Structure-Reduced Dual-Parameter Photonic Crystal Fiber SPR Sensor: Inverse Optimization via Artificial Neural Network and Monte Carlo Hybrid Approach

This work presents an efficient design and optimization framework for dual-parameter photonic crystal fiber (PCF) sensors, integrating structural parameter reduction with artificial intelligence (AI)-based optimization. The adopted four-hole symmetric structure features high simplicity and manufacturability. By conducting correlation analysis to establish coupling relationships among structural parameters, all geometric variables are controlled jointly through the central variable of air-hole spacing, thereby significantly reducing input dimensionality and improving modeling efficiency. Based on this, an artificial neural network (ANN) model is established for predicting optical properties, and a Monte Carlo sampling method is employed for inverse structural optimization to identify structural combinations with optimal sensing characteristics. The resulting structures are applied to the simultaneous measurement of magnetic field and temperature, where a magnetic fluid is utilized for magnetic field response, and an ethanol-chloroform mixture serves as a temperature-sensitive medium. The sensing ranges span 0-70 mT for magnetic field and 10-40 °C for temperature, showing high sensitivities of 8.6 nm/mT and 6.16 nm/°C, respectively. This design framework not only effectively reduces computational complexity and improves the design efficiency and performance of multi-parameter PCF sensors, but also provides a general methodological reference for rapid inverse optimization and intelligent integration of other photonic devices.