Hybrid Deep Learning-Assisted Dual-Core D-Shaped PCF-SPR Sensor for High-Performance Refractive Index and Temperature Sensing

Abstract

This paper designs a surface plasmon resonance dual-parameter sensor based on a dual-core D-type photonic crystal fiber, and proposes a hybrid deep learning prediction model. The sensor constructs two independent sensing channels for refractive index and temperature detection. The upper channel uses an Au-TiO\(_2\) composite coating for refractive index detection, and the lower channel uses an Au-Polydimethylsiloxane (PDMS) composite structure for temperature detection. From the structural perspective, the sensor achieves complete physical decoupling of the two parameters and the crosstalk between the channels is less than 1 nm. The sensor has a maximum sensitivity of 19000 nm/RIU within the refractive index range of 1.30 1.41 RIU, and a maximum sensitivity of 27.1 nm/\(^\circ \)C within the temperature range of -19 - 100 \(^\circ \)C. Moreover, the structural parameters such as the radius of the air holes, the hole spacing, and the polishing depth all have excellent tolerance. In response to the low computational efficiency problem of the traditional finite element method, this paper further proposes a hybrid deep learning model named HybridOpticalNet, which integrates a residual multi-layer perceptron and a one-dimensional convolutional neural network. The model adopts a dual-branch parallel architecture to extract global structural features and local spectral features, achieving deep complementary fusion of the two types of information. The prediction coefficient \(R^2\) of the model for the dual-channel loss spectrum is above 0.998, the average absolute error is less than 30 dB/m, and the spectral prediction can be completed within milliseconds, with a significant improvement in efficiency compared to traditional numerical methods. This sensor has high sensitivity, low crosstalk, and good process tolerance. Combined with the deep learning model, it can complete the rapid design and performance optimization of multi-parameter PCF-SPR sensors, and has broad application prospects in biochemical detection and environmental monitoring fields.