Presentation SBO

Amel Jaoua, Ing., Ph.D.
Full Professor
Simulation-based Industrial Engineering Department
National Engineering School of Tunis (ENIT)
Optimization for
Smart Industry Head of the Industry 4.0 Laboratory
Email: amel.jaoua@enit.utm.tn
amel.jaoua@polymtl.ca
Phone: (+216) 25 889 434

1
Professor Amel Jaoua, Ph.D. amel.jaoua@polymtl.ca
About the Speaker
Professional Experience
2013 – Today
Full Professor - Industrial Engineering Department
@ National Engineering School of Tunis
Co-Founder and Head of the Industry 4.0 Laboratory
2009 – Today
Consultant and Expert in Simulation and Optimization for Decision-Making and
Control in Manufacturing and Service Systems
2009 – 2013
Assistant Professor – Mathematic and Industrial Engineering Department
Speciality Simulation of Industrial Systems.
@ Ecole Polytechnique Montreal-Canada
Professor Amel Jaoua, Ph.D. amel.jaoua@polymtl.ca 2

About the Speaker
Network & Collaborations

About the Speaker
Collaborating with the MITACS Canada Research Program since 2018,

proposing and contributing to 1-3 projects per year
Concordia University : Professor Anjali Awasthi
Jaoua, A., Ammar, M. B., & Awasthi, A. (2019). A Decision Support System for On-Demand Goods Delivery Using Shared
Autonomous Electric Vehicles. International Journal of Decision Support System Technology (IJDSST), 11(2), 72-88. (
Real data from Ville de Montréal )
Polytechnic School of Montreal : Professor Soumaya Yacout, Professor Michel Gamache,
Oukassi, H., Jaoua, A., Negri, E, Yacout, S. : towards a standardized integration of a Digital Twin in manufacturing
systems. In Winter Simulation Conference (2024)
University of Québec : Professor Ali Gharbi.
Didi, M., Jaoua, A., & Gharbi, A, Production control of failure-prone manufacturing-remanufacturing systems using
Deep Reinforcement Learning. In Process.

About the Speaker
How the MITACS Canada Research Program Offers Funding for International Students ?
5
Agenda
1. Place of Simulation-based Optimization in Smart Industry
• 1.1 Definition
• 1.2 An example of Simulation Project at Pratt & Whitney Canada
2. Discrete Event Simulation Modelling (DES)
• 2.1 Low-Immersion in a Warehouse System
• 2.2 Simulation by Hand of Queueing behavior
• 2.3 DES model development
3. Coupling Simulator with Optimizer
• 3.1 Simulation-based Optimization (SBO) theory
• 3.2 UML sequencing and searching Algorithms
4. Application for Flexible Manufacturing System (FMS)
• 4.1 Complexity of the FMS
• 4.2 Formulation of the SBO problem
• 4.3 Implementation of the SBO for Makespan Minimization
5. Application in an Industry 4.0 Laboratory
• 5.1 Low-Immersion in the Laboratory 4.0
• 5.2 Definition of the Optimization Problem
• 5.3 Developing the Simulation model with ARENA
• 5.4 Solving the Optimization Problem by coupling ARENA with OptQuest
6
1.Place of Simulation-based optimization in Digital Twin concept
The objective of Industry X.0 is to integrate

intelligent digital twins (DTs) based on discrete
event simulation (DES) models, coupled with
an Optimizer, to enable efficient real-time
monitoring and control of complex systems.
The Industry 4.0 have focused on cyber-
physical systems and the Internet of Things
(IoT) to establish connectivity and integration
of physical systems.
Industry X.0 builds on the connectivity
established in Industry 4.0 by focusing on how
to leverage the data generated for intelligent
monitoring, decision-making, and Lu, Y., Liu, C., Kevin, I., Wang, K., Huang, H., & Xu, X. (2020). Digital Twin-driven smart
manufacturing: Connotation, reference model, applications and research issues. Robotics
optimization. and computer-integrated manufacturing, 61, 101837.
7
1.2 A Simulation Project
Design, manufacture and service of aircraft engines:
8
1.2 An example of Simulation Project
Plant 5 for aircraft engine repair at St-Hubert, QC, Canada
Use plasma spray

techniques to restore
grooves.
(Rep.12, Rep.13, Rep.22)

9
1.2 An example of Simulation Project
Problem : Average Engine Lead Time of 5 days.

