Du Thurian Le, Hartog Arthur, Hilton Graeme, Didelet Roman
FOSINA, 23 rue du port, 92000 Nanterre, France.
Sensors (Basel). 2025 Jul 4;25(13):4180. doi: 10.3390/s25134180.
Distributed acoustic sensing (DAS) is a highly effective method of monitoring all kinds of intrusions on railway tracks. These intrusions represent a real problem in the railway sector, as they can lead to human deaths or damage to railway tracks, and these intrusions may be human or animal. A fiber was deployed along 12 km of track in a railway test center, enabling us to acquire data day and night. A data acquisition campaign was carried out in April 2023 to capture the signatures of several scenarios (walking, digging, falling rocks, etc.) in order to train machine learning models and prevent any intrusion by detecting and classify these intrusion. The study shows the diversity of signals that fiber can acquire in the rail sector and the machine learning model performance. Signals associated with the presence of animals are also presented.
分布式声学传感（DAS）是一种监测铁路轨道上各类入侵情况的高效方法。这些入侵情况在铁路领域是一个实际问题，因为它们可能导致人员死亡或铁路轨道损坏，而且这些入侵可能是人为的或动物造成的。在一个铁路测试中心，沿着12公里的轨道部署了一根光纤，使我们能够日夜采集数据。2023年4月开展了一次数据采集活动，以捕捉多种场景（行走、挖掘、落石等）的特征，以便训练机器学习模型，并通过检测和分类这些入侵行为来防止任何入侵。该研究展示了光纤在铁路领域能够采集到的信号多样性以及机器学习模型的性能。还呈现了与动物存在相关的信号。
