Application-oriented standard-free methods for calibration transfer
|Titel||Application-oriented standard-free methods for calibration transfer|
Within the past few decades, the combination of spectroscopic measurement techniques and multivariate calibration methods has become increasingly prominent for the extraction of (bio-) chemical information in various application fields. While the obtained results via such an approach are very satisfying in general, the process of data collection, model calibration and model optimization is a rather time-consuming and cost-intensive one. As such, one certainly aims to prevent the need for a repetition of these steps. In certain cases, though, changes in the environmental conditions, the measurement setup or the measured substance itself may occur and render the calibrated model invalid. In such a situation, either a new model needs to be developed or mathematical operations, referred to as (calibration) transfer methods, can be performed to transfer knowledge from the original to the new setting. Within this contribution we introduce and discuss a row of application-oriented transfer approaches that are both easy to comprehend and implement. The proposed methods do not assume the availability of transfer standards (i.e. a set of samples measured under the old and new condition) and require the measurement of only a few reference values in the new setting. We evaluate the introduced transfer methods on data from melamine formaldehyde (MF) resin production and show that these techniques can achieve considerable improvement of predictions for data from three forms of changes in the given application context.