Data Mining of the Caltrans Pavement Management System (PMS) Database

Since 1977, Caltrans has been routinely collecting performance information for its pavement network and using a pavement management system (PMS) to manipulate this information in order to aid in the management of the network. This report details the " mining " of this database to extract environmental performance indicators for the various climate regions in California, and to extract section information for portland cement concrete (PCC) pavements that have been overlaid with asphalt concrete (with the specific goal of obtaining information regarding the reflection cracking performance). The first objective of this study is to provide the best possible estimate of the performance of the standard Caltrans strategy for asphalt concrete overlays of PCC pavements, to provide an estimate to Caltrans, and to provide calibration data for the development of mechanistic-empirical models for reflection cracking. The second objective of this study is to provide recommendations for short-term and long-term changes to the Caltrans PMS and the operations that support and maintain it. Before performance models could be developed, the PMS database required the following work: • Conversion from the data records that were collected following the Caltrans " dynamic segmentation " procedure to fixed-length pavement segments. Dynamic segmentation means that for each year, the nodes that define pavement segments change depending upon the observed condition of the pavement surface. Dynamic segmentation is a very good approach for maintenance scheduling, but does not lend itself to predicting performance over time, which is necessary for any optimization of expenditure and the extraction of a statistical sample for performance modeling. v • Conversion of data to appropriate formats for analysis (primarily conversion of text strings to numeric values). • Removal of empty and redundant fields. • Establishment of a primary index to link records across time. This was necessary because the " sequence numbers " used to identify sections over time were found to be non-unique and not uniformly applied. • Linking of the several databases of condition survey data with the traffic database and the awards database. Problems encountered while converting the data to fixed segments and merging of the several PMS databases developed between 1978 and 2000 were as follows: • The biggest problem with extracting pavement performance information from the database is that the only information contained in the database that relates to the pavement structure is whether the pavement surface is flexible (asphalt concrete) or rigid (portland cement concrete). This …