Model-Based Decision Making
Through
Simulation-Optimization
Decision-Support Systems (DSS)
Allen Greenwood & Travis Hill, Mississippi State University
1
Model-based Decision Making Through Simulation-Optimization DSS
Outline of Presentation
Context – MSU model & application development activities
Basic Concepts -- Decision Support Systems (DSS)
Examples of Model-based DSS (MB-DSS)
Panel Shop
Pipe Shop
Lean Manufacturing Flight Simulator
2
Model-based Decision Making Through Simulation-Optimization DSS
Mississippi State University
The Department of Industrial and
Systems Engineering (ISE) provides
education, research, and outreach in
order to design, analyze, and manage
systems of people, materials,
information, equipment, and energy.
The Center for Advanced Vehicular
Systems (CAVS) provides new
capabilities for Mississippi industry in
order to reduce development and
production costs while improving
quality.
3
Model-based Decision Making Through Simulation-Optimization DSS
Key Projects to Provide Context
4
Model-based Decision Making Through Simulation-Optimization DSS
Why Model-based DSS?
Enhance Organizational Performance
put sophisticated engineering tools into the hands of the decision
makers to improve decision making
Cost/Benefit
Cost/Benefit
increase the useful life of models; get a greater return on model
investment
Model
Time
Model
+
DSS
5
Model-based Decision Making Through Simulation-Optimization DSS
“Traditional” Decision Making
Real System
understand, analyze,
design, manage, and
make decisions about …
Decision Maker
… complex operational
processes (interdependencies
+ stochastic + dynamic)
6
Model-based Decision Making Through Simulation-Optimization DSS
“Model-based” Decision Making
Real System
Models
7
Model-based Decision Making Through Simulation-Optimization DSS
Model-Based Decision Making
Data
Base
Data
Base
Models
BaseBase
Models
Base
Models
Model-based
Model-based
Model-based
Model-based
Decision
Making
Decision
Making Making
Decision
Making
Decision
45%
45%
45%
Real
System
Real
System
Real
System
45%
5%
20%
5%
20% 10%
5%
20%
10%
5%
10%
20% 10%
20%
20%
20%
20%
Enterprise
Enterprise
Enterprise
Enterprise
Level
Level
Level
Level
Decision
Decision
Department
Decision
Department
Department
makers
makers Decision
Department
makers
(Shop)
(Shop)
(Shop)
makers
(Shop)
Level
Level
Level
Use
Cases
Level
5
5
4
4 5
3
3 4
2
1 22 3
1
1
1
Initial Parents
Parent 1
6 8 3 1 4 7 2 5
Parent 2
5 3 4 7 8 2 1 6
Crossover
Point
6 8 3 1 4 5 7 2
Analysis Base
5 3 4 7 8 2 1 6
5 3 4 7 8
6 8 3 1 4 7 2 5Initial Parents Final Offspring
Parent 1
Resulting Offspring
Offspring 1
6 8 3 1 4 5 7 2Parent 2
Offspring 2
5 3 4 7 8 6 1 2
Analysis Base
3
4
5
Level
Neighbor Swap
Mutation
5 3 4 7 8 6 1 2
Sequence Extracted Crossover
6 8 3 1 4
2
Workstation
Workstation
Workstation
(Cell)
(Cell)Workstation
(Cell)
Level
Level (Cell)
Level
Random Swap
Mutation
Crossover
6 8 3 1 46 78 23 15 4 7 5 2 Point
6 8 3 1 4 5 7 2
Neighbor Swap
Mutation
5 3 4 7 85 23 11 76 8 6 4 2
5 3 4 7 8 6 1 2
Sequence Extracted Crossover
6 8 3 1 4
5 3 4 7 8 2 1 6
5 3 4 7 8
6 8 3 1 4 7 2 5
Resulting Offspring
Offspring 1
6 8 3 1 4 5 7 2
Offspring 2
5 3 4 7 8 6 1 2
Random Swap
Mutation
Final Offspring
6 8 3 1 4 7 5 2
5 3 1 7 8 6 4 2
8
Model-based Decision Making Through Simulation-Optimization DSS
What changes need to be made?
