Wavelength selection enables robust quantification of oil content with near-infrared spectroscopy in pea protein gels produced under varying heating conditions

Thermal processing influences the near-infrared (NIR) spectra of food products, causing inaccuracy when quantifying the composition of the product. In this study, the effect of thermal processing on NIR measurements of the oil content in pea protein isolate (PPI) gels was investigated. Analysis of variance showed that the heating temperature (30°C–120 °C) during the production of PPI gels influenced large parts of the NIR spectra and time (2.5–15 min) only had a limited effect. Stepwise multiple linear regression was used to select a combination of 5 wavelengths that was suitable to create a robust model for the prediction of oil content (Q 2 = 0.9928 ± 0.002, root mean standard error of prediction = 0.2806 ± 0.0423 wt%). This combination of wavelengths was shown to reduce the effect of the history of thermal processing on the measurement, improving the accuracy of oil content quantification. This approach can be applied for in-line quantification of the oil content of food products subjected to varying heating conditions using specialized sensors. • NIR spectroscopy is suitable to measure oil content of heated food products. • Heating conditions needs to be considered during calibration for oil content measurement. • Temperature and oil content have an interrelated effect on parts of NIR spectra. • Wavelength selection with SW-MLR results in robust calibration. • Wavelength selection allows the development of sensors for inline analysis.