In-situ monitoring of defect formation remains a significant challenge in the laser powder bed fusion (LPBF) process. Recent advances have enabled real-time defect detection with machine learning and in-situ sensing technologies; however, most studies focus on binary classification of keyhole pores, limiting nuanced multi-class pore differentiation and formation mechanisms. This work introduces a multi-class pore detection framework (no pore, small pores < 15 µm, and large pores > 15 µm) by leveraging photodiode sensor data alongside high-fidelity synchrotron X-ray imaging. The 15 µm threshold is selected to distinguish between two fundamentally different defect mechanisms, following the physical size-mechanism boundary established by prior high-resolution synchrotron X-ray characterization of Al6061 LPBF. Distinguishing these classes is critical because large keyhole pores are structurally detrimental, whereas small gas pores are often benign, requiring different process control strategies. Thermal emission monitoring data collected simultaneously with high-speed X-ray imaging at the Stanford Synchrotron Radiation Lightsource (SSRL), are correlated with subsurface melt pool dynamics to establish ground truth. Continuous Wavelet Transform (CWT) with optimized parameters converts the photodiode time-series signals into time–frequency images, facilitating feature extraction. Convolutional Neural Networks (CNN) are then applied for real-time multi-class pore classification in an average inference time of 1 ms per signal window. It achieves 79% accuracy and an Area Under the Receiver Operating Characteristic curve (AUC ROC) score of 0.89 with five-fold cross-validation. The results demonstrate that coupling CWT-based feature engineering with CNN architecture enables reliable multi-class pore detection in Al6061 builds using affordable in-situ sensors. This approach advances scalable and affordable quality assurance in additive manufacturing by moving beyond binary defect detection toward more nuanced classification of porosity mechanisms with in-situ sensors and machine learning.

Laser powder bed fusion (LPBF) enables the fabrication of high-precision metal parts, but spatter can disturb the powder bed and contribute to defects. This study compared single-track LPBF of SUS316L stainless steel in air and low-oxygen N₂ (O₂ < 1%) using high-speed imaging with pulsed illumination. The temporal evolution, apparent size, and spatial distribution of detected particles were examined at different laser powers and scanning speeds. Scanning electron microscopy coupled with energy-dispersive X-ray spectroscopy (SEM–EDS) was used to assess the laser-affected regions. The atmosphere-dependent trend varied with processing conditions: low-oxygen N₂ did not consistently reduce the detected particle count, whereas at 200 W and 500 mm/s, it showed fewer large-particle observations than air. The two atmospheres also produced different particle distributions relative to the moving laser. SEM–EDS showed lower O and Mn signals and a higher relative Fe fraction in the regions examined under N₂. These observations establish condition-dependent differences between the two gas environments, although the contributions of oxygen chemistry, gas flow, and optical differences between the experimental configurations cannot be separated.

In the alloy design and development process, a wealth of atomically resolved structural high-resolution transmission electron microscopy (HRTEM) images are produced. Identifying the different nano-precipitate phases and tracking their evolution under various compositions and during manufacturing or post-processing requires hundreds of HETEM images and thousands of precipitates. The nanoscopic phase information labeling and analysis purely with human are prohibitively costly and time-consuming, sometimes not reliable because of the lack of authoritative knowledge. Here, we develop a novel unsupervised machine learning approach coupled with adaptive computer vision techniques with features in the Fourier space to automatically determine the number of phases and segment/quantify the phases with nanoscale resolution, allowing for quantitative correlation between nanostructure formation, processing and functional properties. To automate the phase extraction/quantification and ascertain its applicability, we have applied the developed framework on the HRTEM images from several alloy systems, processing conditions, phase morphologies and phase types for verification. This study paves the road for compression, visualization, and translation of raw image structural data into physically relevant information in real-time with minimal human supervision. It shows the promise for enabling the high-throughput materials characterization for acceleration of alloy manufacturing and design.