from modeler/analyst centric to
decision-maker centric
from point solutions to
optimization
from work-cell
enterprise
supply-chain applications
from closed, local analyses to
open, distributed analyses
SD
PRO
Professional Workstation 6000
from PC/server based to
web/GRID based
9
Model-based Decision Making Through Simulation-Optimization DSS
Applying the MB-DSS Approach at NGSS
10
Model-based Decision Making Through Simulation-Optimization DSS
Sim-Opt DSSs for NGSS
6/7/00
D26
C27
C26
D65
D64
D63
D72
D71
D61
D25
D42
D52
D24
D23
D51
D41
C25
C55
C45
C54
C53 C44
C43
C52
D31
C34 C24
C15
C14
B45
B44
B35
B34
B43
C33 C23
C13
C12
C42
C32
D32
C22
P:\Const_Supp\DDG 89 BLOCK
C11
B42
B33
B25
B15
B24
B14
B31
A45
A44
D12
A27
A46 A36
A35
A34
A26
A25
A24
A17
A16
A15
B23
B32
B22
B41
D22
D21 D11
B21
B13
B12
B11
A43
A42
A41
A33
A32
A31
A23
A22
A21
Panel
Pipe
Details
11
Model-based Decision Making Through Simulation-Optimization DSS
General MB-DSS Approach
Decision
Variables
System
Performance
USERS
Users
Art Intel
Math Prog
Heuristics
Analysis/Optimizaton Algorithms
Analyses
Schedule /
Forecast
4 0
3 5
3 0
Product
Information
Decision Support System
2 5
2 0
1 5
1 0
5
0
1
2
3
4
5
6
7
8
9
1 0
Pro c e s
n
i g Time
Process
Information
Simulation Models
Flexsim
ProModel
QUEST
12
Model-based Decision Making Through Simulation-Optimization DSS
Panel Shop Problem:
Every Panel is Unique Æ Large variability
Backside SAW
SSAW
1800
1600
1400
1200
1000
800
600
400
200
0
1400
1200
1000
800
600
400
200
0
1
11 21 31 41 51 61 71 81 91 101 111 121 131 141 151 161
3000
2500
2000
1500
1000
500
0
1
11 21 31 41 51 61 71 81 91 101 111 121 131 141 151 161
Edge Grind
700
11 21 31 41 51 61 71 81 91 101 111 121 131 141 151 161
Layout
2500
1200
2000
1000
800
1500
400
1
Topside SAW
600
500
FCAW
3500
600
300
1000
400
200
500
100
0
200
0
0
1
11 21 31 41 51 61 71 81 91 101 111 121 131 141 151 161
1
11 21 31 41 51 61 71 81 91 101 111 121 131 141 151 161
1
11 21 31 41 51 61 71 81 91 101 111 121 131 141 151 161
13
Model-based Decision Making Through Simulation-Optimization DSS
Models Used to Increase Panel Shop Throughput
Simulation model to
represent behavior of
production system and
evaluate sequence
Model runtime: approximately
5 seconds to process 154 panels
(~13 weeks in real time)
Initial Parents
Optimization algorithm
to intelligently generate
possible sequences
There are 1.3 trillion possible ways to
sequence a 15-panel set
Parent 1
6 8 3 1 4 7 2 5
Parent 2
5 3 4 7 8 2 1 6
Crossover
Point
6 8 3 1 4 5 7 2
5 3 4 7 8 6 1 2
Sequence Extracted Crossover
6 8 3 1 4
5 3 4 7 8 2 1 6
5 3 4 7 8
6 8 3 1 4 7 2 5
Neighbor Swap
Mutation
Random Swap
Mutation
Final Offspring
Resulting Offspring
6 8 3 1 4 7 5 2
Offspring 1
6 8 3 1 4 5 7 2
5 3 1 7 8 6 4 2
Offspring 2
5 3 4 7 8 6 1 2
Evolutionary Strategy
14
Model-based Decision Making Through Simulation-Optimization DSS
Example Impact – Swapping 1 of 42 Panels
12%
improvement
Similar solution to adding one welder to first shift for TopSide/BackSide
15
Model-based Decision Making Through Simulation-Optimization DSS
