# Vector Bending Sensor Based on Power-Monitored Tapered Few-Mode Multi-Core Fiber

## TL;DR

A new bending sensor uses a special fiber to detect how and where it bends, with high accuracy and low cost.

## Contribution

A tapered few-mode multi-core fiber sensor with power differential and DNN for accurate bending direction and curvature detection.

## Key findings

- The sensor achieves a sensitivity of at least 0.14/m−1 for curvature detection.
- The DNN predicts bending radius and rotation angle with errors under 0.038 m and 3.087°.
- The method enables vector bending sensing with high accuracy for small curvatures.

## Abstract

We propose a vector bending sensor based on a tapered few-mode multi-core fiber (FM-MCF). A seven-core six-mode fiber is tapered with an optimized taper ratio, enabling bending sensing through power monitoring. When the tapered FM-MCF bends, coupling occurs between the central core and side cores in the tapered region. By monitoring the power of all cores and employing a power differential method, the bending direction and curvature can be reconstructed. The results show that within a curvature range of 2.5 m−1 to 10 m−1, the sensitivity of the ratio of the side core’s power to the middle core’s power with respect to curvature is not less than 0.14/m−1. A deep fully connected feedforward neural network (DNN) is used to demodulate all power information and predict the bending shape of the optical fiber. The algorithm predicts the bending radius and rotation angle with mean absolute errors less than 0.038 m and 3.087°, respectively. This method is expected to achieve low-cost, high-sensitivity bending measurement applications with vector direction perception, providing an effective solution for scenarios with small curvatures that are challenging to detect using conventional sensing methods.