Digital Twins Technology – IDA Mechanical & IPU
2019-03-26 Digital Twins Technology - IDA Mechanical & IPU 1
17:00 – 17:05 Welcome & Introductions
17:05 – 17:20 Introducing IPU & Digital Twin Technology
– Søren Merit, CEO
17:20 – 17:45 Digital Twins for Condition Based Maintenance of
Refrigeration Containers
– Ragnar Ingi Jónsson, Specialist Engineer
17:45 – 18:15 Break – Sandwich & Networking
18:15 – 18:35 Model-in-Loop Software Development for
Automation of Heavy Duty Machinery
– Kevin Rice, Senior R&D Engineer
18:35 – 18:55 Virtual Models for Product Analysis and
Manufacturing Processes
– Nikolas Aulin Paldan, Specialist Engineer
18:55 – 19:00 Final Remarks & Questions
Introducing IPU & Digital Twin Technology
Søren Merit
CEO at IPU,
Technology Driven
Business Innovation,
M.Sc., B.Com.
(+45) 40 90 46 30
We are a spin-off of the Technical University of Denmark
70% Technology development for
industry
30% Research projects
50% of projects are we working
with DTU researchers
130 million DKK in donations
supporting DTU research
Started in
1956 by 4 DTU professors
Independent commercial
foundation
Purpose to facilitate use of
new technology in Danish
industry
2019-03-26 Introducing IPU & Digital Twin Technology, Søren Merit, [email protected]
We Develop Solutions to Complex Technology Challenges
We help our clients speed up development
… and reduce technical risks
… and manufacturing uncertainties
Discovery Basic Research Applied Research
Product and
Manufacturing
Development
Production
• Technology Search
• Proof of Concept
• Feasibility Study
• Test setup
• Data analysis
• Prototyping & development
• Modelling & Simulations
• Digital Twin
• Specialists in multi-disciplinary product- and manufacturing technology development
• Team of international specialists
2019-03-26 Introducing IPU & Digital Twin Technology, Søren Merit, [email protected]
IPU Key Expertises
2019-03-26 Introducing IPU & Digital Twin Technology, Søren Merit, [email protected] 5
We can help you…
developing and optimizing materials
and surfaces processes for specific
purposes, including analyzing
problems and malfunctions.
Our core strengths are metallic and
polymer materials as well as
electroplating and corrosion
Materials Types and Choices
Surface Treatment
Corrosion and Wear
Protection
Software and simulation
Modelling of Cooling
Processes
Energy Optimisation
We can help you…
analyzing thermodynamic and heat
transfer processes and their
components.
We optimize system performance
and efficiency using tailored
simulation models and the latest
R&D expertise.
Digital twins
Fault detection
FEM analysis
Machine learning
We can help you…
developing digital models of
physical system in order to perform
simulations in a fast and safe
environment.
We develop digital twins, perform
big data analyses, decision
algorithms (AI), machine learning
and visual pattern recognition
Advanced materials
and surfaces
Thermodynamics
and energy
Physical systems
modelling
Software & Algorithm
Development
Condition-Based Monitoring
System Analysis
We can help you…
modelling, analyzing and
developing complex autonomous
systems, robotics and automation
of systems and processes.
Combining development of
mechanical design and hardware
with control systems and software
Autonomous
systems and
automation
How We Work with Complex Systems
2019-03-26 Introducing IPU & Digital Twin Technology, Søren Merit, [email protected] 6
Understand
physics
Multi physics
modelling Prototyping
and tests
Data
analysis Machine
Learning
Understand business case and process
We Develop Solutions to Complex Technology Challenges
2019-03-26 Introducing IPU & Digital Twin Technology, Søren Merit, [email protected] 7
IPU has developed a
software solution
offering predictive
maintenance and
conditions-based fault
detection of refrigerated
containers enabling
significant savings in
costs related to physical
inspection.
IPU has developed an
automation concept for
the new ESO telescope,
using a safe chemical
cleaning process, that
strips the mirror coating
during planned
maintenance, without
altering the fragile mirror
substrate.
IPU has developed an autonomous
systems solution for heavy duty
construction machinery. IPU developed
automation control system, retrofit
hardware components, develop control
software and operator user interface.
Digital twin based hardware and software
in the loop development
Digital Twin Technology
Welcome to the Most Hyped Technology!
