Compaction Prediction for Asphalt Mixtures Using Wireless Sensor and Machine Learning Algorithms

Compaction is a critical step in asphalt roadway construction to determine the pavement’s quality and service life. Current field compactions are mainly based on test strips and engineers’ experiences. Such empirical-based approaches sometimes cause compaction problems especially when new materials are implemented. Intelligent compaction (IC) technology was invented to improve compaction quality. The real-time collection and visualization of the pavement responses and temperature can greatly improve compaction uniformity. However, the material viscoelastic property and the complex pavement structure still hinder the establishment of the correlation between the Intelligent Compaction Measurement Value (ICMV) and the compactability of asphalt pavement. This paper aims to establish compaction prediction models based on machine learning (ML) algorithms and sensing technology. Given the strong correlation between the mixture’s compactability and the particle compaction characteristics, a particle-size wireless sensor, SmartRock, was used to collect the particle kinematic behaviors during compaction. 11 asphalt mixtures were compacted with an embedded SmartRock sensor. Three ML models and three dataset scenarios were applied to predict the compaction condition, and a higher than 98% prediction accuracy was achieved. The results revealed that combining the ML approach and sensing data is appropriate and practical for predicting the compaction condition of the asphalt mixture. Future studies shall further evaluate the algorithm based on field compaction data so that intelligent sensing technology can be implemented to improve compaction quality.