Determination of alcohol content in water alcohol hydrates using near-infrared spectroscopy and segmentation regression model

Autor:	Xiaolei Yu, Zhenhua Zhao, Long Zhang, X. D. Jin
Rok vydání:	2010
Předmět:	Measurement method Materials science business.industry Near-infrared spectroscopy Analytical chemistry Alcohol Regression analysis chemistry.chemical_compound Optics chemistry Alcohol content Segmentation business Spectroscopy Instrumentation
Zdroj:	Transactions of the Institute of Measurement and Control. 33:479-493
ISSN:	1477-0369 0142-3312
Popis:	This paper puts forward a measurement method of alcohol content in the mixture of water and alcohol based on near-infrared spectroscopy and segmentation regression model. The wavelengths, which are 1580, 2081 and 2309 nm, are picked out as three characteristic absorption wavelengths of alcohol, whereas 1206, 1456 and 1936 nm are picked out as three characteristic absorption wavelengths of water. The characteristic absorbance has a linear relationship to the alcohol concentration in the sample according to the principle of Lambert. Multiple linear regression equations 1 and 2 are built respectively based on the characteristic absorbance of alcohol and water, and the alcohol concentration is calculated using these two equations. However, the representative characteristic of overtone and combination absorption will change as the concentration of composing component changes so that the measuring effects of different regression equations, which are built by characteristic absorption values of different component, are different in the measurement of different concentration samples. In order to improve the prediction accuracy, this paper adopts the segmentation regression model equation to calculate the alcohol concentration in the sample, ie, we use equation 2 to calculate the alcohol concentration for the low alcohol content samples and use equation 1 to calculate the alcohol concentration for the samples of alcohol concentration between 50% and 70%; on the other hand, we use the mean calculation of equations 1 and 2 as prediction for the high alcohol concentration samples. The adoption of the regression model equation can reduce the mean prediction error for the samples in different alcohol concentrations without increasing the complexity of the equation. Nevertheless, we need to obtain the approximate value of alcohol concentration first in order to determine the distribution range of alcohol concentration.
Databáze:	OpenAIRE
Externí odkaz:	https://explore.openaire.eu/search/publication?articleId=doi_________::160466ce6060fd12ea1435e590f94fc1 https://doi.org/10.1177/0142331210364780 Zobrazit plný text záznamu