U‐Net‐Based Prediction of Concealed Volume Loss in AZ31 From Optical Corrosion Imprints and As‐Exposed Surface Topography

High‐throughput screening of small organic molecules as dissolution modulators (accelerators and inhibitors of corrosion) for the magnesium alloy AZ31 using a multiwell exposure protocol resulted in 842 images of individual corrosion imprints whereas their profilometric evaluation constitutes a major bottleneck. Moreover, corrosion products formed during exposure obscure the underlying surface morphology and need to be removed with hazardous chromic acid prior to the volume‐loss determination. In this study, concealed volume loss in AZ31 corrosion imprints was predicted directly from their as‐exposed state. To this end, a U‐Net‐based regression model was trained using a four‐channel input comprising RGB images of the as‐exposed imprints and corresponding surface topographies. Profilometric depth maps acquired after cleaning served as experimental ground‐truth labels, and the model’s generalization capability for unseen modulators was assessed via cross‐validation. The presented approach lays the foundation for a quantitative estimation of corrosion volume loss and inhibition efficiency without prior removal of corrosion products, providing a faster, safer, and more scalable workflow for high‐throughput corrosion modulator screening.