Australia’s National Science Agency
Data-efficient Machine Learning for
Geometry Prediction in Cold Spray
Additive Manufacturing
Daiki Ikeuchi, Alejandro Vargas-Uscategui, Xiaofeng Wu, Peter C. King | September 2020
For Further Information:
“Data-Driven Overlapping-Track Profile Modeling
in Cold Spray Additive Manufacturing”
Citation: Ikeuchi, D., Vargas-Uscategui, A., Wu, X. et al.
Data-Driven Overlapping-Track Profile Modeling in Cold
Spray Additive Manufacturing. J Therm Spray Tech 33, 530–
539 (2024). https://doi.org/10.1007/s11666-024-01733-3
Our Related Work (Open Access):
“Neural Network Modelling of Track Profile in
Cold Spray Additive Manufacturing”
Citation: Ikeuchi, D.; Vargas-Uscategui, A.; Wu, X.; King,
P.C. Neural Network Modelling of Track Profile in Cold
Spray Additive Manufacturing. Materials 2019, 12, 2827.
https://doi.org/10.3390/ma12172827
“Data-Efficient Neural Network for Track Profile
Modelling in Cold Spray Additive Manufacturing”
Citation: Ikeuchi,D.; Vargas-Uscategui, A.; Wu, X.; King,
P.C. Data-Efficient Neural Network for Track Profile
Modelling in Cold Spray Additive Manufacturing. Appl. Sci.
2021, 11, 1654. https://doi.org/ 10.3390/app11041654
Introduction
osition – The basic concept
Additive Manufacturing | Dr. Alejandro Vargas-Uscategui
ted, high pressure gas is accelerated to supersonic velocities through a de Laval
zle.
allic Particles (< 50 µm) are fed into the gas stream.
particles impact onto a substrate, forming a deposit.
Why Cold Spray Additive Manufacturing (CSAM)?
• Solid-state deposition with low-oxygen contents
• Avoidance of melting-induced microstructure changes
• Stable fabrication with high deposition rate with robotic system
• Large-scale manufacturing without a protective atmosphere
Background and Challenge
Challenge: Geometric Modelling and Control
Approach Mathematical Data-driven
Advantage
• Complete profile prediction
• Physically consistent
• Potential for higher
predictive accuracy
• Learning capability
Disadvantage
• Limited predictive accuracy
• May require multiple forms
of underlying math models.
• Limited to prediction of
geometric features only
• Need large dataset
Question: Can we address the problem of data-
scarcity in CSAM for data-driven modelling?
Objectives
Investigate data-efficient machine learning modelling
for an overlapping-track deposit profile in CSAM
Complete
overlapping-track
profile prediction
Include deposition
surface information
explicitly
Comparative study
against previously
proposed models
Data-efficiency
achieved without
more experiments
Method: Experimental
Fabrication of Two-Track Profiles:
• A commercial Impact Innovation 5/11 cold spray gun
• Purity grade-2 titanium feedstock
Design of Experiments:
• Traverse speed, and Standoff distance and Overlapping ratio
designed in a full factorial manner (48 overlapping tracks)
CSAM Overlapping Two-Track Profile
Method: Modelling Strategy (1)
𝑥 =
𝑇𝑟𝑎𝑣𝑒𝑟𝑠𝑒 𝑠𝑝𝑒𝑒𝑑
𝑃𝑜𝑙𝑎𝑟 𝑎𝑛𝑔𝑙𝑒
𝑆𝑢𝑏𝑠𝑡𝑟𝑎𝑡𝑒 𝑡𝑦𝑝𝑒
𝑃𝑜𝑙𝑎𝑟 𝑙𝑒𝑛𝑔𝑡ℎ
𝑦 = ∆ 𝑝𝑜𝑙𝑎𝑟 𝑙𝑒𝑛𝑔𝑡ℎ
Method: Modelling Strategy (2)
𝑦 𝑥 = 𝑓 𝑥 + ℎ 𝑥
Gaussian Process Regression (GPR) Explicit Mean Function
• 𝑚 𝑥 = 0
• Kernel: 3/2 Matern function
• K-fold cross validation (k = 5)
• 48 overlapping-track profiles
• 151 sampling points taken
between ± 5 mm around Tool
Centre Point
• Substrate type represented as -1
or 1
• Acts as an approximate solution
• Can be based on domain
knowledge or approximate model
• Mathematical Gaussian
superposing model used here
Simply added numerically
Result: Modelling Validation
Performance of ƒ (x):
• MSE = 0.003573
• Evaluated on testing dataset
• Cross-validated
Performance of y(x):
• MSE = 0.0002183
• Evaluated on testing dataset
• Cross-validated
Result: Analysis and Comparison
• Both pure GPR and data-
