Describing Changes in Road Condition by Numerically Described Multi-dimensional Probability Distributions

Abstract

With regard to a systematic pavement maintenance, it is necessary to hold information about the current state of roads and to forecast its changes precisely. Therefore, a Pavement Management System (PMS) for highways was developed in the past few decades, which is shared meanwhile by mostly all road agencies in Germany. Most PMS consist of deterministic performance predicition models, which were developed based on historical empirical data. These deterministic prognoses are roundly approved and accepted, due to their simple practicability and comprehensibleness even though they contain any information concerning the quality of prediction. This lack of information includes the necessity to improve probabilistic prediction models to handle different kinds of maintenance strategies. Therefore, the aim of this thesis was to verify, if describing changes in road condition by numerically described multi-dimensional probability distributions is feasible and beneficial to improve the systematic management of maintenance. First, the following possible explanatory variables like age, traffic loads, design and width of lane were analyzed concerning their power of influence on rut depth and cracks. The results of these analyses of variance (ANOVA) confirmed that the age of roads is the most effective variable to predict the prospective condition of roads. Furthermore, the wide distributions of road condition represent that its development over time is influenced by a large number of predictors. As a consequence, a probabilistic forecast model was drafted in dependence of road age. Based on pavement condition survey data of more than 500km of Bavarian roads, a set of age-related transition probability matrices was developed for the use of recurrent Markov Chains. The practicability of this section based forecast model was tested with an example of use to forecast the prospective condition of an exemplary road section and pointed out its importance and profit in PMS. In this research a probabilistic age-related forecast model was developed which is able to indicate the variance of predicted road conditions. The developed forecast model gives a fundament to discuss several alternatives for maintenance strategies, which can be implemented in modern PMS. Furthermore, the consideration of variance of service life gives an expansive potential to be integrated in construction and maintenance contracts.