Piezoresistivity assessment of self-sensing asphalt-based pavements with machine learning algorithm
Zhizhong Deng, Quang Dieu Nguyen, Aziz Hasan Mahmood, Yu Pang, Tianxing Shi, Daichao Sheng
Abstract
Due to its fast-curing process, asphalt binder has been increasingly used to replace conventional cementitious materials and to produce asphalt-based self-sensing sensors (ASS). The asphalt mixture does not require consideration of curing age, as asphalt-based samples can be cured at room temperature. This is because the asphalt binder transitions from a Newtonian liquid to a solid state upon cooling. The relationship between strain and electrical response is one of the factors that influences the self-sensing performance. In this study, four-electrode method was applied to ASS and 10 V of applied voltage was used to measure the piezoresistivity. The percolation thresholds of ASS, in the range of 0.5–1.0 wt%, was found based on study results. In order to analyse the strain of ASS, digital image correlation (DIC) method was applied, and the cyclic loading process was used to simulate the practical pavement situation. Furthermore, to enable the application of ASS and enhance the efficiency of the self-sensing system, a machine learning approach was applied in this study to establish the relationship between strain changes and the electrical response of ASS. The trained algorithm, exhibiting a high determination coefficient (reached 0.965), can be utilized to predict strain changes in self-sensing sensors based on fractional resistance changes. Artificial intelligence significantly enhanced the application potential of self-sensing sensors. • Carbon fibers reduce electrical resistance in asphalt-based samples; optimal content range is 0.5‐1.5 wt%. • Conductive paths improve due to asphalt shape change, enhancing piezoresistivity. • Machine learning algorithm predicted strain changes with R² higher than 0.96, demonstrating high accuracy. • Study confirms improved self-sensing in asphalt via machine learning for practical use. • Future research should focus on enhancing machine learning and self-sensing methods cooperation.