A data-centric review on machine learning of corrosion prediction

This Review provides a data-oriented perspective of the novel research field on predictive machine learning (ML) applied to corrosion. We aim to determine which ML models have been used and how well they perform, thus providing a general guidance on how to successfully apply ML to corrosion. Bibliometric data mining led to creating a 'ML for corrosion database' [1], where the data-centric approaches of 34 works presenting comparable performance metrics were scrutinised. EDA demonstrated the positive effects of increasing data dimension and including time as an input on the models’ performance. Comparative performance analysis segmented the references by corrosion topic, orientation strategy, type of ML model and data targets. Expanding the types of variables likely increase the models’ robustness. Accurate modelling requires large sets of training data of high quality, and the labelling of corrosion features should be based on specific domain knowledge.