Collapse prediction of karst sinkhole via distributed Brillouin optical fiber sensor

Karst sinkhole collapse is one of the most serious failure modes in railway subgrade, which often leads to dangerous operation and disaster throughout the world. Karst sinkhole has the characteristics of strong insidious and wide spatial distribution, the common collapse monitoring and prediction methods cannot work well. In this paper, the distributed Brillouin optical fiber sensor was proposed to measure the sinkhole collapse strain, a long-term monitoring model test for sinkhole collapse was conducted which can provide the distributed collapse time-series strain. Based on the time-series data, an AR(2) sinkhole collapse model was constructed, and test results shows that the AR(2) model can predict the sinkhole collapse well from the test results.