Real-time surface shape sensing for monitoring flexible structures using fiber Bragg grating

Abstract

An optical shape sensing system determines the position and orientation of an optical fiber along its entire length. An optical fiber is minimally intrusive, light-weight and can be used to monitor the dynamic 3D shape of a structure to which it conforms. Due to its unique advantages over conventional electric sensors, for example, its immunity to electromagnetic inference, fiber optic sensing has raised great research interest to the community. Some researchers demonstrated the ability of temperature and humidity sensing, 1D bending curvature tracking, and vibration sensing, etc. However, currently there are no fiber optic sensors developed for shape sensing on a 3D deformed surface with the consideration of in-plane stretching, which is especially useful for measuring the morphological changes in the field of soft robotics. In this research, fiber Bragg gratings (FBGs) are embedded off-center in a soft and flexible sheet for sensing shape changes in 3D deformed surfaces. With FBGs offset from the neutral plane of the silicone film, both extensions and bending curvatures can be captured by the changes in strain (the shifts in wavelength) of bent fiber. With a proper fiber routing design, surface shapes in 3D could be reconstructed. Finite Element Analysis (FEA) is implemented to predict sensor performances by a series of parametric studies. Strain vectors are extracted at selected deformation mode shapes. Strain patterns are visualized with clustered strain vector groups. Leveraging the sensitivity, the capability of discriminating deformation patterns and the ease of fabrication, a circular fiber configuration (Ø35 mm) is selected. The effect of fiber offset distance h, sensor thickness t, and fiber rigidity E are briefly discussed based on the selected fiber configuration. Taking in-plane stretching into consideration, the final design of the sensor (45mm × 45mm × 5mm) is in a dual-layer and circular fiber configuration. The sensor is modeled by an artificial neural network (ANN). By evaluating the trained network, uniqueness of strain-wavelength relation is validated. High regression index R = 0.99 is obtained with a root mean square error of 0.9234 mm and a maximum error of 4.81 mm. Accuracy could be improved by denser gratings. By repeating the sensor unit or changing the fiber configuration, the sensor can be customized for different applications. A brief discussion on fiber routing configurations is given to improve overall sensor stretchability. The goal of this research to develop a standard workflow of soft sensor design and a real-time surface shape sensor for soft robotics applications. Future studies include optimization of grating parameters and fiber routing.