Optical fiber shape sensing has diverse applications in medical and industrial fields. However, commercially available fiber shape sensors are costly and complex. The development of eccentric fiber Bragg grating (eFBG) sensors provides a cost-effective alternative with unique capabilities. Existing eFBG shape sensing methods calculate curvature using Bragg signal intensity variations. Yet, uncontrolled bending and polarization-dependent losses cause spectral distortions affecting eFBG intensity ratios. To overcome this, we developed a data-driven deep-learning technique for accurate shape prediction. Our approach significantly improves shape prediction, achieving millimeter-level accuracy for curvatures of 3 cm to 70 cm in a 30 cm eFBG sensor. This promising research advances low-cost and accurate fiber sensors, impacting medical and industrial sectors requiring precise and cost-effective shape sensing.
|