Machine-Learning-Assisted Dual Fiber Bragg Grating-Based Flexible Direction Sensing

Flexible sensors have significant potential applications in various fields, such as bionic robots and clinical surgical operations. Compared with the conventional electrical flexible sensors, flexible sensors based on fiber Bragg gratings offer remarkable advantages, such as lightweight, small size, low transmission loss, and strong antielectromagnetic interference, and excellent flexibility and compatibility. To address the issue of isotropic strain sensing in fiber Bragg gratings (FBGs), this article proposes a spiral surrounding structure design, where two FBGs are implanted in a high-toughness resin matrix groove. When subjected to stress in different transverse directions, the FBGs exhibit varying irregular reflection spectra due to deformation. Then, a 1-D convolutional neural network (1-D CNN) model is constructed, and a machine-learning technique is used to identify the lateral force-bearing angle of the sensor. The experimental results show that the flexible fiber optic directional sensing system designed in this article achieves an angle recognition error of less than 2 degrees within a 360 degrees range, which can meet the application requirements of biomimetic robots and clinical surgical operations.

Automated summary

This article presents a flexible sensor based on fiber Bragg gratings and a spiral surrounding structure design. By using a convolutional neural network and machine learning techniques, the sensor can achieve high-precision angle recognition. Experimental results demonstrate that this sensing system can meet the application requirements of biomimetic robots and clinical surgical operations.