Temperature and refractive index sensing by Brillouin scattering in liquid-core PCF assisted by CNN for...

In this paper, we propose a liquid-core photonic crystal fiber (LC-PCF) based on the Brillouin scattering for temperature and refractive index sensing, and introduce a convolutional neural network (CNN) to achieve high-precision prediction of temperature and refractive index. The fiber's optical and acoustic energy is highly concentrated in the liquid-core region, significantly enhancing the interaction between the light and the solution. As a result, the fiber demonstrates excellent performance in both temperature and refractive index sensing. Simulation results show that the temperature sensitivity of the Brillouin frequency shift (BFS) can reach up to 6.78 MHz/°C, and the refractive index sensitivity can reach as high as 15.00 GHz/RIU, which provides a considerable advantage over existing research. Since the sensitivity varies in real-time during the sensing process, this limits the use of the conventional equation solving method (CESM) for extracting sensing parameters. To overcome this challenge, we construct an optimized CNN for predicting temperature and refractive index corresponding to the Brillouin gain spectrum (BGS) under different signal-to-noise ratios (SNRs). During the testing phase of CNN, when the SNR is 25 dB, the temperature prediction has a root mean square error (RMSE) of 0.144°C and a coefficient of determination (R²) of 99.99%, while the refractive index prediction has an RMSE of 1.14 × 10^-4 RIU and an R² of 99.98%. Even under high noise conditions with SNR reduced to 10 dB, the CNN model still achieves excellent results, with the temperature prediction RMSE of 0.712°C and R² of 99.69%, and the refractive index prediction RMSE of 4.50 × 10^-4 RIU and R² of 99.69%. These simulated results highlight the potential of the proposed method in high-precision sensing applications, making it a promising candidate for advanced monitoring systems in fields such as environmental sensing.