Production and Operations Management

Chapter 3- Scheduling
Asma Ben Yaghlane Riahi
Tunis Business School
University of Tunis
2023/2024
 Scheduling is the final step in production
Definitions planning. It’s often the only step in small and
medium-sized companies. It involves
determining
• The sequence of work execution or the
production program schedule.
• The chronology of the use of company
resources or workload, aiming to meet
customer needs in terms of quantity, quality,
time, location, and costs.
 • Effective and efficient scheduling constitutes a
Importance of
competitive advantage.
scheduling
• The sequence of work execution or the
production program schedule.

• A faster flow of goods through a facility enables

better efficiency and cost control.
Examples
Organization scheduling
Use of operating rooms
Hospital Patients’ admission
Assignment of nurses, security staff, maintenance staff
External patients’ treatment
Classes and audiovisual equipments
University Faculty & students schedules
Bachelor & master courses
Production of goods
Plant Equipment’s purchase
Employees
Café Waiters
Food delivery
Aircraft maintenance
Airline Departure schedule
Flight crew & ticketing staff
Catering
 Two methods can be used
• Progressive scheduling begins as soon as the
Scheduling requirements are known.

methods
Present Deadline

• Regressive scheduling begins from the deadline

date.

Present Deadline
• With progressive scheduling, you get a feasible schedule that
may not meet the deadlines.

• Regressive scheduling is used in various fields, such as

catering.

• Both methods are often combined to develop a compromise

between a feasible schedule while meeting deadlines.
 Among scheduling models, we distinguish

Scheduling models • Static models

Once per period
A new order implies a new scheduling
• Dynamic models
All orders are considered on a rolling basis
More complex techniques (Queuing &
simulations)
The most commonly • Minimize execution time
used criteria in
scheduling • Maximize facility utilization

• Minimize semi-finished product storage

• Minimize customer waiting time

 Workload
• The workload document describes the list of tasks or orders that different
workstations or operating centers must execute.
• Two situations are possible:
 Unlimited or infinite capacity: one or more products can be executed
simultaneously.
 Limited or finite capacity: a restricted number of products at the same time.
Deciding which product to prioritize for execution. A product can only be
executed if the resource to be used is available, and the product has passed
through all the previous operations.
 Gantt chart
The simplest tool to illustrate workloads is the Gantt chart. It allows visualizing both:
• The temporal utilization of productive resources (departments, machines,
facilities, employees, etc.).
• The progress of task execution. In particular, it allows monitoring ongoing
work.
• All Gantt charts must be regularly updated to account for changes.
• The Gantt chart is constructed by plotting time on the x-axis and workstations on
the y-axis.
Example

Day Monday Tuesday Wednesday Thursday Friday

Workshop

A1 Task A Task D

A2 Task B Task F

A3 Task C Task E

A4 Task F Task C Task E

Occupied available unavailable

 • The assignment algorithm (also known as the
Assignment Hungarian method) allows us to determine
algorithm how to distribute tasks to available resources
optimally.

• The objective is generally the minimization of

cost or execution time.

• A single task (or employee) is assigned to a

machine (or project).
 Example
We have 4 products (P1 to P4) to be made, and we have 4 employees (A, B, C, and
D) capable of making them, each being paid 18 dinars per hour. The table below
summarizes the time required for each employee to complete each product.

A B C D
P1 8 6 2 4
P2 6 7 11 10
P3 3 5 7 6
P4 5 10 12 9
The steps of the algorithm are as follows:

1. Subtract the smallest value from each row.

2. Subtract the smallest value from each column.

3. Cover all zero values with a minimum number of horizontal or vertical lines.

• If the number of lines is strictly less than the number of columns, proceed to
step 4.

• If the number of lines is equal to the number of columns, proceed to step 5.

4. Identify the smallest value among the uncovered values. Then subtract it from
all uncovered values and add it to the values at the intersections of lines.
Return to step 3.

5. Assign to each element in the first column the element in the first row
corresponding to a zero pivot value. Start with rows and columns having only
one zero.
 Minimum
Step 0 of rows Step 1
8 6 2 4 2 6 4 0 2
6 7 11 10 6 0 1 5 4
3 5 7 6 3 0 2 4 3
5 10 12 9 5 0 5 7 4
Minimum of
0 1 0 2
columns
Steps 2& 3 6 3 0 0
0 0 5 2
0 1 4 1
0 4 7 2

Less than 4 lines are used to cover all zeros.

