Feasibility study for transforming spectral and instrumental artifacts for multivariate calibration maintenance

Area of Science:

• Analytical Chemistry
• Spectroscopy

Background:

• Multivariate calibration models are often instrument-specific.
• Transferring models to new instruments (calibration maintenance) is crucial for practical applications.

Purpose of the Study:

• To develop a novel, efficient calibration maintenance approach.
• To reduce prediction errors when using models on secondary instruments without extensive recalibration.

Main Methods:

• A new hybrid method utilizing spectral differences between instruments.
• Piecewise or full spectrum fitting of spectral differences with modeling terms.
• Incorporation of a sample weighting scheme.

Main Results:

• Demonstrated effectiveness on a near-infrared (NIR) benchmark dataset.
• Reduced prediction errors for samples on secondary instruments compared to no transfer.
• Achieved prediction error comparable to full calibration on the secondary instrument.

Conclusions:

• The new method offers a practical alternative for calibration maintenance.
• It avoids the need for standardization sample sets measured under primary conditions.
• Potential for extension to address time-varying measurement conditions is discussed.