Calibration updating using Multivariate Curve Resolution Methods

The normal process of building a spectral calibration model typically requires a large set of samples such that all variances are included in future predictions. However, once such a primary calibration model has been built, circumstances can cause the model to become invalid. For example, instrumental drift or uncalibrated spectral features appearing in new secondary samples can occur later in time. For these and other situations, the instrument must be recalibrated to accommodate new conditions. In theory, the remedy would be to include a large number of new calibration samples along with their reference measurements. In practice, though, the inclusion of such samples is often costly and lengthy in terms of laboratory time. Hence, mechanisms are needed to update the current model to include the new chemical, physical, environmental, and/or instrumental effects not spanning the current primary calibration domain. Calibration updating is an adaption process where models are updated from predicting in primary sample and measurement conditions to predict the analyte in new secondary conditions. Many calibration updating methods were used for the analyte quantitation in secondary situations and were compared with each other [1]. Multivariate Curve Resolution (MCR) strategies are powerful tools allowing the description, species identification and system understanding, of totally or partly unknown chemical processes and reactions where species cannot be easily isolated and where unknown intermediate species may be present. These methods are also powerful techniques for quantification of complex mixtures. Recently MCR-ALS was used for the analyte quantitation in first-order data sets and was compared with PLS regression [2, 3]. In this study we have used MCR first-order calibration for evaluating its ability to calibration updating in comparison with other updating methods. In order to evaluate our proposed strategy for calibration updating, the publicly available benchmark near-infrared spectra dataset consists of NIR spectra of 80 corn samples was utilized which were measured by different separate near-infrared instrumentals (M5, Mp5 and Mp6) [4]. The M5 instrument is considered as the primary condition and the Mp5 instrument is considered as the secondary condition and number of validation samples in the secondary condition is predicted with the updated model. The prediction error for the model that has not been updated (primary model) is RMSEV= 16.68 and the prediction error of the validation samples for the updated model is RMSEV= 0.54. The results show the ability of the MCR in calibration updating.