Prediction of the rhodinol content in Java citronella oil using NIR spectroscopy in the initial stage developing a spectral smart sensor system - Case report

Abstract

The rhodinol content is an essential component in determining the citronella oil qualities. This study aimed to develop a model calibrated to predict the rhodinol content in citronella oil using near-infrared (NIR) spectroscopy. This research is the initial stage in developing a spectral smart sensor system that predicts the rhodinol content of citronella oil in the distillation and fractionating process. Citronella oil samples were scanned by NIRFlex liquid N-500 with a wavelength of 1 000-2 500 nm having an absorbance value (log 1/T). The accuracy of the prediction was achieved using the partial least square (PLS) model. Based on the NIR spectrum at a peak of around 1 620 nm, the rhodinol content in the citronella oil was estimated. The finest model to predict the rhodinol content was y = 0.9874x + 15.6439 with a standard error of the calibration set (SEC) = 2.78%, a standard error of the prediction set (SEP) = 2.88%, a ratio of the performance to the deviation (RPD) = 9.23, a coefficient of variation (CV) = 16.81%, and the correlation coefficient (r) = 0.99. The NIR and PLS models are possible to use for the initial stage in developing a spectral smart sensor system to determine the rhodinol content of citronella oils.

Keywords

• calibration
• fractional distillation
• partial least square
• process control
• spectra