Details

This study aims to develop a high-precision distributed refractive index sensor based on a single-mode–multimode–single-mode (SMS) fiber interferometer, integrated with machine learning algorithms for signal demodulation. The primary objective is to address the challenges of achieving distributed refractive index measurements with high sensitivity and spatial resolution in complex environments. The research methodology involved the design and optimization of an SMS fiber interferometer, where theoretical calculations and experimental validations were conducted to evaluate mode coupling efficiency and transmission power. Machine learning algorithms, including Gaussian process regression (GPR) and principal component analysis (PCA), were employed to analyze interference spectra and predict refractive index variations. The system utilized Optical Spectral Interferometry  (OSI) for distributed sensing, with data preprocessing techniques such as Fast Fourier Transform (FFT) applied to enhance feature extraction.  The GPR model exhibited superior prediction accuracy, yielding a root mean square error (RMSE) of 0.0271 on validation data. Additionally, the optimized SMS structure showed distinct interference pattern shifts under varying refractive index conditions, validating its sensitivity. These results underscore the effectiveness of combining SMS interferometry with machine learning for precise and distributed refractive index sensing.