Step 4

7 4 0 0
0 0 4 1
0 1 3 0
0 4 6 1

Step 5:
• Column 3 : x13 = 1
• Column 4 : x34 = 1
• Column 2 : x22 = 1
• Column 1: x41 = 1
Therefore, P4 to A, P2 to B, P3 to D et P1 to C.
 Remarks

• To solve an assignment problem where the goal is to maximize the

objective function, multiply the profit matrix by (-1) and solve the
problem as a minimization problem.

• If the number of rows and columns in the cost matrix is unequal, then
the assignment problem is unbalanced. Any assignment problem must
be balanced (by adding one or more dummy points) before being
solved by the Hungarian method.
 Many priority rules can be adopted to decide the
Scheduling order of performing multiple tasks on the same
multiple tasks on a machine. The most commonly used rules are:

machine • FIFO (first in, first out): goods and services are
executed in the order of arrival.
• MOT (minimum operating time): priority is given
to the task with the shortest execution time.
• PD (promised date): priority is given to the task
with the closest promised date to the customer.
Performance indicators allow measuring scheduling effectiveness

σ 𝒕𝒊𝒎𝒆 𝒔𝒑𝒆𝒏𝒕 𝒐𝒏 𝒕𝒂𝒔𝒌𝒔

𝑨𝒗𝒆𝒓𝒂𝒈𝒆 𝒕𝒊𝒎𝒆 𝒊𝒏 𝒕𝒉𝒆 𝒔𝒚𝒔𝒕𝒆𝒎 =
𝒏𝒖𝒎𝒃𝒆𝒓 𝒐𝒇 𝒕𝒂𝒔𝒌𝒔

σ 𝒆𝒙𝒆𝒄𝒖𝒕𝒊𝒐𝒏 𝒕𝒊𝒎𝒆
𝑬𝒇𝒇𝒆𝒄𝒕𝒊𝒗𝒆 𝒖𝒕𝒊𝒍𝒊𝒛𝒂𝒕𝒊𝒐𝒏 𝒐𝒇 𝒕𝒉𝒆 𝒔𝒚𝒔𝒕𝒆𝒎 =
σ 𝒕𝒊𝒎𝒆 𝒔𝒑𝒆𝒏𝒕 𝒐𝒏 𝒕𝒂𝒔𝒌𝒔

σ 𝒅𝒆𝒍𝒂𝒚𝒔
𝑨𝒗𝒆𝒓𝒂𝒈𝒆 𝒅𝒆𝒍𝒂𝒚 𝒑𝒆𝒓 𝒕𝒂𝒔𝒌 =
𝒏𝒖𝒎𝒃𝒆𝒓 𝒐𝒇 𝒕𝒂𝒔𝒌𝒔
 Example
Consider tasks A, B, C, D, E, and F; apply FIFO, MOT, and PD formulas, then compare
the different approaches.

Execution time Deadline

A 2 7
B 8 16
C 4 4
D 10 17
E 5 15
F 12 18
According to FIFO method, the order of tasks is unchanged A, B, C, D, E, and F

Start Duration Finish Promised date Delay

A 0 2 2 7 0
B 2 8 10 16 0
C 10 4 14 4 10
D 14 10 24 17 7
E 24 5 29 15 14
F 29 12 41 18 23
Total 41 120 54
 σ 𝒕𝒊𝒎𝒆 𝒔𝒑𝒆𝒏𝒕 𝒐𝒏 𝒕𝒂𝒔𝒌𝒔
𝑨𝒗𝒆𝒓𝒂𝒈𝒆 𝒕𝒊𝒎𝒆 𝒊𝒏 𝒕𝒉𝒆 𝒔𝒚𝒔𝒕𝒆𝒎 =
𝒏𝒖𝒎𝒃𝒆𝒓 𝒐𝒇 𝒕𝒂𝒔𝒌𝒔

120
= 6
= 20 days
σ 𝒆𝒙𝒆𝒄𝒖𝒕𝒊𝒐𝒏 𝒕𝒊𝒎𝒆
𝑬𝒇𝒇𝒆𝒄𝒕𝒊𝒗𝒆 𝒖𝒕𝒊𝒍𝒊𝒛𝒂𝒕𝒊𝒐𝒏 𝒐𝒇 𝒕𝒉𝒆 𝒔𝒚𝒔𝒕𝒆𝒎 =
σ 𝒕𝒊𝒎𝒆 𝒔𝒑𝒆𝒏𝒕 𝒐𝒏 𝒕𝒂𝒔𝒌𝒔

