A Multilayer Perceptron-Based Neural Network Model for Predicting Steel Bridge Coating Conditions by Integrating Bridge and Environmental Features

The service environments (e.g., climatic environments, water environments) of steel bridges have severe impacts on the performance of their protective coatings, thus affecting the safety, integrity, and longevity of the steel bridge elements. By leveraging existing data on both the bridge-related features (e.g., bridge age, type of service under bridge) and environmental features (e.g., chloride, moisture, atmospheric pollutants, temperature), we can develop a data-driven understanding of bridge coating deterioration patterns as well as how such patterns are impacted by the service environments of bridges. This paper focuses on presenting a multilayer perceptron (MLP)-based artificial neural network (ANN) model that predicts coating conditions of the girder elements of steel bridges in Florida. The model integrates both the bridge-related features and environmental features. It uses data from multiple sources, including Florida Department of Transportation and national environmental agencies [e.g., National Atmospheric Deposition Program (NADP), Florida Automated Weather Network (FAWN), United States Environmental Protection Agency (USEPA)]. The results show that the proposed model can effectively predict the coating conditions of steel bridges with a minimum mean absolute error (MAE) of 0.0807. As compared to the MAE of the model that only accounts for bridge-related features, the MAE of the proposed model was significantly reduced by approximately 11.36%. Thus, by offering effective predictions on steel bridge coating performance, the proposed model has the potential to allow bridge owners to better maintain, repair, and/or replace steel bridge coating systems with less effort, time, cost, and risk.