Optimizing and Predicting Performance of Dual-Side Polished SPR Photonic Crystal Fiber using MLR and ANN Models

Abstract

This research presents a surface plasmon resonance (SPR) biosensor that incorporates a dual-side polished photonic crystal fiber (PCF). The biosensor uses an external gold (Au) coating as the plasmonic layer to identify changes in the refractive index (RI) of various analytes. Five critical design parameters, including the diameters of the air holes and the thicknesses of both the analyte and gold layers, were optimized using the Taguchi L₈(2⁵) orthogonal array method. The optimization resulted in outstanding spectral and amplitude sensitivities, achieving 1000 nm/RIU and 98.422 RIU⁻¹, respectively. Additionally, Multiple Linear Regression (MLR) and Multi-Layer Perceptron Artificial Neural Network (MLP-ANN) models were employed to predict the sensor’s confinement loss. The findings demonstrate the efficacy of artificial neural networks in providing quick and accurate predictions for various geometric configurations, showcasing their potential in this advanced application. The designed sensor can detect a wide range of analytes (RI range of 1.28–1.44), making it suitable for applications in organic chemical detection, pharmaceutical analysis, and biosensing.