From Computers in Industry A. Phua, C.H.J. Davies, G.W. Delaney Presenter ：CHEN,YOU-SHENG (Shane) 2022/07/01

1. From Computers in Industry
A. Phua, C.H.J. Davies, G.W. Delaney
Presenter ：CHEN,YOU-SHENG (Shane) 2022/07/01
A digital twin hierarchy for
metal additive manufacturing

2. JCR For Computers in Industry
/27
1

3. Introduction
01
The digital twin
hierarchy
02
Discussion and
outlook
03
TABLE OF CONTENTS
/27
2

4. Purpose Findings
Digital twin hierarchy for
additive manufacturing (AM)
provides a developer framework
for engineering digital twins, both
for AM and other intelligent
manufacturing systems
The development of a metal AM
digital twin can be organised logically
into a hierarchy of four levels of
increasing complexity
Abstract
/27
3

5. Methodology Originality
• Document Analysis
• Build a process diagram
• Provide estimates for their
respective Technology
Readiness Level (TRL)
index
Outcome
Demonstrates how
elements of the digital
twin fit together for part
qualification and
optimisation
• The hierarchical
framework provides a
unified ontology for the
unique needs of AM
• Classify digital twin
hierarchy with estimation
of technology readiness
level
VALUABLE USES OF 3D PRINTING
/27
4

6. Introduction
01
/27
5

7. Digital twin
The first published definition of a digital twin
appeared in NASA’s
Simulation and Modelling Roadmap
published in 2010/11
• Its quickly became the central
framework employed for virtual
spacecraft testing and failure prediction
(Tuegel et al., 2011)
• Also inspired many other applications in
healthcare, smart cities and
manufacturing (Fuller et al., 2020)
Source: https://www.nasa.gov/pdf/501321main_TA11-
MSITP-DRAFT-Nov2010-A1.pdf
/27
6

8. Introduction
Industry 4.0
Smart
Manufacturing
Digital twin
• Industry 4.0 focus on the enterprise scale, which provide a simulated replica of the
manufacturing process, factory, operations and related logistics (Negri et al., 2017; Tao et al., 2018)
• Smart Manufacturing has emerged from Industry 4.0 as a key concept to imbue automated
manufacturing processes with machine intelligence; real-time data monitoring, planning,
control and optimisation (Kusiak, 2019).
/27
7

9. Introduction
• There is still a clear research gap in
the development of an interconnected
digital twin for both prediction, sensing
and control of a metal AM printer
• Authors had formulated a digital twin
hierarchy specific to the complexities of
metal AM
• They focus primarily on powder based
metal AM methods which often deal
with the most complexity
Metal 3D Additive Manufacturing Laser Equipment (Model TSLM by tsemc)
/27
8

10. 02
• The implicit digital twin
• The instantiated digital twin
• The interfaced digital twin
• The intelligent digital twin
The digital
twin hierarchy

11. The digital twin hierarchy
/27
10

12. The digital twin hierarchy
Fig. 2. The number of published articles relating to the core elements of the AM digital twin. Articles were identified using
the labelled combination of key search terms and quantified by the filled area on the graph. /27
11

13. The implicit digital twin
• using multi-physics modelling to provide a way to predict
state variables and make predictions about future states
Work
• often limited by their lengthy computation time of hours or
days (Cummins et al., 2021)
• begin at the large computational cost of a high-fidelity
simulation of the physical print
Challenges
information is inferred from the pre-print data, part geometry, machine settings and
build paths in preparation slicing software (Anon, 2020).
/27
12

14. The implicit digital twin
Fig. 3. Multi-physics simulations for metal additive manufacturing process modelling. (Phua et al., 2021; Khairallah et al., 2016;
Cummins et al., 2021; Koepf et al., 2019; Williams et al., 2018).
/27
13

15. The implicit digital twin
Surrogate modelling can be employed to reduce computational cost by substituting high
fidelity physics models with fast-executing surrogates
Reduced Order
Modelling (ROM)
• ROM projects the
equations onto a
subspace of reduced
dimension (Lucia et al.,
2004)
data-fit models
• data-fit surrogates regress
and inter-polate a set of
data generated from the
original physics model
machine learning
/CNN or RNN
• present a potential way to
make predictions across
a range of physics and
length scales
• leading to the ability to
train larger models
/27
14

