Machine learning approach to surface plasmon resonance bio-chemical sensor based on nanocarbon allotropes for formalin detection in water

This article investigates the design of a surface plasmon resonance (SPR) sensor that utilizes carbon nanotubes (CNT) and graphene to detect formalin concentration in water. The proposed sensor's design optimization and performance evaluation are achieved by implementing Gradient Boosting Regression (GBR), a machine learning (ML) algorithm, and the artificial hummingbird algorithm. An iterative transfer matrix technique is employed to create training and test sets for machine learning analysis, and a dataset of 8505 x 8 is obtained. The optimized thickness of Ag, CNT, and graphene 51.71 nm, 0.489 nm, and 4.32 nm were obtained using the artificial hummingbird algorithm. The results demonstrate that the SPR sensor achieves excellent reflectance curves, leading to a significant increase in detection sensitivity of 340.44 deg./RIU. Other characteristic parameters such as detection accuracy (DA), full width at half maximum (FWHM), and figure of merit (FoM) have also been evaluated.

Automated summary

This article investigates the design of a surface plasmon resonance (SPR) sensor that utilizes carbon nanotubes and graphene to detect formalin concentration in water. By implementing Gradient Boosting Regression and the artificial hummingbird algorithm, the sensor's design optimization and performance evaluation are achieved. The results demonstrate that the SPR sensor achieves excellent reflectance curves, leading to a significant increase in detection sensitivity.