The long-term pavement performance (LTPP) data offered a unique opportunity to explore data with known provenance. A single LTPP site was selected to explore the potential of using the full data set as opposed to the aggregated data represented by its average. Data aggregation tends to remove some of the information available in the full data set. Exploration of the nature and extent of the information lost due to data aggregation indicates that by representing the data with a measure of its central tendency, the probability of misrepresenting the data increases, depending on its uniformity. The use of a new data condensation technique was demonstrated by applying cumulative histograms (cumulative probability distributions) to the full data set. The deterioration history of the selected site is illustrated and documented. A review of the distributions indicated an unexpected major rehabilitation of the site, which was later confirmed by the maintenance records. Current models focus on forecasting the progression of either the severity of properties (e.g. rutting) or extent (e.g. cracking). Extent without severity, or severity without extent, is of limited value for the purpose of performance forecasting. Some models, such as those for roughness, combine the extent of one property with the severity of another one with reasonable success. However, questions may be raised about the logic of mixing variables representing severity with those representing the extent of distress. Recent developments in computer technology combined with the advent of data condensation technologies have opened new possibilities to analyse and explore data previously restricted to organisations with access to statistical packages and complex links to conventional data base packages. Data sets can now be condensed into a single stochastic information package (SIP) stored in an HTML text string. This text string occupies a single cell in Excel or one item in a database. The way is now open to use a data set (full population) instead of a singular representative such as a mean or median value. These tools are readily available in the literature, making the overall procedure both accessible to, and portable between a wide range of end users and applications. Use of the full data set also allows innovative approaches to modelling the distribution of the data as opposed to modelling a singular representation, such as an average, as commonly practised. Modelling distributions may require a novel approach using the new tools available for dealing with complete data sets. Using the full data population offers several benefits, such as: severity and extent can be evaluated to allow accurate targeting of maintenance; errors related to representing data by averages can be reduced; maintenance history can be revealed from the data; local failures of the road can be identified; and, the uniformity of a road section can be determined and monitored. The work presented in this report challenges well-established practices and procedures. The readily available new techniques meeting these challenges can offer many benefits to the road construction and maintenance industry. Exploration of data management and modelling opportunities is proposed, using the SIP technique on an Excel platform, by means of a pilot study on a small-scale road network over a 10-year analysis period using LTPP data, or similar. This study can be used to demonstrate the variations and risks in predicted conditions and maintenance works program budgets. (Author/publisher)
Abstract