Investigation of Relationship Between Deterministic and Probabilistic Prediction Models in Pavement Management

Abstract

A good pavement management system should have the capacity to predict pavement structural and functional deterioration versus age or accumulated traffic loading. Basically, there are two types of performance prediction models in pavement management: deterministic and probabilistic. Although both performance models can be used to predict pavement deterioration, the inherent relationship between the two models has not been explored. An investigation was directed to find the relationship in terms of system conversion. Some of the findings related to system conversion, including the concepts and techniques applied in model conversion, the characteristics of model development, comparisons of prediction results between the two models, sensitivity analysis of the probabilistic models, and sample applications in real situations, are highlighted. The deterministic models that are to be converted to probabilistic models are the flexible pavement deterioration model used in the Ontario Pavement Analysis of Costs system and the flexible pavement design model recommended in the 1993 AASHTO design guide. The converted probabilistic models are time-related (nonhomogeneous) Markov processes, which are represented by a set of yearly transition probability matrices (TPMs). TPMs can be established for any individual pavement section in a road network.