Research of ACM timeseries data imputation and atmospheric corrosion factors analysis of carbon steel

Atmospheric corrosion represents a significant threat to the longevity of infrastructure, especially under complex environmental conditions. Conventional methods for corrosion assessment are hindered by their time-intensive nature and fragmented data collection. This study presents a self-attention-based time series imputation model (SAITS) to impute missing data in the corrosion monitoring of 11 sites in Xinjiang, China. By training on incomplete atmospheric corrosion monitor data, SAITS exhibits superior performance relative to other imputation techniques, achieving the minimum mean absolute error (MAE) of 0.1750 and root mean square error (RMSE) of 0.3438 on the test set. Seasonal and environmental analyses reveal strong correlations between corrosion current and humidity, followed by concentrations of SO₂, NO₂, and temperature. This data-driven approach offers new insights into atmospheric corrosion behavior and facilitates proactive corrosion management for critical infrastructure.