Keywords: low volume traffic roads, deterioration models, neural network analysis models.

The major proportion of the road network in Australia is made up of low volume traffic (LVT) roads that carry a substantial proportion of the total freight load; often with a seasonal harvesting loading, together with a regular lighter loading pattern of regular logistic support to regional centres. These roads are generally single carriage way, two lane roads with flexible pavements and thin seals. This is in contrast to the highways and main roads that carry a high proportion of heavy axle loads to satisfy continual demand for supplies of goods between major centres.

Deterioration models have been derived for the heavily loaded roads but the models for LVT roads have not been paid the same amount of attention. The LVT road deterioration is more affected by environmental deterioration and generally less influenced by axle loads, making the existing models in need of adaption. The development of models for LVT roads is also limited by the availability of databases for the network, recording scientific measurements of road condition and performance over a long period of time. Application of existing models to the LVT road network has shown a wide range of relationships between roughness and age, showing an opportunity to improve the correlation.

An examination of databases of road performance and condition held by Australian Local Government Authorities, who are responsible for a substantial proportion of the LVT roads revealed some "snapshot" databases, complied by examining the pavement condition at a single time point. These data sets had a mixture of objective and subjective measurements, with some missing data.

A region North East of Melbourne was chosen for testing. A linear regression analysis did not provide a good fit to the data with a very low correlation. It was decided to trial a neural network analysis (NNA) approach. Two networks were examined. One used the input of pavement age with an output of surface roughness, which mimicked the relationship developed for the heavily loaded pavements. The other used more inputs such as traffic count, subgrade strength, pavement thickness and age. The first did not provide a good relationship with prediction, whereas the second produced a better fit, encouraging further development of this approach to estimate roughness.

It appeared that the existing deterioration models required improvement to be applied with any confidence to the LVT road network. An impediment to their application is the scarcity of substantial databases. An NNA approach that can accommodate missing data and that can incorporate a wide range of influencing factors has shown promise in producing LVT road deterioration models.

purchase the full-text of this paper (price £20)

go to the previous paper
go to the next paper
return to the table of contents
return to the book description
purchase this book (price £70 +P&P)