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

For more than a decade, Fourier-transform infrared (FTIR) spectroscopy combined with partial least squares (PLS) regression has been used as a fast and reliable method for simultaneous estimation of multiple parameters in wine. In this study, different FTIR instruments (single bounce attenuated total reflection, transmission with variable and defined pathlength) and different variable selection techniques (full spectrum PLS, genetic algorithm PLS, interval PLS, principal variable PLS) were compared on an identical sample set of international wines and ten wine parameters. Results suggest that the single bounce attenuated total reflection technique is well suited for the analysis of ethanol, relative density and sugars, but less accurate in the analysis of organic acid content. The transmission instrument with variable pathlength shows good validation results for the analysis of organic acids, but less accurate results for the analysis of ethanol and relative density as compared to the other instruments. The transmission instrument with defined pathlength was well suited for the analysis for all parameters investigated in this study. Variable selection improved model robustness and calibration results, with genetic algorithm PLS being the most effective technique.