Modelling near infrared instrument differences by chemometric methods: testing for near infrared forage analysis

摘要

The possibilities of using the two-block Partial Least Squares (PLS2) method to allow the standardization between the spectra of forage measured in different types of Near Infrared (NIR) instruments, tilting filters and scanning monochromators, are presented. Procedures are developed to transfer spectra from one instrument to another in a network of instruments of different types. PLS1 plus repeatability control is used to build a calibration equation of the network which minimizes the influence of the spectral residuals (remaining after standardization) on the predicted concentrations among the standardized instruments. The validation of the standardization matrix is obtained from representative samples (feed and raw materials) used as standards - yielding a general external validation - or from real samples of Catalonian origin - yielding a specific external validation. With the proposed procedures, the repeatability of concentration measurements among instruments is improved and the number of reference analysis needed is decreased. The possibility is afforded for instruments of lower resolution (e.g., tilting-filter instruments) to participate in the same network as higher resolution instruments (e.g. monochromators) taking advantage of the calibration spectra measured in the latter. (C) 1997 Elsevier Science B.V.