960s - Systems J ust for Inventory Control
1970s - MRP Material Requirement Planning
(Inventory with material planning & procurement)
1980s - MRP II Manufacturing Resources
Planning (Extended MRP to shop floor &
distribution Management)
Mid 1990s - ERP Enterprise Resource Planning
(Covering all the activities of an Enterprise)
2000 onwards ERP II Collaborative Commerce
(Extending ERP to external business entities)
Improvement for MRP-II: ERP
Definition: Software solution that addresses the
enterprise needs, taking a process view of the overall
organization to meet the goals, by tightly integrating
all functions and under a common software
platform
ERP is a solution, which facilitates company-wide
integrated information systems, covering all
functional areas;
ERP performs core corporate activities and increases
customer service augmenting corporate image.
Homework
P346, Q6
P357, Q17
P358 Q28
The End!
Production Planning & Control
Dr. Na GENG
Prof. Zhibin JIANG
Contents
Introduction
Job Shop Scheduling Terminology
Sequencing Rules
Sequencing Theory for a Single Machine
Sequencing Theory for Multiple Machines
Assembly Line Balancing
Advanced Topics for Operations scheduling
Introduction-What is Operations Scheduling ?
Implement the production orders
generated in MRP under given
objectives ;
Allocate production resources
(machine, workers et al.) to
production orders (jobs or tasks
and their due dates) in an
optimized manners;
The results are time allocations of
production resources to different
jobs (job sequences on each
production resources);
All the orders can be completed
while all production resources are
utilized with their loads being
balanced.
Aggregate plan
purchase order
Introduction- Optimal schedule rules
Significant issues for determining optimal or approximately
optimal scheduling rules are the following.
1) The job arrival pattern
2) Number and variety of machine in the shop.
3) Number of workers in the shop.
4) Particular flow patterns.
5) Evaluation of alternative rules.
Introduction-Objectives of Job Shop Scheduling
Objectives of operations scheduling
1) Meet due date;
2) Minimize WIP inventory;
3) Minimize the average flow time through the systems;
4) Provide for high machine/worker (time) utilization
(minimize idle time);
5) Provide for accurate job status information;
6) Reduce setup cost;
7) Minimize production and worker costs
Proper trade off between cost and quality is one of the most
challenging strategic issues facing a firm today.
Introduction-Objectives of operations Scheduling
Some of these objectives conflicts,
Reduce WIP inventory Worker idle time may increase or
machine utilization may decrease;
Reasons: differences in the throughput rate from one part of
the system to another may force the faster operations to wait.
Fig 8-3 A Process Composed of Two Operations in Series
As an example, if there is no buffer for WIP between 1 and 2,
what happens?
Finding the proper mix between WIP inventory and worker
idle time is equivalent to choosing a point on the trade-off
curve of these conflicting objectives.
A true improvement in the overall system would mean that
the entire trade-off curve undergoes a downward shift.
Introduction-Functions of Scheduling and
Control
The following functions must be performed in scheduling and
controlling a shop floor:
Allocating orders, equipment, and personnel to work centers
or other specified location-Short term capacity planning;
Determining the sequence of orders (i. e. job priorities);
Initializing performance of the scheduled work, commonly
termed the dispatching of jobs;
Shop-floor control, involving
Reviewing the status and controlling the progress of
orders as they are being worked on;
Expediting the late and critical orders;
Revising the schedules in light of changes in order status.
Introduction-Elements of the Shop Floor
Scheduling Problems
The classic approaches to shop floor scheduling focuses on the
following six elements:
J ob arrival patterns: static or dynamic
Static: jobs arrive in single or batch;
Dynamic: jobs arrive over time interval according to some
statistical distribution.
Numbers and variety of machines in the shop floor
If there is only one machine or if a group of machines can
be treated as one machine, the scheduling problem is much
more simplified;
As number of variety of machines increase, the more
complex the scheduling problems is likely to become.
Scheduling Problems
(Continued)
Ratio of workers to machines
Machine limited system: more workers than machine
or equal number workers and machines;
Labor-limited system: more machines than worker.
Flow pattern of jobs: flow shop or job shop
Flow shop: all jobs follow the same paths from one
machine to the next;
J ob shop: no similar pattern of movement of jobs from
one machine to the next.
Scheduling Problems
(Continued)
J ob sequencing
Sequencing or priority sequencing: the process of
determining which job is started first on some machines or
work center by priority rule;
Priority rule: the rule used for obtaining a job sequencing;
Priority rule evaluation criteria
To meet corresponding objectives of scheduling;
Common standard measures:
Meeting due date of customers or downstream operations;
Minimizing flow time (the time a job spends in the shop flow);
Minimizing WIP;
Minimizing idle time of machines and workers (Maximizing
utilization).
Introduction-Elements of the Job Shop
Scheduling Problems
An assembly line is a classic example of flow shop
Every cars go through all the stations one by one in the same sequences;
Same tasks are performed on each car in each station;
Its operations scheduling is simplified as assembly line balancing;
An assembly balancing problem is to determine the number of stations and to
allocate tasks to each station.
Flow shop:
Each of then jobs must
be processed through the
m machines in the same
order.
Each job is processed
exactly once on each
machine.
Introduction-Job Shop
A job shop is organized by machines which are grouped
according to their functions.
Introduction-Job Shop
Not all jobs are assumed to require exactly the same number of operations, and
some jobs may require multiple operations on a single machine (Reentrant
system, J ob B twice in work center 3 ).
Each job may have a different required sequencing of operations.
No all-purpose solution algorithms for solving general job shop problems ;
Operations scheduling of shop floor usually means job shop scheduling;
Job A
Job B
Contents
Introduction
Sequencing Rules
1. Parallel processing versus sequential processing
Sequencing Processing: the m machines are distinguishable, and different
operations are performed by different machines.
Parallel processing: The machines are identical, and any job can be
processed on any machine.
M
1
M
2
M
3
M
4
Job
A
Job
B
M
1
, M
2
, M
3
, and M
4
are different;
J ob A has 2 operations which should be
processed on different Machines: M
1
and
M
2
;
J ob B has 3 operations which should be
processed on different Machines: M
3
, M
2
and M
4
;
M
1
M
2
M
3
M
4
Job
A
Job
B
M
1
, M
2
, M
3
, and M
4
are identical;
J obs A and B can be processed on
any one of the 4 machines
2 Flow time
The flow time of job i is the time that elapses from the initiation of that
job on the first machine to the completion of job i.
The mean flow time, which is a common measure of system
performance, is the arithmetic average of the flow times for all n jobs
Job 1 Job 2
Job 1
Job 3
Job 2 Job 3
Machine
s
M
1
M
2
Time
F
1
: FT of Job 1
F
2
: FT of Job 2
F
3
: FT of Job 3
Mean Flow Time=(F
1
+F
2
+F
3
)/3
3. Make-span
The make-span is the time required to complete a group of jobs (all n jobs).
Minimizing the make-span is a common objective in multiple-machine
sequencing problems.
Job 1 Job 2
Job 1
Job 3
Job 2 Job 3
Machines
M
1
M
2
Time
F
1
: FT of Job
1
F
2
: FT of Job
2
F
3
: FT of Job
3
Make-span of the 3
jobs
4. Tardiness and lateness
Tardiness is the positive difference between the completion time and the
due date of a job.
Lateness refers to the difference between the job completion time and its
due date and differs from tardiness in that lateness can be either positive or
negative.
If lateness is positive, it is tardiness; when it is negative, it is earliness
Due date
of Job i
Completion
time of Job i
Tardines
s
of Job i
Due date
of Job i
Completion
time of Job i
When the completion of J ob is earlier than due date, the tardiness is 0
Lateness>0---
Tardiness
Lateness<0---
Earliness
In-class discussion
Discuss each of the following objectives listed and the relation
each has with job shop performance:
1. Reduce WIP inventory
2. Provide a high level of customer service
3. Reduce worker idle time
4. Improve factory efficiency
Why are 1 and 3 conflicting? Why are 2 and 4 conflicting?
The End!
Dr. Na GENG
Prof. Zhibin JIANG
Contents
Introduction
Sequencing Rules
Sequencing Rules
FCFS (first come-first served)
J obs are processed in the sequence in which they entered the shop;
The simplest and nature way of sequencing as in queuing of a bank
SPT (shortest processingtime)
J obsaresequencedinincreasingorder of their processingtime;
Thejobwithshortest processingtimeisfirst, theonewiththenext
shortest processingtimeissecond, andsoon;
Sequencing Rules
EDD (earliest due date)
J obs are sequenced in increasing order of their due dates;
The job with earliest due date is first, the one with the next
earliest due date is second, and so on;
CR (Critical ratio)
Critical ratio is the remaining time until due date divided by
processing time;
Scheduling the job with the smallest CR next;
Sequencing Rules
Processing time of J ob i
Due date of Job i
Current time
Remaining time of Job i
CR
i
=Remaining time of J ob i/Processing time of J obi
=(Due date of J ob i-current time)/Processing time of J ob i
Sequencing Rules
CR provides the balance between SPT and EDD, such that the
task with shorter remaining time and longer processing time takes
higher priority;
CR will become smaller as the current time approaches due date,
and more priority will given to one with longer processing time;
For a job, if the numerator of its CR is negative ( the job has
been already later), it is naturally scheduled next;
If more than one jobs are later, higher priority is given to one
that has shorter processing time (SPT).
Sequencing Rules
Example8.1
A machinecenter in ajob shop for alocal fabrication company has
fiveunprocessedjobs remainingat aparticular point intime. Thejobs
arelabeled1, 2, 3, 4, and5intheorder that theyenteredtheshop. The
respectiveprocessingtimesandduedatesaregiveninthetablebelow.
Sequencethe5jobs by above4rules andcompareresults basedon
meanflowtime, averagetardiness, andnumber of tardyjobs
Job number Processing Time Due Date
1
2
3
4
5
11
29
31
1
2
61
45
31
33
32
Sequencing RulesFCFS
1
2
3
4
5
11
29
31
1
2
61
45
31
33
32
Job Completion Time Due Date Tardiness
1 11 61 0
2 40 45 0
3 71 31 40
4 72 33 39
5 74 32 42
Totals 268 121
Mean Flow time=268/5=53.6
Average tardiness=121/5=24.2
No. of tardy jobs=3.
Sequencing RulesSPT
1
2
3
4
5
11
29
31
1
2
61
45
31
33
32
Job Processing Time Completion Time Due Date Tardiness
4 1 1 33 0
5 2 3 32 0
1 11 14 61 0
2 29 43 45 0
3 31 74 31 43
Totals 135 43
Sequencing RulesEDD
1
2
3
4
5
11
29
31
1
2
61
45
31
33
32
Job Processing Time Completion Time Due Date Tardiness
3 31 31 31 0
5 2 33 32 1
4 1 34 33 1
2 29 63 45 18
1 11 74 61 13
Totals 235 33
Sequencing RulesCR
Current time: t=0
Job number Processing Time Due Date Critical Ratio
1
2
3
4
5
11
29
31
1
2
61
45
31
33
32
61/11(5.545)
45/29(1.552)
31/31(1.000)
33/1 (33.00)
32/2 (16.00)
t=31 after processing the job 3
Sequencing RulesCR
Current time: t=31
Job number Processing Time Due Date-Current Time Critical Ratio
1
2
4
5
11
29
1
2
30
14
2
1
30/11(2.727)
14/29(0.483)
2/1 (2.000)
1/2 (0.500)
Current time should be reset after scheduling one job
Sequencing RulesCR
Current time=60
Job number Processing Time Due Date-
Current Time
Critical Ratio
1
4
5
11
1
2
1
-27
-28
1/11(0.0909)
-27/1<0
-28/2<0
Both J obs 4 and 5 are later, however J ob 4 has shorter
processing time and thus is scheduled first; Finally, job 1 is
scheduledlast.
Sequencing RulesCR
Job number Processing Time Completion Time Tardiness
3
2
4
5
1
31
29
1
2
11
31
60
61
63
74
0
15
28
31
13
Totals 289 87
Sequencing RulesSummary
Rule Mean Flow
Time
Average
Tardiness
Number of
Tardy Jobs
Maximum
Tardiness
FCFS
SPT
EDD
CR
53.6
27.0
47.0
57.8
24.2
8.6
6.6
17.4
3
1
4
4
42
43
18
31
Discussions
SPT results in smallest mean flow time;
EDD yields the minimum maximum tardiness;
Always true? Yes!
Objectives in J ob Shop Management: An Example
Example8.2
An air traffic controller is facedwith theproblemof schedulingthe
landing of five aircraft. Based on the position and runway
requirementsof eachplane, sheestimatesthefollowinglandingtimes:
Only one plane may land at a time. The problemis essentially the
same as that of scheduling five jobs on a single machine, with the
planes corresponding to the jobs, the landing times to the processing
times, andtherunwaytothemachine.
Discussthepossibleschedulingobjectivesfor thisproblem.
Plane: 1 2 3 4 5
Time (in minutes) 26 11 19 16 23
Example8.2 Solutions:
1. Withthegiveninformation, tworeasonableobjectiveswouldbeto
minimizethetotal timerequiredto landall planes(i.e., themakespan)
or theaveragetimerequired to land theplanes (themean flowtime).
The makespan is fixed, whereas the mean flow time is sequence
dependent andtheSPT minimizesthemeanflowtime.
Plane: 1 2 3 4 5
Sequence of Plane 2 4 3 5 1
Flow time 11 27 46 69 95
Average Flow time 49.6
Example8.2
2. An alternative objective might be to land as many people as
quickly as possible. Theappropriateobjectiveinthis casemight beto
minimizetheweightedmakespanor theweight sumof thecompletion
times, where the weights would correspond to the number of
passengersineachplane.
A possiblesolution:
Plane: 1 2 3 4 5
Number of passengers 180 12 45 75 252
Plane: 1 2 3 4 5
Number of passengers 180 12 45 75 252
Weighted time 6.92 1.09 2.37 4.69 10.96
Flow time 23 49 65 84 95
Average flow time 63.2
Example8.2
3. Anissuethat wehavenot yet addressedisthetimethat eachplane
isscheduledtoarrive. Assumethefollowingdata:
Somepossibleobjectives related to duedates includeminimizing the
averagetardinessandminimizingthemaximumtardiness.
Plane: 1 2 3 4 5
Scheduled arrival time 5:30 5:45 5:15 6:00 5:40
Scheduled arrival time 5:15 5:30 5:40 5:45 6:00
Landing time 5:15 5:34 6:00 6:23 6:34
Delay 0 4 20 38 34
Average delay 19.2
Example8.2
4. Thus far we have ignored special conditions that favor some
planesover others.
Someplanehasacriticallylowfuel level.
Planesarescheduledfor continuingflights.
Planescarrypreciousor perishablecargo.
It is difficult to choose the objective function for job sequencing
problems.
The optimal sequence is highly sensitive to the choice of the
objective, andtheappropriateobjectiveisnot alwaysobvious.
Contents
Introduction
Sequencing Rules
Sequencing Theory for A Single Machines
Assuming that n jobs are to be processed through one machine.
For each job i, define the following quantities:
t
i
=Processing time for job i, constant for job i;
d
i
=Due date for job i, constant for job i;
W
i
=Waiting time for job i, the amount of time that the job must wait before its
processing can begin.
When all the jobs are processed continuously, W
i
is the sum of the processing
times for all of the preceding jobs;
t
1
t
2
t
3
t
4
W
4
=t
1
+t
2
+t
3
F
i
=Flow time for job i, the waiting time plus the processing time: F
i
= W
i
+ t
i
;
L
i
=Lateness of job i , L
i
=F
i
- d
i
, either positive or negative;
T
i
=Tardiness of job i, the positive part of L
i
, T
i
=max[L
i
,0] ;
E
i
=Earliness of job i, the negative part of L
i
, E
i
=max[- L
i
,0]
F
4
=W
4
+t
4
Maximum Tardiness I
mux
= mox I
1
, I
2
, , I
n
Mean Flow Time F =
1
n
F
n
=1
Mean Waiting Time w
i
=
1
n
w
n
=1
=
1
n
(F
-t
)
n
=1
Mean Lateness L
i
=
1
n
I
n
=1
=
1
n
(F
-J
)
n
=1
The following measures are equivalent:
Mean flow time
Mean waiting time
Mean Lateness
For only a single machine, every schedule can be represented by a
permutation (ordering) of the integers 1, 2, 3, , n.
There are totally n! (the factorial of n) different permutations.
Suppose that 4 jobs J
1
, J
2
, J
3
, J
4
need to be scheduled
For example a schedule
is J
3
-J
2
-J
1
-J
4
Considered as a permutation of
integers 1, 2, 3, 4: 3, 2, 1, 4.
A permutation of integers 1, 2, , n is expressed by [1], [2], , [n],
which represents a schedule;
[i] denotes the integer that put in the i
th
place in the permutation;
In case of a schedule 3, 2, 1, 4, [1]=3, [2]=2, [3]=1, and [4]=4;
Theorem 8.1 The scheduling rule that minimizes the mean flow
time F is SPT.
F
|k]
= t
|]
k
=1
The mean flow time of all jobs on the schedule is given by
F
i
=
1
n
F
|k]
n
=1
=
1
n
t
|]
k
=1
n
k=1
Suppose a schedule is [1], [2], [k], [k+1], [n], the flow time of the
job that is scheduled in position k is given by, say job in position 3:
F
[2]
=t
[1]
+t
[2]
=t
2
+t
1
t
[1]
(t
2
) t
[2]
(t
1
) t
[3]
(t
4
) t
[4]
(t
3
)
1.Shortest-Processing-Time Scheduling
Theorem 8.1 The scheduling rule that minimizes the mean flow
time F is SPT
The mean flow time is given by F
i
=
1
n
F
|k]
n
=1
=
1
n
t
|]
k
=1
n
k=1
The double summation term may
be written in a different form.
Expanding the double summation,
we obtain
k=1:t
[1]
;
k=2:t
[1]
+t
[2]
;
;
k=n:t
[1]
+t
[2
+t
[n]
SPT sequencing rule: the job with shortest processing time t is set first
By summing down the column
rather than across the row, we may
rewrite F in the form
nt
[1]
+(n-1)t
[2]
++t
[n]
Clearly, it is minimized by setting
t
|1]
t
2
t
n
The End!
Dr Na GENG
Prof. Zhibin JIANG
Contents
Introduction
Sequencing Rules
Sequencing Theory for A Single Machine
Assuming that n jobs are to be processed through one machine.
For each job i, define the following quantities:
t
i
=Processing time for job i, constant for job i;
d
i
=Due date for job i, constant for job i;
W
i
=Waiting time for job i, the amount of time that the job must wait before its
processing can begin.
When all the jobs are processed continuously, W
i
is the sum of the processing
times for all of the preceding jobs;
t
1
t
2
t
3
t
4
W
4
=t
1
+t
2
+t
3
F
i
=Flow time for job i, the waiting time plus the processing time: F
i
= W
i
+ t
i
;
L
i
=Lateness of job i , L
i
=F
i
- d
i
, either positive or negative;
T
i
=Tardiness of job i, the positive part of L
i
, T
i
=max[L
i
,0] ;
E
i
=Earliness of job i, the negative part of L
i
, E
i
=max[- L
i
,0]
F
4
=W
4
+t
4
Maximum Tardiness I
mux
= mox I
1
, I
2
, , I
n
Mean Flow Time F =
1
n
F
n
=1
Mean Waiting Time w
i
=
1
n
w
n
=1
=
1
n
(F
-t
)
n
=1
Mean Lateness L
i
=
1
n
I
n
=1
=
1
n
(F
-J
)
n
=1
The following measures are equivalent:
Mean flow time
Mean waiting time
Mean Lateness
1. SPT: The scheduling rule that minimizes the mean flow time
F is SPT.
SPT minimizes mean flow time, mean waiting time, and mean
lateness for single machine sequencing.
2. Earliest-Due-Date Scheduling: If the objective is to minimize
the maximum lateness, then the jobs should be sequenced
according to their due dates. That is, d
[1]
d
[2]
d
[n]
.
3. For a single machine, if there is a job with T
max
=0, then the
minimum average flow time can be derived with a new
T
max
=0 by the following way: put the job with the longest
processing time in the last position among the unscheduled
jobswhoseduedateisnosmaller thanthesumof processing
timeof unscheduledjobs.
Sum the processing
time of all the
unscheduled jobs
Sum the processing
time of all the
unscheduled jobs
Find the jobs with due
date longer than the
sum
Find the jobs with due
date longer than the
sum
Put the job with
longest processing
time in the last place
Put the job with
longest processing
time in the last place
Example:
Makespan: F
max
=20
Job 1 2 3 4 5
Processing Time 3 7 1 5 4
Due Date 23 20 8 6 14
Sequence Average flow time Tmax
SPT 3-1-5-4-2 9.2 7
EDD 4-3-5-2-1 11.6 0
SPT st. Due date 3-4-1-5-2 9.8 0
4. Minimizing the number of Tardy J obs: An algorithm from
Moore(1968) that minimizes the number of tardy jobs for the
single machine problem.
Step1. Sequence the jobs according to the earliest due date to
obtain the initial solution. That is d
[1]
d
[2]
,, d
[n]
;
Step2. Find the first tardy job in the current sequence, say job
[i]. If none exists go to step 4.
4. Minimizing the number of Tardy J obs: An algorithm from
Moore(1968) that minimizes the number of tardy jobs for the single
machine problem.
Step3. Consider jobs [1], [2], , [i]. Reject the job with the longest
processing time. Return to step2. (Why ?)
Reason: It has the largest effect on the tardiness of the J ob
[i]
.
Step4. Form an optimal sequence by taking the current sequence and
appending to it the rejected jobs. (Can be appended in any order?)
Yes, because we only consider the number of tardiness jobs
rather than tardiness.
Job 1 2 3 4 5 6
Due date 15 6 9 23 20 30
Processing time 10 3 4 8 10 6
Example 8.3
Solution
Job 2 3 1 5 4 6
Due date 6 9 15 20 23 30
Completion
time
3 7 17 27 35 41
Longest processing time
Job 2 3 5 4 6
Due date 6 9 20 23 30
Processing time 3 4 10 8 6
Completion time 3 7 17 25 31
Example 8.3 :Solution (Cont.)
Longest processing time
Job 2 3 4 6
Due date 6 9 23 30
Processing time 3 4 8 6
Completion time 3 7 15 21
The optimal sequence: 2, 3, 4, 6, 5, 1 or 2, 3, 4, 6, 1, 5. In each case
the number of tardy jobs is exactly 2.
Precedence constraints: Lawlers Algorithm
g
i
is any non-decreasing function of the flow time F
i
( )
i i i i i
g F F d L
) 0 , max( ) (
i i i i
d F F g
Minimizing maximum lateness
Minimizing maximum tardiness
1
( )
max min
i i
i n
g F

