Discriminating Temperature, Strain, and Bending Effects in Fiber Bragg Grating Fabry- Perot Interferometer Sensors Using Traditional Machine Learning Techniques

Abstract

<p>In this study, for the simultaneous measurement of axial forces, temperature, and bending curvature, we analyzed Fiber Bragg Grating (FBG) and Fiber Bragg Grating Fabry- Perot Interferometer (FBGFPI) sensors embedded in 170 μm thin optical fibers, which are suitable for medical applications such as implant strain measurements, catheter ablation, endovascular guidewires, and temperature sensing during thermal therapies. The evaluation of the sensor signals was conducted using both traditional machine learning methods and conventional methods. Multiple datasets were generated for the training of the models, each containing labeled sensor signals with a predefined parameter range of pressure, bending, and temperature. Our findings indicate that we were able to simultaneously detect axial tip pressure with a root-mean-square error (RMSE) of 0.053 bar (R2 = 0.99), local temperature with an RMSE of 0.026 °C (R2 = 1.00), and bending with an RMSE of 0.0019 1/cm (R2 = 1.00). We found that machine learning consistently outperformed conventional approaches. Moreover, for simultaneous measurement of bending, temperature, and pressure, small (< 5mm) Fiber Bragg Gratings without additional mechanical or structural augmentation were sufficient, paving the way for innovative biomedical devices incorporating fiber optic sensing technology. </p>