Network Level Analysis of Pavement Maintenance in Norway

Abstract

There has been identified a need for a network level Pavement Management System (PMS) to carry out strategic analyses for the Norwegian road network. As part of the initial effort to achieve this a research project has been carried out parallel to the development of a project level PMS. A Finnish system using Markov-chain modeling of deterioration for different main classes of roads and dynamic programming optimization techniques has been used with data about the Norwegian national road network (26,500 km). This paper describes the methodology used and the latest results (as of 2001) of the ongoing research effort to evaluate a network level pavement management system in Norway. In addition a short presentation of the existing PMS at the project level is given. The main tasks done as part of this work were: The road network was divided into 15 sub-networks according to climatic zone, road type and AADT-classes. Deciding which condition parameters to use. In this first research project only rutting and roughness were included, mainly because these were the only two which there was enough good quality data for. Bearing capacity and surface defects could have been included in the analysis, like they are in e.g. Finland, but the lack of enough good quality data together with the fact that this was a first trial of the methodology made us decide not to include these two parameters at this stage. They may be included at some future stage. Each condition parameter was divided into 3 classes (good, medium and poor), making a total of 9 discrete condition states. Markov chain probability matrixes were developed using data from all the 26,500 km from pairs of consecutive years, filtering out data for road sections which had M/R-actions applied to them in either of the two years. In this way the data should express the degradation of the road network. Four main categories of M/R-works were defined with costs depending on the condition state of the road and sub-network. The effect of M/R-works was also modeled by using Markov-chain transition probability matrixes. Road User costs were modeled for each condition state for every sub-network. Dynamic programming techniques were then used to find the theoretical optimal long-term, steady state condition distribution for each sub-network. These were then used to find how various short-term M/R-strategies (taking into account annual budget restrictions) affected the total condition distribution over a 10-year analysis period. Although the approach chosen in this research project shows a lot of promise, the models used need to be improved. A continuation of the research project is planned for 2002. Continued research and development is needed in the years to come in order to achieve a fully functional PMS at all levels of the NPRA.