Objective Function
Precedence constraints: Lawlers Algorithm
The Algorithm
First schedulesthejobtobecompletedlast, thenthejobtobe
completed next to last, and so on. At each stage, we
determine the set of jobs not required to precede any other.
Call thisset V. amongtheset V, choosethejobk that satisfies
n
i
i
i
v i
k
t
g g
1
)) ( ( min ) (

The processing time of the current
sequence
e.g.: the job among V that has smallest
tardiness, if arranged on position [n].
The Algorithm (Cont.)
Consider the remaining jobs and again determine the set
of jobs that are not required to precede any other
remaining job.
The value of is then reduced by t
k
and the job scheduled
next to last is now determined.
The process is continued until all jobs are scheduled.
Note: As jobs are scheduled, some of the precedence
constraintsmay berelaxed, sotheset V islikely tochangeat
eachiteration.
Example 8.4
Job 1 2 3 4 5 6
Due date 3 6 9 7 11 7
Example 8.4
Job 1 2 3 4 5 6
Due date 3 6 9 7 11 7
Step1: find the job scheduled last(sixth)
Not predecessor
=2+3+4+3+2+1=15
3 5 6
Tardiness 15-9=6 15-11=4 15-7=8
Job 1 2 3 4 6
Processing time 2 3 4 3 1
Due date 3 6 9 7 7
Step2: find the job scheduled fifth
=15-2=13
3 6
Tardiness 13-9=4 13-7=6
Not predecessor
Example 8.4
Job 1 2 4 6
Processing time 2 3 3 1
Due date 3 6 7 7
Step3: find the job scheduled fourth
Not predecessor
=13-4=9
2 6
Tardiness 9-6=3 9-7=2
Job 1 2 4
Processing time 2 3 3
Due date 3 6 7
Step4: find the job scheduled third
=9-1=8
2 4
Tardiness 8-6=2 8-7=1
Not predecessor
Because job3 is
no longer on the
list, Job 2 now
because a
candidate.
Because job 6 has been
scheduled, Job 4 now
because a candidate
along with Job 2.
Example 8.4
Job 1 2
Processing
time
2 3
Due date 3 6
Step5: find the job scheduled second
Not predecessor
The optimal sequence: 1-2-4-6-3-5
Job Processing
time
Flow
time
Due date Tardiness
1
2
4
6
3
5
2
3
3
1
4
2
2
5
8
9
13
15
3
6
7
7
9
11
0
0
1
2
4
4
Maximum
tardiness
The End!
Dr Na GENG
Prof. Zhibin JIANG
Contents
Introduction
Sequencing Rules
Assume that n jobs are to be processed through m machines.
The number of possible schedules is astonishing, even for
moderate values of both n and m.
For each machine, there is n! different ordering of the jobs; if
the jobs may be processed on the machines in any order, there
are totally (n!)
m
possible schedules. (n=5, m=5, 25 billion
possible schedules)
Even with the availability of inexpensive computing today,
enumerating all feasible schedules for even moderate-sized
problems is impossible or, at best, impractical.
Machine 1 Machine 2
Job I 4 1
Job J 1 4
Gantt chart
Suppose that two jobs, I and J, are to be scheduled on two
machines, 1 and 2, the processing times are
Assume that both jobs must be processed first on machine 1
and then on machine 2. There are four possible schedules.
Schedule Total flow time Mean flow time Mean idle time
1 9 (5+9)/2=7 (4+4)/2=4
2 6 5.5 1
3 10 8 5
4 10 9.5 5
1. Scheduling n Jobs on Two Machines
Theorem 8.2 The optimal solution for scheduling n jobs on two
machines is always a permutation schedule.
One can restrict attention to schedules in which the sequence of
jobs is the same on both machines.
Because the total number of permutation schedules is exactly n!,
determining optimal schedules for two machines is roughly of the
same level of difficulty as determining optimal schedules for one
machine.
A very efficient algorithm for solving the two-machine problem
was discovered by Johnson(1954).
Denote the machines by A and B
The jobs must be processed first on machine A and then on
machine B.
Define
A
i
=Processing time of job i on machine A
B
i
=Processing time of job i on machine B
Rule: Job i precedes job i+1 if min(A
i
, B
i+1
)<min(A
i+1
,B
i
)
Procedures:
List the values of Ai and Bi in two columns.
Find the smallest remaining element in the two columns. If it
appears in column A, then schedule that job next. If it appears
in column B, then schedule that job last.
Cross off the jobs as they are scheduled. Stop when all jobs
have been scheduled.
Rule: Job i precedes job i+1 if min(A
i
, B
i+1
)<min(A
i+1
,B
i
)
Alternative Procedures:
Sequence the jobs in the increasing order of Ai for those with
Ai Bi, which is called Sequence A.
Sequence the jobs in the decreasing order of Bi for those with
AiBi, which is called Sequence B.
The optimal sequence is formed by appending Sequence B
with Sequence A.
Example 8.5
Job Machine A Machine B
1 5 2
2 1 6
3 9 7
4 3 8
5 10 4
Optimal sequence : 2 4 3 5 1
2. Extension to Three Machines
The three-machine problem can be reduced to a two-machine
problem if the following condition is satisfied
min A
i
max B
i
or min C
i
max B
i
It is only necessary that either one of these conditions be satisfied.
If that is the case, then the problem is reduced to a two-machine
problem.
Define A
i
=A
i
+B
i
, B
i
=B
i
+C
i
Solve the problem using the rules described above for two-
machines, treating A
i
and B
i
as the processing times.
The resulting permutation schedule will be optimal for the
three-machine problem.
If the condition are not satisfied, this method will usually give
reasonable, but possibly sub-optimal results.
Example 8.6 Consider the following job times for a three-machine
problem. Assume that the jobs are processed in the sequence A-B-
C.
Job A B C
1 4 5 8
2 9 6 10
3 8 2 6
4 6 3 7
5 5 4 11
Machine
Min A
i
=4
Max B
i
=6
Min C
i
=6
Example 8.6. Solution
Job A' B'
1 9 13
2 15 16
3 10 8
4 9 10
5 9 15
Machine
The optimal sequence:
1-4-5-2-3
3. The Two-Job Flow Shop Problem: assume that two jobs are to
be processed through m machines. Each job must be processed
by the machines in a particular order, but the sequences for the
two jobs need not be the same.
Draw a Cartesian coordinate system with the processing times corresponding
to the first job on the horizontal axis and the processing times corresponding
to the second job on the vertical axis.
Block out areas corresponding to each machine at the intersection of the
intervals marked for that machine on the two axes.
Determine a path from the origin to the end of the final block that does not
intersect any of the blocks and that minimizes the vertical movement.
Movement is allowed only in three directions: horizontal, vertical, and 45-
degree diagonal. The path with minimum vertical distance corresponds to the
optimal solution.
Example 8.7
A regional manufacturing firm produces a variety of household
products. One is a wooden desk lamp. Prior to packing, the lamps
must be sanded, lacquered, and polished. Each operation requires
a different machine. There are currently shipments of two models
awaiting processing. The times required for the three operations
for each of the two shipments are
Job 1 Job2
Operation Time Operation Time
Sanding (A) 3 A 2
Lacquering (B) 4 B 5
Polishing( C ) 5 C 3
Minimizing the flow time is the same as maximizing the time that both jobs are
being processed. That is equivalent to finding the path from the origin to the end of
block C that maximizes the diagonal movement and therefore minimizes either the
horizontal or the vertical movement.
or 10+(3+2)=15
or 10+6=16
Example 8.8
Reggie Sigal and Bob Robinson are roommates who enjoy spending Sunday
mornings reading the Sunday newspaper. Reggie likes to read the main section
first, followed by the sports section, then the comics, and finally the classifieds.
Bob also starts with the main section, but then goes directly to the classifieds,
followed by the sports section and finally the comics. The time required (in
tenth of an hour) for each to read the various sections are
Reggie Bob
Required Sequence Time Required Sequence Time
Main section (A) 6 Main section (A) 4
Sports (B) 1 Classifieds ( D ) 3
Comics (C) 5 Sports (B) 2
Classifieds ( D ) 4 Comics (C) 5
The End!
Dr. Na GENG
Prof. Zhibin JIANG
Contents
Introduction
Sequencing Rules
Stochastic Scheduling: Static Analysis
Summary
The issue dealt with is the uncertainty of the processing times.
Single Machine
Suppose that n jobs are to be processed through a single machine.
Assume that the job times, t
1
, t
2
,, t
n
, are random variables with
known distribution functions.
The objective is to minimize the expected average weighted flow
time,
Hin E
1
n
u
n
=1
Where u
i
are the weights and F
i
is the random flow time of job i.
Single Machine
Rothkopf (1966) has shown that the optimal solution is to
sequence the jobs so that job i precedes job i+1, if
E t
<
E t
+1
u
+1
If u
i
=1, then this rule equals with SPT.
Banerjee (1965) shows that if the objective is to minimize the
maximum over all jobs of the probability that a job is late, then
the optimal schedule is to order the jobs according to EDD (or
earliest expected due date).
Multiple Machine
An assumption that is usually made for the multiple-machine
problem is that the distribution of job times is exponential.
This assumption is important and necessary because the
memoryless property.
The problem: n jobs are to be processed through two identical
parallel machines. Each job needs to be processed only once on
either machine.
The objective is to minimize the expected time that elapses from
time zero until the last job has completed processing (expected
makespan).
Multiple Machine
Assume that the n jobs have processing times t
1
, t
2
, , t
n
, which
are exponential random variables with rates
1
,
2
,,
n
. This
means that the expected time required to complete job i,
E(t
i
)=1/
i
.
In parallel processing, the jobs need to be processed on only one
machine, and any job can be processed on either machine.
Assume that at time t=0, machine 1 is occupied with a prior job,
job 0, and its remaining processing time is t
0
.
The remaining jobs are processed as follows: Let [1], [2],, [n]
be a permutation of the n jobs.
Multiple Machine
Let T
0
T
1
T
2
T
n
, be the completion times of successive
jobs. The makespan is the time of completion of the last job,
which is T
n
.
The expected value of the makespan is minimized by using the
longest-expected-processing-time-first rule.
The optimality of scheduling the jobs in decreasing order of their
expected size rather than in increasing order (SPT rule) is more
likely a result of parallel processing than of randomness of the job
times).
Multiple Machine
I
n
+I
n-1
= t
n
=0
, I
n-1
= I
n
-I
Therefore, 2I
n
= t
n
=0
+I
Min T
n
= t
n
=0
+Min I
Because I is minimized when the processing time of last job is
minimized, we schedule the jobs in order of decreasing expected
processing time.
The Two-Machine Flow Shop Case
Scheduling n jobs on two machines in a flow shop setting, that is,
where each job must be processed through machine A first, then
through machine B.
Suppose that A
1
, A
2
,, A
n
and B
1
, B
2
,, B
n
are exponential
random variables with respective rates a
1
, a
2
,, a
n
and b
1
, b
2
,,
b
n
.
We now wish to minimize the expected value of the makespan.
Johnsons Rule: Job i precedes job i+1
if min(Ai, Bi+1)<min(Ai+1,Bi)
The minimum of two exponential random variables has a rate
equal to the sum of the rates.
F
mn
z = P min A, B z
= 1 - P min A, B z
= 1 -P A z, B z
= 1 - 1 -P A z 1 -P B z
= 1 - 1 -F
A
z 1 -F
B
z
= 1 - 1 - 1 -c
-uz
1 - 1 -c
-bz
= 1 -c
- u+b z
when z u
As the minimum of two exponential random variables has a rate
equal to the sum of the rates, it follows that
E min A
, B
+1
=
1
u
i
+b
i+1
, E min A
+1
, B
=
1
u
i+1
+b
i