Solution : Develop a Simulator with Arena and
couple it with an Optimizer to find
prioritization strategy better than First in First
Out that Minimizes the Waiting time before
bottleneck repair workstations.
Results : Placing the engine according to the

OptQuest found strategy decreases the Lead
time to an Average of 2.5 days.
10
Agenda
2. Discrete Event Simulation (DES) Modelling

• 2.1 Low-Immersion in a Warehouse System
• 2.2 Simulation by Hand of Queueing behavior
• 2.3 DES model development

11
2. Discrete Event Simulation Modelling
▪ Discrete-event simulation (DES) involves modelling a system as it

changes over time, where state variables change instantly at specific
points.
▪ Although DES could theoretically be done using hand calculations, the

large amount of data needed for most real-world systems means that
these simulations should be performed using simulation software.
▪ Many simulation software options are now available on the market,

such as Simulink, Arena, AnyLogic, FlexSim, Simio, etc.
12
Can you spot warehouse
areas where queues may
2.1 Low-Immersion in a Warehouse System appear?
❑ Packed products are received from

workstations through this conveyor
❑ Automated Guided Vehicles (AGV) are

transporting these packages to the
storage areas waiting for the shipping
❑ When the shipping truck arrives, AGV

picks the required package from the
storage and deliver them to the
shipping Area
13
2.2 Simulation by Hand of Queueing behavior
Packed products arrive every 5 minutes, i.e. deterministic arrival rate

λ = 12 𝑝𝑟𝑜𝑑𝑢𝑐𝑡/ℎ.
This AGV, has dedicated pathway its total transportation time, is also
deterministic and equal to 6 minutes i.e. service rate
𝜇 = 10 𝑝𝑟𝑜𝑑𝑢𝑐𝑡/ℎ.
If the AGV is busy, these products will be placed in a conveyor, herein

the Queue, with limited capacity equal to 9 packages.
To do:
1. Execute this simple queuing system for a period of 30 minutes

and plot the instantaneous number of product in system
denoted L(t)
2. Deduce the Average Waiting Time in Queue, denoted 𝑊𝑄 .
14
Service
Delay
L(t) Exit of Pr.3 &
Exit of Pr.1 & Exit of Pr.2 & Exit of Pr.4 & Exit of Pr.5 &
Entry of Pr.2 in Server Entry of Pr.3 in Server Entry of Pr.4 in Server Entry of Pr.5 in Server Entry of Pr.6 in Server
0
5 6 10 12 15 18 20 24 25 30
Entry of Pr.6 in Queue
Time (minutes)
Entry of Pr.2 in Queue Entry of Pr.3 in Queue Entry of Pr.4 in Queue Entry of Pr.5 in Queue
Entry of Pr.1 in Server
Number in system, L(t), at time t.
15
Deduce over this period of 30 minutes the following estimates :
𝐿෠ : the time-weighted-average number of product in the system

𝑊෢𝑄 : average time spent in queue per product
∞
1 1 30
𝐿෠ = ෍ 𝑖𝑇𝑖 = න 𝐿 𝑡 𝑑𝑡 = 1.5 𝑝𝑟𝑜𝑑𝑢𝑐𝑡𝑠
𝑇 30 0
𝑖=0
𝑁 𝑄 𝑄 𝑄 𝑄 𝑄 𝑄
1 𝑄 𝑊1 + 𝑊2 + 𝑊3 + 𝑊4 + 𝑊5 + 𝑊6
෢
WQ = ෍ 𝑊𝑖 = = 2.5 𝑚𝑖𝑛𝑢𝑡𝑒𝑠
N 6
𝑖=1
As we are interested by the steady-state performances of 𝐿 and 𝑊𝑄 we let 𝑇 → ∞
16
Limitation of analytical
modeling?
2.3 DES model development
Interconnected conditional
Non-Markovian Processes
flow behavior
Estimate 𝑳 and 𝑾𝑸
for the Warehouse
system?
Warehouse process
Returns Arrive: Quality control No Disposal
Transport Time~𝑇𝑟𝑖𝑎𝑛𝑔𝑢𝑙𝑎𝑟 0.5,1.5,6

