Lucas Yiew Yuting Jin, Allan Magee - nfas.autonomous-ship.org
Development of Smart Digital Twin for Maritime Industry...
Transcript of Development of Smart Digital Twin for Maritime Industry...
Development of Smart Digital Twin for Maritime Industry via O2DES Framework
1
Dr. LI HaobinNational University of SingaporeDepartment of Industrial Systems Engineering and ManagementCentre of Excellence in Modelling and Simulation for Next Generation Ports
Outline
Smart Digital TwinsBackground – Industry 4.0Complexity of SystemsSimulation as an Exploration ToolThe Four Dimensions
Proposed MethodologiesClassification of SimulationFormalisms for Discrete-Event Systems (DES)Object-Oriented DES (O2DES) Framework
Future TrendsHigh-Level ArchitectureContainerization with Microservice
2
3
Smart Digital Twins
Background – Industry 4.0
Explore(Mining)
Summarize(Refining)
Propagate(Transporting)
3 Scenario Variables
Responding D
ecision Variables
Unknown
Unknown
Reinforcement Learning + Deep Learning
4
Smart Digital Twins
Complexity of Systems
The unknown is not about individual components,
but how they are combined and interact.Same principle applies for a complex industrial system, e.g., a mega container port
5
Smart Digital Twins
Simulation as an Exploration Tool
Simulation
Bridging Physical & Digital World
Commercial off-the-shelf simulation software
- For evaluating specific design / configuration
- Manually configured
Simulation as an Exploration / Experiment Tool
An object -oriented framework / paradigm to- Facilitate DES model building- Automate model configuration- Integrate DES with optimization- Bridge DES to other AI components, e.g.,
deep learning & data analytics
Interface with Human Users
Interface with Machine / Algorithms
6
Smart Digital Twins
The Four Dimensions
v0v1
v2v3
To draw the “game- board” for industrial systems to explore unknowns
To train a “grandmaster” for complex industrial system
Connectivity
Visibility
SMARTDIGITAL TWIN
Analyzability
Granularity
Conceptual models
Planning models
Operational models
Exact protocols
mor
e ev
ents
/ m
odul
ar d
etai
ls
closer to physical systems
Hard-coded / console input
parameters
Simple GUI configuration
Port XMLGraphical
configuration tools
Interface to database
Sync with IoT sensors
Simulation analytics with learning
Optimization via large scale search
Optimization embedded simulation
Simulation evaluation
tow
ards
real
-tim
e de
cisio
ns
closer to human perception
Console displayfor analytical measures
Dashboard statisticalcharts
2D/3D animation
Multi-screendisplayVR/AR/
Mix-Reality
Ranking and SelectionSimulation Analytics
Simulation Modelling
7
Proposed Methodologies
Classification of Simulation
Electro-mechanical
Systems
Enterprise Systems
Socio-Economic, Ecological Systems
Subatomic World
Cosmological Systems
Discrete-Event SystemContinuous System Continuous System
- Finite element simulation- Physical / mechanical /
chemical behavior- Fluid dynamics- Material science- Equipment design
- System Dynamics- Cause loop analysis- Large scale system
engineering
- Simulation of man-made process / protocols
- Discretized decision makers- Event-based decision
making procedures
Scale of System
Quantum System
Quantum System
(by State Variables)
8
Proposed Methodologies
Classification of Simulation (by Time Representation)
Computer Simulation
Analytic Simulation(Time-stamp simulation)
VE Simulation(Time-delay simulation)
Autonomous Constructive (HIL) Virtual Constructive (HIL) Live (HIL+MIL)
Might also include MIL (either hardware and/or software), it is also referred as an emulation
Most common Animation / movie
Game Monitoring / surveillance system / control tower
9
Proposed Methodologies
Formalisms for Discrete-Event Systems (DES)
In the classical activity-cycle diagram an activity typically represents the interaction between an entity* and active resources. An entity or an active resource is either in a passive state called a queue or in an active state called an activity. Queue nodes and activity nodes are connected by arcs.
