30 May 2026 to 2 June 2026
ESA HQ-N
Europe/Paris timezone

Using Multiple, Interconnected Machine Learning Approaches to Enable and Control 3D Dendrite Simulations of Metal-Alloy Solidification

1 Jun 2026, 12:40
5m
ESA HQ-N

ESA HQ-N

Paris, France
Poster Lunch

Speaker

Jonathan Mullen (European Space Agency - Human and Robotic Exploration - Science Utilisation)

Description

The crystal microstructure of a metal alloy has a significant impact on its performance, with consequences for any components manufactured using it. Over the course of decades, ESA has researched the impact that microgravity can have on these crystals, referred to as dendrites, via the use of in-situ X-ray radiography imaging of the solidification process. As experiments move away from model alloys, chosen for their contrast within X-ray videos, the challenge of isolating, measuring and comparing visible dendrites will only increase. Thus, a new, 3D modelling based approach is being developed to allow for simulation-based-inference to overcome previously insurmountable image properties. This poster demonstrates the core of this new approach, two interconnected machine learning networks which learn the necessary parameters for the 3D modeller to successfully simulate unclear or obscured areas of highly valuable and unique experimental data.

Author

Jonathan Mullen (European Space Agency - Human and Robotic Exploration - Science Utilisation)

Co-author

Dr Wim Sillekens (European Space Agency - Human and Robotic Exploration - Science Utilisation)

Presentation materials

There are no materials yet.