A Deep-Learning-Assisted Fiber-Optic Sensor Capable of Simultaneously Measuring Temperature and Vector Magnetic Field

A deep-learning-assisted fiber-optic sensor was proposed for simultaneous measurement of temperature and vector magnetic field. The sensor employs an asymmetric structure to generate distinct spectral responses for varying magnetic field directions. By incorporating deep learning, the sensor effectively extracts spectral features and overcomes the limitations of traditional wavelength demodulation methods, enabling accurate identification of magnetic field directions across the 0°-360° range. Additionally, the sensor utilizes matrix demodulation to achieve simultaneous measurement of temperature and magnetic field strength. Experimental results demonstrated that the sensor achieved a 97.3% probability of predicting magnetic field angle errors below 0.8° on the test set. The sensor exhibited detection sensitivities of -0.249 nm/Gs and 1.031 nm/°C for magnetic field strength and temperature, respectively. The sensor offers the advantages of compact size, high sensitivity, and the ability to accurately identify the direction of the magnetic field, introducing a novel approach to the field of simultaneous measurement of temperature and vector magnetic field.