Abstract: Atmospheric corrosion poses a critical threat to the durability of infrastructure, particularly under complex environmental conditions. Traditional corrosion assessment methods are limited by time-consuming procedures and fragmented data. This study introduces a self-attention-based time series imputation model (SAITS) to impute missing data in corrosion monitoring of 11 sites in Xinjiang, China. By training on incomplete atmospheric corrosion monitor data, SAITS demonstrates superior performance compared to other imputation methods, achieving a minimum MAE of 0.1750 and RMSE of 0.3438 on the test set. Seasonal and environmental analyses reveal strong correlations between corrosion current and humidity, followed by SO₂ and PM₂.₅ concentrations. This data-driven approach provides new insights into atmospheric corrosion behavior and supports proactive corrosion management for critical infrastructure.