It follows that Johnsons condition translates in the stochastic case
to the condition that
o
-b
o
+1
-b
+1
So that the jobs should be scheduled in the order of decreasing
values of the difference in the rates.
Example 8.9 Consider Example 8.5, let us assume that the job
times are random variables having the exponential distribution
with mean times given in the example.
According to the decreasing values of differences, the sequence
should be 2-4-3-5-1.
A B
1 5 2
2 1 6
3 9 7
4 3 8
5 10 4
Job
Expected times
A B A B
1 5 2 0.20 0.50 -0.30
2 1 6 1.00 0.17 0.83
3 9 7 0.11 0.14 -0.03
4 3 8 0.33 0.13 0.21
5 10 4 0.10 0.25 -0.15
Job
Expected times Rates
Differences
Contents
Introduction
Sequencing Rules
Summary
The problem of balancing an assembly line is a classic
industrial engineering problem.
The problem is characterized by a set of n distinct tasks that
must be completed on each item.
The time required to complete task i is a known constant t
i.
The goal is to organize the tasks into groups, with each group of
tasks being performed at a single workstation.
In most cases, the amount of time allotted to each workstation is
determined in advance, based on the desired rate of production
of the assembly line. This is known as the cycle time and is
denoted by C.
The problem of balancing an assembly line is a classic
industrial engineering problem.
There are a variety of factors that contribute to the difficulty of
the problem.
Precedence constraints: some tasks may have to be
completed in a particular sequence.
Zoning restriction: Some tasks cannot be performed at the
same workstation.
Let t
1
, t
2
, , t
n
be the time required to complete the respective
tasks.
The total work content (time) associated with the production
of an item, say T, is given by I = t
n
=1
.
For a cycle time of C, the minimum number of
workstations possible is [T/C], where the brackets
indicate that the value of T/C is to be rounded to the next
larger integer.
Ranked positional weight technique:
Places a weight on each task based on the total time
required by this task and all of the succeeding tasks;
Tasks are assigned sequentially to stations based on
these weights-the bigger the weight is, the higher the
priority is.
In-class Quiz
An independent accountant is planning to prepare tax
returns for six of her clients. Prior to her actually
preparing each return, her secretary checks the clients
file to be sure that all necessary documentation is there
and obtains all the tax forms needed for the preparation
of the return. Based on the past experience with the
clients, her secretary estimates that the following times
(in hours) are required for preparation of the return and
for the accountant to complete the necessary paperwork
prior to filling each return:
Client Secretary Time Accountant Time
1 1.2 2.5
2 1.6 4.5
3 2 2
4 1.5 6
5 3.1 5
6 0.5 1.5
Answer
Consider the certain case: 6-1-4-2-3-5, or 6-1-4-2-5-3,
Consider the uncertain case: 6-4-1-2-5-3
Client Secretary Time Accountant Time
1 1.2 2.5
2 1.6 4.5
3 2 2
4 1.5 6
5 3.1 5
6 0.5 1.5
Client Secretary Time Accountant Time Secretary Service Rate Accountant Service Rate Difference
1 1.20 2.50 0.83 0.40 0.43
2 1.60 4.50 0.63 0.22 0.40
3 2.00 2.00 0.50 0.50 0.00
4 1.50 6.00 0.67 0.17 0.50
5 3.10 5.00 0.32 0.20 0.12
6 0.50 1.50
2.00 0.67 1.33
The End!
Dr. Na GENG
Prof. Zhibin JIANG
Contents
Introduction
Sequencing Rules
Summary
Example 8.11
TheFinal assembly of Nonamepersonal computers, ageneric mail-order PC
clone, requiresatotal of 12tasks. Theassembly isdoneat theLubbock, Texas,
plant using various components imported from the Far East. The network
representationof thisparticular problemisgiveninthefollowingfigure.
Task Immediate
Predecessors
Time
1 _ 12
2 1 6
3 2 6
4 2 2
5 2 2
6 2 12
7 3, 4 7
8 7 5
9 5 1
10 9, 6 4
11 8, 10 6
12 11 7
Sum 70
Precondition
The job times and precedence
relationships for this problemare
summarizedinthetablebelow.
Question
Suppose that the company is
willing to hire enough workers to
produce one assembled machine
every15minutes.
T/C=4.67
Five-station OK?
Task Positional Weight
1 70
2 58
3 31
4 27
5 20
6 29
7 25
8 18
9 18
10 17
11 13
12 7
The solution precedence requires determining the positional
weight of each task. Thepositional weight of task i is defined as
the time required to performtask i plus the times required to
performall taskshavingtaski asapredecessor.
t
3
+t
7
+t
8
+t
11
+t
12
=31
The ranking
1, 2, 3, 6, 4, 7, 5, 8, 9, 10, 11, 12
Profile 1 C=15
Station 1 2 3 4 5 6
Tasks 1 2, 3, 4 5, 6, 9 7, 8 10, 11 12
Idle time 3 1 0 3 5 8
The ranking
1, 2, 3, 6, 4, 7, 5, 8, 9, 10, 11, 12
Task Immediate
Predecessors
Time
1 _ 12
2 1 6
3 2 6
4 2 2
5 2 2
6 2 12
7 3, 4 7
8 7 5
9 5 1
10 9, 6 4
11 8, 10 6
12 11 7
T2=6
Profile 1 C=15
Station 1 2 3 4 5 6
Tasks 1 2,3,4 5,6,9 7,8 10,11 12
Idle time 3 1 0 3 5 8
Cycle Time=15
T1=12
T2=6 T3=6 T4=2
T5=2 T6=12 T9=1
T5=2
T8=5 T7=7 T10=4
T10=4 T11=6 T12=7
T12=7
15
Evaluate the
balancing results by
the efficiency
t
i
/NC;
The efficiencies
for Profiles 1 is
77.7%.
The ranking
1, 2, 3, 6, 4, 7, 5, 8, 9, 10, 11, 12
Profile 2: Increasing cycle time from 15 to 16
Station 1 2 3 4 5
Tasks 1 2,3,4,5 6,9 7,8,10 11,12
Idle time 4 0 3 0 3
Increasing the cycle time from 15 to 16, the total idle time has
been cut down from 20 min/units to 10; resulting in a
substantial improvement in balancing rate.
However, the production rate has to be reduced from one
unit/15 minutes to one unit/16minute;
Alternative 1: Change cycle time to ensure 5 station
balance
Profile 2 C=13
Station 1 2 3 4 5 6
Tasks 1 2,3 6 4,5,7,9 8,10 11,12
Idle time 1 1 1 1 4 0
Alternative2: Staying with 6 stations, seeif asix-station
balance could be obtained by cycle time less that 15
minutes
13 minutes appear to be the minimum cycle time with six
station balance.
Increasing the number of stations from 5 to 6 results in a
great improvement in production rate;
The efficiencies for profile 1~3 are 77.7%,
87.5%, and 89.7%. Thus the profile 3 is the
best one.
Contents
Introduction
Sequencing Rules
Summary
Summary
Aggregate plan
purchase order
Summary
Objectives of operations scheduling
1) Meet due date;
2) Minimize WIP inventory;
3) Minimize the average flow time through the systems;
4) Provide for high machine/worker (time) utilization
(minimize idle time);
5) Provide for accurate job status information;
6) Reduce setup cost;
7) Minimize production and worker costs
Proper trade off between cost and quality is one of the most
challenging strategic issues facing a firm today.
Summary
A job shop is organized by machines which are grouped
according to their functions.
Operations scheduling of shop floor usually means job
shop scheduling;
Flow shop:
Each of then jobs must be processed through the m
machines in the same order.
Each job is processed exactly once on each machine.
An assembly balancing problem is to determine the
number of stations and to allocate tasks to each station.
Summary
Terminology:
1.Parallel processing versus sequential processing
2.Flow time
3.Make-span
4.Tardiness and lateness
Sequencing Rules
1.FCFS (first come-first served)
2.SPT (shortest processing time)
3.EDD (earliest due date)
4.CR (Critical ratio)
Summary
Sequencing Theory for A Single Machines:
1. The scheduling rule that minimizes the mean flow time F is
SPT.
2.Earliest-Due-Date Scheduling: If the objective is to minimize
the maximum lateness, then the jobs should be sequenced
according to their due dates. That is, d
[1]
d
[2]
d
[n]
.
3.Minimizing the number of Tardy J obs
4.Precedence constraints: Lawlers Algorithm
Summary
Sequencing Theory for Multiple Machines:
1.Scheduling n J obs on Two Machines
Theorem 8.2 The optimal solution for scheduling n jobs on two
machines is always a permutation schedule.
J ohnsons Rule: J ob i precedes job i+1 if min(A
i
, B
i+1
) <
min(A
i+1
,B
i
)
The three-machine problem can be reduced to a two-
machineproblemif thefollowingconditionissatisfied
min A
i
max B
i
or min C
i
max B
i
Define A
i
=A
i
+B
i
, B
i
=B
i
+C
i
Summary
Sequencing Theory for Multiple Machines:
3.The Two-J ob Flow Shop Problem: assume that two jobs are to
be processed through m machines.
Draw a Cartesian coordinate system with the processing times
corresponding to the first job on the horizontal axis and the processing
times corresponding to the second job on the vertical axis.
Block out areas corresponding to each machine at the intersection of
the intervals marked for that machine on the two axes.
Determine a path from the origin to the end of the final block that does
not intersect any of the blocks and that minimizes the vertical
movement. Movement is allowed only in three directions: horizontal,
vertical, and 45-degree diagonal. The path with minimum vertical
distance corresponds to the optimal solution.
Summary
Stochastic Scheduling: Static Analysis:
Single Machine
Tominimizetheexpectedweightedflowtime, theoptimal
solution is to sequencethejobs so that job i precedes job
i+1, if
L t
i
u
i
<
L t
i+1
u
i+1
Multiple Machine
The expected value of the makespanis minimized by
using the longest-expected-processing-time-first rule.
J ohnsons conditiontranslates inthestochastic caseto the
conditionthat o
-b
o
+1
-b
+1
Summary
Assembly Line Balancing:
Ranked positional weight technique:
Places aweight oneachtask basedonthetotal timerequired
bythistaskandall of thesucceedingtasks;
Tasks are assigned sequentially to stations based on these
weights-thebigger theweight is, thehigher the priorityis.
Homework
P400: Q4
P407: Q8Q9
P415 :Q13
The End!
Dr. Na GENG
Prof. Zhibin JIANG
Just-in-time
Just-in-time
Contents
Introduction
Just-in-time (JIT)
Kanban Introduction
Preface Toyota Automobile
In 50s last century, J apanese Toyota Automobile was not known
at all since its annual output was even lower than daily output of
Ford Automobile.
20 years later, J apanese automobile industry rapidly raised such
that it subverted American leading position in worldwide
manufacturing industries.
American learnt from their poor lessons, and a group of scholars
of MIT conducted a insight investigation in Toyota and wrote a
well-known book The Machine that Changed the World and
have proposed a profound manufacturing mode, that is Lean
Production (LP).
The core philosophy of LP is Just in Time (JIT)
Dr James P. Womack Visited IE&M Dept. of SJTU
Introduction
J IT - produce exactly what we need
when we need it.
J IT is a philosophy whereby we
continuously attempt to drive all waste
out of our production system.
J IT Goal Continuous improvement
Zero inventory
Zero defects
Zero setups
Zero breakdowns
Zero handling
Zero lead time
Lean Production
Through Waste Elimination
Operational Stability
JIT
Quality
at the
Source
People
Introduction
Traditional Mode
(Push)
Seller Driven
Selling
Price
Profit = Cost +
(SELLER)
LP Mode
(Pull)
Customer Pull
Selling
Price
-
Cost
= PROFIT
(MARKET)
Process
downtime
Scrap
Setup
time
Late deliveries
Quality
problems
Introduction
Process
downtime
Scrap
Setup
time
Late deliveries
Quality
problems
Minimizing Waste: Inventory Hides Problems
Introduction
Types of wastes
Overproduction
Transportation
Waiting time
Inventory waste Processing Motion/movement
Product defects
1. Overproduction
Produce not to the marketing.
High WIP level.
Increasing inventory level
Backlog of funds
Difficult to realize First-In-First-Out.
The increasing equipment cost, for example, turnover boxes
The undiscovered quality problem

