Design and Prediction of a Wide Range Sensor Based On SPR-PCF for Simultaneous Measurement of Temperature and Magnetic Field by Taguchi Approach and Machine Learning

Abstract

A novel surface plasmon resonance–photonic crystal fiber (SPR-PCF) sensor is proposed for simultaneous measurement of temperature and magnetic field. The sensor is designed with two orthogonal polarized sensing channels filled with polydimethylsiloxane (PDMS) and magnetic fluid (MF), respectively, which enables simultaneous detection of temperature and magnetic field. Taguchi optimization method is employed to optimize the key structural parameters of the sensor. The proposed sensor is designed to operate over a wide temperature sensing range from − 20 °C to 80 °C, with a maximum sensitivity of 2.849 nm/°C. A magnetic field range of 0 to 50 mT is also covered, with a sensitivity of 1.42 nm/mT. Furthermore, multiple linear regression (MLR) and multilayer perceptron (MLP) machine learning models are applied to predict sensor performance. The MLP model demonstrates strong predictive performance. The mean squared errors for the resonance wavelength under different temperatures and magnetic fields are 3.7478 and 1.3392, respectively, while the coefficients of determination (R²) are 0.9955 and 0.9982, demonstrating the excellent predictive capability of neural network. This work bridges plasmonic, photonic engineering, and machine learning, offering a rapid, highly sensitive tool for industrial temperature and magnetic field monitoring. The proposed sensor exhibits excellent performance and has a wide range of applications, especially in the field of low-temperature industry.