Establishing a predictive model for fast online determination of relative density and refractive index of e-cigarette liquids using near-infrared spectroscopy

Abstract

Relative density and refractive index are two fundamental physical properties of e-cigarette liquids to indicate their uniformity and batch stability. These parameters are mainly determined by a density meter and refractometer respectively, which is tedious and the analysis results are not readily available for massive measurements. A rapid determination of the two parameters is important for quality inspection and control of e-cigarette liquids, and a lot efforts have been devoted to establishing a predictive model for these parameters. In this study, a novel near-infrared spectroscopy (NIR) combined with particle swarm optimization-support vector regression (PSO-SVR) algorithm was applied to build a prediction model. The experimental results showed that comparing with the traditional partial least squares regression (PLSR) model and the principal component regression (PCR) model, the PSO-SVR model had superior prediction performance.