MANUFACTURING

MODELS

Manufacturing models
In manufacturing models:
Resource is called a machine
Task is called as job
A job may be a single operation or a collection of
operations to be done in several different machines.
There are five classes of manufacturing models:
Project planning and scheduling
Job shop models
Flexible manufacturing systems
Lot scheduling
Supply-chain models
27

1
Project planning and scheduling
Project scheduling is important for large projects.
A large project consists of a number of jobs with
precedence constraints.
Examples:
construction of an aircraft
large consulting project
Goal: minimize completion time of last job (makespan)
The critical path (set of jobs that determine the
makespan) can be identified.

28

Job shop models

Job shop scheduling (include single machine and
parallel machine models).
Jobs Machine Sequence
1 1, 2, 3
2 2, 1, 4, 3
3 1, 2, 4

Minimize makespan or the number of late jobs.

Mostly for make-to-order manufacturing systems.
Also in services.

29

2
Flexible manufacturing systems
Production systems with automated material handling.
Material handling or conveyor system controls the
movement of jobs and timing of their processing.
Mostly for mass production systems.
Maximize throughput.
Examples:
automotive industry
consumer electronics industry

30

Lot scheduling
For medium and long term production planning.
Processes are continuous.
Switching between products incurs a setup cost.
Minimize total inventory and setup costs.
Examples:
process industries, e.g. oil refineries
paper mills

31

3
Supply-chain models
In general, are an integration of job-shop and lot
scheduling, including transportation costs.

Material Flow Information Flow

32

Manufacturing models revisited

Discrete models: project scheduling, job shop or
flexible assembly systems.
Formulated as an integer programming or disjunctive
programming.

Continuous models: lot scheduling.

Formulated as a linear or nonlinear programming

33

4
Jobs and machines
machines i = 1,,m
jobs j = 1,,n
Static data:
Processing time pij processing time of job j on machine i
Qj = 1/pj , machines production rate of job j
Release date rj earliest time at which job j can start its
processing
Due date dj committed shipping or completion date of
job j
Weight wj importance of job j relative to the other jobs in
the system

34

Modeling parameters
Dynamic data:
Starting time Sij time when job j starts its processing
on machine i
Completion time Cij time when job j is completed on
machine i

Model representation:
machine configuration | characteristics | objectives
| |

35

5
Completion time

Job 2
Resources
r2 = 7
d2 = 11
1 1 4 2 3 w2 = 4 (?)

2 4 1 2 3
i Si2 pi2 Ci2

1 8 2 10
3 3 4 1 2
2 7 1 8

3 10 1 11
Time
C4 C1 C2, C3

36

Completion time

Job 3
Resources
r3 = 0
d3 = 11
1 1 4 2 3 w3 = 4 (?)

2 4 1 2 3
i Si3 pi3 Ci3

1 10 1 11
3 3 4 1 2
2 8 2 10

3 0 2 2
Time
C4 C1 C2, C3

37

6
Machine environment ( )
Single machine 1
Identical machines in parallel Pm
Job j can be processed in any of the m machines
Machines in parallel with different speeds Qm
Unrelated machines in parallel Rm
Open shop Om
Jobs have to be processed in all m machines but in some
the processing time may be zero

38

Machine environment ( )
Flow shop Fm
m machines in series and each job has to be processed
on each of the m machines
Flexible Flow shop FFc
Generalization of Fm and Pm
c stages in series with a number of parallel machines in
each stage

39

7
Machine environment ( )
Job shop Jm
m machines and each job has its own predetermined
route to follow
If job may visit a machine more than once, it has
recirculation
Flexible job shop FJc
Generalization of Jm and Pm
c work centers with a number of parallel machines in
each work center

40

Constraints ( )
Release date rj
job j may not be processed before rj
Sequence dependent setup times sjk
Setup time between job j and job k
Preemptions prmp
It is possible to interrupt job j
Recirculation recrc
In FJc, when a job visits a machine more than once

41

8
Objective functions ( )
Lateness Lj Lj Cj dj

Tardiness Tj Tj max( L j ,0)

Makespan Cmax C max max(C1 , ,C j )

Maximum lateness Lmax Lmax max( L1 , , Lj )

Total weighted tardiness jTj

42

Dispatching solutions
Scheduling solutions are found using heuristics:
Algorithm that is able to produce a simple and
acceptable solution
Usually it provides a fast good solution, but not the best
one (optimal solution)

43

9
Single machine
Single machine:

When there is a single bottleneck in a multi-machine

environment, that bottleneck is scheduled first.
Earliest Due Date (EDD) orders the jobs in increasing
order of their due dates.
Minimize maximum lateness among all jobs
Short Processing Time first (SPT) minimize the
average number of jobs waiting for processing.

