Analysis methods used for quality and authenticity control of any raw materials and their derivatives in industry and research laboratories have been mostly conventional so far. They are characterized by steps such as sampling, sample preparation for extraction of target compounds and quantitative determination of analyte content using chemical reagents. Most methods are standardized and are used mainly for monitoring and optimizing the production process. However, the industry has an increasing need for fast and non-destructive methods. The most common is spectrometry, which is performed in the near and mid-infrared range and is mainly used for the analysis of biochemical parameters such as the content of total fats, polyphenols, methylxanthines, etc.
The aim of the master's thesis was to develop new methods on a NIR (near infrared) spectrometer for rapid evaluation of the quality of cocoa mass and cocoa butter. 40 samples of cocoa mass and 40 samples of cocoa butter were analysed, of which 30 samples were used to build calibration models and 10 samples were used for validation of newly established NIR calibration lines. The total fat and theobromine content of cocoa mass, and the total free fatty acid content, expressed as oleic acid, and the iodine value of cocoa butter were determined using classical methods. The total fat content was determined by extraction using Soxhlet distillation. Theobromine was determined by extraction of the analyte into water and the extract was analysed by HPLC (high performance liquid chromatography) with UV detection. The content of total free fatty acids, expressed as oleic acid, was determined by neutralization of acids by titration with potassium hydroxide, and the iodine value was determined using the Wijs method. To verify the correctly determined values, the results of the same parameters of the same samples were compared to the results obtained in an external accredited laboratory.
For method development, NIR spectra were recorded and calibration lines were established based on the results obtained with classical methods, thus indirect methods were developed. The calibration lines were fitted using PerCal® Plus, where different models were fitted by testing PLS and HR regressions and pretreatments harmonization and the Savitzky-Golay filter in combination with detrendization and SNV. The models were compared based on model value, correlation, R$^2$, RMSE, SECV, Q$^2$ and R$^2$ of validation.
|