Torsion and bending sensing based on the specklegrams from a coupled few-mode multi-core fiber

A novel optical torsion and bending dual-parameter sensor based on the detection of a coupled few-mode multi-core fiber (FM-MCF) specklegrams is proposed and experimentally demonstrated. A 3mm-long multimode fiber (MMF) is used to couple the light from the injection single mode fiber into all cores of the FM-MCF, and thus the fiber structure is formed by the ultra-short MMF and FM-MCF. The specklegrams of the FM-MCF can be detected and segmented to infer the change of the FM-MCF. The torsion angle of the FM-MCF from 0-360 can be determined accurately by detecting the overall rotation angles of the specklegrams. And since the modes in each core of the FM-MCF are strongly coupled and the mode interference is highly sensitive to external disturbance, bending sensing based on the analysis of sub-specklegrams from each outer core is verified within the curvature range of 2.57-12.83m-1 with the help of deep learning. Furthermore, based on a dual-output layer classification network, the capability of the sensor for torsion and bending sensing simultaneously is tested. The recognition accuracy for 36 combinations of 4 torsion angles and 9 bending states are 100% and 97.12%, respectively. The sensor here proposed has the advantages of low cost, simple structure, ease of interrogation and good accuracy.