High-throughput scanning electron microscopy (SEM) is essential for efficient microstructural characterization. This is particularly important when characterizing additive manufactured (AM) metal compotents that are centimeter-sized, but where defects such as pores and cracks at the nanometer scale critically affect material performance.

Here, we present a high-throughput high-resolution SEM framework for automated defect detection that uniquely integrates preprocessing, feature identification visualization, and statistical quantification to improve efficiency and accuracy of large-area SEM analysis.

We aquired large-area secondary electron (SE) micrographs of AM Fe7336 samples at different pixel sizes (48.8 nm to 390.6 nm) and dwell times (10 µs to 100 ns) and systematically evaluated the robustness of defect detection under these SEM imaging conditions. Defect detection accuracy was evaluated by comparing results against high-resolution benchmark images (our “ground truth”) and manually labeled datasets. We specifically focused on the effects of pixel size and dwell time on the accuracy of defect characterization, and found that shorter dwell times lead to increased noise and therefore obscure small defect details. To improve detection consistency at shorter dwell times, we then introduced filtering constraints based on peak signal-to-noise ratio (PSNR) trends.

Our results provide important insights into how different SEM image acquisition speed affects defect characterization accuracy and identify key trade-offs that enable significant reductions in data collection time up to a factor of four while preserving reliable feature detection at around 90%). By optimizing SEM acquisition parameters and automating defect identification, our framework not only improves the scalability and reproducibility of microstructural analysis in metal AM, but alos provides a customizable approach for high-throughput materials characterization that can be adapted to other materials and manufacturing processes.