Intuitive User Interface for Planners & Shop Floor
Model
Selection
run in ProModel or QUEST
Operations
parameters
Panel
selection
Optimal
sequence
16
Model-based Decision Making Through Simulation-Optimization DSS
Various Output Reports for Planners & Shop Floor
Resource
Allocation
Performance
Schedule
17
Model-based Decision Making Through Simulation-Optimization DSS
Panel Shop DSS
Effectively Links Users With Models & Data
Decision
Variables
System
Performance
Users
USERS
Art Intel
Math Prog
Heuristics
Analysis/Optimizaton Algorithms
Analyses
Schedule /
Forecast
4 0
3 5
3 0
Product
Information
Decision Support System
2 5
2 0
1 5
1 0
5
0
1
2
3
4
5
6
7
8
9
1 0
Pro c e s
n
i g Tim
e
Process
Information
Simulation Models
Flexsim
ProModel
QUEST
18
Model-based Decision Making Through Simulation-Optimization DSS
Pipe Details
Pipe Shop Problem:
Large Variability in Demand + Unknown Capacity
Capacity ???
Time
Capacity
Time
19
Model-based Decision Making Through Simulation-Optimization DSS
Basic Issue: What, Where, & When to Produce
Each PD with an Effective Use of Resources
Multiple
programs
Multiple
yards/shops
Pascagoula
Pascagoula
(Pipe Fabrication)
(Ship Construction)
Outsource
(Pipe Fabrication)
Avondale
Avondale
(Pipe Fabrication)
(Ship Construction)
20
Model-based Decision Making Through Simulation-Optimization DSS
Pipe Shop Operations & DSS Approach
Bottleneck
21
Model-based Decision Making Through Simulation-Optimization DSS
DSS Integrates Two-Levels of Models
Planning Model (high-level)
for bottleneck capacity
planning and analysis
high-level operational tradeoffs and production decisions
considers each PD as a
resource-constrained project
scheduling using project
management methodologies
Operations Model (low-level)
assess plan at shop level,
including all operations
considers interactions,
variability, and dynamics
establish capacity for high-level
model
22
Model-based Decision Making Through Simulation-Optimization DSS
Architecture
Outputs
Inputs
Shop & Operational
Characteristics
SPSDSS
Main
User
Interface
Bill/Pipe
Characteristics
Model Data
Model Inputs /
Outputs
Model Inputs /
Outputs
Processing Time
Models
Output
Viewer
Project
Wrapper
Project Management
Software (MSProject)
Flexsim
Wrapper
Discrete-Event Simulation
Software (Flexsim)
23
Model-based Decision Making Through Simulation-Optimization DSS
Pipe Shop DSS Primary Interface
Current model representation
Basic
system
outputs
System
inputs
Model chart or graph output
Basic Information Flow
24
Model-based Decision Making Through Simulation-Optimization DSS
Interface to Extract Data from Central Database
The Date
Range menu
item allows
users to specify
the data to
bring into the
system.
25
Model-based Decision Making Through Simulation-Optimization DSS
Interface to Modify Pipe Detail Attributes
Doubleclicking on an
input brings
up related
information
about the
selected
object.
26
Model-based Decision Making Through Simulation-Optimization DSS
Example Output
The Output
Viewer is used
to display all
graphs and
reports and
each are
printable and
exportable to
Excel.
