Comparison of different measurement techniques and variable selection methods for FT-MIR in wine analysis

Area of Science:

• Analytical Chemistry
• Food Science
• Spectroscopy

Background:

• Fourier-transform infrared (FTIR) spectroscopy coupled with partial least squares (PLS) regression has been a standard for over a decade in wine analysis.
• Accurate and simultaneous estimation of multiple wine parameters is crucial for quality control and research.

Purpose of the Study:

• To compare the performance of different FTIR instruments and variable selection techniques for simultaneous wine parameter estimation.
• To identify the optimal combination of FTIR hardware and chemometric methods for robust wine analysis.

Main Methods:

• Utilized a diverse set of international wine samples.
• Evaluated three FTIR instrument types: single bounce attenuated total reflection (ATR), transmission with variable pathlength, and transmission with defined pathlength.
• Compared four variable selection techniques: full spectrum PLS, genetic algorithm PLS (GA-PLS), interval PLS (iPLS), and principal variable PLS (PV-PLS).

Main Results:

• Single bounce ATR was suitable for ethanol, density, and sugars but less accurate for organic acids.
• Variable pathlength transmission showed good results for organic acids but was less accurate for ethanol and density.
• Defined pathlength transmission demonstrated suitability for all ten investigated wine parameters.
• Variable selection enhanced model robustness, with GA-PLS being the most effective technique.

Conclusions:

• The defined pathlength transmission FTIR instrument is highly effective for comprehensive simultaneous wine parameter analysis.
• Genetic algorithm PLS is the superior variable selection method for improving model performance in wine analysis.
• Instrument choice and variable selection strategy significantly impact the accuracy and robustness of FTIR-based wine analysis.