Types of wastes
2. Transportation
The wasted caused by the placement, movement, and
cleaning of material
The waste of place needed by the movement of material;
The waste of human resources and tools;
The waste of time;
The movement of cart
Types of wastes
3. Product defects
The rework of defective goods
The waste of materials
The waste of equipment and human resources
The additional workers
The impact on the production
The delay of delivery
The lower price of products
Types of wastes
4. Inventory
The more inventory, the more financial investment.
The waste of movement, packing, and searching.
The difficulty of FIFO rule caused by huge inventory
The wasted place
The wasted funds
The wasted overhead
Types of wastes
5. Processing
The wrong processing leads to the waste of human and materials.
The additional processing time
The waste of production resources
The waste of additional human resources
The increasing of overtime
The impact on production
Types of wastes
6. Motion/movement
The wasted caused by the unreasonable motions.
The pickup, placement, and rotation are unreasonable.
The unreasonable turn around.
The arm higher or lower
Types of wastes
7. Waiting
Shortage of material, unbalance of tasks, unreasonable plan.
Switch of production line.
Unbalanced tasks among different work stations
The breakdown of machine
The shortage of material
The idleness of human and machines by production
plan.
.
Types of wastes
Just-in-time
Contents
Introduction
Just-in-time (JIT)
Kanban Introduction
JIT Fundamentals
Features Advantages Disadvantages
Small WIP
inventories
Kanban
information
system
Coordinated
inventory and
purchasing
1. Decrease inventory cost;
2. Improved production
efficiency;
3. Identify quality and take
immediate action;
1. Provides for efficient lot
tracking;
2. Inexpensive implementation;
3. Allow for predetermined level
of WIP
1. Inventory reduction;
2. Improved coordination of
different systems;
3. Improved relationships with
vendors
1. May result in increased
worker idle time
2. May decrease production
rate
1. Slow to react to changes in
demand;
2. Ignores known information
about future demand pattern
1. Decreased opportunity for
multiple sourcing;
2. Suppliers must react more
quickly;
3. Improved reliability
required of suppliers.
Kanban Pull
Demand pull
Back flush
Reduce batch/lot sizes
Work with suppliers
Reduce lead times
Frequent deliveries
Reliability
Quality expectations
Reduce inventory in
Stores
Transit
Conveyors
Calculation from MRP & EOQs
JIT Implementation
Leveling production
Mixed production line
Balance of product
types, output, machine
hours, and equipment
workload.
Frequent change of
production line
People Focus
Respect
Team work
Regular
discussion
Authorization
Training
JIT Implementation
Reducing setup time
More flexibility
Adjust the machine
during operation
Reduce the adjusting
time when stopping
Training
Production Synchronization
No buffer
U-shape layout
One StreamProduction
Less production cycle
Lower WIP level
Less defective products
JIT Implementation
Product Design Improvement
DFM & process design
Modules & fewer parts
Standardization design
Robust design
Quality control
Worker responsibility
Error proving
Automatic inspection
SQC (Statistic Quality
Control)
Total Quality Management
Quality control
Worker responsibility
Error proving
Automatic inspection
SQC (Statistic Quality
Control)
Total Quality Management
JIT Requirements
In-class Discussion
1. Can J IT realize zero-inventory?
2. Which layout is better for J IT, job shop or flow
shop? Why?
3. One prerequisitefor J IT is stable production plan,
why?
The End!
Dr. Na GENG
Prof. Zhibin JIANG
Just-in-time
Just-in-time
Contents
Introduction
Just-in-time (JIT)
Kanban Introduction
Kanban
Kanban
Final
assembly
Work
cell
Kanban
Material/Parts
Supplier
Finished
goods
Customer
order
Kanban Introduction
Kanban( Kan=card, Ban=signal) developed at Toyota 1950s to
manage material flow ;
Kanbanis
an information network used to control production quantities;
a communication process that controls the movement of
material;
a tool to implement J IT.
Simple movement system
cardsto signal & communicate reorder information
boxes/containers to take lotsof parts from one work station
to another (client-server).
Kanban Introduction
1. Parts are never to be produced or conveyed without a Kanban;
2. Kanbancomes from downstream process;
3. The upstream process produces items in the quantity and the
sequence indicated by the Kanban;
4. The downstream process draws only what is required by the
Kanbanfrom upstream processes;
5. Kanbanmust be attached to the actual containers of parts;
6. Never send defective products to downstream processes.
Rules of Kanban
Communicates production and transportation information.
Communicates production information. It tells what to
produce and how many are required. Kanbanmust be
produced in the order received.
Provides pick-up and/or transportation information, which
includes where from and where to locations. It can also,
in some cases, indicate schedule times for pick-up.
Prevents overproduction and excessive transportation by
restricting in-bound raw material.
Function of Kanban
A tool for visual management
Allows ordering of multiple lot quantities only to the extent of
Kanbanavailable to be replenished.
Maintain the production quantity of a process to specific pre-
determined lot quantity indicated on the Kanban.
Serves as a work order indicating what has been produced and
identifies the physical product.
Improvement tool
Revealsexisting production problems by reducing the number of
Kanban.
Helps to reduce passage of defects from one area to another by
identifying preceding processes.
Function of Kanban
Standardization of
processes
Synchronized
production
Leveling production
Quick Changeover
procedures
Preconditions of Kanban
Four pre-conditions
for the successful
implementation of
Kanban
Four pre-conditions
for the successful
implementation of
Kanban
Example25 days in May and 7h of working time
every day.
Leveling production:
Multi-product mix production;
Smooth and balanced production;
Balanced not only in yield, but also in product mix, time and load.
In order to reduce inventory level of final products.
Product Demand Output/day Takt Time (min)
A 400 16 26.2
B 300 12 35
C 200 8 52.5
D 100 4 105
Sum 1000 40 10.5
Mix 1: A: 400-B:300-C:200-D:100, cycle only once per
month;
Mix 2: A:100-B:75-C:50-D:25, cycle 4 times per month;
Mix 3: A:4-B:3-C:2-D:1, cycle 100 times per month;
Mix 4: ABCABCABAD, A cycle 400 time per month, B 300
time per month, C 200 times per month, and D 100 times per
month.
Product Demand Output/day Takt Time (min)
A 400 16 26.2
B 300 12 35
C 200 8 52.5
D 100 4 105
Sum 1000 40 10.5
Bin
Part A
Bin
Part A
Machine
Center
Assembly
Line
Material Flow
Card (signal) Flow
Parts Withdrawal kanban
Production instructionkanban
Minimizing Waste: Kanban Control System
Production
kanban
Signal
kanban
Intra-process kanban
Supplier kanban
JIT Fundamentals
Fig.7-8 Kanban System for Two Production Centers
1. An accumulated number of WKs
along with empty containers are taken
to store location, where full containers
with PWs attached are stored.
JIT Fundamentals
2. The worker removes the PKs from
full containers, places them into PKs
post, and places WKs in the full
containers.
JIT Fundamentals
3.When a specified number of PKs have
accumulated, PKs are attached to empty
containers, and work center 1 proceeds
with production.
JIT Fundamentals
4. Fished parts produced in work center 1 are put into
containers with PKs; and when a container with PK is
full, it is sent to the store, where PKs attached to full
containers are exchanged with WKs as stated in step 2.
JIT Fundamentals
5. Full containers with WKs
attached are sent from store to
holding area of work center
2 ,waiting to be processed.
JIT Fundamentals
6. When full containers with WKs attached enter
production at work center 2, the worker detach WKs
from the containers and put into WKs post. Parts are
taken from the containers for processing and the
containers become empty. Then repeat from step 1.
C
S dL ) (1
k
capacity size
Expected demand during lead time + safety stock
k = No. of kanbans in card set

d = Average No. of units demanded by the downstream over the lead
time
L = Lead time to replenish order (same units of time as demand)
S = Safety stock as % of demand during lead time
C = Container size
The number of parts in each container (Container size)
= minimum replenishment lot size .
Calculate lead time required to produce a container
How many Kanban is needed?
A switch is assembled in batches of 4 units at an
upstream work area.
Delivered in a bin to a downstream control-panel
assembly area that requires 5 switch assemblies/hour.
The switch assembly area can produce a bin of switch
assemblies in 2 hours.
Safety stock = 10% of needed inventory.
k =
dL(1+S)
C
=
521.1
4
= 2.7S = S
Example
Whats the difference between J IT and MRPII?
The backgrounds are different.
The elements are different.
The management methods are different.
The capacity planning is different.
In-class Discussion
The End!
Dr. Na GENG
Prof. Zhibin JIANG
Project Scheduling
In this age of time-based competition,
successful project management canbringnew
products to the marketplace sooner, thereby
preemptingrivalsandcapturingmarket share.
2
Content
The nature of project management
Techniques for project scheduling
Resource constraints
Activity crashing
Incorporating uncertainty in activity times
Problems with implementing critical path analysis
Monitoring projects
3
A project
Planned set of one-time, interrelated tasks to be executed
over a fixed period and within certain cost and other
limitations.
Characteristics of projects
1) One-time
2) Objectives
3) Interdependencies
4) Conflict
4
Project management
Planning, scheduling, and controlling project activities
to achieve timely project completion
within budget and meeting performance expectations.
History of project management
1930s~1950s: Gantt chart and Mile stone system
1950s~1970s: Critical path method (CPM), program
evaluation and review technique (PERT)
1970s~ : interdisciplinary
5
Project management process
Planning
A clearly defined statement of work and objectives
Aid: work breakdown structure (WBS)
6
Planning
1.0Movethehospital (Project)
1.1Movepatients(Task)
1.1.1Arrangefor ambulance(subtask)
1.1.1Preparepatientsfor themove(Workelement)
1.1.2Boxpatients personal effects(Workelement)
1.2Movefurniture
1.2.1Contract withmovingcompany

