MLP-optimized arc-PCF surface plasmon resonance sensor for refractive index detection

This study presents a refractive index (RI) sensor utilizing an arc-side-polished photonic crystal fiber integrated with surface plasmon resonance (PCF-SPR) architecture, along with a Multi-layer Perceptron (MLP)-genetic algorithm (GA) optimization framework for performance enhancement. The MLP mode...

Full description

Saved in:

Bibliographic Details
Published in:	Optics communications Vol. 599; p. 132603
Main Authors:	Li, Pengxiang, Wang, Hao, Tong, Zhengrong, Zhang, Weihua, Ma, Jing
Format:	Journal Article
Language:	English
Published:	Elsevier B.V 01.01.2026
Subjects:	Genetic algorithm optimization Multi-layer perception Optical fiber sensors Photonic crystal fiber Surface plasmon resonance Photonic crystal fiber Optical fiber sensors Surface plasmon resonance Multi-layer perception Genetic algorithm optimization
ISSN:	0030-4018
Online Access:	Get full text
Tags:	Add Tag No Tags, Be the first to tag this record!

Description
Summary:	This study presents a refractive index (RI) sensor utilizing an arc-side-polished photonic crystal fiber integrated with surface plasmon resonance (PCF-SPR) architecture, along with a Multi-layer Perceptron (MLP)-genetic algorithm (GA) optimization framework for performance enhancement. The MLP model, trained using finite element simulation datasets, accurately predicts sensitivity based on five key structural parameters, while the GA identifies optimal parameter combinations. The optimized sensor demonstrates 7,387 nm/RIU peak sensitivity within the 1.33-1.37 RI range, achieving merely 0.176% prediction error and 5.53% improvement over conventional approaches. Extended evaluation across 1.30-1.40 RI range reveals outstanding performance with 22,200 nm/RIU peak sensitivity. Both univariate analysis and SHAP interpretation confirm the physical soundness of the optimization strategy. This approach establishes an effective paradigm for multi-parameter optimization of high-performance optical fiber sensors. •A new arc-side-polished PCF-SPR sensor is proposed.•MLP is used to predict sensitivity from structural parameters.•GA optimization improves sensitivity from 7000 to 7387 nm/RIU.•Peak sensitivity of 22,200 nm/RIU in RI = 1.30–1.40.•SHAP analysis reveals the influence of key parameters.
ISSN:	0030-4018
DOI:	10.1016/j.optcom.2025.132603