44

Exercise single machine

45

10
Solution
a) Brut force:
P4 = 4! = 24 possible different sequences
Consider Job Sequence #1: 1 2 3 4
Job 1, w = 6 Job 2, w = 11 Job 3, w = 9 Job 4, w = 5
3 8 15 19

Total Cost of Sequence #1 is given by

JJS#1 = 6 x 3 + 11 x 8 + 9 x 15 + 5 x 19 = 336

Follow the same procedure for the remaining and find

out the optimal sequence such that the cost is minimal

46

Solution
b) Higher Priority (HP) first
Job Sequence HP: 2 3 1 4

Job 2, w = 11 Job 3, w = 9 Job 1, w = 6 Job 4, w = 5

5 12 15 19

Total Cost of Sequence HP is given by

JHP = 11 x 5 + 9 x 12 + 6 x 15 + 5 x 19 = 348

Worst than previous case!

47

11
Solution
c) Shortest Processing Time (SPT) first
Job Sequence SPT: 1 4 2 3

Job 1, w = 6 Job 4, w = 5 Job 2, w = 11 Job 3, w = 9

3 7 12 19

Total Cost of Sequence SPT is given by

JSPT = 6 x 3 + 4 x 7 + 11 x 12 + 9 x 19 = 349

Worst than previous cases!

48

Parallel machine

Machines do not have to be identical.

Machines can be old or new, or can be people.

49

12
Exercise parallel machine

50

Solution
a) Considering combinations of all possible different
sequences , gives C511 = 462

b) Longest Processing Time (LPT) first: whenever a

machine is free, the next job with the longest
processing time is assigned. Reasoning: it is easier to
distribute short processing time jobs in the long run.

51

13
Solution
c) Possible solution
M1: Job 1, pj = 9 M1: Job 9, pj = 5 M1: Job 11, pj = 5
M2: Job 2, pj = 9 M2: Job 10, pj = 5
M3: Job 3, pj = 8 M3: Job 7, pj = 6
M4: Job 4, pj = 8 M4: Job 8, pj = 6
M5: Job 5, pj = 7 M5: Job 6, pj = 7

CmaxLPT = 19

d) If CmaxOptimal = 15, then CmaxLPT = 19 is worse.

52

Solution
c) Possible Solution pj / #machines = 75 / 5 = 15
:-) Fill in each machine at a time to complete 15 time
units
M1: Job 1, pj = 9 M1: Job 7, pj = 6
M2: Job 2, pj = 9 M2: Job 8, pj = 6
M3: Job 3, pj = 8 M3: Job 5, pj = 7
M4: Job 4, pj = 8 M4: Job 6, pj = 7
M5: Job 9, pj = 5 M5: Job 10, pj = 5 M5: Job 11, pj = 5

CmaxLPT = 15

53

14
Other machine configurations
Flow shop:

Flexible flow shop:

54

Other machine configurations

Job shop: (can have recirculation)

Flexible job shop (several machines in parallel)

Supply- chain

55

15
Processing characteristics
Precedence constraints one or more jobs may have
to be completed before another job is allowed to start
its processing

Machine eligibility constraints in parallel machine

environment, Mj denotes the subset of machines that
can process job j

56

Processing characteristics
Workforce constraints
workforce consists of several pools
each pool consists of operators with a specific skill set
number of operators in pool l is denoted by Wl
Routing constraints specify the operations for each
job and the machines at which these operations must
be processed.
Material handling constraints are fixed for
automated (e.g., robotized) work centers and
adjustable for manual tasks.

57

16
Processing characteristics
Sequence-dependent setup times and costs
length of setup (reconfiguration or cleaning) depends
on jobs
sijk: setup time for processing job k after job j on
machine i
cijk: setup costs, e.g. waste of raw material, labor
Storage space and waiting time constraints amount
of space available for Work-In-Progress (WIP) is
limited.

Completion time Start time Buffer space

58

Processing characteristics
Make-to-Stock and Make-to-Order
Dj demand rate fixed and constant. Items are
produced for inventory and do not have tight dates.
Make-to-Order jobs have specific due dates; amount
produced is determined by the costumer.
Preemptions interrupt the processing of a job to
process another one with higher priority.
Preemptive resume processing done is not lost
Preemptive repeat processing must be completely
repeated
Transportation constraints, etc.

59

17
Performance measures and objectives
Througput is frequently determined by the
bottleneck machines, for which the utilization should
be maximized.
Makespan the time when the last job leaves the
system
Cmax max(C1 , C2 , , Cn )

where Cj is the completion time of job j.

Minimizing makespan tends to maximize throughput
and balance load.

60

Performance measures and objectives

Due date related objectives
Lateness
Lj Cj dj

where dj is the due date of job j.

Maximum lateness (minimize worst performance)

Lmax max( L1 , , Ln )

61

18
Due date related objectives
Tardiness
Tj max(C j d j ,0)

Objective function
n
Tj
j 1
Weighted Tardiness
n
w jT j
j 1

62

Due date related objectives

Lateness Tardiness

Lj Tj

Cj Cj
dj dj

Late or Not In practice

Uj
1
Cj Cj
dj dj

63

19
Performance measures and objectives
Work-In-Process (WIP) inventory costs
Minimizing WIP also minimizes average throughput
(lead) time, which is the time it takes a job to transverse
the system.
Equivalent to minimize the average number of jobs in
the system.
Minimizing average throughput time is closely related
to minimize the sum of completion times:
n n
Cj wjC j
j 1 j 1

64

Other costs and concepts

Setup costs
Finished goods inventory costs
Transportation costs
In Just-In-Time (JIT) concepts, it is important to
minimize the total earliness.
A job should be completed just before its committed
shipping, avoiding inventory and handling costs.
Robustness. A schedule is robust when the necessary
changes in case of disruption (e.g. machine
breakdown, rush order) are minimal.