Passive State
Passive State
Passive State
Active State Active State
Arc (for resource)
Arc (for entities)
Passive State
Arc (for resource)
Duration
1 Token
Infinite Tokens
Zero Token
Multiple Tokens
Marking
Intuitive,
but less flexible
Activity-based Formalism
10
Proposed Methodologies
Formalisms for Discrete-Event Systems (DES)
Event-based Formalism
An event graph is a graphical formal model consisting of a set of event nodes and a set of directed edges. It provides a complete description of a discrete event system (DES) in a concise and clear manner, and its execution rules have to be described unambiguously.
A graphical model can be specified in algebraic form (to be analyzed by a human logically) as well as in computer-readable form (to be executed on a computer).
Originating Event Destination EventDestination Event
Scheduling Edge
Canceling Edge
Time Delay ConditionCondition
State Change State Change State Change
Precise,
but Sophisticated
11
Proposed Methodologies
Formalisms for Discrete-Event Systems (DES)
State-based Formalism
Generator Queue Server
State:How many arrivals are generated?
State:How many are in the queue?- When the queue has finite capacity, how many are requested but pending to enqueuer?
Arrive
Request toEnqueue
Dequeue
EnqueueRequest toStart
Start
Ready toDepart
Depart
State:How many being served?- When the queue has finite capacity, how many are requested but pending to start?- How many are “stuck” due to subsequent procedures?
12
Proposed Methodologies
Object-Oriented DES (O2DES) Framework
Object-Oriented Formalism
Input 1(parameters)
Output 1(parameters)
Input 2(parameters)
Output 2(parameters)
Input M(parameters)
Output N(resource, quantity)
……
……
Parent Module
Sub Module1
In-1
In-2
Out-1
Sub Module2
In-1
In-2
Out-1
Evt1
Evt2
Evt3
Evt5
Evt4
Evt6
Evt7 Evt8
Out-2
13
Proposed Methodologies
Object-Oriented DES (O2DES) Framework
Big Data Analytics / Deep Learning
Optimal Configuration
SIMULATION ANALYTICS
Random Seed
SIMULATION IN OPTIMIZATION
Sim.-based Optimization (Search Algorithm / Ranking & Selection)
Event 1
Event 2
Event 3
Event 1
Executor
Entity 1
Entity 3
Entity 2
Load 1
Load 2
STATE
EVENTS
ConfigurationEn
viro
nmen
t /
Initi
al S
tate
Obj
ectiv
es
Event 1
Event 2
Event 4
Event 5
Event 3
Event 3(OPT.)
Event 1
Event 2
Event 4
Event 5
OPTIMIZATION
OPTIMIZATION IN SIMULATION
Conditions
ASSETS( Static(
SANDBOX( Dynamic(
Seamless Integration betweenSimulation and Optimization
14
Proposed Methodologies
Object-Oriented DES (O2DES) Framework
Hierarchical Modelling with Modularization
15
Future Trends
High-Level Architecture (HLA)
Run-Time Infrastructure (RTI)
High-Level Architecture (HLA)
Federate 4 Federate 5 Federate 6
SubscribePublish SubscribePublish SubscribePublish
Sub-Federation
Run-Time Infrastructure (RTI)High-Level Architecture (HLA) conforms to IEEE 1516-2010
Federate 1 Federate 2 Federate 3
SubscribePublish SubscribePublish SubscribePublish
Federation
16
Future Trends
High-Level Architecture (HLA)
Multi-Module Distributed Run
Multi-Formalism Synchronized Run
Integrated Simulation + Visualization + Optimization + Learning
Simulation – Emulation Seamless Transition
Digital Twins Development and Experiment
17
Future Trends
Containerization with Microservice
Monolithic Application Scaling Approach
without
Functional layers such as
front endbusiness logic
and data storage
A simple update can have unforeseen effects on the rest of the tier, and so a change to an application component requires its entire tier to be retested and redeployed.
(Simulation Modules)
18
Future Trends
Containerization with Microservice
Microservices Application Scaling Approach
Microservices are independent components that work together to deliver the application's overall functionality. The term microservice emphasizes that applications should be composed of services small enough to reflect independent concerns.
with
scale + agility + reliability
(Simulation Modules)
19
Development of Smart Digital Twin for Maritime Industry via O2DES Framework
Thank you!
Q&A