.png){kind=icon}
.png){kind=icon}
Return
A PQF Refractive Index and Temperature Sensor Optimized by Parallel Neural Network Combining PatchTST and LSTM
.png){kind=icon}
PRE
AI
X
W
Z
H
P
DOI：10.1007/s11468-026-03424-x.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
A dual-parameter sensor based on lossy mode resonance (LMR) using a D-shaped offset-core photonic quasicrystal fiber (PQF) is proposed in this paper, and a parallel-designed neural network structure combining the Patch Time Series Transformer (PatchTST) and Long Short-Term Memory Network (LSTM) is employed to optimize its structure. This sensor detects refractive index (RI) by coating a layer of titanium dioxide (TiO2) and a layer of indium tin oxide (ITO) on a D-shaped polished surface, and by coating a layer of ITO and filling polydimethylsiloxane (PDMS) on the pores on both sides of the core for temperature detection. Without the optimization of structural parameters by the neural network, the detection ranges of RI and temperature are 1.370 ~ 1.442 and − 25 ℃~40 ℃, respectively, with the maximum RI sensitivity of 17,500 nm/RIU and the maximum temperature sensitivity of 2.8 nm/℃. The applied neural network predicts the sensitivity under different structural parameters with coefficient of determination (R²) scores of 0.99978 and 0.99992, respectively. After optimization of the structural parameters, the detection ranges of RI and temperature are 1.360 ~ 1.442 and − 30 ℃~60 ℃, respectively, with maximum sensitivities of 73,000 nm/RIU and 2.9 nm/℃. This achieves a significant improvement in sensor performance and an effective reduction in computational complexity, providing a more efficient and reliable design method for sensor development.
Keywords:
Photonic quasicrystal fiber (PQF)
Lossy mode resonance (LMR)
PatchTST-LSTM
Dual-parameter
Journal
IF:
4.3
Papers:
4.3K
Citations:
7.5K
Organization
Cited Papers