In this study, the corrosion of carbon steel immersed in CO2 saturated aqueous solutions, in the presence and absence of sand deposits, were investigated by electrochemical noise measurement and recurrence quantification analysis. Uniform corrosion occurred at samples without sand deposit, while localized corrosion took place at the sand-covered steel samples. These two different corrosion types can be accurately predicted by random forest and principal component models based on recurrence quantification analysis of either electrochemical potential or current noise data, regardless of threshold values. The study provides a potential automated online corrosion monitoring scheme to ensure the integrity of pipelines. (C) 2018 Elsevier Ltd. All rights reserved.