Calibration Maintenance and Transfer Using Tikhonov Regularization Approaches

GABRIEL SIANO; JH. KALIVAS; E ANDRIES; GOICOECHEA, HÉCTOR C

APPLIED SPECTROSCOPY

Whiley

Lugar: N York; Año: 2009 vol. 63 p. 800 - 809

0003-7028

Maintaining multivariate calibrations is essential and involves keepingmodels developed on an instrument applicable to predicting new samplesover time. Sometimes a primary instrument model is needed to predictsamples measured on secondary instruments. This situation is referred toas calibration transfer. This paper reports on using a Tikhonovregularization (TR) based method in both cases. A distinction of the TRdesign for calibration maintenance and transfer is a defined weightingscheme for a small set of new (transfer or standardization) samplesaugmented to the full set of calibration samples. Because straightapplication of basic TR theory is not always possible with calibrationmaintenance and transfer, this paper develops a generic solution to alwaysenable application of TR. Harmonious (bias/variance tradeoff) andparsimonious (effective rank) considerations for TR are compared withthe same TR format applied to partial least squares (PLS), showing thatboth approaches are viable solutions to the calibration maintenance andtransfer problems.