Gartner’s Hype Cycle, August 2018
2019-03-26 Introducing IPU & Digital Twin Technology, Søren Merit, [email protected]
What is a Digital Twin?
Digital model of the elements and
dynamics of how a product, process or
service operates
10
..applied in development
..and in operations
2019-03-26 Introducing IPU & Digital Twin Technology, Søren Merit, [email protected]
What is Digital Twin?
Digital model of the elements and
dynamics of how a product, process or
service operates
11
Development Operations
• CAD interacting with multi-
physical simulations
• Developing and testing
software (Model-in-Loop)
• Developing and testing
components
(Hardware-in-loop)
• Installation: Calibrating and
adjusting
• Monitoring: Comparing
sensor data with simulation
results
• Optimizing: Adjusting
system for wear and
external conditions
2019-03-26 Introducing IPU & Digital Twin Technology, Søren Merit, [email protected]
Why use Digital Twins
12
Automotive
Aerospace
Benefits of digital twins
• Faster insights – Fail fast succeed faster
• Cheaper and faster tests
• Feasible to explore extreme conditions
• Understand dynamics better
• Ability to predict and adjust – during operations
• Faster update with minimal stops in operations
2019-03-26 Introducing IPU & Digital Twin Technology, Søren Merit, [email protected]
Digital Twins for Condition-Based
Maintenance of Refrigeration Containers
Ragnar Ingi Jónsson
Specialist Engineer at IPU,
Physical System Modelling &
Conditions-Based Monitoring,
M.Sc., Ph.D.
(+45) 45 25 41 86
Maersk Motivation & Goals
2019-03-26 Digital Twins for Condition-Based Maintenance of Refrigeration Containers, Ragnar Ingi Jónsson, [email protected] 14
• Remote Container Management (RCM)
• Connectivity and transparency – being in control
• Improved customer experience – documentation
• Monetary savings – maintenance and operation
How can we improve the efficiency of the
reefer maintenance operations, cargo
safety and energy consumption?
• Vast funds spent on pre-trip inspections (PTI)
approx. US$ 750 yearly, per unit.
Timeline & Overview
2019-03-26 Digital Twins for Condition-Based Maintenance of Refrigeration Containers, Ragnar Ingi Jónsson, [email protected] 15
• 2010 IPU/Maersk collaboration on Reefer Alarm
Prediction System (RAPS/ePTI) of the RCM system
• 2012 Agreement with Ericsson and AT&T for
hardware and data infrastructure.
• Satellite communication installed on 400 vessels
• Local GSM communication between container and
vessels
• 2015 RCM system launch may 1st.
• 2018 IPU to update RAPS/ePTI with new reefer
models, detections updates and other features.
RAPS / ePTI – Step-by-Step
2019-03-26 Digital Twins for Condition-Based Maintenance of Refrigeration Containers, Ragnar Ingi Jónsson, [email protected] 16
1. Sensor readings are collected and saved on
each reefer.
2. Reefer data is send to vessel via on-board
local GSM or commercial GSM while in land
3. Satellite communication sends reefer data
to head quarters
4. Individual reefer data is processed through
simulation models and fault detection
algorithms.
5. Alarms and warnings are reviewed and
appropriate actions taken service
ordered if needed.
Reefer Models for Fault Detection
2019-03-26 Digital Twins for Condition-Based Maintenance of Refrigeration Containers, Ragnar Ingi Jónsson, [email protected] 17
• Model requirements
• Fast calculation models
• Include all major properties
• Basic refrigeration system
• Heat uptake, heat release, power
consumption
• High and low pressure parts
• Internal temperatures
• Other temperatures included
(ambient, cooling water, reefer, set point)
• The general refrigeration system model
• Refrigeration type determines, compressor type, whether there is e.g. internal heat exchange,
economizer etc.
Fault Detection Algorithm Overview
2019-03-26 Digital Twins for Condition-Based Maintenance of Refrigeration Containers, Ragnar Ingi Jónsson, [email protected] 18
Normal operation Fault
Normal operation Fault
Model
Residual
(difference)
Statistical
Detection
Threshold
Comparison
Detection
Threshold
Normal
Behavior
Faulty
Behavior
Detection Results
measured data
time stamp
operational data
other…
measured data
not used in
model
simulated
values component-wise
residual data
setpoint
ambient conditions
cargo info
etc.
