
Volumetric defects in additively manufactured components remain a major barrier to their safe adoption in critical applications. This talk shows how we can guide machine learning (ML) models by encoding the underlying wave physics, allowing us to quantify volumetric porosity from raw ultrasonic signals and to image defects with unprecedented resolution. By combining physics-based models with ML, we move toward faster, more reliable, and more automated quality control.
Michail Skiadopoulos | Engineering Science and Mechanics
