Multivariate detection limits of on-line NIR model for extraction process of chlorogenic acid from Lonicera japonica

A methodology is proposed to estimate the multivariate detection limits (MDL) of on-line near-infrared (NIR) model in Chinese Herbal Medicines (CHM) system. In this paper, Lonicera japonica was used as an example, and its extraction process was monitored by on-line NIR spectroscopy. Spectra of on-line NIR could be collected by two fiber optic probes designed to transmit NIR radiation by a 2 mm-flange. High performance liquid chromatography (HPLC) was used as a reference method to determine the content of chlorogenic acid in the extract solution. Multivariate calibration models were carried out including partial least squares regression (PLS) and interval partial least-squares (iPLS). The result showed improvement of model performance: compared with PLS model, the root mean square errors of prediction (RMSEP) of iPLS model decreased from 0.111 mg to 0.068 mg, and the R-2 parameter increased from 0.9434 to 0.9801. Furthermore, MDL values were determined by a multivariate method using the type of errors and concentration ranges. The MDL of iPLS model was about 14 ppm, which confirmed that on-line NIR spectroscopy had the ability to detect trace amounts of chlorogenic acid in L. japonica. As a result, the application of on-line NIR spectroscopy for monitoring extraction process in CHM could be very encouraging and reliable. (c) 2013 Published by Elsevier B.V.