Grouping Factors in Bridge Component Deterioration: A Data-Driven Comparative Analysis Using Multifaceted Similarity Measures

Abstract

Assessing bridge conditions using deterioration models is crucial for bridge management programs to establish reliable maintenance strategies. Grouping bridges at the component or element level is essential for developing statistically sound deterioration models. However, conventional grouping approaches for deterioration models are based on professional experience or customary practices, presenting a need for more empirical evidence validating their effectiveness. This study addresses this gap by providing a data-driven foundation for developing more reliable deterioration curves. This study aims to explore the similarities in deterioration patterns across various attributes of grouping factors at the bridge component level, seeking to identify relevant grouping factors, develop deterioration curves, and conduct a similarity analysis. We applied the regression nonlinear optimization method to produce deterioration curves using historical condition data from 1992 to 2022 and employed three similarity methods—Euclidean distance, Manhattan distance, and dynamic time warping—for cross-validation. The results exhibit consistent clustering patterns among the grouping factors of designs and materials with similar structural characteristics. Moreover, the study indicates that grouping bridges for deterioration models should consider the interplay of multiple grouping factors. This study establishes a data-driven foundation when grouping bridges to develop more accurate predictive deterioration curves, refining conventional approaches with empirical evidence. The findings offer practical insights for transportation agencies to understand the most influential attributes of grouping factors on bridge deterioration and could trigger a reevaluation of bridge design and construction practices, considering a broader range of operational and environmental factors.