% IMPORTANT: The following is UTF-8 encoded. This means that in the presence
% of non-ASCII characters, it will not work with BibTeX 0.99 or older.
% Instead, you should use an up-to-date BibTeX implementation like “bibtex8” or
% “biber”.
@INPROCEEDINGS{Colliard:1038802,
author = {Colliard, Andre and Malek, Kourosh and Eikerling, Michael
and Eslamibidgoli, Mohammad Javad},
title = {{E}nhanced 3{D} {B}ubble {C}haracterization in {V}anadium
{R}edox {F}low {B}atteries {U}sing {S}ynchrotron {X}-ray
{I}maging and {D}eep {L}earning},
reportid = {FZJ-2025-01642},
year = {2024},
abstract = {The analysis of gas bubbles in vanadium redox flow
batteries (VRFBs) plays a crucial role in understanding and
improving their performance [1,2]. Traditional methods have
struggled with the direct visualization and quantitative
analysis of these bubbles, a challenge overcome by recent
advances in Synchrotron X-ray tomography. However, the large
volume of data and the complexity of interpreting the images
necessitate an innovative approach. We present a novel deep
learning framework designed for the automated analysis of 3D
bubble dynamics in VRFBs, leveraging high-resolution
Synchrotron X-ray imaging. This framework employs an
optimized U-Net architecture for semantic segmentation,
trained on a dataset of 2294 annotated images across diverse
battery configurations. Our methodology not only automates
the identification and quantification of bubbles but also
introduces advanced descriptors for improved understanding
of bubble behaviour within the battery's electrolyte and
electrode matrices. The optimized model demonstrates
exceptional performance on test dataset, with a precision of
$98\%,$ recall of $97\%,$ and an F1-score of $97\%$ on a
diverse validation set. Beyond segmentation, the framework
offers tools for the detailed examination of electrode
interfaces, volumetric and morphological analysis of
bubbles, and the generation of comprehensive bubble
distribution maps. It also provides metrics for evaluating
the impact of bubble presence on membrane efficiency and
potential blockages. By providing a robust, scalable
solution for the high-throughput analysis of complex 3D
datasets, our framework paves the way for deeper insights
into the operational challenges of VRFBs. The interactive 3D
visualization capabilities further enhance the utility of
this analysis, offering researchers an intuitive means to
explore and interpret the multifaceted data.[1] K. Köble et
al., “Insights into the hydrogen evolution reaction in
vanadium redox flow batteries: A synchrotron radiation based
X-ray imaging study,” Journal of Energy Chemistry, vol.
91, pp. 132–144.[2] L. Eifert et al., “Synchrotron
X‐ray Radiography and Tomography of Vanadium Redox Flow
Batteries—Cell Design, Electrolyte Flow Geometry, and Gas
Bubble Formation,” ChemSusChem, vol. 13, no. 12, pp.
3154–3165.},
month = {Jun},
date = {2024-06-12},
organization = {Helmholtz AI Conference 2024,
Düsseldorf (Germany), 12 Jun 2024 - 14
Jun 2024},
subtyp = {After Call},
cin = {IET-3},
cid = {I:(DE-Juel1)IET-3-20190226},
pnm = {1231 - Electrochemistry for Hydrogen (POF4-123)},
pid = {G:(DE-HGF)POF4-1231},
typ = {PUB:(DE-HGF)24},
url = {https://juser.fz-juelich.de/record/1038802},
}