α× d f× α × d f ×p ×α× d
Functional?
Yes
Acceptance Update finished item

Update return inventory xF(t) and
f × (1 - p) × α × d Remanufacturing
inventory xR(t) AOQ
Average outgoing
quality AOQ
Manufacturing process
Yes
Manufacturing at
Row materiels Check the stock
demand
arrive level xF(t)
d / ( 1 -AOQ)
No
Sufficient?
Manufacturing at
max
AGV charging Time~1 + 4 ∗ 𝐵𝐸𝑇𝐴(1.97, 1.01)
17
Steps in Simulation Project?
Problem Definition
Setting of Objectives Design of Experiments to optimize decision
Conceptual Model Data Collection

Documentation
Model Translation
Verified ?
Implementation
Non
Validated ?
Non Non
18
Steps in Simulation project
•Problem definition: formulate the problem to be addressed.

•Setting of objectives :The objectives define the specific questions that the simulation is designed
to address
To do :
Draw an analogy to the previously exposed Pratt & Whitney Simulation Project for engine maintenance ?
•Conceptual Model : Define the boundaries and assumptions, then use a well-established formalism such
as UML activity diagrams or BPMN to represent the flow of processes.
•Data collection: Collect historical data from existing systems. This could include operational logs,
maintenance records, production data, etc. Then convert the collected data into a form that can be directly
input into the simulation model. This often includes standardizing formats, cleaning up datasets, fitting data
to appropriate probability distributions.
19
Steps in Simulation project
•Model translation: coding the model using general purpose ( C#, Phython, etc.) or specific
simulation languages (Arena, Anylogic, Simulink, Simio, etc.)
• Verified?: Debug the computer program by using tracing to detect potential bugs.
•Validated?: Compare the model's performance indicators with the actual system performance data
gathered. Utilize statistical comparison methods, such as the Mann-Whitney U Test, for validation. If
inconsistencies are revealed by the statistical tests, the conceptual model and/or the data collection phases
must be revised.
•Design of Experiments to Optimize decision: Formulate the simulation-based optimization problem
to find efficient decision i.e. that leads to better system performances. Coupling the simulator with an
Optimization tools for the execution of the designed experiments.
20
Design of Experiments
Transport Time~𝑇𝑟𝑖𝑎𝑛𝑔𝑢𝑙𝑎𝑟 0.5,1.5,6

• Estimate the Average Time in 𝜎 = 1.2, 𝐶𝑉 = 0.44
system if we add a third AGV?
• Estimate the Average Time in

system if we change the capacity
of the conveyor?
AGV charging Time~1 + 4 ∗ 𝐵𝐸𝑇𝐴(1.97, 1.01)
• Estimate the Average Time in

system if we assign specific AGVs
to serve a set of storage racks ?
• Etc.
21
Agenda

• 3.1 Simulation-based Optimization (SBO) theory
• 3.2 UML sequencing and searching Algorithms

22
3. Coupling Simulation with Optimizer 3.1 SBO Theory
• In Simulation-based Optimization (SBO) the

Objective function estimates
performance is the Output of a simulation
model.
• Based on the objective function estimated by

the simulation model, optimization method
iteratively generates a set of values which are
passed to simulation model as Inputs.
• The aim of this process is to find values of a

vector of input variables with the highest Input for Evaluation
contribution to the objective function’s value.
23
The Optimization components are :
• Decision Variables
• Objective Function
• Constraints
The problem is formulated as : opt 𝑓 𝑥 = 𝐸 𝐿(𝑥, 𝑤)

𝑥∈𝑋
Where 𝑥 is a 𝑥 -dimensional vector with each position representing a problem decision

variables from the 𝑋 matrix of constraints. 𝑓 𝑥 is the Objective function. The value of 𝑓
is an Expected function of 𝐿 𝑥, 𝑤 which is the output performance estimates of the
Simulation replication, from the vector 𝑥 with a random function 𝑤 that gives the
stochasticity of the system. 24
An estimator of the expected value can be obtained by the sample mean over N
replications:
𝑁
1
𝑓ҧ ≡ ෍ 𝐿(𝑥, 𝑤𝑖 )
𝑁
𝑖=1
Where the number of replication N is defined according to the desired width of the
confidence interval, generally fixed at significance level (α) of 5%.