Scheduling:
begins with time and cost estimates for
each work element or activity and
determines the precedence relationship
between activities.
A project network diagram is prepared to
provide a visual picture of the project
schedule.
7
Planning
Scheduling
Controlling
Controlling is concerned with
making sure that all aspects of
project implementation are carried
out according to both time and
budget.
8
Planning
Scheduling
Controlling
9
Manager
Credibility
Sensitivity
Ability to
handle
stress
Leadership
10
Project
Team
Storming
Norming
Performing
Forming
Principle of effective project management
Direct people individually and as a team
Reinforce the excitement of the project
Keep everyone informed
Build agreements that vitalize team
member
Empower yourself and team members
Encourage risk taking and creativity
11
Content
Activity crashing
Monitoring projects
12
Establishing objectives
Identifying available resources
Sequencing activities
Identifying precedence relationships
Determining activity times & costs
Estimating material & worker requirements
Determining critical activities
Project Scheduling Techniques
Gantt chart
Critical Path Method (CPM)
Program Evaluation & Review Technique (PERT)
Gantt charts
Developed by Henry Gantt in 1916, a Gantt chart is used to
determine the timing of individual activities in a project.
This chart plots a time line for each activity against a
calendar.
The first step in using Gantt chart is to break down the
project into discrete activities.
A Gantt chart can be used to schedule a periodic or repetitive
project.
J F M A M J J
Time Period
Activity
Design
Build
Test
J F M A M J J
Time Period
Activity
Design
Build
Test
Techniques for project management
A critique of Gantt project charts
1) Visual, easy to construct, and easy to understand
2) Inadequate for large-scale, complex projects
3) Not show clearly the interdependence of activities
4) Ineffective and cumbersome to use with large,
complicated projected
17
PERT & CPM project charts
Network techniques
Developed in 1950s
CPM by DuPont for chemical plants, 1957
PERT by U.S. Navy for Polaris missile, 1958
Consider precedence relationships & interdependencies
Each uses a different estimate of activity times
PERT & CPM steps
Identify activities
Determine sequence
Create network
Determine activity times
Earliest & latest start times
Earliest & latest finish times
Slack
Find critical path
Constructing a project network
A network consists of a set of circles called nodes and also a set of
arrows.
The arrows connect the nodes to give a visual presentation of the
sequence of activities.
Activity on node (AON):
the nodes: project activities
the arrows: the activity sequence
Activity on arrow (AOA):
the arrow: project activities
the nodes: events, which are the starts or completions of activities
20
Activity
(time)
ii jj
A(t) A(t)
B(t) B(t)
All critical path analyses use AON convention.
Assumption: an activity cannot begin until all immediate predecessor
activities are completed.
AON charts are connected and acyclic.
Connected: it is possible to get to any network node by following
arrows leaving the start node
Acyclic: the sequence of activities progresses without interruption from
the start node to the end node without looping around in the circles
21
Ex. 10.1 Tennis tournament Project network AON
22
Activity description Code
Immediate
predecessor
Estimated
duration (days)
Negotiate for location A - 2
Contact seeded players B - 8
Plan promotion C A 3
Locate officials D C 2
Send RSVP invitations E C 10
Sign player contacts F B,C 4
Purchase balls and trophies G D 4
Negotiate catering H E,F 1
Prepare location I E,G 3
Tournament J H,I 2
Table 10.1 Tennis tournament activities
AOA
23
2
3
1
A
C
B
D
5
4
2
3
1
A
C
B
D
4
Add a phantom arc for clarity.
Ex. 10.2 Tennis tournament Project network
24
Immediate
predecessor
Estimated
duration (days)
Tournament J H,I 2
In-class Quiz
Event
ID
Pred. Description
Time
(Wks)
A None Prepare Site 1
B A Pour fdn. & frame 6
C B Buy shrubs etc. 3
D B Roof 2
E D Do interior work 3
F C Landscape 4
G C Move In 1

Critical Path
A
E D B
C F
G
1
6
2
3
1
4 3
The End!
Dr. Na GENG
Prof. Zhibin JIANG
Project Scheduling
Content
Activity crashing
Monitoring projects
2
Critical Path Analysis
Critical path method:
The critical path method (CPM) is an approach to determine the
start and finish dates for individual activities in a project.
A result of this method is the identification of a critical path.
Procedures:
Draw the network
Provides activity information
Earliest (ES) & latest (LS) start, Earliest (EF) & latest (LF) finish
Slack (S): Allowable delay
Identifies critical path
Longest path in network
Shortest time project can be completed
Activities have 0 slack
Any delay on activities delays project
4
Item Symbol Definition
Expected activity duration t The expected duration of an activity
Early start ES
The earliest time an activity can begin if all previous
acivities began at their earliest times
Early finish EF
The earliest time an activity can be completed if it is
started at its early start time
Late start LS
The latest time an activity can begin without delaying
the completion of the project
Late finish LF
The latest time an activity can be completed if it is
started at its latest start time
Total slack TS
The amount of time an activity can be delayed without
delaying the completion of the project
Table 10.2 Notation for critical path method
ES = EF
predecessor
(max)
EF = ES + t
LF = LS
successor
(min)
LS = LF - t
TS = LF - EF or
TS = LS - ES
Critical path method
Ex. 10.3 Tennis tournament - critical path analysis
5
6
7
Critical Path:
A-C-E-I-J
8
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
A
B
C
E
D
G
F
H
I
J
Outline
Activity crashing
Monitoring projects
9
Resource constraint: staff, equipment, space ect.
Beginning with the bottom, movable activities
Move the farthest steps
10
Resource constraint
Immediate
predecessor
Estimated
duration (days)
Tournament J H,I 2
ID Activity Days
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
A Negotiate for location 2
B Contact seeded players 8
C Plan promotion 3
D Locate officials 2
E Send RSVP invitations 10
F Sign player contacts 4
G Purchase balls and troph 4
H Negotiate catering 1
I Prepare location 3
J Tournament 2
2 2 2 2 2 3 3 3 3 3 3 2 1 1 1 2 1 1 1 1
critical path activities
Activities with slack
Personnel required
ID Activity Days
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
C Plan promotion 3
J Tournament 2
2 2 2 2 2 3 3 3 3 3 3 2 1 1 1 1 1 2 1 1
Personnel required
11
Resource constraint
Immediate
predecessor
Estimated
duration (days)
Tournament J H,I 2
ID Activity Days
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
C Plan promotion 3
J Tournament 2
2 2 2 2 2 3 3 2 2 2 2 3 2 2 2 1 1 2 1 1
Personnel required
12
Resource constraint
Immediate
predecessor
Estimated
duration (days)
Tournament J H,I 2
ID Activity Days
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
C Plan promotion 3
J Tournament 2
2 2 2 2 2 3 3 2 1 1 1 2 2 3 3 2 2 2 1 1
Personnel required
13
Resource constraint
Immediate
predecessor
Estimated
duration (days)
Tournament J H,I 2
ID Activity Days
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
C Plan promotion 3
J Tournament 2
2 2 2 2 2 2 2 2 1 2 2 2 2 3 3 2 2 2 1 1
Personnel required
14
Resource constraint
Immediate
predecessor
Estimated
duration (days)
Tournament J H,I 2
ID Activity Days
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
C Plan promotion 3
J Tournament 2
2 2 2 2 2 2 2 2 2 2 2 2 2 3 2 2 2 2 1 1
Personnel required
15
Resource constraint
Immediate
predecessor
Estimated
duration (days)
Tournament J H,I 2
Outline
Activity crashing
Monitoring projects
16
Activity crashing
Construction projects often are undertaken with target completion dates that
are important to the client.
An activity is considered to be crashed when it is completed in less time
than is normal by applying additional labor or equipment.
17
Activity crashing
Ex. 10.4 Tennis Tournament activity crashing
1. Calculate the expedite-cost slope:
18
Expedite-cost slope:
S

- C= activity cost (* for crash cost)

- D=activity duration (* for crash duration)
Table 10.3 Cost-time estimates for tennis tournament
Normal Crash Normal Crash
A 2 1 5 15 10
B 8 6 22 30 4
C 3 2 10 13 3
D 2 1 11 17 6
E 10 6 20 40 5
F 4 3 8 15 7
G 4 3 9 10 1
H 1 1 10 10 -
I 3 2 8 10 2
J 2 1 12 20 8
115
Activity
Time estimate (days) Direct cost($) Expedite-
cost slope
Activity crashing
The crashing procedure can be summarized as
1) Calculate the expedite-cost for each activity;
2) List all the paths in the project network and their normal duration;
3) Crash by 1 day the least costly activity on the critical path or the
least costly combination of activities on common critical paths.
Record the cost of the crashed schedule;
4) Update the duration for each path in the project network;
5) If an activity has reached its crash time, note with an asterisk and
do not consider it as a further candidate;
6) If a critical path contains activities all noted with asterisk, STOP;
otherwise, GO to 3.
19
Activity crashing
20
Project
duration
Activity
crashed
Direct
cost $
Indirect
cost $
Opportunity
cost $
Total
cost $
20 Normal 115 45 8 168
Table 10.4 Total cost calculations
A-C-D-G-I-J 16
A-C-E-I-J 20
A-C-E-H-J 18
A-C-F-H-J 12
B-F-H-J 15
Table 10.5 Project path
durations following
crashing
Project paths
Normal
duration
2. List all the paths and the normal duration
Activity crashing
21
Project
duration
Activity
crashed
Direct
cost $
Indirect
cost $
Opportunity
cost $
Total
cost $
20 Normal 115 45 8 168
19 I* 117 41 6 164
Duration after crashing activity
I*
A-C-D-G-I-J 16 15
A-C-E-I-J 20 19
A-C-E-H-J 18 18
A-C-F-H-J 12 12
B-F-H-J 15 15
Table 10.5 Project path durations following crashing
Project paths
Normal
duration
A 2 1 5 15 10
B 8 6 22 30 4
C 3 2 10 13 3
D 2 1 11 17 6
E 10 6 20 40 5
F 4 3 8 15 7
G 4 3 9 10 1
H 1 1 10 10 -
I 3 2 8 10 2
J 2 1 12 20 8
115
Activity
cost slope
A-C-D-G-I-J 16
A-C-E-I-J 20
A-C-E-H-J 18
A-C-F-H-J 12
B-F-H-J 15
Table 10.5 Project path
durations following
crashing
Project paths
Normal
duration
Activity crashing
22
Project
duration
Activity
crashed
Direct
cost $
Indirect
cost $
Opportunity
cost $
Total
cost $
20 Normal 115 45 8 168
19 I* 117 41 6 164
18 C* 120 37 4 161
17 E 125 33 2 160
16 E 130 29 0 159
15 E 135 25 -2 158
14 J * 143 21 -4 160
13 E*,B 152 17 -6 163
12 A*,B* 166 13 -8 171
I* C* E E E J * E*,B A*,B*
A-C-D-G-I-J 16 15 14 14 14 14 13 13 12
A-C-E-I-J 20 19 18 17 16 15 14 13 12
A-C-E-H-J 18 18 17 16 15 14 13 12 11
A-C-F-H-J 12 12 11 11 11 11 10 10 9
B-F-H-J 15 15 15 15 15 15 14 13 12
Table 16.5 Project path durations following
Project paths
Normal
duration
Duration after crashing activity
A 2 1 5 15 10
B 8 6 22 30 4
C 3 2 10 13 3
D 2 1 11 17 6
E 10 6 20 40 5
F 4 3 8 15 7
G 4 3 9 10 1
H 1 1 10 10 -
I 3 2 8 10 2
J 2 1 12 20 8
115
Activity
cost slope
In-class Quiz
Event
ID
Pred. Description
Time
(Wks)
A None Prepare Site 1
B A Pour fdn. & frame 6
C B Buy shrubs etc. 3
D B Roof 2
E D Do interior work 3
F C Landscape 4
G C Move In 1

Critical Path
A
E D B
C F
G
1
6
2
3
1
4 3
Activity ES EF LS LF Slack
A 0 1 0 1 0
B 1 7 1 7 0
C 1 4 4 7 3
D 7 9 7 9 0
E 9 12 9 12 0
F 4 8 7 12 3
G 12 13 12 13 0