efficient GPR models
performing better than
mathematical Gaussian
model in mean error %
• The data-efficient GPR
model showed narrower
range up to 75% quartile
• Maximum error % lower for
GPR models than
mathematical Gaussian
model
Mean % 2.189 1.449 0.5748
Result: Visual Prediction Results
Conclusion
Complete
overlapping-track
profile prediction
Include deposition
surface information
explicitly
Comparative study
against previously
proposed models
Data-efficiency
achieved without
more experiments
Future works:
• Development of in-situ online sensing system to extend the modelling
approach for overlayer manufacturing
• Demonstration of the modelling approach on more complex
deposition scenarios (e.g., on curved or complex surface for repairing)
Australia’s National Science Agency
Daiki Ikeuchi
The University of Sydney
CSIRO Manufacturing
daiki.ikeuchi@sydney.edu.au
Thank you
For Further Information:
“Data-Driven Overlapping-Track Profile Modeling
in Cold Spray Additive Manufacturing”
Citation: Ikeuchi, D., Vargas-Uscategui, A., Wu,
X. et al. Data-Driven Overlapping-Track Profile
Modeling in Cold Spray Additive Manufacturing. J
Therm Spray Tech 33, 530–539 (2024).
https://doi.org/10.1007/s11666-024-01733-3
Our Related Work (Open Access):
“Neural Network Modelling of Track Profile in
Cold Spray Additive Manufacturing”
Citation: Ikeuchi, D.; Vargas-Uscategui, A.; Wu, X.; King,
P.C. Neural Network Modelling of Track Profile in Cold
Spray Additive Manufacturing. Materials 2019, 12, 2827.
https://www.mdpi.com/1996-1944/12/17/2827
“Data-Efficient Neural Network for Track Profile
Modelling in Cold Spray Additive Manufacturing”
Citation: Ikeuchi,D.; Vargas-Uscategui, A.; Wu, X.; King,
P.C. Data-Efficient Neural Network for Track Profile
Modelling in Cold Spray Additive Manufacturing. Appl. Sci.
2021, 11, 1654. https://www.mdpi.com/2076-3417/11/4/1654

Data-efficient Machine Learning for Geometry Prediction in Cold Spray Additive Manufacturing

  • 1.
    Australia’s National ScienceAgency Data-efficient Machine Learning for Geometry Prediction in Cold Spray Additive Manufacturing Daiki Ikeuchi, Alejandro Vargas-Uscategui, Xiaofeng Wu, Peter C. King | September 2020 For Further Information: “Data-Driven Overlapping-Track Profile Modeling in Cold Spray Additive Manufacturing” Citation: Ikeuchi, D., Vargas-Uscategui, A., Wu, X. et al. Data-Driven Overlapping-Track Profile Modeling in Cold Spray Additive Manufacturing. J Therm Spray Tech 33, 530– 539 (2024). https://doi.org/10.1007/s11666-024-01733-3 Our Related Work (Open Access): “Neural Network Modelling of Track Profile in Cold Spray Additive Manufacturing” Citation: Ikeuchi, D.; Vargas-Uscategui, A.; Wu, X.; King, P.C. Neural Network Modelling of Track Profile in Cold Spray Additive Manufacturing. Materials 2019, 12, 2827. https://doi.org/10.3390/ma12172827 “Data-Efficient Neural Network for Track Profile Modelling in Cold Spray Additive Manufacturing” Citation: Ikeuchi,D.; Vargas-Uscategui, A.; Wu, X.; King, P.C. Data-Efficient Neural Network for Track Profile Modelling in Cold Spray Additive Manufacturing. Appl. Sci. 2021, 11, 1654. https://doi.org/ 10.3390/app11041654
  • 2.
    Introduction osition – Thebasic concept Additive Manufacturing | Dr. Alejandro Vargas-Uscategui ted, high pressure gas is accelerated to supersonic velocities through a de Laval zle. allic Particles (< 50 µm) are fed into the gas stream. particles impact onto a substrate, forming a deposit. Why Cold Spray Additive Manufacturing (CSAM)? • Solid-state deposition with low-oxygen contents • Avoidance of melting-induced microstructure changes • Stable fabrication with high deposition rate with robotic system • Large-scale manufacturing without a protective atmosphere
  • 3.