3. Coupling Simulation with Optimizer 3.2 UML sequencing
OptQuest: SNDModel.p:
User
Optimizer Simulator
Optimize (Variance)
Optimization loop loop
[[TimeUp]or[AutomaticStop]]
alt Simulate(Path)
[Variance == 0]
PerformanceIndicators (Cost)
Destroy
[Else]
Simulation loop loop
[ConfLevel Met]
Simulate (Path, Stream) CMRG:
Disk
RNG
Request RN (Stream)
New Random Numbers
PerformanceIndicators (Cost)
Evaluate(ConfLevel)
Compute Half-width
Destroy
SimulationSummary (Performance Indicators)

Layeb, S. B., Jaoua, A., Jbira, A., &
ApplySearchAlgorithm(Path, Cost ) Makhlouf, Y. (2018). A simulation-
SearchAlgorithm returns new Path to optimization approach for scheduling in
Simulate
Optimization can advance to next iteration
stochastic freight transportation.
Computers & Industrial
OptimizationSummary (Optimal Path, Cost) Engineering, 126, 99-110.
26
3. Coupling Simulation with Optimizer 3.2 Searching Algorithms
In the optimization module, various searching algorithms can be applied, classified by

the nature of the feasible region:
• Finite and Small Feasible Region: Ranking and selection methods can be effectively
utilized.
• Combinatorially Large Feasible Region: Metaheuristic algorithms are more suitable
for optimization.
Various metaheuristics have been suggested such as genetic algorithms, simulated

annealing, tabu search, and neural networks.

3. Coupling Simulation with Optimizer 3.2 Searching Algorithms
Commercial software Optquest (provided by OptTek System Inc.), is powerful

Optimizer module already added to many simulation software Arena, Anylogic,
Flexsim, Simio etc.
The Search Algorithm used in Optquest is based on a combination of the

metaheuristics of tabu search, neural networks, and scatter search.
Optquest also handle multi-objective optimization using the weighted sum method.

Agenda

• 4.1 Complexity of the FMS
• 4.2 Formulation of the SBO problem
• 4.3 Implementation of the SBO for Makespan Minimization

29
4. Application for FMS 4.1 Complexity of the FMS
❑ FMS consists of a group of processing

workstations interconnected by means of an
automated material handling and storage system
and controlled by integrated computer control
system.
❑ FMS is based on automatization and flexibility of

the process. It can process various part types at
the workstation, handling flexible routing of parts
through machines and quantities of production
are often adjusted in response to changing
demand patterns.
❑ DES is considered the most powerful modeling

approach for robust estimates of FMS S.B. Saoud, A. Jaoua, N.B. Saoud, Efficient Simulator Based on Meta-Heuristic for FMS and
AGV systems Design and Control, Journal of Computer Science, vol. 4, n°1, pp. 8-14, 2005
performance
❑ The FMS consists of five workstations that can produce

three different types of products (P1, P2, P3).
❑ The intralogistics is performed with a fleet of automated

guided vehicles (AGV) following an unidirectional path.
❑ Parts arrive at the system's entrance and are stored in a

main stock. The upstream stock at each workstation
allows for waiting until the station becomes available.
The downstream stock enables the AGV to wait.
31
At first a DES model is developed for the FMS in

which:
❑ The number of vehicles and their speeds are
controllable parameters that can variates
respectively between one to 12 AGV; and 1 and 40
zones/seconds.
❑ There are three possible AGV dispatching rule :
▪ 1st rule : select closest free AGV : when a
workstation requests an AGV, the scheduler
selects all free vehicles and then chooses the
closest.
▪ 2nd rule: Cyclical AGV : in order to avoid an
over-exploitation of AGVs.
▪ 3rd rule: Random : choosing randomly
between available AGVs during dispatching
request.
32
❑ Explore the DES model and analyze the Makespan under

different traffic conditions.
❑ In FMS the Makespan refers to the total time required to

complete the production tasks. Specifically, it is the time
elapsed from the start of the first operation to the
completion of the last operation for all jobs in the system.
❑ Minimizing the makespan is a key objective in

manufacturing, as it directly impacts overall productivity
and efficiency.

