Transportation Infrastructure Performance Modeling through Seemingly Unrelated Regression Systems

Abstract

Deterioration modeling plays a central role in performance prediction, a key component in transportation infrastructure system management. The deterioration of a facility can be expressed in terms of numerous indicators. For instance, in the case of highway pavements, these indicators range from panel ratings (such as pavement serviceability rating) to distress variables (such as percentage of surface cracking and rut depth). Different indicators capture different deterioration mechanisms; however, they may be correlated to each other. Numerous studies have been conducted involving modeling the deterioration of transportation facilities. Empirical regression has proved to be an effective way to characterize the relationship between the explanatory variables and dependent variable (performance indicator). The majority of these models were established with the focus on each individual performance indicator separately and independently. However, one type of indicator does not sufficiently capture the performance of a facility. For instance, while a softer binder may improve fatigue cracking performance, it could possibly be detrimental to rutting performance. As a result, the facility deterioration mechanism may not be fully characterized by single models. Since decision making in infrastructure management usually relies on a series of different performance indicators, the systematic considerations of these indicators is imperative. In this study, a flexible and statistically sound modeling approach is presented to jointly account for the performance of a transportation facility based on two indicators. A system of nonlinear structured econometric models is established to capture different deterioration processes in terms of these indicators. Due to the fact that the selected indicators represent the performance of a single section from different aspects, the correlation among the models in the system is particularly addressed. A case study is presented to demonstrate the feasibility of the proposed methodology with focus on in-service highway pavements. Two major indicators, international roughness index and rut depth, widely used in characterizing flexible pavement performance, are adopted and analyzed in the model system. The results demonstrate that: (1) different indicators are associated with different deterioration mechanisms reflected through their respective parameters; (2) the correlation of different deterioration mechanisms and relation of parameters across the models can be captured through relevant statistics; and (3) model estimation is efficient when accounting for the correlation among the models.