摘要: This study presents an automated, data-driven framework for optimising sensor data sampling frequency in pavement response prediction (rutting-related surface displacement). Addressing the challenges of excessive data volume and costly storage and processing in road digitalisation, the framework uses data thinning techniques and machine learning to determine the optimal sampling rate. The framework is implemented through real-world experiments at the National Buried Infrastructure Facility. The validation results show that sensor data could be reduced by up to 95% while maintaining predictive accuracy (e.g., R² ≥ 80%). This substantial reduction lowers storage, energy, and operational costs, supporting more sustainable and efficient monitoring of digital road infrastructure. The framework provides practical guidance for balancing data reliability with efficient resource allocation, promoting intelligent, scalable practices in pavement infrastructure asset management.