Number of AGV Speed Dispatching rule Makespan Half-width Lower bound Makesapn Upper bound Makespan
3 10 1 676.94 10.671 666.269 687.611
3 15 1 671.75 10.788 660.962 682.538
5 15 1 671.62 10.696 660.924 682.316
9 15 1 671.6 10.71 660.89 682.31
9 15 3 682.47 18.15 664.32 700.62
9 15 2 672.89 10.607 662.283 683.497
10 15 2 672.14 10.793 661.347 682.933
34
4. Application for FMS 4.2 Formulation of the SBO problem
The Optimization problem consist on finding the optimal/suboptimal combination of the number of AGV, their
speed and the dispatching rule that leads to minimize the Makespan.
The SBO problem components are :

Decision Variables 𝑥1 : 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝐴𝐺𝑉𝑠
𝑥2 : 𝑆𝑝𝑒𝑒𝑑 𝑜𝑓 𝐴𝐺𝑉𝑠
𝑥3 : 𝐷𝑖𝑠𝑝𝑎𝑡𝑐ℎ𝑖𝑛𝑔 𝑟𝑢𝑙𝑒
Objective Function Makespan
Constraints 𝑥𝑖 ∶ 𝑑𝑖𝑠𝑐𝑟𝑒𝑡𝑒 𝑣𝑎𝑙𝑢𝑒𝑠

𝑥1 ∈ 1,12
𝑥2 ∈ 1,40
𝑥3 ∈ 1,3
35
4. Application for FMS 4.2 Implementation of the SBO
To implement the Simulated Annealing algorithm, we need to define the problem-specific parameters and to
determine the annealing parameters.
Problem-specific parameters are the evaluation function and the neighbourhood structure
Select an initial state X0;
Select an initial temperature T0; 1 2 3
Select the epoch length E; X at searching step (i-1)
Repeat 3 8 3
Set i = 0
Repeat Random choice of three range
Generate a neighbourhood state Xi;
Evaluate by simulation mechanism fi = f(Xi);
Calculate δ := fi - fi-1; 1 2 3
New X to test at
If δ<0 then Xi-1 := Xi 9 30 1
Step (i)
else if random(0,1) <exp(-δ /T) then Xi-1 := Xi ;
i = i+1;
until i=E;
T := α T; Generation of neighborhood move
until T= Tf
Jaoua, A., Gamache, M., & Riopel, D. (2012). Specification of an
intelligent simulation-based real time control architecture:
Application to truck control system. Computers in Industry, 63(9),
882-894.
36
4. Application for FMS 4.2 Formulation of the SBO problem
SBO searching results based on the Simulated Annealing

Tested combination Optimal combination
𝑥1 𝑥2 𝑥3 ҧ
𝑓(𝑥) 𝑥1 𝑥2 𝑥3 ҧ
𝑓(𝑥)
3 30 2 1437 3 30 2 1437
3 8 3 5252,5 3 30 2 1437
9 30 3 573 9 30 3 573
9 30 1 419 9 30 1 419
9 30 2 531 9 30 1 419
8 30 1 533 9 30 1 419
5 30 1 897 9 30 1 419
6 35 2 405,3871 9 31 1 405,3871
8 35 2 482,5714 9 31 1 405,3871
8 39 3 523,8571 9 31 1 405,3871
4 30 2 522,2857 9 31 1 405,3871
4 10 2 2202 9 31 1 405,3871
4 31 2 667,9677 9 31 1 405,3871
7 35 3 652,7143 9 31 1 405,3871
3 35 2 1233,286 9 31 1 405,3871
7 28 2 826,4286 9 31 1 405,3871
7 35 1 404,4286 7 35 1 404,4286
5 35 1 726,4286 7 35 1 404,4286
5 35 2 623,7143 7 35 1 404,4286
7 35 2 402,22 7 35 2 402,22
12 35 2 406,29 7 35 2 402,22
12 39 1 405,32 7 35 2 402,22 37
3 39 1 670 7 35 2 402,22
Agenda

• 5.1 Low-Immersion in the Laboratory 4.0
• 5.2 Formulation of the SBO Problem
• 5.3 Developing the Simulation model with ARENA
• 5.4 Solving the Optimization Problem with OptQuest
38
5. Application in an Industry 4.0 Laboratory 5.1 The Industry 4.0 Laboratory
At this Laboratory, the Festo Modular Production

System, MPS® system 403-1, from Festo®
Didactic is used as an Industry 4.0 learning
system.
The MPS 403-1 is composed of three fully
automated workstations: Distribution, Joining and
Sorting.
The role of these workstations is to produce the

customer order received at the Manufacturing
Execution System (MES) to mimic the make-to
order approach in the e-commerce context.
We consider three types of customized order

quantity of Tokens corresponding to a batch of
2, 4 or 6 workpieces.