Compute Slack
Critical Path
A
E D B
C F
G
1
6
2
3
1
4 3
The End!
Dr. Na GENG
Prof. Zhibin JIANG
Project Scheduling
Content
Activity crashing
Monitoring projects
2
Estimating activity duration distribution
In PERT model, the duration for each activity is assume to
follow Beta Distribution with pdf
x =
p +q
p q
x -t
o
p-1
t
p
-x
q-1
t
p
-t
o
p+q-2
where
t
o
, t
p
the lower and upper control point
=
]
x
q-1
c
-x
Jx
0
p, q - the shape parameters
3
Expected value:
p = t
o
+ t
p
- t
o
p
p + q
Variance:
S = t
p
- t
o
2
pq
p + q
2
p + q + 1
A typical Beta distribution is commonly used to describe the
duration of uncertain activities.
Optimistic time (A): the duration of an activity if no
complications or problems occur.
Most likely time (M): the duration is most likely to occur.
Pessimistic time (B): the duration of an activity if
extraordinary problems arise.
4
Expected value:
t =
A +4H +B
6
Standard Variance:
o =
B -A
6
Project completion time distribution
1) For every activity, obtain estimates of A, M, B.
2) Use t=(A+4M+B)/6 to calculate the expected activity durations,
and perform critical path analysis using the expected activity
durations t.
3) The expected project completion time T is assumed to be the
sum of the expected durations of activities on critical path.
4) The variance of project completion time o
2
1
is assumed to be
the sum of the variances of activities on the critical path.
5) Probabilities regarding project completion time can be
determined from standard normal tables.
5
Ex. 10.5 Tennis tournament project
completion time distribution
The expected project completion time T is
determined by the critical path A-C-E-I-J , i.e., 20
days.
The variance of the project completion time is
o
2
1
= S.2
The Z value for the standard normal deviate is
calculated
Z =
X -p
o
=
24 -2u
S.2
= 1.7S
From the standard normal table with Z=1.75, we
find the probability of completing the project within
24 days to be approximately 0.96.
6
A M B
A 1 2 3 0.11 2
B 5 8 11 1.00 8
C 2 3 4 0.11 3
D 1 2 3 0.11 2
E 6 9 18 4.00 10
F 2 4 6 0.44 4
G 1 3 11 2.78 4
H 1 1 1 0.00 1
I 2 2 8 1.00 3
J 2 2 2 0.00 2
Activity Variance
Expected
duration
Time estimates
able 10.6 Variances and expected activity duratio
A critiqueof theproject completiontimeanalysis
The key assumption underlying our analysis leading to a project
completion time distribution is that the critical path as calculated
fromexpectedactivitydurationswill actuallybethetruecritical path.
In reality, the critical path itself is a random variable that is not
knownfor certainuntil theproject iscompleted.
If theexpecteddurationof thecritical pathismuchlonger thanthat of
anyother path, thentheestimateslikelywill begood.
If theproject network contains noncritical paths with very littletotal
slack time, thesepaths may well affect project completiontime. This
situation is called merge node bias. That is, the project completion
nodehas several paths cominginto it, any oneof whichcouldbethe
critical paththat determinestheproject completiontime.
7
A few days into the project, we discover that activity
G may take longer than expected, with revised
estimates of A =2, M=3, B=28. What effect does
this have on the probability of completing the
project in 24 days?
t=7, o
2
=
676
36
= 18.78 for activity G
Initially, activity G had a 4-day expected duration
and total slack of 4 days.
With a revised duration of 7 days, activity G is still
noncritical, with a TS=1.
However, the large variance of activity G, will have
an impact on the likelihood of this path A-C-D-G-I-J
becoming critical.
8
A M B
A 1 2 3 0.11 2
B 5 8 11 1.00 8
C 2 3 4 0.11 3
D 1 2 3 0.11 2
E 6 9 18 4.00 10
F 2 4 6 0.44 4
G 1 3 11 2.78 4
H 1 1 1 0.00 1
I 2 2 8 1.00 3
J 2 2 2 0.00 2
Activity Variance
Expected
duration
Time estimates
A few days into the project, we discover that activity
G may take longer than expected, with revised
estimates of A =2, M=3, B=28. What effect does
this have on the probability of completing the
project in 24 days?
Duration distribution of the path A-C-D-G-I-J :
T=19, o
2
1
= 2u
z =
X-
c
=
24-19
20
= 1.12
With z=1.12, we found that the probability of
completing the project within 24 days is
approximately 0.87.
9
A M B
A 1 2 3 0.11 2
B 5 8 11 1.00 8
C 2 3 4 0.11 3
D 1 2 3 0.11 2
E 6 9 18 4.00 10
F 2 4 6 0.44 4
G 1 3 11 2.78 4
H 1 1 1 0.00 1
I 2 2 8 1.00 3
J 2 2 2 0.00 2
Activity Variance
Expected
duration
Time estimates
Outline
Activity crashing
Monitoring projects
10
As the project is implemented, the project network is
subject to review and possible revision, which may be
needed because some activities get off schedule or
resourcesmaynot beavailablewhenneeded.
The second concern in using network models is eliciting
timeestimatesfor activities.
Another problemisthat of biasintroducedintoestimatesof
activitydurations.
11
Outline
Activity crashing
Monitoring projects
12
Monitoring projects
Three steps in monitoring projects:
Project objective setting
Performance evaluation
Deviation correction
Earned value chart
Time variance =STWP ATWP
Cost variance =BCWP ACWP
Schedule variance =BCWP BCWS
13
STWP: Scheduled time for work performed
ATWP: Actual time for work performed
BCWP: Budgeted cost for work performed
ACWP: Actual cost for work performed
BCWP: Budgeted cost of work performed
BCWS: Budgeted cost of work scheduled to
be performed to date
Monitoring projects
Project Termination
Extinction: successful completed or killed
Addition: Successful project becomes institutionalized as
part of parent organization.
Integration: Successful project is dismantled and distributed
throughout parent organization.
Starvation: slow death by budget cuts.
14
Monitoring projects
Project history report
Project performance
Administrative performance
Organizational structure
Project and administrative teams
Techniques of project management
15
Review
A project
Planned set of one-time, interrelated tasks to be executed
over a fixed period and within certain cost and other
limitations.
Characteristics of projects
1) One-time
2) Objectives
3) Interdependencies
4) Conflict
16
Gantt chart
Critical Path Method (CPM)
Program Evaluation & Review Technique (PERT)
Review
A network consists of a set of circles called nodes and also a set of
arrows.
The arrows connect the nodes to give a visual presentation of the
sequence of activities.
Activity on node (AON):
the nodes: project activities
the arrows: the activity sequence
Activity on arrow (AOA):
the arrow: project activities
the nodes: events, which are the starts or completions of activities
18
Activity
(time)
ii jj
A(t) A(t)
B(t) B(t)
Review
Critical path method:
The critical path method (CPM) is an approach to determine the
start and finish dates for individual activities in a project.
A result of this method is the identification of a critical path.
Procedures:
Draw the network
Provides activity information
Earliest (ES) & latest (LS) start, Earliest (EF) & latest (LF) finish
Slack (S): Allowable delay
Identifies critical path
Longest path in network
Shortest time project can be completed
Activities have 0 slack
Any delay on activities delays project
Review
20
Item Symbol Definition
Expected activity duration t The expected duration of an activity
Early start ES
The earliest time an activity can begin if all previous
acivities began at their earliest times
Early finish EF
The earliest time an activity can be completed if it is
started at its early start time
Late start LS
The latest time an activity can begin without delaying
the completion of the project
Late finish LF
The latest time an activity can be completed if it is
started at its latest start time
Total slack TS
The amount of time an activity can be delayed without
delaying the completion of the project
Table 10.2 Notation for critical path method
ES = EF
predecessor
(max)
EF = ES + t
LF = LS
successor
(min)
LS = LF - t
TS = LF - EF or
TS = LS - ES
Review
Beginning with the bottom, movable activities
Move the farthest steps
21
Immediate
predecessor
Estimated
duration (days)
Tournament J H,I 2
ID Activity Days
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
C Plan promotion 3
J Tournament 2
2 2 2 2 2 3 3 3 3 3 3 2 1 1 1 2 1 1 1 1
Personnel required
Review
The crashing procedure can be summarized as
1) Calculate the expedite-cost for each activity;
2) List all the paths in the project network and their normal duration;
3) Crash by 1 day the least costly activity on the critical path or the
least costly combination of activities on common critical paths.
Record the cost of the crashed schedule;
4) Update the duration for each path in the project network;
5) If an activity has reached its crash time, note with an asterisk and
do not consider it as a further candidate;
6) If a critical path contains activities all noted with asterisk, STOP;
otherwise, GO to 3.
22
Review
A typical Beta distribution is commonly used to describe the
duration of uncertain activities.
Optimistic time (A): the duration of an activity if no
complications or problems occur.
Most likely time (M): the duration is most likely to occur.
Pessimistic time (B): the duration of an activity if
extraordinary problems arise.
23
Expected value:
t =
A +4H +B
6
Standard Variance:
o =
B -A
6
Review
Project completion time distribution
1) For every activity, obtain estimates of A, M, B.
2) Use t=(A+4M+B)/6 to calculate the expected activity durations,
and perform critical path analysis using the expected activity
durations t.
3) The expected project completion time T is assumed to be the
sum of the expected durations of activities on critical path.
4) The variance of project completion time o
2
1
is assumed to be
the sum of the variances of activities on the critical path.
5) Probabilities regarding project completion time can be
determined from standard normal tables.
24
Review
Homework 1
A consulting firm is planning a reengineering project for a client. The
following activities and time estimates have been identified
Draw a project network diagram.
Calculate the scheduling times and total slack for each activity.
List the critical path activities and project duration.
Assuming that one worker is required for each activity, prepare a
resource-leveled schedule. What is the maximum number of workers
required to finish the project on time?
Activity Time, days Immediate precedecessor
A 1 -
B 2 -
C 2 -
D 2 A,B
E 4 A,C
F 1 C
G 4 D
H 8 G,E,F
Homework 2
Crash the completion time to the minimum level according to the
information in the following table.
Activity Time, weeks Immediate precedecessor Cost Crash time, weeks Cost
A 4 - 2500 2 6000
B 5 - 4000 4 5000
C 2 A 3000 1 5000
D 2 A 2000 1 3000
E 3 A 3000 4 4000
F 6 B,C 2000 1 5000
G 1 D 2000 1 2000
Homework 3
You have been asked to plan the following convert operation for the AIC:
Draw a network diagram for this project.
Calculate the expected time and variance for each activity
Determine the critical path and the expected project completion time.
What is the probability of the project taking more than 25 days to
complete?
Optimistic Most likely Pessimistic
A - 1 2 3
B A 3 3 3
C B 4 6 8
D A 2 8 8
E A 6 9 12
F D 4 7 10
G C 10 10 16
H D,E 4 5 6
I F,G,H 2 2 2
Activity Immediate precedecessor
Activity duration,days
The End!
Dr. Na GENG
Prof. Zhibin JIANG
Advancement in Production
Planning and Control
Content
Optimized Production Planning
Theory of Constraint (TOC)
Advanced Planning System (APS)
Mass Customization and its Production Planning
2
The most prevalent approach in the production planning is
based on the concept of material requirement planning
(MRP).
The release time is obtained by shifting the expected
output time back along the time scale by a period of the
estimated average lead time;
The release quantity is derived by dividing the expected
output by the estimated average product yield.
However, MRP-based methods have three major drawbacks:
The lead time not only needs to be pre-specified but also is
assumed to be static over the entire planning horizon;
The capacity is assumed to be infinite, which means the
derived production planning may not be realized;
The production system is made nervous. Little adjustment in
MPS changes the due date, requiring the recalculation of
MRP.
Optimized Production Planning -Introduction
New methods need to be developed for production planning
based on mathematical programming
Time Dimension
Space Dimension:
Corporate level planning: production planning
Shop floor level planning: production lot planning
Mathematical programming based optimized Production
Planning Commonly used in production planning:
Linear Programming (LP)- the most widely used methods;
Stochastic Programming (SP)-coping with the uncertainty.
Optimized Production Planning - LP
1
(or ) ( ), ( , )
. . ( ) 0, 1,2,
= =
s =

