The Potential of Sparse X-ray Imaging to Advance Inkjet Printing with AI
(2025) PHYM01 20242Synchrotron Radiation Research
Department of Physics
- Abstract
- Tetra Pak, a global leader in packaging solutions, seeks to optimize inkjet printing as a more flexible alternative to traditional methods. Although inkjet printing offers rapid design customization and reduced waste, its performance is hindered by issues such as unwanted satellite droplets. Capturing the formation of these droplets at microsecond timescales demands specialized experimental methods, precisely what high-speed X-ray imaging at cutting-edge synchrotron facilities like MAX IV can provide.
However, typical beamline setups at synchrotrons are constrained to only one to three two-dimensional (2D) X-ray projections, far fewer than the thousands conventionally used in three-dimensional (3D) tomographic reconstruction, due to the... (More) - Tetra Pak, a global leader in packaging solutions, seeks to optimize inkjet printing as a more flexible alternative to traditional methods. Although inkjet printing offers rapid design customization and reduced waste, its performance is hindered by issues such as unwanted satellite droplets. Capturing the formation of these droplets at microsecond timescales demands specialized experimental methods, precisely what high-speed X-ray imaging at cutting-edge synchrotron facilities like MAX IV can provide.
However, typical beamline setups at synchrotrons are constrained to only one to three two-dimensional (2D) X-ray projections, far fewer than the thousands conventionally used in three-dimensional (3D) tomographic reconstruction, due to the need for stationary X-ray source and non-rotating samples. In this thesis, a simulation-based approach was employed, leveraging computational fluid dynamics (CFD) to generate synthetic inkjet droplet datasets. X-ray propagation through these droplets was then simulated to mimic real experiments under sparse projection conditions. By adapting machine learning (ML) strategies to these novel constraints, the reconstructions were able to capture critical droplet breakup features, including the formation of satellite droplets, even from very few projections.
Although the single-projection scenario posed the greatest challenges, the results underscore both the feasibility of conducting MHz-scale imaging at MAX IV and the potential for further performance gains using more experiments and thus data available for the model to learn the underlying dynamics. Collectively, this work provides a roadmap for future experimental efforts, offering insights to improve droplet control and enhance both the efficiency and quality of next-generation inkjet printing processes. (Less) - Popular Abstract
- Inkjet printing is rapidly reshaping how products are packaged and labeled by allowing custom designs without the traditional need for bulky printing plates. Tetra Pak, a global leader in packaging, aims to harness inkjet’s potential to print faster and more flexibly, while still maintaining high quality. However, creating thousands of tiny droplets per second comes with a major challenge: small, unintended “satellite droplets” can form and blur the final image.
To better understand and control this phenomenon, we can use extremely bright X-ray sources, like those at the MAX IV synchrotron in Lund, to “see” inside the printing process as it happens. Under ideal conditions, you might gather hundreds of X-ray views (projections) to... (More) - Inkjet printing is rapidly reshaping how products are packaged and labeled by allowing custom designs without the traditional need for bulky printing plates. Tetra Pak, a global leader in packaging, aims to harness inkjet’s potential to print faster and more flexibly, while still maintaining high quality. However, creating thousands of tiny droplets per second comes with a major challenge: small, unintended “satellite droplets” can form and blur the final image.
To better understand and control this phenomenon, we can use extremely bright X-ray sources, like those at the MAX IV synchrotron in Lund, to “see” inside the printing process as it happens. Under ideal conditions, you might gather hundreds of X-ray views (projections) to reconstruct a detailed 3D image over time, effectively making a “4D movie” of the droplets. But capturing hundreds of projections is tough when the droplets move at microsecond speeds, and rotating the sample or X-ray source is not feasible in fast processes. Instead, scientists have turned to new methods that only require a few angles, sometimes just one.
This thesis explores the potential of analyzing inkjet printing using X-ray imaging at facilities like MAX IV. The scope includes simulations, reconstruction, and validation. Machine learning techniques are employed to reconstruct high-fidelity images from a limited number of X-ray snapshots. Through computer simulations of inkjet droplets, including the flow and formation of satellites, neural networks are trained to fill in the missing details. The result is a model capable of handling extremely sparse data while preserving crucial features, such as the formation and breakup of droplets.
With further refinement, these reconstructions could help Tetra Pak and other industries quickly analyze and optimize printing conditions, reducing waste and improving print clarity. Ultimately, this blend of cutting-edge X-ray imaging, fluid simulations, and machine learning offers a glimpse into the future of high-speed, high-resolution analysis of processes involving individual fluid droplets. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/9186176
- author
- Olofsson, Fredrik LU
- supervisor
- organization
- course
- PHYM01 20242
- year
- 2025
- type
- H2 - Master's Degree (Two Years)
- subject
- keywords
- Inkjet Printing, Sparse Projection Reconstruction, X-ray Imaging, X-ray Multi-Projection Imaging, AI
- language
- English
- id
- 9186176
- date added to LUP
- 2025-03-10 08:38:53
- date last changed
- 2025-03-10 08:38:53
@misc{9186176,
abstract = {{Tetra Pak, a global leader in packaging solutions, seeks to optimize inkjet printing as a more flexible alternative to traditional methods. Although inkjet printing offers rapid design customization and reduced waste, its performance is hindered by issues such as unwanted satellite droplets. Capturing the formation of these droplets at microsecond timescales demands specialized experimental methods, precisely what high-speed X-ray imaging at cutting-edge synchrotron facilities like MAX IV can provide.
However, typical beamline setups at synchrotrons are constrained to only one to three two-dimensional (2D) X-ray projections, far fewer than the thousands conventionally used in three-dimensional (3D) tomographic reconstruction, due to the need for stationary X-ray source and non-rotating samples. In this thesis, a simulation-based approach was employed, leveraging computational fluid dynamics (CFD) to generate synthetic inkjet droplet datasets. X-ray propagation through these droplets was then simulated to mimic real experiments under sparse projection conditions. By adapting machine learning (ML) strategies to these novel constraints, the reconstructions were able to capture critical droplet breakup features, including the formation of satellite droplets, even from very few projections.
Although the single-projection scenario posed the greatest challenges, the results underscore both the feasibility of conducting MHz-scale imaging at MAX IV and the potential for further performance gains using more experiments and thus data available for the model to learn the underlying dynamics. Collectively, this work provides a roadmap for future experimental efforts, offering insights to improve droplet control and enhance both the efficiency and quality of next-generation inkjet printing processes.}},
author = {{Olofsson, Fredrik}},
language = {{eng}},
note = {{Student Paper}},
title = {{The Potential of Sparse X-ray Imaging to Advance Inkjet Printing with AI}},
year = {{2025}},
}