PCF-SPR Sensor for Carbonated Drinks Detection Using Random Forest

This paper explores the design and performance of a photonic crystal fiber-based surface plasmon resonance (PCF-SPR) sensor for detecting different types of carbonated drinks. The sensor geometry was optimized and analyzed using the Finite Element Method (FEM), with Random Forest employed to predict and enhance its sensitivity. Input parameters such as core diameter, cladding structure, and gold layer thickness were varied, generating a dataset for training the RF model. Experimental simulations revealed that the sensor achieved a maximum sensitivity of 22,000 nm/RIU, with a minimum wavelength resolution of 4.55 × 10⁻⁵RIU. Accurately detecting carbonation levels in drinks ensures quality, consistency, and customer satisfaction in the food and beverage industry. Moreover, monitoring alcohol content in beverages is essential for regulatory compliance and product labeling, as well as for meeting consumer expectations. The ability to analyze alcohol levels alongside carbonation adds a new dimension to beverage quality assessment, offering a comprehensive solution for manufacturers. Random Forest demonstrated superior predictive capabilities, reducing computational overhead by 93% compared to traditional optimization techniques. This work demonstrates a novel approach for integrating machine learning into PCF-SPR sensor design, paving the way for applications in beverage quality control and aligning with Industry 5.0 and supporting sustainable technology.