T
n
i
min max z f x x x x
s t g x i m

Common LP model:
x: Decision variables;
f(x): Objective function;
g
i
(x)s0: constraints.
Commonly used terms:
Objective function, constraints, right-hand side, feasible
region, feasible solution, optimal solution.
Features of LP
Linearity: the objective and all constraints can be expressed
as a linear function of the decisions variables;
Continuity: the decision variables should be continuous
Optimized Production Planning -LP
Example: Make the production planning of milk product
One barrel of milk can be made into 3kg A
1
by 12 hours or 4kg
A
2
by 8 hours. The profit of A
1
and A
2
are $24/kg and $16/kg,
respectively. The supply of raw material, milk, is 50 barrels per
day. Capacity is 480 hours per day and the production limit of A
1
is 100kg at most.
50 barrels of milk/ day, 480 hours available/day, and 100kg A
1
at most
1 barrel
of milk
12 hours
8 hours
3 kg A
1
4 kg A
2
Profit: $16/kg
Profit: $24/kg
Decision variables
x
1:
barrels of milk
to produce A
1
x
2:
barrels of milk to
produce A
2
Constraints
Raw material: 50
2 1
s + x x
Work hours:
480 8 12
2 1
s + x x
Requirement
constraints
100 3
1
s x
0 ,
2 1
> x x
Objective function Total profit/day:
1 2
24*3 16*4 Max z x x = +
Model Analysis:
Features of LP: Linearity and Continuity:
Xi is continuous:
The barrels of
milk is real
number
Proportion: The contributions of
x
i
to objective function and
constraints are separately
proportional to x
i
.
Addition: The contributions of x
i
to objective function and
constraints are separately
independent of x
j
.
The profit/ kg of A
1
, A
2
is
constant, and the production
quantity and time of A
1
,A
2
from one barrel of milk are
constant.
The profit/ kg of A
1
, A
2
is
constant, and the production
quantity and time of A
1
, A
2
from one barrel of milk are
constant.
Conventional LP based production planning:
A fixed upper bound is placed on production, assuming
instantaneousproductionregardlessof WIPlevel.
Notations:
i: Product index
t: Periodindex
X
it
: Amount of product i producedinperiodt
W
it
: Amount of WIPof product i at theendof periodt
R
it
: Amount of releasefor product i at thebeginningof periodt
I
it
: Amount of inventoryfor product i at theendof periodt
B
it
: Amount of backordersfor product i at theendof periodt
Notations:
i
: Unit productioncost for product i
e
i
: Unit WIPholdingcost for of product i
i
: Unit Releasecost for product i
t
i
: Unit inventoryholdingcost for product i
|
i
: Unit backorder cost for product i
d
it
: Demandof product i inperiodt
c
i
: Unit processingtimefor product i
C: Aggregatecapacityof themachineinunitsof time
Min z =
+p
X
t
+n
I
t
+[
B
t
,t
S.t.
I
t-1
+X
t
+B
t
-B
t-1
-I
t
= J
t
foi ony i, t
e
X
t
C for anyt
I
t
, X
t
, B
t
u for anyi, t
Objective: Min (Production cost + Inventory cost +
Backorder cost)
Constraint 1: inventory balance
Constraint 2: capacity constraint
Improved NLP based production planning:
Capacityisamuchmorecomplexconstruct.
Theactual output that canbeachievedduringexecutionishighly
dependent on operational dynamics, e.g. WIP, set-ups, routings,
dynamicallychangingbottleneck, etc.
These dynamics, in turn, depend on planning decisions such as
product mix, thetimingandquantity of releases, batchsizes and
sequencing.
Tosolvethisproblem, therearetomainstreams:
Capacity parameters are estimated and progressively
correctedby iteratingbetweena deterministic optimization
model and a simulation.
A capacity function akin to an operating curve that
explicitlyrepresentsthecircularityphenomenonistriedtobe
developed off-line and used in the production planning
model. Clearingfunctionisthemost successful study.
Clearing function:
IE =
C wIP
H +wIP
IE = C 1 -c
-MwIP
Where, TH is throughput,
C: the maximum possible output in a period
M: a user-specified parameter controlling the curvature of
the clearing function
Minz =
+p
X
t
+n
I
t
+[
B
t
,t
S.t.
I
t-1
+X
t
+B
t
-B
t-1
-I
t
= J
t
foi ony i, t
e
X
t
IE
t
for anyt
w
t
= w
t-1
-X
t
+R
t
, for anyi, t,
I
t
, X
t
, B
t
u for anyi, t
Objective: Min (Production cost + Inventory cost +
Backorder cost)
Constraint 1: inventory balance
Constraint 2: capacity constraint
Constraint 3: WIP balance
Optimized Production Planning -Stochastic
Programming (SP)
SP is a framework for
modeling optimization
problems that involve
uncertainty;
The most widely applied and
studied stochastic
programming models are
two-stage linear programs;
Multi-stage linear programs
have been extended,
SP can tackle the uncertainty
of future demand.
The decision makers take some
actions at the first stage; A
recourse decision can then be
made in the second stage that
compensates for any bad effects
that might have been
experienced as a result of the
first-stage decision.
Each stage consists of a decision
followed by a set of observations
of the uncertain parameters
which are gradually revealed
over time.
Optimized Production Planning -SP
Example (SP): Make the production planning of milk product
Onebarrel of milk canbemadeinto3kgA
1
by 12hours or 4kgA
2
by 8 hours. The profit of A
1
and A
2
are $24/kg and $16/kg,
respectively. The supply of raw material, milk, is 50 barrels per
day. Capacity is 480 hours per day. Suppose the unit inventory
holdingcost andbackorder cost are1. Thedemands for A
1
andA
2
inthenext 3daysare:
1 2 3 1 2 3 1 2 3
A1 60 90 120 75 75 105 90 120 135
A2 120 100 140 108 120 120 100 140 200
Product\Period
Probability=0.3 Probability=0.5 Probability=0.2
Scenario 1 Scenario 2 Scenario 3
( ) ( )
1 2 1 2 1 2
24*3 16*4
t t s st st s st st
t t s t s t
Max z x x p I I p B B = + + +

Decision variables:
x
1t
: barrels of milk to produce A
1
in period t
x
2t
: barrels of milk to produce A
2
in period t
I
1st
: inventory of product A
1
in scenario s at the end of period t
I
2st
: inventory of product A
2
in scenario s at the end of period t
B
1st
: Backorder number of product A
1
in scenario s in period t
B
2st
: Backorder number of product A
2
in scenario s in period t
Constraints
Raw material: x
1t
+x
2t
Su
Working hours: 12x
1t
+8x
2t
48u
Demand balance:
I
1st-1
+Sx
1t
+B
1st
-B
1st-1
= J
1st
I
2st-1
+Sx
2t
+B
2st
-B
2st-1
= J
2st
x
1t
, x
2t
, I
1st
, I
2st
, B
1st
, B
2st
>=0
In-class discussion
What the difference between the two-stage SP and
theLP model withtheaveragedemandconsidered?
If we use the LP model and different demand
scenarios to deriveseveral first-stagedecisions, are
theaveragedecisions thesamewiththosefromtwo-
stageSP model?
The End!
Dr. Na GENG
Prof. Zhibin JIANG
Content
2
Theory of Constraints (TOC)-Introduction
The theory was first described by Israels physicist Dr.
EliyahuM. Goldratt in his book The Goal as a way of
managing the business to increase profits in 1980s .
TOC is a proven method that can be
used by existing personnel to increase
throughput (sales), reliability, and quality
while decreasing inventory, WIP, late
deliveries, and overtime.
Successful organizations also adopt
TOC to help make tactical & strategic
decisions for continuous improvement.
The Theory of Constraints is based on the premise that:
Every real system, such as a business, must have within it at
least one constraint. If this were not the case then the system
could produce unlimited amounts of whatever it was striving for,
profit in the case of a business..
Dr Eli Goldratt
TOC-Introduction
TOC-Drum Buffer Rope
DrumBuffer Rope (DBR) is a planning and scheduling solution derived
fromtheTheoryof Constraints(TOC).
Thefundamental assumptionof DBR isthat withinany plant thereisoneor
alimitednumber of scarceresourceswhichcontrol theoverall output of that
plant. Thisisthedrum, whichsetsthepacefor all other resources.
In order to maximize the output of the system, planning and execution
behaviors are focused on exploiting the drum, protecting it against
disruption through the use of time buffers, and synchronizing or
subordinating all other resources and decisions to the activity of the drum
throughamechanismthat isakintoarope.
buffer
drum
rope
TOC-the Steps for Implementation
Step1: Identifythesystem'sconstraint(s)
Step 2: Decide how to exploit the systems
constraint(s)
Step 3: Subordinate everything else to the
abovedecision
Step4: Elevatethesystemsconstraint(s)
Step5: If intheprevious step, aconstraint has
been broken go back to step 1, but do not
allowinertiatobecomethesystemsconstraint
What to change? What to change?
To What to change? To What to change?
How to cause the
change?
How to cause the
change?
TOC-Types of Constraint
A constraint is anythinginanorganizationthat limits it from
movingtowardor achievingitsgoal.
Therearetwo basic types of constraints: physical constraints
andnon-physical constraints.
A physical constraint is somethinglikethephysical capacity
of amachine.
A non-physical constraint might besomething likedemand
for a product, a corporate procedure, or an individual's
paradigmfor lookingat theworld.
The market, capacity, resources, suppliers, finance,
knowledgeor competence, policy
TOC- Thinking process
Reality Tree
To identify the current reality
Leaves are the unsatisfied
phenomenon, and the roots are the
root cause.
Current Reality Tree, CRT
To answer what to change
Identify the undesirable effects
Evaporating CloudEC (conflict
resolution diagram )
Win-Win to solve the conflicts in
the company
Identify the break through point
An Example of CRT
An Example of EC
TOC- Thinking process
Future Reality, FRT
Identify to what to change
Negative Effect Branches
Trimming the negative branches
Prerequisite Tree
How to realize the improvement
Transition Tree
Improve
TOC-Applications
1. Production Planning and Scheduling
2. Distribution and Supply Chain
3. Financial Management
4. Marketing
5. Strategic Planning
6. Project Management
TOC-the Principles of Applying TOC in Production
Scheduling
Factory production rate is production rate of bottleneck work
center;
Most of the buffer WIP should be waiting at bottleneck;
Bottleneck implies certain amount of idle time at other work
stations;
Important to regulate bottleneck workload; other work centers
should serve the bottleneck, not optimize themselves.
TOC-Capacity and Bottlenecks in Production
Capacity is defined as the available time for production
(excludingmaintenanceandother downtime)
A bottleneck (constraint) is defined as any resource whose
capacityislessthanthedemandplaceduponit
A non-bottleneck is aresourcewhosecapacity is greater than
thedemandplacedonit
A capacity-constrained resource (CCR) is one whose
utilizationis closeto capacity andcouldbebottleneck if it is
not scheduledcarefully
TOC- An Example in Semiconductor Manufacturing
SLIM (Short Cycle Time and Low Inventory in Manufacturing) is a project
carried by Prof. Leachmanat University of California at Berkeley.
SLIM is a set of methodologies and scheduling applications for managing
cycle time in semiconductor and its main method is TOC.
Between 1996 and 1999, Samsung Electronics Corp., Ltd., implemented
SLIM in all its semiconductor manufacturing facilities and achieve great
success:
During the presentation at the 2001 Franz Edelman Award Competition,
Yoon-Woo Lee, president of Samsungs semiconductor business, stated:
The financial impact of SLIM was significant. We increased revenue
almost $1 billion dollars through five years without any additional capital
investment, and our global DRAM market share increased from 18 percent
to 22 percent.
Thepiperepresents theproduction line, and its width represents themaximum
flow rateor capacity at various process steps. SEC fabs were designed so that
the photo machines are the bottlenecks, and other machines have surplus
capacity. Thus the pipe in the figure narrows at each photo step. When all the
machines are up and the process is in control, the photo machines are the
bottlenecks, andthelargest concentrationsof WIPisat thosepoints.
Fab Process
Photo
Layer j-1
Photo
Layer j
Photo
Layer j+1
Photo
Layer j+2
WIP
level
WIP between photo steps at normal time
Photo
Layer j-1
Photo
Layer j
Photo
Layer j+1
Photo
Layer j+2
Fab Process
WIP
level
Photo Layer j+1running out
Process trouble
The case of process or equipment trouble at a non-photo manufacturing area of the
fab is depicted by a constriction of the pipe at a non-photo step. If this condition
persists, the WIP at a photo machine can become exhausted. A buffer is needed,
proportional to the risk of trouble in that layer. For example, if Layer j of the
process experiences more trouble than Layer j+1 the bottleneck step immediately
following Layer j should be awarded a larger buffer than the bottleneck step
following Layer j+1.
WIP between photo steps at disturbance
Thephilosophy underlyingSLIM is to distributeWIP to
put the fab in the best position to cope with the next
disruption. That is, whiletheequipment and process are
workingwell, thefabshouldstrivetomoveasmuchWIP
aspossibletothephotobottleneck, tobepreparedfor the
next disruption.
The way to solve the problem
Content
17
APS-Overview of Planning and Scheduling
Generally speaking, planning and scheduling jointly determine
how, when, andinwhat quantity productswill bemanufacturedor
purchased.
In essence, planning establishes what should be done and
schedulingdetermineshowtodoit;
Thereisnoagreeddefinitionof planningversusscheduling.
Many believethat theright andonly way toachieveaccurate
duedatesistoperformverydetailedscheduling.
Others believe that it is much more important to put more
effort intheplanningprocess.
But most agree that the distinction between planning and
scheduling is thetrade-off of timehorizon versus thelevel of
detail.
APS- Definition
Advanced Planning and Scheduling (APS) is a
software system that uses intelligent analytical
tools to perform finite scheduling and produce
realistic plans.
APS-Overview (1)
Its most important advantage over traditional planning
approaches is that material and capacity are simultaneously
considered as elements that may constrain production. This is
different fromthe conventional MRP approach of independently
planning material and then subsequently checking this plan
against capacitytoidentifyviolations;
APS systems are able to generate plans and schedules very
quickly. AnAPSenginecanbedesignedtoeither look over along
timehorizon(afewmonths) withless details or moredetails over
ashorter period(afewweeks).
APS covers various capabilities such as finite capacity
scheduling or constraint-basedscheduling at shopfloor level;
Quiteintuitiveto say that theAPS systems resolvetheshortfall
of theERPsystemasaplanningtool;
APS-Overview (2)
APS does Advanced Planning
Consider business objectives
Consider the organization of machines and work cells
APS does Advanced Scheduling
Consider plant capacity
Consider business limitations
APS-the Scope (1)
The scope of APS is not limited to factory planning and scheduling,
but has grown rapidly to include the full spectrum of enterprise and
inter-enterprise planning and scheduling functions:
Strategic and long-term planning
Demand planning and forecasting
Sales and operations planning (SOP)
Inventory planning
Supply chain planning (SCP)
Available-to-promise (ATP)
APS-the Scope (2)
Manufacturing planning
Distribution planning
Transportation planning
Production scheduling
Shipment scheduling
Inter-company collaboration
Source: Bermudez, John. Advance Planning and
Scheduling: Is It as Good as It Sounds? The Report on
Supply Chain Management. Advanced Manufacturing
Research, Inc., March, 1998.)
APS- the Four-Part Model
AMRs APS Model
Enterprise Wide
Planning Available to
Supply Promise
Chain
Planning Capable to
Production Promise
Scheduling
Source: Advanced Manufacturing Research, Inc.
APS- Mathematical Technologies
Linear Programming
Genetic Algorithms
Heuristics
Constraint Based Programming (CBP)
Source: Shires, Nigel, (2005). Optimization Techniques
and Their Application to Production Scheduling , White
Paper, Preactor International, published on website:
http://www.preactor.com/whitepapers.asp.
APS-APS/ERP Integration
The comprehensive nature of todays APS algorithms
drives theneed for copious amounts of data--data that
typically reside in an ERP system. This means that
attainingthefull benefitsof APSislargelypredictedon
how well it is integrated with ERP. When done well,
bothsystemsbenefit.
Source: Musselman, K., and Uzsoy, R. (2001), Advanced
Planning and Scheduling for Manufacturing, in Handbook of
Industrial Engineering, 3
rd
Ed., G. Salvendy, Eds., J ohn Wiley
& Sons, New York.
APS/ERP Integration
APS/ERP Integration
Source: Musselman, K., and Uzsoy, R. (2001), Advanced
Planning and Scheduling for Manufacturing, in Handbook of
Industrial Engineering, 3
rd
Ed., G. Salvendy, Eds., John Wiley &
Sons, New York.
APS- Software Providers
Preactor
SAP - Advanced Planner and Optimizer (APO)
Oracle - APS
Manugistics
i2

