REQUIREMENTS FOR RELIABLE PREDICTIVE PAVEMENT MODELS

The general requirements for reliable pavement prediction models are presented. Performance models are essential for efficient management of pavements. Experience has shown that they can best be derived from a data base of in-service pavements. The major requirements of a reliable model for predicting performance, herein defined as serviceability index and distress occurrence over time, are (a) an adequate data base built from in-service pavements, (b) the inclusion of all variables (including mechanistic variables) that significantly affect performance, and (c) an adequate functional form of the model that considers shape, nonlinearities, and interactions; meets boundary conditions; and provides reasonable sensitivity of variables. The model must also meet statistical criteria for precision (e.g., error of prediction, R squared and regression coefficients). (Author)