A Hybrid Approach Based on Regression Analysis and ANN for Non-destructive Asphalt Road Density Measurement

Abstract

The performance characteristics of asphalt pavement, including durability and resistance to deformation, are linked to its density. Accurate measurement of density is, therefore, critical for the evaluation of asphalt pavement performance, which is commonly performed with the coring method (CM) and the Pavement Quality Indicator (PQI). The former provides high accuracy, but it is destructive, inefficient and requires additional repairs to the pavement after the cores are taken. In contrast, the PQI-based method is non-destructive and efficient, but its accuracy is comparatively lower. The accuracy of the PQI-based method can be improved by applying data processing analysis techniques such as regression analysis and artificial neural network (ANN). This paper proposes a hybrid approach that combines both regression models and ANN models. The density and temperature measured with a PQI are input into the regression models for optimisation. In addition, the optimised regression-model-predicted density is then used to train ANN models. The effectiveness of the proposed approach is validated by the results of the field study.