ML for Manufacturing

Conducted with the Control and Optimization team at General Electric Research, this project developed online machine-learning methods for industrial manufacturing, spanning both subtractive and additive workflows.

For additive manufacturing, the work developed methods for real-time compression, management, and analysis of downbeam camera data. By integrating signal processing and data-driven analytics into the manufacturing pipeline, the approach enables scalable online monitoring of melt-pool dynamics and process anomalies while reducing storage and transmission demands.

For subtractive manufacturing, the work developed model-free reconstruction pipelines that drive CNC and voxel-based manufacturing directly from sensor data such as computed tomography scans. Together, these methods support closed-loop quality assurance and process optimization in industrial manufacturing systems.

Selected outcomes. A granted U.S. patent on downbeam camera data methods (Yu & others, 2021); a related published U.S. patent application (Yu & others, 2021); and earlier published work on voxel-based digital subtractive manufacturing (Lynn et al., 2018) and model-free CT-driven manufacturing (Yu et al., 2017).

Timeline. 2017 – 2018.

References

2021

Systems and methods for downbeam camera data for additive machines

J. Yu and others

2021

U.S. Patent Application 16/818,650, Published

BIB

@misc{yu2021patentdownbeam,
  title = {Systems and methods for downbeam camera data for additive machines},
  author = {Yu, J. and others},
  year = {2021},
  note = {U.S. Patent Application 16/818,650, Published}
}

Composition for cleaning and assessing cleanliness in real-time

J. Yu and others

2021

U.S. Patent Application 17/289,488, Published

BIB

@misc{yu2021patentcleaning,
  title = {Composition for cleaning and assessing cleanliness in real-time},
  author = {Yu, J. and others},
  year = {2021},
  note = {U.S. Patent Application 17/289,488, Published}
}

2018

Direct Digital Subtractive Manufacturing of a Functional Assembly using Voxel-based Models

R. Lynn, M. Dinar, N. Huang, and 5 more authors

ASME Journal of Manufacturing Science and Engineering. 2018

BIB

@article{lynn2018voxel,
  title = {Direct Digital Subtractive Manufacturing of a Functional Assembly using Voxel-based Models},
  author = {Lynn, R. and Dinar, M. and Huang, N. and Collins, J. and Yu, J. and Greer, C. and Tucker, T. and Kurfess, T.},
  journal = {ASME Journal of Manufacturing Science and Engineering},
  year = {2018}
}

2017

Model-free Subtractive Manufacturing from Computed Tomography Data

J. Yu, R. Lynn, T. Tucker, and 2 more authors

ASME Manufacturing Letters. 2017

BIB

@article{yu2017modelfree,
  title = {Model-free Subtractive Manufacturing from Computed Tomography Data},
  author = {Yu, J. and Lynn, R. and Tucker, T. and Saldana, C. and Kurfess, T.},
  journal = {ASME Manufacturing Letters},
  year = {2017}
}