Investigating the effects of machine learning generalization for enhancing accuracy in fast X-ray computed tomography for industrial metrology

Industrial X-ray computed tomography enables comprehensive inspections, but the measurement accuracy can be compromised at high scanning speeds due to reduced image quality. Machine learning shows promise in enhancing tomographic reconstructions under challenging conditions, yet its generalization across varying factors is crucial for maintaining accuracy. This research proposes an experimental methodology to investigate the sensitivity of various scanning conditions to the degree of machine learning generalization and evaluates the accuracy improvements achievable with increasing generalization. Findings lay the foundations for efficiently integrating machine learning into fast tomography workflows to bridge the accuracy-speed gap for industrial metrology applications.