41
= = 34,17%
120

σ 𝒅𝒆𝒍𝒂𝒚𝒔
𝑨𝒗𝒆𝒓𝒂𝒈𝒆 𝒅𝒆𝒍𝒂𝒚 𝒑𝒆𝒓 𝒕𝒂𝒔𝒌 =
𝒏𝒖𝒎𝒃𝒆𝒓 𝒐𝒇 𝒕𝒂𝒔𝒌𝒔

54
= 6
= 9 days
According to MOT method, the order of tasks is A, C, E, B, D, and F

Start Duration Finish Promised Date Delay

A 0 2 2 7 0
C 2 4 6 4 2
E 6 5 11 15 0
B 11 8 19 16 3
D 19 10 29 17 12
F 29 12 41 18 23
Total 41 108 40
 σ 𝒕𝒊𝒎𝒆 𝒔𝒑𝒆𝒏𝒕 𝒐𝒏 𝒕𝒂𝒔𝒌𝒔
𝑨𝒗𝒆𝒓𝒂𝒈𝒆 𝒕𝒊𝒎𝒆 𝒊𝒏 𝒕𝒉𝒆 𝒔𝒚𝒔𝒕𝒆𝒎 =
𝒏𝒖𝒎𝒃𝒆𝒓 𝒐𝒇 𝒕𝒂𝒔𝒌𝒔

108
= 6
= 18 days
σ 𝒆𝒙𝒆𝒄𝒖𝒕𝒊𝒐𝒏 𝒕𝒊𝒎𝒆
𝑬𝒇𝒇𝒆𝒄𝒕𝒊𝒗𝒆 𝒖𝒕𝒊𝒍𝒊𝒛𝒂𝒕𝒊𝒐𝒏 𝒐𝒇 𝒕𝒉𝒆 𝒔𝒚𝒔𝒕𝒆𝒎 =
σ 𝒕𝒊𝒎𝒆 𝒔𝒑𝒆𝒏𝒕 𝒐𝒏 𝒕𝒂𝒔𝒌𝒔

41
= = 37,96%
108

σ 𝒅𝒆𝒍𝒂𝒚𝒔
𝑨𝒗𝒆𝒓𝒂𝒈𝒆 𝒅𝒆𝒍𝒂𝒚 𝒑𝒆𝒓 𝒕𝒂𝒔𝒌 =
𝒏𝒖𝒎𝒃𝒆𝒓 𝒐𝒇 𝒕𝒂𝒔𝒌𝒔

40
= 6
= 6,67 days
According to PD method, the order of tasks is C, A, E, B, D, and F

Start Duration Finish Promised Date Delay

C 0 4 4 4 0
A 4 2 6 7 0
E 6 5 11 15 0
B 11 8 19 16 3
D 19 10 29 17 12
F 29 12 41 18 23
Total 41 110 38
 σ 𝒕𝒊𝒎𝒆 𝒔𝒑𝒆𝒏𝒕 𝒐𝒏 𝒕𝒂𝒔𝒌𝒔
𝑨𝒗𝒆𝒓𝒂𝒈𝒆 𝒕𝒊𝒎𝒆 𝒊𝒏 𝒕𝒉𝒆 𝒔𝒚𝒔𝒕𝒆𝒎 =
𝒏𝒖𝒎𝒃𝒆𝒓 𝒐𝒇 𝒕𝒂𝒔𝒌𝒔

110
= 6
= 18,33 days
σ 𝒆𝒙𝒆𝒄𝒖𝒕𝒊𝒐𝒏 𝒕𝒊𝒎𝒆
𝑬𝒇𝒇𝒆𝒄𝒕𝒊𝒗𝒆 𝒖𝒕𝒊𝒍𝒊𝒛𝒂𝒕𝒊𝒐𝒏 𝒐𝒇 𝒕𝒉𝒆 𝒔𝒚𝒔𝒕𝒆𝒎 =
σ 𝒕𝒊𝒎𝒆 𝒔𝒑𝒆𝒏𝒕 𝒐𝒏 𝒕𝒂𝒔𝒌𝒔

41
= = 37,27%
110

σ 𝒅𝒆𝒍𝒂𝒚𝒔
𝑨𝒗𝒆𝒓𝒂𝒈𝒆 𝒅𝒆𝒍𝒂𝒚 𝒑𝒆𝒓 𝒕𝒂𝒔𝒌 =
𝒏𝒖𝒎𝒃𝒆𝒓 𝒐𝒇 𝒕𝒂𝒔𝒌𝒔

38
= 6
= 6,33 days
The results are summarized in the following table:

Average time Average

Utilization
on the system delay
FIFO 20 34,17 9
MOT 18 37,96 6,67
PD 18,33 37,27 6,33

• MOT is the best in terms of average time in the system and utilization indicators.
• PD is th best according to average delay.
• No rule excels in all criteria.

