| Salt spray testing plays a critical role in evaluating the corrosion resistance of materials and protective coatings in environments exposed to aggressive saline conditions. Conventional methods for monitoring salt spray concentration of-ten suffer from interference, which limits measurement accuracy and compromises result reliability. Recent advances in artificial intelligence offer effective pathways to automate and enhance corrosion monitoring by improving measurement consistency, strengthening predictive capability, and adapting to dynamic environmental conditions. This study applied explainable artificial intelligence techniques such as Integrated Gradients, GradientShap, LIME, and Occlusion to quantify feature importance. The developed model achieved a mean absolute error of 1.9384, a root mean squared error of 2.4375, and a coefficient of determination of 0.9938, demonstrating excellent predictive performance. The model predicted the time to 99.9% corrosion as 17.19 cycles, closely aligning with the labelled value of 16.85 cycles, thereby confirming re-liable and interpretable corrosion prediction. |
*** Title, author list and abstract as submitted during Camera-Ready version delivery. Small changes that may have occurred during processing by Springer may not appear in this window.