Machine Learning Assisted Optical Fiber Vector Bending Sensor Based on MZI

This paper presents a Mach-Zehnder interference-based axisymmetric misalignment structure and analyzes the curvature response in the 0°, 90°, and 180°. Breaking through the limitations of the wavelength linear fitting demodulation method, Support Vector Machine (SVM) can accurately distinguish between the three directions, meanwhile, Artificial Neural Network (ANN) can precisely predict the curvature magnitude in each direction with the Mean Absolute Error (MAE) only 0.021 m<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">-1</sup>. With the assistance of machine learning algorithms, simultaneous and accurate demodulation of both vector and curvature is achieved.