Development of a polarization strain sensor system using a neural network

We propose a polarization strain sensor system based on the principle that the polarization state of light propagating in a single-mode fiber changes when external strains change. We use a three-layer feedforward neural network for data processing. Learning is performed by training a designed neural network using the experimental data as training data. In addition, the output obtained for the test data using the trained neural network is in good agreement with the experimental data used as the test data. This result demonstrates the feasibility of both the sensor system and data processing method.