Image-based intelligent detection of road cracks with high accuracy and efficiency is vital to the overall condition assessment of the pavement. However, significant problems of continuous cracks interruption and background discrete noise misidentification are frequently observed in current semantic segmentation of pavement cracks, which mainly caused by traditional segmentation convolutional neural networks. This paper proposes a skip-level round-trip sampling block structure with the implementation of convolutional neural networks, thereby constructed a novel pixel level semantic segmentation network called CrackW-Net. After that, two datasets, including the widely recognized Crack500 dataset and a self-built dataset, were used to train two versions CrackW-Net, FCN, U-Net and ResU-Net. Meanwhile, comparative experiments are conducted among all these network models for crack detection. Results show that CrackW-Net without residual block performs the best in the task of pavement crack segmentation.

Neural network-based models have been implemented to predict various health indicators of asphalt pavement using pavement historical detection data. Unfortunately, their accuracy and reliability are not acceptable owing to their shallow architecture. To solve the issue, this study proposed an improved Transformer network to predict asphalt pavement health, called the Transformer with forward and reversed time series (Transformer FRTS). In terms of the input data, Transformer FRTS uses a new data form, so-called the random difference time series, to reduce the time dependency of the network prediction. In terms of the network architecture, the proposed network uses its encoder and decoder to obtain the data association from the forward and reverse time series. In addition, Transformer FRTS uses a post-processing decision criterion to improve the accuracy and reliability of prediction. The numerical experiment using the detection data from RIOHTrack full-scale track demonstrates that the proposed network has state-of-the-practice performance in asphalt pavement health prediction.

Real-time structural health monitoring (SHM) is important for the performance evaluation and maintenance management of asphalt pavement. This study applies SmartRock, a wireless embedded sensor widely used in track health monitoring, to the long-term SHM of asphalt pavement to achieve its residual life assessment. According to the structural characteristics of asphalt pavement, a novel processing method of SmartRock monitoring data is proposed. Firstly, An improved genetic algorithm for back-calculation of pavement dynamic modulus using SmartRock monitoring data is proposed. Then, the evaluated dynamic moduli are used to predict the residual rutting life and residual fatigue life of the pavement structure. Finally, the two types of residual life are combined to achieve the pavement residual life. This research uses the real-time monitoring data of a test road in Shandong Province, China, to carry out a case study to verify the proposed approach.

Rutting development in semi-rigid asphalt pavements can be predicted using models such as mechanistic-empirical (M-E) and data-driven methods. However, the prediction accuracies of the M-E methods in the laboratory may not be generalised due to the boundary condition differences, while inappropriate calibration in the data-driven methods may reduce the reliability owing to the lack of theoretical support. This study has proposed a combined framework of an M-E method and an artificial neural network (ANN) for rutting development prediction. This framework, namely the mechanistic-empirical and artificial neural network method, first calibrated the existing M-E model by adding a term of the time-varying hardening characteristics of asphalt mixture and optimised the parameters in the term using a genetic algorithm. The new M-E model was used to predict the possible range of rutting depth. An ANN then predicted the rutting depth that was normalised by the predicted range of the new M-E model. Thus, the proposed approach combined the reliability of the M-E method and the prediction accuracy of the ANN method. Field test results showed that the proposed framework with a 97.5% accuracy outperformed the ANN-based method.