Surface texture evolution of weathering steel in corrosive marine atmospheres: Geospatial variability and explainable machine learning models

Abstract

The surface texture of 144 weathering steel samples were analyzed under marine atmospheres to assess the effects of the exposure conditions and structural factors. Comprehensive methods were used to characterize the corrosion evolution and pitting performance. Geospatial variability was introduced as a new method for quantifying non-uniform and irregular surface substrates. The results showed significant variations in corrosion severity under identical environmental conditions. Closely spaced pits formed clustered patterns with an increased range, whereas isolated sparse pits demonstrated the converse. Sill effectively captured serrated surface variations. Explainable machine learning models were developed to regenerate and predict the steel surface texture.