PLS Subspace-Based Calibration Transfer for Near-Infrared Spectroscopy Quantitative Analysis

Area of Science:

• Chemometrics
• Spectroscopy
• Analytical Chemistry

Background:

• Calibration models are often instrument-specific, limiting their transferability.
• Differences in spectral measurements between instruments necessitate calibration transfer methods.
• Existing methods may require extensive data or lack robustness.

Purpose of the Study:

• To develop a novel feature transfer model for robust calibration transfer across multiple instruments.
• To address spectral differences between master and slave instruments efficiently.
• To enable accurate calibration transfer using a minimal set of standard samples.

Main Methods:

• Proposed a Partial Least Squares (PLS) subspace calibration transfer (PLSCT) model.
• Constructed a PLS subspace using feature vectors from the master instrument.
• Projected master and slave spectra into the PLS subspace for feature extraction and transfer using ordinary least squares.

Main Results:

• The PLSCT method demonstrated superior stability and prediction accuracy compared to SBC, PDS, CCACT, GLSW, and MSC.
• Statistical validation using the Wilcoxon signed rank test confirmed the significance of PLSCT's performance.
• Effective calibration transfer was achieved with a reduced number of standard samples.

Conclusions:

• The PLS-based subspace transfer (PLSCT) offers an efficient and accurate solution for calibration transfer.
• PLSCT provides a robust method for harmonizing spectral data across different instruments.
• This approach facilitates the wider application of chemometric models in diverse analytical settings.