Reefer Data
Statistical Change Detection Overview
• Statistical model of normal and faulty behavior, and
signal-to-noise ratio affects the detection time
• Cumulative log-likelihood is typically used with a
cumulative threshold
2019-03-26 Digital Twins for Condition-Based Maintenance of Refrigeration Containers, Ragnar Ingi Jónsson, [email protected] 19
Normal operation Fault
15:00 18:00 21:00 00:00 03:00 06:00 09:00 12:00 15:00 18:00 21:00 00:000
20
40
60
80
100
120
140
160
180
200
time [hour:minute]
Evap. decis.
Pdisc decis.
Normal operation Fault
Statistical Change Detection Example
• Reduced flow of air in evaporator
2019-03-26 Digital Twins for Condition-Based Maintenance of Refrigeration Containers, Ragnar Ingi Jónsson, [email protected] 20
Normal Fault
Normal Fault
Normal Fault
IPU Deliveries & Results for Maersk
• Backend computational core of the RAPS/ePTU systems, processing approximately 200.000
hourly updates providing alarms and warnings to monitoring systems.
• Calibrated high performance models of all reefer refrigeration system and their variants
• The majority of technical issues detected before cargo is affected ($), efficiency of the reefer
maintenance operations could be improved by over 40%
2019-03-26 Digital Twins for Condition-Based Maintenance of Refrigeration Containers, Ragnar Ingi Jónsson, [email protected] 21
ePTI which was developed in close coordination with IPU, has not only resulted in significant direct cost
reductions. ePTI also gives Maersk Line full cold chain transparency, improves and optimizes operational
processes and offers faster turn times benefiting our customers. All this whilst ensuring that the equipment we
release for our customer is in fully cargo worthy condition suitable for transport of temperature sensitive cargo.
We look forward to further develop the ePTI algorithm together with IPU to get the full benefits from the huge
amount of data now made available through RCM.
— Lars-Henrik Jensen, Operations Manager, Remote Container Management, Maersk Line
“
“
Lessons Learned & Recommendations
• Technical Recomentations
• Exploit your physical understanding of the problem or data-set – especially when developing the digital
twin of refrigeration system.
• Balance the amount of time used on cleaning up data and improving the detection algorithm.
• Time is on your side, a longer timeline will limit false predictions.
• Ensure constant alignment with the process – physical experiments are needed.
• Organization and project alignment
• Secure the easy wins first and build from there based on risk assessment, FMEA, experience.
• Business case with Digital twin – ensure the efforts brings value to the organization.
• Align with stakeholders and involve their knowledge, inputs and ideas if applicable.
• It will most likely never be perfect, set reasonable goals.
2019-03-26 Digital Twins for Condition-Based Maintenance of Refrigeration Containers, Ragnar Ingi Jónsson, [email protected] 22
Break – Sandwich & Networking
Model-in-Loop Software Development for
Automation of Heavy Duty Machinery
Kevin Rice
Senior R&D Egineer at IPU,
Autonomous System,
Mechatronics and Controls,
M.Sc.
(+45) 29 93 47 93
Automation Project Overview (On-going & Confidential)
• Costumer Owns and Operates Heavy-Duty Machinery
• Task to develop automation solution to ensure high machine efficiency.
2019-03-26 Model-in-Loop Software Development for Automation of Heavy Duty Machinery, Kevin Rice, [email protected] 25
Operator & GUI
Automation
Controller Construction
Machine
Sensor &
Positioning
Actuator
Input
System Data
& User Input
Difficult to find skilled operators. Training takes time
with high chance of new operators leaving.
• Automation Controls
System Architecture
• Retrofit Machine with
new Hardware
• Automation software
and User Interface
Software development with Digital Twin
2019-03-26 Model-in-Loop Software Development for Automation of Heavy Duty Machinery, Kevin Rice, [email protected] 26
• Large efforts on automation software
• Adaptive/learning path planning depending on soil hardness.
• Sensor algorithms – fusion techniques, global/local location,
redundant sensors for safety.
• Safe-zone of operation – not damaging itself or surrounding
objects.
• Software complexity, and low availability of machine,
demands development with Digital Twin
• Hardware-in-Loop (HiL)
development.
• Model-in-Loop (MiL)
development.