The exit of each workstation is equipped with an

RFID write/read device, and each workpiece has a
tag inserted into it.
Based on the OPC-UA protocol embedded within

the Turck's RFID system each workpiece is
Workpiece with
tracked, and RFID data are stored in real-time. an RFID_Tag
RFID AMR
read/write Camera
the AMR is equipped with camera. The REST API

streamlines operations by sending requests to the
Robotino for navigating to specific “Manual
Station” and to the camera sensor for activating
QR code scanning. QR code
A specific QR code is associated to each manual

workstation for delivery and also another QR code
is placed at the dispatching point
41
Assembled Order of 2 workpieces
42
5. Application in an Industry 4.0 Laboratory 5.1 Low-immersion
43
5. Application in an Industry 4.0 Laboratory 5.2 Formulation of the SBO problem
The optimization problem consists of finding the optimal number of operators to assign to
“Assembly_workstation_1” and to the “Packaging workstation”, along with the AMR
allocation rate for each assembly workstation, in order to Minimize the Makespan.
• At least two operators must be assigned to Assembly_workstation_1.

• We have a total of 6 operators.
• The AMR must serve each Assembly_workstation for at least 10% of the time.
To do:
Formulate the SBO problem ?

5. Application in an Industry 4.0 Laboratory 5.2 Formulation of the SBO problem
The SBO problem components are :
Decision Variables 𝑥1 : 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑜𝑝𝑒𝑟𝑎𝑡𝑜𝑟 𝑖𝑛 𝐴𝑠𝑠𝑒𝑚𝑏𝑙𝑦 1

𝑥2 : 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑜𝑝𝑒𝑟𝑎𝑡𝑜𝑟 𝑖𝑛 𝑃𝑎𝑐𝑘𝑎𝑔𝑖𝑛𝑔
𝑥3 : 𝐴𝑀𝑅 𝑎𝑙𝑙𝑜𝑐𝑎𝑡𝑖𝑜𝑛 𝑟𝑎𝑡𝑒 to 𝐴𝑠𝑠𝑒𝑚𝑏𝑙𝑦 1
Objective Function Makespan
Constraints 𝑥𝑖 ∶ 𝑑𝑖𝑠𝑐𝑟𝑒𝑡𝑒 𝑣𝑎𝑙𝑢𝑒𝑠

𝑥1 ∈ 2,4
𝑥2 ∈ 1,3
𝑥3 ∈ 10,90
𝑥1 + 𝑥2 ≤ 5
45
5. Application in an Industry 4.0 Laboratory 5.3 Developing the Simulation Model
46
5. Application in an Industry 4.0 Laboratory 5.4 Solving with OptQuest

5. Application in an Industry 4.0 Laboratory 5.4 Solving with OptQuest
48
A glimpse into the future of SBO
Training the Reinforcement Learning Algorithms on the Simulator

towards intelligent Digital Twin framework considering the real-time
state of the Physical System.
[https://github.com/neprev].
49
A glimpse into the future of SBO
50
References
• Jaoua, A., Masmoudi, S., & Negri, E. (2024). Digital twin-based reinforcement learning framework: application
to autonomous mobile robot dispatching. International Journal of Computer Integrated Manufacturing, 1-24.
• Layeb, S. B., A. Jaoua, A. Jbira, and Y. Makhlouf. 2018. “A Simulation-Optimization Approach for Scheduling in
Stochastic Freight Transportation.” Computers & Industrial Engineering 126:99–110.
https://doi.org/10.1016/j.cie.2018. 09.021 .
• Glover, F., M. Laguna, and R. Marti. 2000. “Scatter Search and Path Relinking: Advances and Applications.”
Control and Cybernetics 29 (3): 653–684.
• Gosavi, A. 2015. Simulation-Based Optimization, 47–55. Berlin: Springer.
• S.B. Saoud, A. Jaoua, N.B. Saoud, Efficient Simulator Based on Meta-Heuristic for FMS and AGV systems Design
and Control, Journal of Computer Science, vol. 4, n°1, pp. 8-14, 2005
• Fu, M. C. 2002. “Feature Article: Optimization for Simulation: Theory Vs. Practice.” INFORMS Journal on
Computing 14 (3): 192–215. https://doi.org/10.1287/ijoc.14.3.192.113
51