    Background and Challenge Challenge:Geometric Modelling and Control Approach Mathematical Data-driven Advantage • Complete profile prediction • Physically consistent • Potential for higher predictive accuracy • Learning capability Disadvantage • Limited predictive accuracy • May require multiple forms of underlying math models. • Limited to prediction of geometric features only • Need large dataset Question: Can we address the problem of data- scarcity in CSAM for data-driven modelling?
  • 4.
    Objectives Investigate data-efficient machinelearning modelling for an overlapping-track deposit profile in CSAM Complete overlapping-track profile prediction Include deposition surface information explicitly Comparative study against previously proposed models Data-efficiency achieved without more experiments
  • 5.
    Method: Experimental Fabrication ofTwo-Track Profiles: • A commercial Impact Innovation 5/11 cold spray gun • Purity grade-2 titanium feedstock Design of Experiments: • Traverse speed, and Standoff distance and Overlapping ratio designed in a full factorial manner (48 overlapping tracks) CSAM Overlapping Two-Track Profile
  • 6.
    Method: Modelling Strategy(1) 𝑥 = 𝑇𝑟𝑎𝑣𝑒𝑟𝑠𝑒 𝑠𝑝𝑒𝑒𝑑 𝑃𝑜𝑙𝑎𝑟 𝑎𝑛𝑔𝑙𝑒 𝑆𝑢𝑏𝑠𝑡𝑟𝑎𝑡𝑒 𝑡𝑦𝑝𝑒 𝑃𝑜𝑙𝑎𝑟 𝑙𝑒𝑛𝑔𝑡ℎ 𝑦 = ∆ 𝑝𝑜𝑙𝑎𝑟 𝑙𝑒𝑛𝑔𝑡ℎ
  • 7.
    Method: Modelling Strategy(2) 𝑦 𝑥 = 𝑓 𝑥 + ℎ 𝑥 Gaussian Process Regression (GPR) Explicit Mean Function • 𝑚 𝑥 = 0 • Kernel: 3/2 Matern function • K-fold cross validation (k = 5) • 48 overlapping-track profiles • 151 sampling points taken between ± 5 mm around Tool Centre Point • Substrate type represented as -1 or 1 • Acts as an approximate solution • Can be based on domain knowledge or approximate model • Mathematical Gaussian superposing model used here Simply added numerically
  • 8.
    Result: Modelling Validation Performanceof ƒ (x): • MSE = 0.003573 • Evaluated on testing dataset • Cross-validated Performance of y(x): • MSE = 0.0002183 • Evaluated on testing dataset • Cross-validated
  • 9.
    Result: Analysis andComparison • Both pure GPR and data- efficient GPR models performing better than mathematical Gaussian model in mean error % • The data-efficient GPR model showed narrower range up to 75% quartile • Maximum error % lower for GPR models than mathematical Gaussian model Mean % 2.189 1.449 0.5748
  • 10.
  • 11.
    Conclusion Complete overlapping-track profile prediction Include deposition surfaceinformation explicitly Comparative study against previously proposed models Data-efficiency achieved without more experiments Future works: • Development of in-situ online sensing system to extend the modelling approach for overlayer manufacturing • Demonstration of the modelling approach on more complex deposition scenarios (e.g., on curved or complex surface for repairing)
  • 12.
    Australia’s National ScienceAgency Daiki Ikeuchi The University of Sydney CSIRO Manufacturing daiki.ikeuchi@sydney.edu.au Thank you For Further Information: “Data-Driven Overlapping-Track Profile Modeling in Cold Spray Additive Manufacturing” Citation: Ikeuchi, D., Vargas-Uscategui, A., Wu, X. et al. Data-Driven Overlapping-Track Profile Modeling in Cold Spray Additive Manufacturing. J Therm Spray Tech 33, 530–539 (2024). https://doi.org/10.1007/s11666-024-01733-3 Our Related Work (Open Access): “Neural Network Modelling of Track Profile in Cold Spray Additive Manufacturing” Citation: Ikeuchi, D.; Vargas-Uscategui, A.; Wu, X.; King, P.C. Neural Network Modelling of Track Profile in Cold Spray Additive Manufacturing. Materials 2019, 12, 2827. https://www.mdpi.com/1996-1944/12/17/2827 “Data-Efficient Neural Network for Track Profile Modelling in Cold Spray Additive Manufacturing” Citation: Ikeuchi,D.; Vargas-Uscategui, A.; Wu, X.; King, P.C. Data-Efficient Neural Network for Track Profile Modelling in Cold Spray Additive Manufacturing. Appl. Sci. 2021, 11, 1654. https://www.mdpi.com/2076-3417/11/4/1654