Developing the Digital Twin
2019-03-26 Model-in-Loop Software Development for Automation of Heavy Duty Machinery, Kevin Rice, [email protected] 27
• Digital Twin must represent the actual system through Cyber-Physical Modelling
• Excavator dynamics / kinematics, hydraulic actuation system, sensor noise, …
• Certain Components modelled in hardware – electronics of sensor, valve dynamics, …
[1] [2]
Digital Twin and Physical System Mismatch
• Digital Twin will deviate from physical system
• Sufficient accuracy is obtained from system
understanding, engineering intuition and
experience.
• Digital Twin accuracy vs. Modelling Efforts vs.
Simulation Time should be considered
• Digital model mismatch results in higher
software quality
• Algorithms are implemented to support model and
physical system behavior.
• Typically through calibration options, resulting in
software supporting variations
(Manufacturing, mechanical wear, etc.)
2019-03-26 Model-in-Loop Software Development for Automation of Heavy Duty Machinery, Kevin Rice, [email protected] 28
[3]
Model-in-Loop Development Workflow
• System and Integration testing is fast with MiL
development. Limiting time consuming physical
tests.
• Continuous Integration testing with simulated
operating conditions, and failure modes.
• High confidence in algorithms in other software
functionality.
2019-03-26 Model-in-Loop Software Development for Automation of Heavy Duty Machinery, Kevin Rice, [email protected] 29
Requirements
& Specifications
System
Design
Architecture
Design
Module
Design
Implementation
Functional &
Unit Testing
Integration
Testing
System
Testing
Acceptance
Testing
Implementation Unit and
functional testing
System & integration
testing
Model-in-Loop Recommendation & Final Remarks
• Fast development flow, and possible to development physical system in parallel with software.
• Testing all operating modes, including failure modes, is fast with Digital twins and Simulation.
(Expensive and time consuming tests on physical system, safety issues, …)
• Integration and system test should be performed on physical system in parallel with
development on digital twin. Software and control strategies will require online tuning
2019-03-26 Model-in-Loop Software Development for Automation of Heavy Duty Machinery, Kevin Rice, [email protected] 30
Mechanical & Electronics
Software Design
Time – Traditional Development
Mechanical & Electronics
Software Design Smarter Products
Time – Development with MiL
[1] H. Feng, C. B. Yin, W. Weng, W. Ma, J. Zhou, W. Jia & Z. Zhang, Mechanical Systems and Signal Processing, Volume 105, 15 May 2018, Pages 153-168
[2] T. O Andersen, Department of Energy Technology Lecture Notes, Aalborg University, 2nd Edition, 2003.
[2] L. Schmidt, PhD Dissertation, Department of Energy Technology, Aalborg University, 2014
Virtual Models for Product Analysis
and Manufacturing Processes
Nikolas Aulin Paldan
Specialist Engineer at IPU,
Integrated Product &
Process Technology,
M.Sc.
(+45) 45 25 46 16
Topics
• Background & Introduction
• Virtual models for product analysis &
manufacturing processes
• Example of Virtual Process Model
• Summary
2019-03-26 Virtual Models for Product Analysis and Manufacturing Processes, Nikolas Aulin Paldan, [email protected] 32
Background & Introduction I
• At IPU for 15 years, Specialist engineer in Integrated Product & Process Technology
• Background is in mechanical engineering for equipment and tooling.
2019-03-26 Virtual Models for Product Analysis and Manufacturing Processes, Nikolas Aulin Paldan, [email protected] 33
Background & Introduction II
• Heavy user of numerical tools to create virtual models of processes performed by machines
and tooling.
• Many types of virtual models. Examples shown here will be focused on models for Finite
Element Analysis.
2019-03-26 Virtual Models for Product Analysis and Manufacturing Processes, Nikolas Aulin Paldan, [email protected] 34
Calculation of steady state tool temperature in a cyclic polymer welding process
35
Virtual Models for Product Analysis
& Manufacturing Processes
Optimizing Manufacturing Processes Using Virtual Models
2019-03-26 Virtual Models for Product Analysis and Manufacturing Processes, Nikolas Aulin Paldan, [email protected] 36
Design of Process
Equipment
Process
Optimization
& Test
Process
Modelling
Possibility to Speed-Up Development and Reducing Testing Efforts
2019-03-26 Virtual Models for Product Analysis and Manufacturing Processes, Nikolas Aulin Paldan, [email protected] 37
Design Verification Testing
Less than 100% workload
Virtual Design
Verification
testing
38
Example of Virtual Process Model
• For IPU Use numerical models in many projects.
