Infrastructure Investigation Using Latent Class Cluster Analysis

Abstract

Highway bridges across the United States constitute an infrastructure component that represents an enormous investment at the federal, state, and lo cal levels. Although the poor condition of bridge structures has been identified, monetary constraints due to decreased and competing budgets prevent the complete rehabilitation and/or replacement of these structures. It is becoming increasingly important to use available funds in the optimal way. This requires the best strategies to identify, monitor, and ultimately repair, rehabilitat e, and reduce degradation. This study provides an analysis aimed at identifying distinct and important subgroups of bridges. Stratified samples that represent these subgroups of general bridge types can then be studied in depth, and the findings would apply to the corresponding subpopulation of bridges of similar construction. In particular, such a representative sample of bridge structures could be used for long-term monitoring in order to learn more about the degradation processes and how these processes interact with different bridge structures, material types, and other variables that are associated with bridge perfor mance. A latent class cluster analysis is used to identify similar groups of bridges based on a comprehensive set of variables that are found in the National Bridge Inventory database. Although computationally complex, the analysis provides a comprehensive decomposition of a selected portion of the bridge inventory into groups that are defined in terms of the multivariate relationship of attributes for bridges of the same material and structural type.