Interpretable Machine Learning Reveals Ni–Mo–Cu Synergistic Corrosion Behavior in Low‐Alloy Steels Under Simulated Marine Atmosphere

This study investigates the synergistic corrosion behavior of Ni, Mo, and Cu in the low‐alloy steel under simulated marine atmosphere by integrating interpretable machine learning (IML) with multiscale experimental validation. Results demonstrate that Transformer model can capture the complex, nonlinear, and time‐dependent corrosion behavior from a small−sample time‐series dataset, achieving high predictive accuracy and robust generalization in forecasting corrosion rates. IML analysis reveals that the synergistic corrosion resistance provided by Ni, Mo, and Cu is dynamically governed by their relative elemental concentrations, with Mo playing a critical regulatory role. When the Mo content exceeds both Ni and Cu, molybdenum oxides promote the formation of NiFe 2 O 4 and CuFeO 2 in higher proportions, which can enhance the rust layer's compactness and protectiveness. Conversely, insufficient Mo content leads to the dominance of the oxide phase associated with the most abundant element (Ni or Cu), suppressing the stable coexistence of other protective phases and thereby weakening the Ni–Mo–Cu synergistic effect, thus accelerating corrosion propagation.