Experimental earthquake early detection through polarization changes in intelligent optical networks

In this paper, we replicate the dynamics of an earthquake in a laboratory experiment by employing an optical scrambler and a polarimeter. This experimental setup emulates the impact of a real earthquake-induced ground displacement values on the state of polarization of light propagating through an optical fiber, modeled using a Waveplate computer-based approach. A large dataset of polarization evolution was collected from the experiment to evaluate the performance of our pre-trained machine learning model in detecting the primary earthquake wave which precedes the destructive wave by tens of seconds. The objective is to perform a comparative analysis with our previous findings which were conducted using the same machine learning model, but applied to computer-based simulations of the Waveplate model. The results demonstrate that the model achieved over 95% of accuracy in both computer-based simulations and Laboratory-based experiment, validating the high accuracy and reliability of our system in early earthquake detection despite the inherent challenges of experimental errors.