Date of Award

1998

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Civil Engineering

First Advisor

Dr. Alois J. Adams

Abstract

Fine aggregate shape has been identified as a factor in determining asphalt pavement rutting susceptibility. In an effort to improve test methods for measuring aggregate shape characteristics, a video-based, computer-controlled imaging system was developed. Individual aggregate particles were characterized by shape attributes. A neural network classifier used these attributes to calculate a single number classification index that relates to expected performance in asphalt concrete based on angularity. Expert judges were used to train the neural network classifier by scoring a set of particle outlines. Discrepancies in the expert's scores indicate that considerable uncertainties still exist in the concept of 'angularity'. The neural network classifier performed better than the experts at quantifying the shape of individual aggregate particles and was compared to other methods for measuring particle shape. One of these methods, the NAA Method A test, measures void volume in an uncompacted aggregate sample and is the method used by the Superpave specification for designing asphalt pavement mixtures. Superpave's fine aggregate angularity criteria requires a minimum of 45 percent voids. However, in comparison with the neural network, the NAA test appears to lose accuracy above 43 percent voids, suggesting the Superpave criteria may need reconsideration.

Download

Share

COinS