Investigation of an on-line detection method combining near infrared spectroscopy with local partial least squares regression for the elution process of sodium aescinate

An on-line detection method combining near infrared (NIR) spectroscopy with local partial least squares regression (PLSR) was investigated for the elution process of sodium aescinate. A 2 mm pathlength flow cell which transmitted NIR radiation through two fiber optic probes was placed in bypass of the macroporous resin column to collect real-time spectra of the sodium aescinate eluate. To compare the predictive accuracy, both local and global PLSR were employed to build mathematical models between NIR spectra and reference values. Meanwhile, Mahalanobis distance was introduced to select the appropriate local model for the prediction of unknown samples. Experimental results demonstrated that local PLSR was superior to global PLSR in both calibration performance and predictive accuracy. Moreover, the on-line detection method was proven to be feasible in real application and thereby would be of great value for monitoring the elution process of sodium aescinate in real time as well as determining the start and end points of eluate collection.

Figure optionsDownload as PowerPoint slideHighlights
► We invested an on-line NIR method for elution process of sodium aescinate.
► Local PLSR models were developed and proved to be superior to the global model.
► Local PLSR models can be applied to on-line detection of elution process.
► Start and end points of elution process can be quickly determined by NIR.
► Model updating is efficient in dealing with deceased accuracy of original models.