• MOT is the best in minimizing time spent in the system and in terms of system
utilization.

• However, MOT schedules tasks with the longest execution time last, which can
lead to customer dissatisfaction.

• FIFO is not particularly better than the others, but it is perceived by customers as
fair.

• PD minimizes average delay.

 Limitations of priority rules

1. Scheduling is dynamic, which is why priority rules need to be revised to adapt

to potential changes.

2. These rules don't look upstream or downstream. Bottlenecks in other

resources may not be identified.

3. These rules don't look beyond deadlines. For example, two orders may have
the same deadline. However, one of the orders is initiated just to replenish a
distributor's stock, while the other is designed to prevent a production halt in a
factory.
 Scheduling A.Same order of processing on both machines

multiple tasks on • Two or more tasks to be scheduled on two

different machines following the same sequence.
two machines
• Johnson algorithm can be used to schedule
multiple tasks on two machines in a way that
minimizes the total execution time and the idle
time of both machines.
Johnson algorithm consists on

1. Listing all tasks with their execution times on both machines

2. Finding the task with the shortest execution time. If this time corresponds to the
first machine, schedule this task first. However, if it corresponds to the second
machine, schedule it last.

3. eliminating a task from the list once it is scheduled.

4. If there is more than one task left to schedule, return to step 2; if only one task
remains, its position will be automatically determined.
 Example
Schedule tasks A, B, C, D, and E using Johnson algorithm

Task Machine 1 Machine 2

A 5 2
B 3 6
C 8 4
D 10 7
E 7 12
Task Machine 1 Machine 2

A 5 2
B 3 6
B E D C A
C 8 4
D 10 7
E 7 12
Once the sequence is decided, the production schedule for this sequence needs to
be established. Generally, a Gantt chart is used.

Task Machine 1 Machine 2

A 5 2
B 3 6 B E D C A
C 8 4
D 10 7
E 7 12

Time 0 3 10 20 28 33

M1 B E D C A

M2
 Task Machine 1 Machine 2

A 5 2
B 3 6
B E D C A
C 8 4
D 10 7
E 7 12
Time 0 3 10 20 28 33

M1 B E D C A

M2 B E D C A

Time 0 1 3 5 7 9 10 11 12 13 17 19 21 22 23 25 27 29 31 33 35

B E D C A
B.The non-uniqueness of the order of processing on the two machines
Jackson algorithm consists on
1. Partitioning the set of n tasks into 4 subsets:
• Subset 1 includes all tasks that require only machine 1.

• Subset 2 includes all tasks that require only machine 2.

• Subset 1-2 includes all tasks that require machine 1 followed by machine 2.

• Subset 2-1 includes all tasks that require machine 2 followed by machine 1.
2. Schedule each subset using the Johnson's algorithm

3. Combine the obtained results in the following manner:

• For machine1: optimal sequence from Subset1-2, then the optimal

sequence from Subset1, and finally the optimal sequence from Subset2-1.

• For machine 2: optimal sequence from Subset2-1, then the optimal

sequence from Subset2, and finally the optimal sequence from Subset1-2.
 Example
Schedule tasks A, B, C, D, …, K using Jackson algorithm, then draw the
corresponding Gantt chart.

Tasks to be performed on machine Tasks to be performed on machine

1 and then 2 2 and then 1

Task Machine 1 Machine 2 Task Machine 2 Machine 1

A 50 30 G 90 70
B 80 60 H 20 30
C 10 30 I 10 100
D 50 0 J 40 0
E 30 0 K 10 0
F 70 0
 Tasks to be performed on machine
1 and then 2 Tasks to be performed on machine
2 and then 1
Task Machine 1 Machine 2
Task Machine 2 Machine 1
A 50 30
G 90 70
B 80 60
H 20 30
C 10 30
I 10 100
D 50 0
J 40 0
E 30 0
K 10 0
F 70 0 M1 C, B, A, E, D, F, I, H, G

M2 I, H, G, K, J, C, B, A
Time 0 10 90 140 170 220 290 390 420 490

M1 C B A E D F I H G

M2 I H G K J C B A
 • In some special cases, a scheduling problem
Scheduling involving 3 production centers can be reduced to
a scheduling problem involving 2 production
multiple tasks on centers. This is the situation where the second
three machines production center is completely dominated by
either one of the other two production centers.
• We say that machine B is dominated by one of
the other two machines (A or C) if and only if the
maximum execution time on B is lower (or equal)
than the minimum execution time observed on
the dominating machine.
• When this case occurs, we reformulate the problem into a problem with 2
fictitious production centers.

