Artificial neural network-based prediction of field permeability of hot mix asphalt pavement layers

ABSTRACT

Field permeability of Hot Mix Asphalt (HMA) needs to be controlled to prevent excessive ingress of water into asphalt pavements, which leads to premature failure. Existing literature shows scatter in the prediction of field permeability values and relationships between factors affecting permeability are complex and are not known precisely. The objective of this paper is to present the analysis and modelling of field permeability of HMA with artificial neural networks (ANN). Permeability data from field testing at five sites, along with materials and mix data are presented. Preliminary statistical analysis to identify correlations and significant factors were conducted. A three-layer ANN was built and trained with part of the data and it was validated and also tested on two separate sets of data. The performance results indicated excellent prediction ability of the neural network. Experiments were conducted to improve the model, and the relative importance of the different factors was evaluated. Gradation and air voids were identified as two of the primary factors affecting field permeability. It is recommended that ANNs be considered to be used on a regular basis for predicting field permeability, particularly because field permeability tests are processes that require a considerable amount of time and resources.

KEYWORDS:

• Pavement
• permeability
• machine learning methods
• artificial neural network
• gradation
• air voids

Disclosure statement

No potential conflict of interest was reported by the authors.