Presentation SBO

Uploaded by

Copyright:

Available Formats

Presentation SBO

Uploaded by

Document Information

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

Presentation SBO

Uploaded by

Copyright:

Available Formats

Amel Jaoua, Ing., Ph.D.

Phone: (+216) 25 889 434

Professor Amel Jaoua, Ph.D. amel.jaoua@polymtl.ca 2

Network & Collaborations

Professor Amel Jaoua, Ph.D. amel.jaoua@polymtl.ca 3

Collaborating with the MITACS Canada Research Program since 2018,

Professor Amel Jaoua, Ph.D. amel.jaoua@polymtl.ca 4

The objective of Industry X.0 is to integrate

Design, manufacture and service of aircraft engines:

Plant 5 for aircraft engine repair at St-Hubert, QC, Canada

Use plasma spray

(Rep.12, Rep.13, Rep.22)

Problem : Average Engine Lead Time of 5 days.

Results : Placing the engine according to the

1. Place of Simulation-based Optimization in Smart Industry

2. Discrete Event Simulation (DES) Modelling

3. Coupling Simulator with Optimizer

4. Application for Flexible Manufacturing System (FMS)

5. Application in an Industry 4.0 Laboratory

▪ Discrete-event simulation (DES) involves modelling a system as it

▪ Although DES could theoretically be done using hand calculations, the

▪ Many simulation software options are now available on the market,

❑ Packed products are received from

❑ Automated Guided Vehicles (AGV) are

❑ When the shipping truck arrives, AGV

Packed products arrive every 5 minutes, i.e. deterministic arrival rate

If the AGV is busy, these products will be placed in a conveyor, herein

1. Execute this simple queuing system for a period of 30 minutes

Number in system, L(t), at time t.

Deduce over this period of 30 minutes the following estimates :

𝐿෠ : the time-weighted-average number of product in the system

Transport Time~𝑇𝑟𝑖𝑎𝑛𝑔𝑢𝑙𝑎𝑟 0.5,1.5,6

Acceptance Update finished item

AGV charging Time~1 + 4 ∗ 𝐵𝐸𝑇𝐴(1.97, 1.01)

Setting of Objectives Design of Experiments to optimize decision

Conceptual Model Data Collection

•Problem definition: formulate the problem to be addressed.

Transport Time~𝑇𝑟𝑖𝑎𝑛𝑔𝑢𝑙𝑎𝑟 0.5,1.5,6

• Estimate the Average Time in

• Estimate the Average Time in

1. Place of Simulation-based Optimization in Smart Industry

2. Discrete Event Simulation Modelling (DES)

3. Coupling Simulator with Optimizer

4. Application for Flexible Manufacturing System (FMS)

5. Application in an Industry 4.0 Laboratory

• In Simulation-based Optimization (SBO) the

• Based on the objective function estimated by

• The aim of this process is to find values of a

The Optimization components are :

The problem is formulated as : opt 𝑓 𝑥 = 𝐸 𝐿(𝑥, 𝑤)

Where 𝑥 is a 𝑥 -dimensional vector with each position representing a problem decision

Professor Amel Jaoua, Ph.D. amel.jaoua@polymtl.ca 25

Optimization loop loop

Simulation loop loop

New Random Numbers

SimulationSummary (Performance Indicators)

In the optimization module, various searching algorithms can be applied, classified by

Various metaheuristics have been suggested such as genetic algorithms, simulated

Professor Amel Jaoua, Ph.D. amel.jaoua@polymtl.ca 27

Commercial software Optquest (provided by OptTek System Inc.), is powerful