16. The implicit digital twin
/27
15

17. The instantiated digital twin
• extends on the Implicit Digital Twin by introducing real-time
sensor data (Gaikwad et al., 2020; Yavari et al., 2021)
• addition of in-situ sensors from the printer
Work
• a high acquisition rate data processing and storage
• many state variables cannot be monitored by in-situ
sensing during print such as x-ray imaging, microstructure
and residual stresses
Challenges
• Current computer vision based systems show strong
potential to non-intrusively identify defects within the part or
the powder bed
Potential
/27
16

18. The instantiated digital twin
/27
17

19. The interfaced digital twin
• additions are the use of logic controllers, and interfacing
hardware with the printer to dynamically update print
instructions
• directly use sensor data to proportionally adjust specific
build parameters
Work
• require extensive parameter tuning
• controlling the multiple quantities of interest
Challenges
• For more sophisticated control decisions such as defect
correction or property optimisation, more intentional
decision making is required.
Potential
/27
18

20. The interfaced digital twin
/27
19

21. The intelligent digital twin
• the distinguishing feature of the Intelligent Digital Twin is its
control policy for complex decision making
• the human user is no longer required
Work
• with little to no application for AM systems
• however these intelligent systems still require extensive
hand engineering of reward functions
Challenges
•these algorithms benefit from the cost-effective and
scalable nature of virtual simulation training (DRL)
•can be trained directly from the physical printing systems
•Potential
/27
20

22. The intelligent digital twin
/27
21

23. The intelligent digital twin
/27
22

24. Discussion
and outlook
03
/27
23

25. Discussion and outlook
/27
24
TRL Current NASA usage[14] European Union[15]
1 Basic principles observed and reported Basic principles observed
2 Technology concept and/or application formulated Technology concept formulated
3
Analytical and experimental critical function and/or
characteristic proof-of concept
Experimental proof of concept
4
Component and/or breadboard validation in laboratory
environment
Technology validated in lab
5
Component and/or breadboard validation in relevant
environment
Technology validated in relevant environment
(industrially relevant environment in the case of key
enabling technologies)
6
System/subsystem model or prototype demonstration
in a relevant environment (ground or space)
Technology demonstrated in relevant environment
(industrially relevant environment in the case of key
enabling technologies)
7
System prototype demonstration in a space
environment
System prototype demonstration in operational
environment
8
Actual system completed and "flight qualified" through
test and demonstration (ground or space)
System complete and qualified
9
Actual system "flight proven" through successful
mission operations
Actual system pro

26. Discussion and outlook
/27
25

27. Discussion and outlook
To organises the pertinent AM literature relating to digital twin development and provides
estimates for their respective Technology Readiness Level (TRL) index
TRL 9 TRL 4-6 TRL 3-7
TRL the
lowest 3
/27
26

28. Discussion and
outlook
● Today, commercial metal AM printers already feature basic sensor
suites (Anon, 2022) (TRL 9)
● Computer vision has seen strong industrialisation potential (Scime et
al., 2020) (TRL 7)
● We suggest the use of probabilistic modelling frameworks that
can dynamically update their virtual representation
● Emerging use of DRL offers strong potential as a means of
dynamically constructing high-utility control policies
● DRL need Looking at benchmarking their performance in real
printing environments
● This hierarchy can also be applied to other metal AM technologies
/27
27

29. THANK YOU!

30. RESOURCES
• A. Phua, C.H.J. Davies, G.W. Delaney,A digital twin hierarchy for metal additive
manufacturing, Computers in Industry, Volume 140, 2022, 103667, ISSN 0166-3615,
https://doi.org/10.1016/j.compind.2022.103667.
• PPT template- 3D Printing Day 3D Printing Day | Google Slides & PowerPoint
(slidesgo.com)
• P7 15 Microsoft Stock images (royalty-free images)
• 友晁能源材料股份有限公司-金屬3D積層製造雷射設備 http://www.tsemcorp.com/Metal-
3D-printing.html