Predicting localized corrosion in corrosion-resistant alloys is challenging due to the complexities of pitting evolution. This study explores a data-driven method using time-series analysis, specifically a Long Short-Term Memory (LSTM) neural network, to predict free corrosion potential (*E*_corr) in cobalt-based alloys and duplex stainless steels. The LSTM model, trained on 150 days of immersion testing data, outperformed traditional machine learning methods in predicting *E*_corr for the subsequent 70 days, demonstrating its ability to capture the time-series dependency of pitting propagation.

Xue Jiang, Yu Yan, Yanjing Su