Assessment of pavement structure using multi-source data from a moving vehicle

Abstract

In this paper we propose a novel method for assessing pavement structure based on multi-source data from a moving vehicle. The classification of pavement on subgrade (POS) and pavement on bridge deck (PBD) was used as examples. A mobile acquisition system was designed to collect multi-source data. Vehicle accelerations on PBD and POS were analyzed, and some features were selected to build 15 Support Vector Machine classifiers. Our results show that some features are helpful, including vehicle speed, maximum value, minimum value, standard deviation of acceleration in the time domain, and three peak frequencies in the frequency domain. The accuracy was 91.67%. Then, some sections were tested with a falling weight deflectometer to verify the structural differences between POS and PBD. Our results provide a novel method for analyzing pavement structure via vehicle vibration, and this work will be built on for more detailed analysis of pavement structure in the future.