Bridging the gap: A comprehensive analytical study of traditional, hybrid, and explainable AI for bridge condition prediction

This study introduces a novel hybrid artificial intelligence (AI) framework to predict reinforced concrete (RC) bridge conditions and analyze climate impacts across their service life. The hybrid method integrates supervised learning, such as gradient boosting, random forests, and deep neural networks, with unsupervised learning, such as clustering methods. A total of 142,265 bridges from the U.S. National Bridge Inventory were analyzed. The highest prediction accuracy of 94.33% was achieved in Illinois, with notable improvements in Texas (85.13% to 91.77%) and California (81.30% to 89.25%) using hybrid models. Explainable AI (XAI) was also introduced to analyze the influence of climate factors, revealing distinct effects of humidity, precipitation, and temperature across U.S. climate zones, which insights are not captured by traditional AI models. These findings contribute to the advancement of data-driven automation in the design and maintenance of bridge infrastructure.