NEW TECHNIQUES FOR MODELING PAVEMENT DETERIORATION

Abstract

The ability to model and predict pavement condition accurately is critical to the success of pavement management systems. This paper evaluates the applicability of three mathematical curve-fitting techniques for modeling pavement condition deterioration behavior. The mathematical models investigated are: stepwise regression, B-spline approximation, and constrained least-squares estimation. The best features of each are integrated into an interactive format capable of operating within the PAVER pavement management system. Pavement sections from a given location consisting of the same pavement type, use, and rank are grouped into families. Models that filter obvious errors and statistical outliers from the data are applied to the family data. Both the B-spline approximation and constrained least-squares techniques are used in the screening process. A constrained least-squares curve-fitting technique is used to fit a family prediction model curve to the filtered data. Pavement condition prediction at the section level is accomplished using the position of a section relative to the family prediction model curve. Extrapolation of pavement family condition values beyond the maximum age available in a family data base is accomplished using a backtracking method. These methods were determined to give the most consistent and believable results from among a number of possible methods considered. These procedures constitute a complete method to model and predict pavement family behavior and pavement section behavior accurately. The method is being integrated into the PAVER pavement management system to improve the pavement evaluation process.