EUVIMEDEUVIMED entdecken
Crossref

AI-Enabled SEM Defect Analysis Toolkit for Battery Microstructure Characterization

Qixuan Jiang

ECS Meeting Abstracts · 07.07.2026 · MA2026-01 (2) · 159-159

DOI: 10.1149/ma2026-012159mtgabs ↗

Kurz gefasst

Fachartikel in der Zeitschrift ECS Meeting Abstracts, erschienen am 07.07.2026. Verfasst von Qixuan Jiang.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Characterization is one of the biggest bottlenecks in battery research and manufacturing. Along the manufacturing line we generate large volumes of scanning electron microscopy (SEM) images, but much of the quantitative analysis is still manual and slow. Researchers visually inspect images, annotate a few regions of interest, and run one off scripts with hand tuned parameters. This makes results hard to reproduce and hard to compare across samples, tools, or labs. It also slows the feedback loop between processing conditions, microstructure observations, and performance decisions. This abstract presents an AI enabled defect analysis toolkit that turns SEM images into interpretable overlays and standardized, reusable metrics for battery microstructures. The goal is to provide a stable analysis layer that sits after segmentation or pixel classification and converts model outputs into defect descriptors that battery researchers can trust. Rather than relying on one specific neural network, the toolkit is designed to accept probability maps or masks from many vision models and apply a transparent sequence of post processing steps to produce consistent metrics. Given a grayscale SEM image and a model output, the toolkit supports tiled processing for very large images and performs calibrated thresholding, morphological cleaning, connected component labeling, region based measurements, and optional skeletonization. Tile level results are aggregated into image level summaries and spatial heat maps, enabling comparisons across different sample locations and processing conditions. Because the metric layer is separated from the model layer, researchers can update or replace the segmentation model while keeping defect definitions and metrics consistent across time. I demonstrate the toolkit with two defect focused case studies. The first case study is lithium dendrites and associated voids in lithium metal environments. Starting from pixel wise model outputs, the toolkit separates metallic lithium from internal pores and generates overlays that visualize dendrite geometry and porosity. It computes standardized metrics including lithium area fraction, local porosity, pore count and pore size distributions, connectivity indices, and skeletal length density based on a one pixel wide skeleton of the lithium network. These outputs provide a structured way to compare dendrite growth and pore formation across samples and conditions. The second case study is cracking in nickel rich cathode particles (NMC). The toolkit highlights crack like features using ridge enhancement and thresholding, reduces them to centerlines using skeletonization, and assigns cracks to nearby particle regions after particle segmentation. It reports metrics such as cracks per particle, crack length density, and geometry descriptors that capture whether crack paths are relatively straight or tortuous. These measurements provide a systematic alternative to manual crack counting and support fair comparisons across experiments. As a light deployment example, I reference an Information Commons website where these workflows can be packaged as documented modules with defined inputs and outputs, and where users can obtain overlays and CSV summaries of defect metrics. This illustrates a practical path to move the toolkit from a single lab script to a shared resource, while keeping the core focus on defect analysis and standardized characterization. Overall, this work aims to make SEM based defect characterization more scalable, transparent, and comparable for battery manufacturing and research. By improving how defects are quantified and tracked, it can accelerate process optimization, improve reliability and lifetime, and reduce rework and material waste during development and manufacturing. These outcomes support sustainability by enabling longer lasting energy storage systems, reducing scrap and repeated experimentation, and improving the efficiency of materials and energy use across the battery lifecycle.

Zitieren

APA

Jiang, Q. (2026). AI-Enabled SEM Defect Analysis Toolkit for Battery Microstructure Characterization. ECS Meeting Abstracts, MA2026-01(2), 159-159. https://doi.org/10.1149/ma2026-012159mtgabs

Vancouver

Jiang Q. AI-Enabled SEM Defect Analysis Toolkit for Battery Microstructure Characterization. ECS Meeting Abstracts. 2026;MA2026-01(2):159-159. doi:10.1149/ma2026-012159mtgabs

BibTeX
@article{Jiang2026,
  author = {Jiang, Qixuan},
  title = {AI-Enabled SEM Defect Analysis Toolkit for Battery Microstructure Characterization},
  journal = {ECS Meeting Abstracts},
  year = {2026},
  volume = {MA2026-01},
  number = {2},
  pages = {159--159},
  doi = {10.1149/ma2026-012159mtgabs}
}
Kostenlos starten →