APS-Preactor APS Screenshot
APS-SAP-APO Screenshot
The End!
Dr. Na GENG
Prof. Zhibin JIANG
Content
2
CS-Introduction
With the increasing competition in the global market, the
manufacturing industry has been facing the challenge of
increasingcustomer value;
More importantly, quality means ensuring customer satisfaction
and enhancing customer value to the extent that customers are
willingtopayfor thegoodsandservices;
A well-accepted practice in both academia and industry is the
exploration of flexibility in modern manufacturing systems to
providequick responseto customers withnewproducts catering
toaparticular spectrumof customer needs;
CS-Introduction
The key to success in the highly competitive manufacturing
enterprise often is the companys ability to design, produce,
andmarket high-quality productswithinashort timeframeand
at apricethat customersarewillingtopay;
Inorder to meet thesepragmatic andhighly competitiveneeds
of todays industries, it is imperative to promote high-value-
addedproductsandservices;
Masscustomizationenhancesprofitabilitythroughasynergyof
increasingcustomer-perceivedvaluesandreducingthecosts of
productionandlogistics.
CS-Introduction
Mass customization is producing goods and services to meet
individual customers needs with near mass production
efficiency;
Mass customization is a new paradigm for industries to
provide products and services that best serve customer needs
while maintaining near-mass production efficiency.
Contradicted two sides:
Mass production
demonstrates an advantage in
high-volume production;
Satisfying each individual
customers needs can often be
translated into higher value,
however, economically not
viable;
CS-Introduction
Economic Implication of Mass Customization
Mass customization is capable of reducing costs and lead time
by accommodating companies to garner economy of scale by
repetitions.
With flexibility and programmability, companies with low to
medium production volume can gain an edge over competitors
by implementing MC;
Mass customization can potentially develop customer loyalty,
propel company growth, and increase market share by widening
the product range.
CS-Introduction
Technical challenges
The essence of mass customization lies in the product and
service providers ability to perceive and capture latent market
niches and subsequently develop technical capabilities to meet
the diverse needs of target customers;
To encapsulate the needs of target customer groups means to
emulate existing or potential competitors in quality, cost, quick
response;
Therefore, the requirements of mass customization depend on
three aspects: time-to-market (quick responsiveness), variety
(customization), and economy of scale (volume production
efficiency);
Successful mass customization depends on a balance of three
elements: features, cost, andschedule.
CS-Introduction
Maximizing reusability
Maximal amounts of repetition are essential to achieve the efficiency of mass
production, as well as efficiencies in sales, marketing, and logistics, which is
attained through maximizing commonality in design, which leads to reusable
tools, equipment, and expertise in subsequent manufacturing;
Customization emphasizes the differentiation among products.
An important step toward to this goal is the development and proliferation of
design repositories that are capable of creating various customized products;
Dynamic stability: a firm can serve the widest range of customers and
changing product demands.
To achieve mass customization, the synergy of commonality and modularity
needs to be tackled and needs to encompass both the physical and process
domains of design.
CS- Introduction
Product platform
The effectiveness of a firms new product generation lies in:
Its ability to create a continuous stream of successful new products over an
extended period of time;
The attractiveness of these products to the target market niches;
The essence of mass customization is to maximize such a match of internal
capabilities with external market needs;
A product platform is impelled to provide the necessary taxonomy for
positioning different products and the underpinning structure describing the
interrelationships between various products with respect to customer
requirements, competition information, and fulfillment processes;
This implicates two aspects:
to represent the entire product portfolio, including both existing products and
proactively anticipated ones, by characterizing various perceived customer
needs, and
to incorporate proven designs, materials, and process technologies;
CS- Introduction
Integrated product life cycle
Mass customization starts from understanding customers individual
requirements and ends with a fulfillment process targeting each particular
customer;
The time-to-market can be achieved by telescoping lead time;
Product realization should simultaneously satisfy various product life cycle
concerns, including functionality, cost, schedule, reliability, manufacturability,
marketability, and serviceability, to name but a few;
The realization of mass customization requires not only integration across the
product development horizon, but also the provision of a context-coherent
integration of various viewpoints of product life cycle.
CS-Design for mass customization
Design for mass customization (DFMC) aims at considering
economies of scope and scale at the early design stage of the
product-realization process;
The main emphasis of DFMC is on elevating the current practice
of designing individual products to designing product families;
There two basic concepts underpinning DFMC:
Product family architecture;
Product family design.
CS-Understanding DFMC
CS-Product family
A product family is a set of products that are derived
from a common platform;
A product family targets a certain market segment,
whereas each product variant is developed to address
a specific set of customer needs of the market
segment;
The interpretation of product families depends on
different perspectives.
CS-Modularity and commonality
There two basic issues associated with product families: modularity and
commonality;
Modularity tries to separate a system into independent parts or modules
that can be treated as logical units;
Decomposition is a major concern in modularity analysis;
Modularity is achieved from multiple viewpoints, including
functionality, solution technologies, and physical structures;
There are three types of modularity involved in product realization:
functional modularity, technical modularity, and physical modularity;
The interaction between modules is important in characterizing
modularity.
As for functional modularity, the interaction is exhibited by the
relevance of functional features (FFs) across different customer groups;
CS-Modularity and commonality (Continued)
The commonality reveals the difference of the architecture of
product families from the architecture of a single product;
Corresponding to the three types of modularity, there are
three types of commonality in accordance with functional,
design, and process views;
Functional commonality manifests itself through functional
classification, that is, grouping similar customer requirements
into one class, where similarity is measured by the Euclidean
distance among FF instances;
A class of products is described by modularity and product
variants differentiate according to the commonality among
module instances.
CS-A comparison of modularity and commonality
Issues Modularity Commonality
Focused objects Type (class) Instances (members)
Characteristic of measure Interaction Similarity
Analysis method Decomposition Clustering
Product differentiation Product structure Product variants
Integration/relation Class-member relationship
CS-Product variety
Product variety is defined as the diversity of products that a manufacturing
enterprise provides to the marketplace;
Two types of variety can be observed: functional varietyand technical
variety;
Functional varietyis used broadly to mean any differentiation in the
attributes related to a products functionality from which the customer
could derive certain benefits;
Technical varietyrefers to diverse technologies, design methods,
manufacturing processes, components and/or assemblies, and so on that are
necessary to achieve specific functionality of a product required by the
customer. It may be invisible to customers;
While functional varietyis mostly related to customer satisfaction from the
marketing/sales perspective, technical varietyusually involves
manufacturability and costs from the engineering perspective.
CS-Product variety (Continued)
These two types of variety result in two different variety
design strategies: functional variety strategyand design for
functional variety strategy;
Since functional variety directly affects customer
satisfaction, this type of variety should be encouraged in
product development;
A design for functional variety strategy aims at increasing
functional variety and manifests itself through vast research
in the business community, such as product line structuring,
equilibrium pricing, and product positioning;
A design for technical variety tries to reduce technical
variety so as to gain cost advantages.
CS-Variety Leverage: Handling Variety for
Mass Customization
CS-Product family architecture
A well-planned product family architecture
(PFA) provides a generic umbrella for capturing
and utilizing commonality, within each new
product instantiated and extends so as to anchor
future designs to a common product line
structure.
CS-PFA and Its Relationships with Market Segments
CS-Composition of PFA
The PFA consists of three elements: the common base, the
differentiation enabler, and the configuration mechanism;
Common bases (CBs) are the shared elements among different
products in a product family;
Differentiation enablers (DEs) are basic elements making
products different from one another. They are the source of
variety within a product family;
Configuration mechanisms (CMs) define the rules and means of
deriving product variants. Three types of configuration
mechanisms can be identified: selection constraints, include
conditions, and variety generation.
CS-Composition of PFA (Continued)
Selection constraints specify restrictions on optional features
because certain combinations of options are not allowed or
feasible or, on the contrary, are mandatory;
Include conditions are concerned with the determination of
alternative variants for each differentiation enabler. The
include condition of a variant defines the condition under
which the variant should be used or not used with respect to
achieving the required product characteristics;
Variety generation refers to the way in which the
distinctiveness of product features can be created. It focuses
on the engineering realization of custom products in the form
of product structures.
CS-Basic Methods of Variety Generation
CS-Synchronization of multiple views
The strategy is to employ a generic, unified
representation and to use its fragments for different
purposes, rather than to maintain consistency among
multiple representations through transformation of
different product data models to standard ones.
CS-Representing Multiple Views of Product
Family within a Single Context
CS-Product family design
Under the umbrella of PFA, product family
design manifests itself through the derivation
processes of product variants based on PFA
constructs.
PFA-Based Product Family Design: Variant Derivation
through GPS Instantiation
CS- PFA-Based Product Family Design: Variant Derivation
through GPS Instantiation
CS- Manufacturing and production planning
Competition for mass customization manufacturing is focused on the
flexibility and responsiveness in order to satisfy dynamic changes of global
markets. The future major trends are:
A major part of manufacturing will gradually shift from mass production
to the manufacturing of semi-customized or customized products to meet
increasingly diverse demands;
The made-in-house mindset will gradually shift to distributed locations,
and various entities will team up with others to utilize special capabilities
at different locations to speed up product development, reduce risk, and
penetrate local markets;
Centralized control of various entities with different objectives, locations,
and cultures is almost out of the question now. Control systems to enable
effective coordination among distributed entities have become critical to
modern manufacturing systems.
CS-Managing variety in production planning
Major challenge of mass customization production planning results from the
increase of variety;
Facing such a variety dilemma, many companies try to satisfy demands from
their customers through engineer-to-order, make-to-order, or assembly-to-
order production systems;
The traditional approach to variant handling is to treat every variant as a
separate product by specifying a unique BOM for each variant. This works
with a low number of variants, but not when customers are granted a high
degree of freedom in specifying products. The problem is that a large number
of BOM structures will occur in mass customization production;
To overcome these limitations, a generic BOM (GBOM) concept has been
developed.
The GBOM provides a means of describing, with a limited amount of data, a
large number of variants within a product family, while leaving the product
structure unimpaired. The structure has three aspects:
CS-A Generic Structure for Characterizing Variety
CS-The generic variety structure for souvenir clocks
Structure items
{I
i
}
Variety parameter
{P
j
}
Variety instance {V
j
}
3/hands
Setting type Two-hand setting, three-hand setting
Color White, Grey, etc.
Size Large, medium, small
3/dial
pattern Logo, mosic, scenery, customized photo, etc.
Size Large, medium, small
4/transmission Alarm Yes, no
4/core Alarm Yes, no
3/base
Shape Round, rectangular, hexagonal
Material Acrylic, aluminum, etc.
Color Transparent, red, etc.
3/front plate
Shape Rectangular,round, rhombus
Material Acrylic, aluminum, etc.
Color Transparent, red, etc.
1/label sticker Pattern HKUST, signature, etc.
1/paper box Type Ordinary, deluxe, etc.
CS-The generic variety structure for souvenir clocks
(continued)
To understand the generic concept
underlying such a variety representation,
two fundamental issues need to be
highlighted:
Generic item;
Indirect identification.
CS-Coordination in manufacturing resource allocation
Challenges of manufacturing resource allocation for mass
customization include:
1) The number of product variety flowing through the
manufacturing system is approaching an astronomical scale;
2) Production forecasting for each line item and its patterns is not
often available;
3) Systems must be capable of rapid response to market
fluctuation;
4) The system should be easy for reconfiguration ---- ideally, one
set of codes employed across different agents;
5) The addition and removal of resources or jobs can be done with
little change of scheduling systems.
CS-Major considerations of scheduling for resource
allocation
Decompose large, complex scheduling problem into smaller,
disjointed allocation problems;
Decentralize resource access, allocation, and control mechanisms;
Design a reliable, fault-tolerant, and robust allocation mechanism;
Design scalable architectures for resource access in a complex
system and provide a plug-and-play resource environment such that
resource providers and consumers can enter or depart from the
market freely;
Provide guarantees to customers and applications on performance
criteria.
CS-A Market Structure for Collaborative Scheduling
CS-A Market Structure for Collaborative
Scheduling (Continued)
The satisfaction of multiple criteria, such as costs
and responsiveness, cannot be achieved using solely
a set of dispatching rules;
A price mechanism should be constructed to serve
as an invisible hand to guide the coordination in
balancing diverse requirements and maximizing
performance in a dynamic environment.
CS-The Price Mechanism of a Market Model
CS-Message-Based Bidding and Dynamic Control
CS-High-variety shop-floor control
The requirements of the new control systems
include re-configurability, decomposability,
and scalability to achieve make-to-order with
very short response time.
The End!

0生产计划与控制

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

0生产计划与控制

Uploaded by

Copyright:

Available Formats

1896 1920 1987 2006

Production Planning and Control

Total Cost for buy x units parts: c

as the sum of the squares of the distances from line a+bX to

We may choose a and b to minimize

2.4. Evaluating Forecast

Fig.2-5 Weights in Exponential Smoothing

2.8. Winters Method for Seasonal Problems

2.8. Winters Method for Seasonal Problems

2.4. Evaluating Forecast

(b) Lower inventory level

Relevant Costs- Order cost

Example 4.4: If Weighty uses trash bags at a fairly constant rate

Quantity Discount Model- Optimal Policy for All-

Fig. 4-12 All-Units Discount Average

The order cost function follows

Note that ( ) 1 , the condition is equivalent to

EOQ models for production planning

Optimal lot sizes for each production run

When the predictable variation is more important than

Unbalanced TPDemand > Supply

Suppose that link customer i and

k = No. of kanbans in card set

The nature of project management

- C= activity cost (* for crash cost)

You might also like