• Due to confidentiality agreements IPU examples can not be shown…
• Example from a totally different business area, in which they have fully embraced the use of
virtual models:
• Reduced testing of spot-welds in Automotive manufacture
• The optimization of the spot-welding process normally requires a lot of physical testing –
being able to do testing on virtual models has been widely adapted.
2019-03-26 Virtual Models for Product Analysis and Manufacturing Processes, Nikolas Aulin Paldan, [email protected] 39
Spot-welding in Cars (Tesla Production Line)
2019-03-26 Virtual Models for Product Analysis and Manufacturing Processes, Nikolas Aulin Paldan, [email protected] 40
Youtube.com
• The following slides borrowed from the Company Swantec
• Spin-out from DTU in 1999, software SORPAS for the simulation of resistance welding –
costumers
2019-03-26 Virtual Models for Product Analysis and Manufacturing Processes, Nikolas Aulin Paldan, [email protected] 41
Material Challenges for Automotive Spotwelding Processes
• Conventional steels - Relative simple to
spotweld
• Mild steels
• Interstitial free (IF) steels
• Bake hardenable (BH) steels
• High strength low alloy (HSLA) steels
• Carbon Manganese (CMn) steels
2019-03-26 Virtual Models for Product Analysis and Manufacturing Processes, Nikolas Aulin Paldan, [email protected] 42
• Advanced High strength steels (AHSS) -
Can be very difficult to spotweld
• DP – Dual Phase
• CP – Complex Phase
• TRIP – (TRansformation Induced Plasticity)
• Mart. – Martensite steel
• TWIP – (Twinning Induced Plasticity)
• 3rd Gen AHSS
Principle of Resistance Welding Regarding Current
2019-03-26 Virtual Models for Product Analysis and Manufacturing Processes, Nikolas Aulin Paldan, [email protected] 43
dttItRtQ
t
0
2 )()()(Joule heating:
Principle of Resistance Welding Regarding Force
2019-03-26 Virtual Models for Product Analysis and Manufacturing Processes, Nikolas Aulin Paldan, [email protected] 44
(3)
Weld
F I
(2)
Squeeze
F F
(4)
Hold
(5)
Weldability Lobe – Welding Process Window
2019-03-26 Virtual Models for Product Analysis and Manufacturing Processes, Nikolas Aulin Paldan, [email protected] 45
Welding current
Fo
rce
Minimum
weld size Acceptable
weld size
Expulsion
Zone
Typical
Nugget failures
Importance of Welding Process Optimization
2019-03-26 Virtual Models for Product Analysis and Manufacturing Processes, Nikolas Aulin Paldan, [email protected] 46
Undersized
weld
Acceptable
weld size
Expulsion
Zone
Spot Welding - Three Sheets of Steel
2019-03-26 Virtual Models for Product Analysis and Manufacturing Processes, Nikolas Aulin Paldan, [email protected] 47
Welding Process Optimization – Welding Ranges
• Spot welding mild steel sheets
2019-03-26 Virtual Models for Product Analysis and Manufacturing Processes, Nikolas Aulin Paldan, [email protected] 48
Weld Growth Curve Weldability Lobe
Weld Strength Testing and Failure Modes
2019-03-26 Virtual Models for Product Analysis and Manufacturing Processes, Nikolas Aulin Paldan, [email protected] 49
• Weld strengths
• Tensile shear strength
• Cross tension strength
• Peel strength
• Failure modes
• Plug (button) failure
• Interface failure
Plug failure Interface failure
Weld Strength Testing and Failure Modes
2019-03-26 Virtual Models for Product Analysis and Manufacturing Processes, Nikolas Aulin Paldan, [email protected] 50
• Tensile-shear test – plug failure
Weld Strength Testing and Failure Modes
2019-03-26 Virtual Models for Product Analysis and Manufacturing Processes, Nikolas Aulin Paldan, [email protected] 51
• Weld strengths based on welding process simulation
• Output weld strengths for structural and crash modeling
Final Remarks & Questions
Nikolas Aulin Paldan
Specialist Engineer
(+45) 45 25 46 16
Kevin Rice
Senior R&D Engineer
(+45) 29 93 47 93
Ragnar Ingi Jónsson
Specialist Engineer
(+45) 45 25 41 86
Søren Merit
CEO
(+45) 40 90 46 30
… see more cases at ipu.dk