Landslide monitoring using optical fiber sensing based on polarization

The Master Thesis work has its main focus on Optical Fiber Sensing, applied to the detection of dangerous events in several types of geology monitoring, such as for debris in a mountain scenery. In particular, the monitoring system that has been built is based on the detection of vibrational events through the continuous check of the polarization state of light conveyed inside the fibers. The intensity of the stresses that happen on the fiber is strictly related to the speed with which the polarization state changes with time. The sensor proposed in this Thesis bases its working principle on this concept, in particular on the computation of the angular speed at which the Stokes vector, which conventionally represents the polarization state over the Poincaré sphere, changes along time. This work has been developed in collaboration with a Turinese company (Geosolving srl) and financed as a project of Piedmont Region. All the experiments have been performed on a reduced scale model of a side of a mountain with size of approximately 3 by 0.7 meters. In this way, it has been possible to generate and analyze a lot of events that would be reasonably similar to what a dangerous event on a real scale would look like. Moreover, different fiber configurations buried inside the soil of the model have been tested, changing depth with respect to the surface and disposition along the soil. This scenario gave the possibility to deeply understand the limits and the advantages of this approach, which in the end gave very interesting and optimal results. The first part of the Thesis has been devoted to a brief introduction to the Optical Sensing state of the art scenario, and to the description of how the experimental set up has been built, with a detailed focus on the devices used. A description of the generation of events has been also provided, along with a preliminary but deep analysis of how the fibers reacted to them, and how to improve in the best way possible the processing of the samples acquired, to get the best out of them. The remaining and most important portion of this experimental Thesis has been instead dedicated to the creation and optimization through Matlab of a real time algorithm that would be able to detect in real time the occurrences of dangerous events and, in a real world scenario, potentially protect people from accessing an unsafe area, by triggering an alarm (example: switch on of a red traffic light when something anomalous is detected). It has been found experimentally that the proposed system works perfectly in the reduced scale model: the detection of some dangerous events happens correctly and in real time, with a delay of no more than a few seconds. These experiments have been possible thanks to the interaction between a proper algorithm (written in Matlab), and a polarimeter, which is the actual sensor. The only limits that have been observed are about the inability of the system to distinguish between different events, and also the fact that the current version of the monitoring system does not allow to localize the position of the vibration events along the fiber. The first issue has been actually discussed and partially solved, thanks to the application of Neural Networks that could be able to correctly classify events. A focus on the implementation of these ones in the scenario described up to now is an interesting possible follow-up of this Thesis.