Estimating pavement failure time is essential for pavement management at both network and project levels. Pavement failure is usually characterized as a stochastic event triggered by the deficiency of one of several distress types influencing one another. In practice, pavement failure is often modeled as the deficiency of a specific distress condition or based on a composite condition index regardless of competing risks. This is primarily due to the limitation of parametric modeling on competing risks. To recognize the competing risks involved in pavement failure, this paper describes a nonparametric approach based on cumulative incidence function and conditional probability. The non-parametric approach make inference based on time event data and does not require sufficient causal data as parametric models do. Therefore, it is readily applicable to evaluate pavement failure given the pavement condition data routinely collected as an integral part of pavement management systems. The methodology was applied to evaluate the Floridas flexible pavement conditions represented by three distress types: crack, ride, and rut. It showed that the risks of crack failure, ride failure, and rut failure are significantly different from one another with crack failure as the dominating failure mode. Failure risk of overlaid pavements is generally higher than that of as-built pavements. As expected, the estimated pavement service lives concur with the pavement design lives used by the Florida Department of Transportation for as-built and overlaid flexible pavements.
Abstract