Optimization of the ME Rutting Depth Prediction Model Using Swarm Intelligence Algorithms

Rutting, a common disease of asphalt pavement, not only compromises the road quality and safety, but also plays a critical role in the structural design of asphalt pavement in many countries. To achieve more accurate prediction and evaluation of rutting evolution trends, it is particularly important to improve and optimize the existing rutting depth prediction model. Therefore, based on the long-term observation data of the RIOHTrack full-scale pavement acceleration loading test loop, the mechanical-empirical rutting depth prediction model in the Highway Asphalt Pavement Design Specifications (JTG D50—2017) was comprehensively adjusted and optimized. Three calibration parameters were introduced to calibrate the constant coefficients, temperatures, and cumulative load times, respectively, to improve the prediction accuracy and generalization ability of the model. Subsequently, a multi-strategy adaptive particle swarm optimization (MAPSO) algorithm incorporating a neighborhood mutation strategy and fusing exponential adaptive inertia weights with sinusoidal adaptive learning factors, was proposed. Then this algorithm was used to estimate the values of three calibration parameters to further improve the accuracy of the model. Finally, with the rutting data of 19 types of asphalt pavements in the RIOHTrack as an example, the MAPSO-RME model proposed in this article was applied for rutting depth prediction. The experimental results demonstrate that, compared with the mechanical-empirical rutting depth prediction model in the Highway Asphalt Pavement Design Specifications (JTG D50—2017), the MAPSO-RME model achieves remarkable improvement in fitting performance with a significant reduction in mean squared error (MSE) of prediction.