Prediction of Chlorogenic Acid and Rutin in Tobacco by Multivariate Calibration Model Based on Raman Spectroscopy

In order to meet the needs of on-site batch detection, multivariate calibration models based on Raman spectroscopy were established, and the contents of chlorogenic acid and rutin in tobacco were predicted. Samples of 120 tobaccos (including 90 calibration samples and 30 validation samples) were extracted with 50% (volume fraction) methanol solution, and sample solution obtained was introduced into the Raman spectral liquid pool, and Raman spectra in the range of 800-2 000 cm^-1 were collected at 325 nm excitation wavelength. The original Raman spectra were preprocessed using Savitzky-Golay convolutional smoothing method, and the number of hidden variables was selected by Monte-Carlo interactive testing. Partial least square (PLS) multivariate calibration models were established at the same wavelength interval 1 555.8-1 652.9 cm^-1) to avoid spectral overlap of the chlorogenic acid and rutin spectra around 1 600 cm^-1. As found by results, the predicted mean square root errors (RMSEP) of the chlorogenic acid and rutin models were 0.88 and 0.67, the determination coefficients (R_p²) were 0.948 and 0.970, which was turned that the contents of chlorogenic acid and rutin in tobacco could be accurately and reliably predicted via models built by Raman spectra and PLS.