27
Model-based Decision Making Through Simulation-Optimization DSS
Lean Manufacturing Flight Simulator
Objective: Improve intuition about the implementation of lean
manufacturing within the furniture industry. This will be
accomplished by …
Reviewing basic principles of lean manufacturing
Providing the user with the ability to “experiment” with a
simulated plant (i.e., manufacturing flight simulator).
Enhancing the understanding about relationships between key
plant decision variables and performance measures.
28
Model-based Decision Making Through Simulation-Optimization DSS
Characteristics of a Traditional Plant
Producing to forecasts and mixes that maximize plant efficiencies.
Similar functions grouped together
Individual department measures focus on internal efficiencies
Quality problems are only detected at inspection stations
Set-up times and equipment downtimes are “givens”
Arm
Mill
Chair
Assembly
Frame Mill
Arm
Assembly
Fabric Cutting
Sofa Assembly
Pillow
Sew &
Stuff
Sewing
Love Seat Assembly
29
Model-based Decision Making Through Simulation-Optimization DSS
Characteristics of a Lean Plant
Focus is on throughput
Workers are flexible and move as needed
Dependencies between work station are exploited and managed
Fabrication & Assembly are managed jointly.
Significantly reduced reliance on mechanism for material movement
Assembly
Line 1
Frame Mill
Arm Mill
Assembly
Line 2
Pillow
Assembly
Assembly Line 3
30
Model-based Decision Making Through Simulation-Optimization DSS
Experimentation Parameters
Product Mix – percentage of the forecast / demand assigned to each product type
Setup Time – time required to setup between product types
Defect Rate – percent of products that are defective from a processor
Downtime – frequency and duration of downtimes
Processing Time – mean processing time for a given processor
Queue Capacities – capacity levels of WIP areas
Performance Measures
WIP - overall Work-In-Process of the number of units in process
Lead-time - average time that an order takes to complete all operations
Throughput - actual production rate in terms of completed units for each hour
Efficiency - average across all workstations of the percentage of the actual to
theoretical production rate
Performance to Schedule - percentage of scheduled demand achieved
Travel - average number of feet traveled by each unit produced
First Pass Quality - percentage of units that complete production with no re-work
31
Model-based Decision Making Through Simulation-Optimization DSS
“Points Game”
Users are budgeted a given number of points to spend on improving
each system.
Users can use run experiments to determine where to focus
improvement initiatives.
Users can apply points to key model areas to:
improve set up time
reduce defect rates
reduce downtime
32
Model-based Decision Making Through Simulation-Optimization DSS
Manufacturing Flight Simulator Main Interface
Product
Mix
Settings
Views
and Fly
Paths
Simulation
Model
View
Model
Controls
Status
Window
Output
Panel
33
Model-based Decision Making Through Simulation-Optimization DSS
Manufacturing Flight Simulator Interface
Product Mix Settings – allows users to set the production mix
for the simulation model. If the future and initial mix values
differ, the model changes the product mix required during the
execution of the model.
Views and Fly Paths – allows users to maneuver through the
simulation model
Model Controls – allows users to control the execution of the
simulation model
Status Window – displays status messages from the
simulation (e.g. machine breakdowns, machines coming back
online, and whether production was meet for the day)
34
Model-based Decision Making Through Simulation-Optimization DSS
Simulation Model View
35
Model-based Decision Making Through Simulation-Optimization DSS
Output Panel
Model Statistics – displays
summary statistics as the
model runs
Graphical Data – displays data over time
as the model runs. The user can view
different graphs by selecting them in the
drop down box.
36
Model-based Decision Making Through Simulation-Optimization DSS
System Benefits
Ability to experiment on a more complex system
Ability to compare global metrics such as equipment efficiency, total
WIP and performance to schedule
Ability to see how the systems react to a major change in demand
Demand Change
37
Model-based Decision Making Through Simulation-Optimization DSS
38
Model-based Decision Making Through Simulation-Optimization DSS