This study aims to address the limitations of traditional temperature sensors, such as temperature test limit, low sensitivity and accuracy. A two-dimensional photonic crystal (PC)-based temperature sensor is proposed. It is designed on a hexagonal lattice comprising silicon rods suspended in air.
Rigorous simulations were conducted using OptiFDTD to optimize the structure. Conventional approaches often fail to capture certain variable parameters. Machine learning (ML) offers a robust method for identifying and capturing subtle changes. This manuscript proposes ML-based algorithms for performance monitoring, specifically targeting refractive index (RI) and electric field distribution pattern identification.
The PC ring resonator-based sensor demonstrates excellent stability within the temperature range 0–70 °C. A comparative study was conducted using various ML algorithms. The results indicate that EfficientNet-b0 achieves the highest accuracy among the evaluated algorithms, even at reduced image sizes. Furthermore, the prediction accuracy reaches 97%. These findings suggest that the proposed approach provides an effective tool for intelligent signal evaluation in sensing applications.
In contrast to conventional approaches, the proposed method uses ML algorithms to model the complex, nonlinear relationships among electric field distributions, RI changes and temperature. The primary advantage of this method is its ability to maintain high accuracy under adverse conditions.