65

20
 SERVICE MODELS

Introduction
Impossible to store goods
It is not possible to get back the lost time in a hotel
room.
Resource availability (e.g. people, rooms or trucks)
often varies
May even be part of the objective function
Saying no to a customer is common
No available seats on that flight (even if there are
some!)
Try to book a restaurant for 8 or 9 PM

67

21
Reservation systems and timetabling
Reservation systems
A job j has a duration pj and the starting and completion
times are usually fixed in advance.
Example: in a car rental agency, a job is the reservation
of a car for a given period.
Timetabling (rostering)
A job or activity j with a duration pj, which has to be
scheduled in a time window in the interval:
[earliest possible starting time rj, latest possible completion time dj]
Examples: exam scheduling, scheduling operating
rooms

68

Service models
Tournament scheduling and broadcast television
models
Tournament scheduling parallel machine problem,
where all the jobs have the same processing time.
Transportation scheduling
Examples: airlines, railroads, shipping.
Job trip or flight leg; machine ship, plane or vehicle.
Trip k incurs a cost ck and generates a profit k.
Objective: minimize total cost or maximize total profit.

69

22
Workforce scheduling
Shift scheduling in service facilities
Example: call center
Time interval i requires a staffing of bi (integer).
Objective: minimize total cost.
Crew scheduling in transportation.
Depends on the specific tasks to be done
Crew scheduling is often intertwined with other
schedules (e.g. routing and scheduling of planes or
trucks).

70

Activities
Examples of activities:
meetings to be attended by certain people
game to be played by 2 teams
flight leg to be covered by a plane
personnel position to be occupied in a given time period
Data:
duration processing time pij
earliest possible start time release time rj
latest possible finishing time due date dj
priority level weight wj

71

23
Resources
Machines: classroom, hotel, rental car, stadium,
operating room, plane, ship, airport gate, dock, railroad
track, person (nurse/pilot)
Synchronization of resources may be important
Need a plane and a pilot
Classroom, video projector equipment, professor,
students
Characteristics of resources
Classroom: capacity, equipment, cost, accessibility.
Truck: capacity, refrigeration, speed
Person: specialist (surgeon, nurse) with skills (languages)
72

Operational characteristics
Time windows (release dates and due dates)
Capacity requirements and constraints
Preemptions many activities are difficult to preempt
(e.g. operations flight leg or game).
Setup times and turnaround times
Setups between consecutive meetings or trips.
Example: Runway in an airport is a resource, takeoff
and landing are activities. Necessary idle time between
activities.

73

24
Operational characteristics
Operator and tooling requirements
Workforce scheduling constraints
Shift patterns, break requirements
Union and safety rules

74

Objectives
Combination of two types of objectives: one
concerning the timings and other concerning the
utilization of resources.
General objective: minimize the total cost of all the
assignments.
Makespan
cost
Setup costs
Earliness and Tardiness costs
Personnel costs
time
ideal departure time

75

25
Computational complexity
Easy problems:
Sort n numbers
Solve a system of linear equations
Hard problems:
Schedule a factory, deliver packages, schedule buses,

76

Computational complexity
f (n): the number of basic operations needed to
solve the problem with input size n
Easy: f (n) is polynomial in n
O(n), O(n log n), O(n2),
Hard: f (n) is exponential in n
O(2n),

77

26
Hard vs Easy

O(n) O(n log n) O(n2) O(2n)

1 0 1 2

10 10 100 1024

20 26 400

50 85 2500

100 200 10000

1000 3000 1,000,000

78

Hard vs Easy

O(n) O(n log n) O(n2) O(2n)

1 0 1 2

10 10 100 1024

20 26 400 1048576

50 85 2500 1,125,899,906,842,624

100 200 10000

1000 3000 1,000,000

79

27
Hard vs Easy

O(n) O(n log n) O(n2) O(2n)

1 0 1 2

10 10 100 1024

20 26 400 1048576

50 85 2500 1,125,899,906,842,624

100 200 10000 1.268 X 1030

1000 3000 1,000,000 1.072 X 10301

80

Hard vs Easy
10301 operations required in worst case
Age of universe: 1018 seconds
Fastest Computer today: 1014 operations/sec
If it is going to take 10250 times the age of the universe
to schedule a factory?
May be it is possible to do it in a reasonable time in
most cases!
May be a good (but not the best) solution can be
obtained in a reasonable amount of time!

81

28
P and NP problems
The efficiency of an algorithm is measured by the
maximum (worst-case) number of computational steps
needed to obtain an optimal solution.
Problems which have a known polynomial algorithm
are said to be in class P; the effort of the algorithm is
bounded by a polynomial function of the size of the
problem.
For NP (non-deterministic polynomial problems) no
simple algorithm yields optimal solutions in a limited
amount of computer time.

82

29