 The first group combines production centers A and B into a virtual

production center denoted {AB} with an operating time (tiAB = tiA + tiB)

 the second group combines production centers B and C into a virtual

production center denoted {BC} with an operating time (tiBC = tiB + tiC)
 Example
Schedule these tasks using the Johnson's algorithm, then draw the corresponding
Gantt chart

Taski tiA tiB tiC

1 7 1 6
2 4 3 2
3 3 2 4
4 8 2 1
5 5 1 3
It is easy to verify that machine B is dominated by machine A.
The maximum execution time on B = 3
The minimum observed execution time on A = 3

Taski tiA tiB tiC Taski tiAB tiBC

1 7 1 6 1 8 7
2 4 3 2 2 7 5 3 1 2 5 4
3 3 2 4 3 5 6
4 8 2 1 4 10 3
5 5 1 3 5 6 4
• In case where we cannot reduce the 3 centers to 2 production centers,
the problem can be solved by modeling it as an integer linear program
and using an appropriate technique such as the "Branch and Bound"
method.
 When the order of task processing is identical and
Scheduling n tasks the number of production centers does not exceed
on m machines a few dozen, a solution often close to the optimal
solution can be found by applying the Johnson's
algorithm on successive groupings of production
centers, known as the CDS algorithm (Campbell,
Dudek, and Smith).
 Example
Consider the 4 production centers labeled in their order of operation by the letters
A to D (time in minutes).

Taski tiA tiB tiC tiD

1 50 43 15 4
2 89 99 95 77
3 7 47 20 98
4 8 64 12 94
5 61 19 65 14
6 1 80 66 78
It is necessary to solve the following 3 fictitious problems

• Machines {A} – {D}

• Machines {AB} – {CD}

• Machines {ABC} – {BCD}

The first fictitious problem (machines A and D) leads to the order 6-3-4-2-5-1,
with a total execution time of 512 minutes.
The second fictituous problem: Machines {AB} – {CD}

Task i tiAB tiCD

1 93 19
2 188 172
3 54 118
4 72 106
5 80 79
6 81 144
The second fictitious problem (machines AB and CD) leads to the order 3-4-6-
2-5-1, with a total execution time of 487 minutes.
The third fictitious problem: machines {ABC} – {BCD}.

Task i tiABC tiBCD

1 108 62
2 283 271
3 74 165
4 84 170
5 145 98
6 147 224

This third fictitious problem leads to the same order 3-4-6-2-5-1 with a total
execution time of 487 minutes.
• The solution 3-4-6-2-5-1 is then retained with a total execution time
of 487 minutes.

• By using more advanced mathematical tools (integer programming) or

by explicitly enumerating all possible schedules, a minimum execution
time of 485 minutes was achieved
 Example
A screed and tile installation company has recorded the following 5 orders for the
next month (the duration, in days, assumes that an entire team is assigned to a task) :

Taski C1 C2 C3 C4 C5
screed 2 3 10 1 2,5
tile 5 4 12 3 2

It is necessary to have a 2-day drying period between the completion of the screed
and the start of tiling. Drying is done with a machine, and the company has only one
drying machine.
1. If the company has a team of screed installers and a team of tilers, in what
order should it carry out these contracts to complete the 5 projects in the
minimum amount of time?

2. Represent your solution graphically using a Gantt chart.

3. How many days are needed to successfully complete the five projects?
 Answer
1.
Tâche i C1 C2 C3 C4 C5
Chape 2 3 10 1 2,5 C4 C1 C2 C3 C5
Carrelage 5 4 12 3 2

2. 2 days of drying between the completion of the screed and the start of tiling.

Screed C4
Chape C1 C2 C3 C5 Z

Drying
Séchage Z C4 C1 Z C2 Z C3 Z C5 Z

Tile
Carrelage Z C4 C1 C2 Z C3 C5
3. 32 jours 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33