Simulated Spectral Strategy to Enhance Numerical Tobacco Blending Based on Near-Infrared (NIR) Diffuse Reflectance Spectroscopy and Calibration Transfer

Abstract

Near-infrared diffuse reflectance spectroscopy, with the characteristics of simplicity, speed, and nondestructive analysis, has been frequently used in tobacco blending design and maintenance. Its main function is to identify the appropriate combination from hundreds of types of tobacco leaves to match the target formula. Due to the large number of combinations of tobacco leaves, preparing so many blended samples of tobacco powder with different proportions and measuring their spectra are time-consuming, laborious, and costly. Therefore, a novel strategy of simulated spectra based upon calibration transfer is proposed to substitute for the measured spectra without preparing these tobacco powder blended samples. Five and nine single-grade tobacco powder blending experimental results showed that the substitutive spectra for a massive tobacco powder blended samples may be obtained in a short time according to the established calibration transfer models between measured and simulated spectra of a small number of blended samples. Moreover, with the help of calibration models built on the same spectrometer, the desired properties including reducing sugar, total sugar, potassium, starch, and total nitrogen of these tobacco mixtures are quickly predicted. In addition, transfer based on canonical correlation analysis and independent component analysis were superior to other approaches, respectively. The strategy of simulated spectra partially may replace powder blending and spectral analysis and may be considered to be a valuable and effective tool for computer-aided numerical tobacco blending.

Keywords:

• Calibration transfer
• diffuse reflectance mode
• near-infrared (NIR) spectroscopy
• numerical tobacco blending
• simulated spectral strategy

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This work was supported by the Science and Technology Project of China Tobacco Hunan Industrial Company (Grant No. KY2020JC0034).