Robust data driven condition monitoring of pavements based on temperature data

Summary

Unexpected damage and aging of road infrastructure can severely disrupt transportation and hinder Minnesota's economic vitality. Unplanned repairs often cause major delays in the movement of people and goods. To shift from reactive to preventive and predictive maintenance, cost-effective and reliable pavement monitoring systems are essential. In response to this need, this research proposes a data-driven framework to evaluate the thermal performance of pavements under extreme conditions common in cold regions, such as freeze-thaw cycles and heavy precipitation. These environmental stresses often lead to distresses like thermal and reflective cracking. The framework will utilize surface and sub-surface temperature data collected via thermal cameras and thermocouple trees. Analyzing the evolution of the temperature field from surface to depth will support early detection of unseen distresses in both flexible and rigid pavements. To enhance data quality and interpretability, the project will develop advanced techniques for error correction, data cleaning, and model calibration. Machine learning and signal processing methods ? including compressed sensing and neural networks ? will be used to analyze and reconstruct defective data. Additionally, the research will leverage extensive temperature datasets from both high-traffic and low-volume roads at MnROAD, enabling broad application of the findings across Minnesota's transportation network.