Abstract This work integrates computational simulations with a hybrid machine learning framework to investigate the nonlinear relationships between plasmonic layer geometry, refractive index variations, and spectral response in a photonic crystal fiber (PCF) surface plasmon resonance (SPR) sensor. The proposed approach achieves reliable detection of small refrative index chances from a simple yet optimized PCF SPR sensing structure, reaching competitive sensitivity levels in the refractive index range of 1. 33–1. 39. Accurate predictions were obtained with R^2> 0. 99 and minimal error (< 0. 1). A central contribution of this work is the simultaneous optimization of multiple optical metrics. Beyond maximizing wavelength sensitivity, the methodology balances sensitivity, figure of merit, Q-factor, and FWHM. This multiobjective strategy enables precise tailoring of the plasmonic layer geometry, producing sharp resonances, high-quality factors, and robust performance. Overall, the results demonstrate how plasmonic engineering in photonic crystal fibers can drive high-performance SPR sensing platforms. The methodology provides valuable insights into the geometry–plasmonics interplay while opening avenues for practical implementations in biochemical detection, environmental monitoring, and chemical sensing.
Romeiro et al. (Tue,) studied this question.