Sigmoidal Models for Predicting Pavement Performance Conditions

Abstract

This study presents an approach to develop sigmoidal family pavement performance models (pavement performance ratings versus pavement age) for a pavement management system (PMS). Pavement condition data collected from windshield surveys oftentimes suffer quality issues stemming from human subjectivity, and pavement age sometimes not being properly reset after a treatment. These issues can be systematically addressed by the proposed approach, and nonlinear sigmoidal family performance models can then be developed using the cleaned condition data. In a case study, this approach was successfully applied to a sample data set extracted from the North Carolina Department of Transportation (NCDOT) PMS. Contour plots developed for the raw data and the cleaned data showed that the data cleansing process was effective. Goodness-of-Fit indicators and cross-validation suggest that the resulting nonlinear sigmoidal models fit the condition data well.