Machine Learning-Assisted Simultaneous Measurement of Salinity and Temperature Using OCHFI Cascaded Sensor Structure

A compact offset-coupled hybrid fiber interferometer (OCHFI) is designed and experimentally demonstrated for simultaneous measurement of salinity and temperature. The sensor integrates multimode fiber (MMF) and offset no-core fiber (NCF) through an intermediate single-mode fiber (SMF), producing distinct interference patterns for multi-parameter sensing. The optimal SMF length was determined through COMSOL simulations (version 6.2) and fixed at 50 cm to achieve stable and well-separated interference dips. Fast Fourier Transform analysis confirmed that the modal behavior originates from the single-mode-multimode-single-mode (SMS) and single-mode-no-core-single-mode (SNS) segments. Experimentally, Dip 1 exhibits salinity sensitivity of 0.62206 nm/‰, while Dip 2 shows temperature sensitivity of 0.09318 nm/°C, both with linearity (R2 > 0.99), excellent repeatability, and stability, with fluctuations within 0.15 nm over 60 min. To remove cross-sensitivity, both the transfer matrix method and an Artificial Neural Network (ANN) model were employed. The ANN approach significantly enhanced prediction accuracy (R2 = 0.9999) with RMSE improvement approximately 539-fold for salinity and 56-fold for temperature, compared with the analytical model. The proposed OCHFI sensor provides a compact, low-cost, and intelligent solution for precise simultaneous salinity and temperature measurement, with strong potential for applications in marine, chemical, and industrial process control.