Machine learning-integrated multi-wavelength SPR sensor based on aluminum-doped zinc oxide and 2D amorphous silicon for high-precision salinity measurement

This study proposes a highly sensitive SPR sensor for the rapid and accurate determination of water salinity in the range of 1%-30%. The sensor is based on an aluminum-doped zinc oxide and 2D amorphous silicon material, an MgF 2 prism, and a five-wavelength light source. Four machine learning models, namely RBFNN, ANN, ANFIS, and LSTM, are used for the accurate determination of salinity in water. The proposed SPR sensor has high sensitivity, 472.1 (deg./RIU), 0.71 (deg./%cons.), a high quality factor 116.17 (1/RIU), and a high figure of merit 115.75, low reflectance (<0.05), and high detection accuracy 0.52 (1/deg.), with minimal error (0.17%) standard deviation for the RBFNN model. The proposed SPR sensor is more sensitive, efficient, and accurate compared to recent studies.