Unsupervised dynamic orthogonal projection. An efficient approach to calibration transfer without standard samples.

Area of Science:

• Analytical Chemistry
• Chemometrics
• Spectroscopy

Background:

• Traditional calibration transfer relies on standard samples between instruments.
• Emerging applications extend calibration transfer to diverse domains like product form and seasonality.
• Current methods often require reference analyses, limiting efficiency.

Purpose of the Study:

• To develop an unsupervised calibration transfer method.
• To enable transfer without standard samples or new reference analyses.
• To interpret interferences affecting calibration model performance across domains.

Main Methods:

• Proposed an unsupervised calibration transfer methodology.
• Utilized orthogonalization to address structural differences between domains.
• Applied the method to simulated and real-world datasets.

Main Results:

• Successfully transferred calibration models across different domains.
• Demonstrated the method's effectiveness on simulated and real datasets.
• Identified interferences causing calibration model degradation in new domains.

Conclusions:

• The proposed unsupervised method achieves efficient calibration transfer.
• Orthogonalization effectively handles structural domain differences.
• The approach facilitates understanding of cross-domain calibration challenges.