In-Situ Dynamic Modulus Prediction for Asphalt Pavement Combining Machine Learning Algorithm and Sensing Technology

Abstract

The in-situ dynamic modulus property of asphalt mixtures is critical to aiding the decision-making of pavement maintenance and rehabilitation. With the recent advancement in data science and sensing technologies, embedded sensors have been applied to collect the in-situ signal induced by traffic loading and estimate the traffic information. However, very limited studies have focused on evaluating the mechanical properties of pavement materials and structures using embedded sensors. This paper aims to present a practical approach to performing the in-situ dynamic modulus test and develop an in-situ dynamic modulus predictive model using Artificial Neural Network (ANN) based on real-time sensing data. Particle-size wireless sensors were implemented in several pavement sections to collect data under vehicular loading. An empirical mode decomposition (EMD) method was introduced to calculate the intrinsic modes of the collected data as the ANN model inputs. Laboratory dynamic modulus tests using the same material as the paving projects were also performed with embedded wireless sensors. Those laboratory data, combined with the field sensing data, were used as the training and testing dataset for developing the ANN model. The results show that the developed ANN model, when adequately trained with particle-level sensing data, is feasible and robust for predicting the in-situ dynamic modulus of asphalt pavement.