Publication: Transfer learning assessment of small datasets relating manufacturing parameters with electrochemical energy cell component properties

This publication by CNRS/Université de Picardie Jules Verne (UPJV) examines how Machine Learning (ML) models can help to overcome the obstacles imposed by the current time-consuming and costly approach to improve the performance of electrochemical cells for energy storage and conversion. This is, traditionally, done by optimizing the manufacturing processes. ML models typically require large amounts of data while the number of datasets in the laboratories can be limited. In this publication, UPJV proposes a simple and novel application of a Transfer Learning (TL) approach to address the manufacturing problems. They used pre-existing experimental and stochastically generated datasets consisting of component properties (e.g. electrode density) related to different manufacturing parameters (e.g., solid content, comma gap, coating speed). It turns out that the TL approach is robust and achieves excellent prediction performance or electrodes in lithium-ion batteries and gas diffusion layers in fuel cells.

The full publication is available here.