Predicting the Resilient Modulus of Unbound Granular Materials by Neural Networks

Abstract

The process of pavement design requires the provision of material properties. For mechanistic–empirical design methods, the resilient modulus represents the most suitable alternative for describing the behavior of aggregate materials commonly used in sub-base and base layers. However, the adoption of the resilient modulus has been slow due to the complicated nature of the laboratory test used to obtain the parameter and its cost. Attempts to correlate the resilient modulus to the widely used California Bearing Ratio and other empirical parameters in the past fall short of providing reasonably accurate estimates of the parameter. With the renewed interest in using the resilient modulus as advocated by the AASHTO 2002 Guide, a quick and inexpensive solution to provide accurate estimates of this parameter is needed. This paper presents the artificial neural network (ANN) technique as a promising method that can help designers have a good first-step estimation of the resilient modulus based on data accumulated over the years. The study h ighlights the use of ANN technique, which utilizes simple parameters as input to predict the resilient modulus of unbound granular materials. Results of ANN simulations confirm the potential of the technique to predict the resilient modulus of compacted samples tested at various compaction densities, state of stress and moisture contents. Such a tool represents an attractive alternative to laboratory testing for small jurisdictions with limited budget and personnel.