Real-Time Key Information Extraction for Highway Structural Health Monitoring Data Based on Random Forests

With the rapid advancement of informatization and intelligence in transportation infrastructure, Structural Health Monitoring (SHM) data collected by Fiber Bragg Grating (FBG) strain sensors embedded in highways have become an important foundation for health assessment and maintenance decisionmaking. However, the high sampling frequencies and large-scale deployment of these sensors generate massive amounts of data, placing tremendous pressure on data transmission to cloud-based analytical systems. These sensors lack computational and storage capabilities, yet the cloud-based analytical systems rely heavily on this data. To address this challenge, we propose a real-time key information extraction method for SHM data in edge computing environments, utilizing Random Forests to efficiently identify and transmit key information from SHM data while filtering out irrelevant information. Experimental evaluations based on real highway SHM data demonstrate that our method reduces data transmission volume to 1% of the original size while achieving a high recall rate, ensuring network efficiency and real-time performance. This study highlights the potential of integrating edge computing and machine learning (particularly Random Forests) for efficient SHM data processing in large-scale highway infrastructures.