Hızlı Gıda Analizlerine Yönelik Yakın Kızılötesi Spektroskopisi (NIR) Sistemi Geliştirilmesi

Abstract

In the food production chain, provision of food quality and safety from raw material to final consumer becomes compulsory in terms of providing healthy and reliable products to people. The quality of food is ensured by getting under control the physical, chemical and microbiological parameters of food in every step of the production processes. In particular, the identification of a profit-making adulteration in the food industry is of great importance as it has negative implications for human health. Reference methods used to determine quality parameters are preferred due to their reproducible and accurate results. However, these methods have disadvantages such as the need of specialists and the long analysis steps. Therefore, fast, reliable and alternative food analysis methods are needed. Near infrared spectroscopy (NIR) is a frequently used method in food analysis, allowing rapid and easy analysis without destroying the sample. Within the scope of the thesis, it is aimed to develop 3 different NIR spectrometers which enable fast and reliable analysis of NIR spectroscopy which can be operated with reflection, transmittance and transflectance measurement modes. Another aim of the thesis was to develop three different food analysis in the field of food adulteration with the NIR spectrometers developed together with the software that allows to use the prediction ability of the models developed by the partial least squares (PLS) method. In the first study, pistachio adulteration with pea and spinach was determined by using developed reflectance spectrometer. Pistachio, green pea, spinach and their binary mixtures were successfully classified with the developed principal component analysis (PCA) model. The performance of PLS models developed with different pretreatments was investigated in order to determine the amount of green peas and spinach that added into pistachio. The high determination of coefficient (R2) of calibration and validation (green pea: 0.992-0.995, spinach: 0.998-0.991) and low limit of detection (LOD) and limit of quantification (LOQ) (green peas: 1.05% - 5.27%, spinach: 0.19% - 0.95%) values were determined as they were two best performing models. In the second study, the developed reflectance spectrometry was used to determine whether heat treatment was applied to the milk samples, and afterwards type determination was performed among samples of sheep, goat and cow milks as well as their binary and triple mixtures. Raw and pasteurized milk samples have been successfully classified through the developed partial least squares-discriminant analysis (PLS- DA) model. Milk samples were then assigned as pure sheep, goat, cow or binary and ternary mixture of these milks by different PLS-DA models. In the last study, developed NIR spectrometers were used to determine the amount of erucic acid in canola oil. The effects of different pretreatments on the developed PLS models were examined and the most appropriate PLS models were determined. The R2 values for the calibration and validation were found to be 0.994-0.978 for the reflectance, 0.993-0.990 for the transmittance, and 0.993-0.991 for the transflectance spectrometers. The LOD and LOQ values were 0.15% -0.75%, 0.11% -0.56% and 0.26% -1.30% for the reflectance, transmittance and transflectance spectrometers, respectively. It was shown that the developed NIR systems can predict the amount of erucic acid successfully, while the transmittance and transflectance spectrometers perform better than the reflectance spectrometer.