Fiber Bragg grating-integrated soil settlement three-dimensional trajectory pipe sensor: dynamic soil subsidence evolution and stage prediction

Traditional soil settlement monitoring fails to capture dynamic evolution accurately, endangering engineering safety. Here, a fiber Bragg grating-integrated 3D trajectory pipe sensor (SST-3D) is developed for soil settlement strain monitoring, combined with machine learning for loess settlement stage prediction. Air tests verify FBG wavelength shift has a good linear strain relationship; soil burial tests identify four strain response stages during drainage, namely initial rapid response, relative stability, pre-collapse dynamic stage, and post-collapse sustained response. The Frenet frame enables accurate 3D morphological reconstruction of SST-3D in both conditions. Among four machine learning algorithms, Random Forest performs best, with 95.65% classification accuracy for the loess collapse stage and only 4.02% regression relative error. The successful development and experimental verification of the SST-3D sensor has achieved precise perception and dynamic monitoring of the multi-directional settlement morphology of soil.