Directional bending monitoring using a multimode elliptical-core fiber and a machine learning algorithm

Abstract

An approach for directional bending monitoring based on a multimode fiber and a machine learning algorithm is presented. The sensor if formed by splicing a single mode fiber to a multimode elliptical-core fiber. Using this elliptical-core fiber, multimode interference generates an interferogram with non-uniform amplitude and non-periodic shape. These characteristics are important to process the sensing signal using a machine learning algorithm. The machine learning algorithm implemented is the well-known random forest algorithm. In the reported experiments, the fiber is bended in different directions and different magnitudes of bending, generating a specific interferogram in each position, then each bending position is identified by the random forest algorithm. Once the position is identified, the trajectory of the sensor can be calculated. Experimental demonstration for directional bending monitoring, based on a machine learning algorithm, is presented.