Deep Learning for the Automation of Particle Analysis in Catalyst Layers for Polymer Electrolyte Fuel Cells

Abstract

The rapidly growing use of imaging infrastructure in the energy materials domain drives significant data accumulation in terms of their amount and complexity. The applications of routine techniques for image processing in materials research are often \textit{ad hoc}, indiscriminate, and empirical, which renders the crucial task of obtaining reliable metrics for quantifications obscure. Moreover, these techniques are expensive, slow, and often involve several preprocessing steps. This paper presents a novel deep learning-based approach for the high-throughput analysis of the particle size distributions from transmission electron microscopy (TEM) images of carbon-supported catalysts for polymer electrolyte fuel cells. Our approach employs training an instance segmentation model, called StarDist [Schmidt et al. Medical Image Computing and Computer-Assisted Intervention – MICCAI 2018, Lecture Notes in Computer Science, vol 11071. Springer, Cham], which resolves the main challenge in the pixel-wise localization of nanoparticles in TEM images: the overlapping particles. The segmentation maps outperform models reported in the literature, and the results on particle size analyses agree well with manual particle size measurements, albeit at a significantly lower cost.

nNPipe: a neural network pipeline for automated analysis of morphologically diverse catalyst systems

“…Deep learning (DL) based solutions have been applied in various contexts for nanoparticle analysis in TEM imaging, including Convolutional Neural Networks (CNNs) for object detection and semantic segmentation at atomic 30,31 and lower resolution [32][33][34][35] , analysis of the performance of the U-Net neural network [36][37][38] , as well as liquid-cell experiments 39,40 . The former are appealing to studies of heterogeneous catalysts as they allow a statistically significant determination of relevant material properties once the respective networks are trained.…”

A Data‐Driven framework for prediction the cyclic voltammetry and polarization curves of polymer electrolyte fuel cells using artificial neural networks

“…Furthermore, artificial neural networks are the most preferred methods over other ML algorithms because of their generalization capabilities. Recently, some learning algorithms based on neural learning have been applied for processing complex and multi‐scale structural features such as ink imaging data, 18 selecting electrocatalyst for CO2 reduction reactions, 19 and predicting particle size distributions from transmission electron microscopy (TEM) images of carbon‐supported catalysts for polymer electrolyte fuel cells 20 . The findings highlight the significance of model pre‐training and data augmentation in selecting the best materials.…”