An Efficient and Time-Effective Experimental and Numerical Methodology for Mapping Atmospheric Corrosivity and Assessing Corrosion-Related Failures in Power Transmission Lines and Substations

In the electric sector, power transmission lines, substations, and associated equipment are highly susceptible to environmental aggressiveness, necessitating updated and efficient monitoring systems. This study analyzed the corrosion rate of metals in a 1,150 km power transmission line spanning Ceará, Piauí, and Maranhão in northeastern Brazil. Carbon and galvanized steel coupons were exposed for one year, with corrosion assessed through mass and thickness loss. The results, combined with local meteorological data such as temperature, humidity, and precipitation rates, were used to train, test, and validate a feed-forward neural network model. The model showed good prediction accuracy, particularly for galvanized steel, with low mean absolute and mean squared errors, enabling reliable forecasts. For carbon steel, although predictions exhibited slightly lower accuracy, they were still adequate for classifying local atmospheric corrosivity levels. Furthermore, the model successfully generated corrosion maps for both metals, enabling detailed visual analysis across the entire transmission line.