Browsing by Subject "chemometrics"

Now showing 1 - 2 of 2
  • Thumbnail Image
    Item
    Development of an FT-NIR Method to Predict Process Cheese Functionality
    (2021-02) Chou, Lisa
    Process cheese (PC) production involves using natural cheese, other dairy and non-dairy ingredients, heating, mixing, and cooling to form a final product. The properties of natural cheese, in particular, can be difficult to measure or control, leading to process cheese products with undesirable functional properties that may not be apparent until after cooling. Fourier-Transform Near Infrared (FT-NIR) spectroscopy methods exist for measuring fat and moisture in process cheese (Kapoor & Metzger, 2008) and could be a promising tool for predicting PC properties later in shelf life. In our study, calibrations were developed to correlate FT-NIR spectra of rapidly-cooled PC samples made at benchtop and pilot level production to several functional properties later in shelf life. Properties included heated sauce viscosity, oven melt area, firmness, and fat droplet size. Benchtop and pilot level PC samples were made with cheddar cheese of two different ages, cooked at two mixing speeds, and held at two hold times once final temperature was reached. Sample spectra were collected using a BUCHI NIRFlex N-500 FT-NIR spectrometer (BUCHI Labortechnik AG, CH). NIRCal 5.2 Chemometric Software (BUCHI Labortechnik) was used to correlate the spectra to functional properties using a cross-validation model and partial least squares regression. Across both the benchtop level and pilot level production of PC, 5 and 14 FT-NIR calibration models were created, respectively. Calibration models were developed using the chemometric software calibration wizard and optimized to achieve highest correlation coefficient and minimize standard error by adjusting pretreatments, spectral regions, and quantity of principal components. Many calibration models of different properties (melt area, viscosity, and firmness at different time points) achieved correlation coefficients above 0.6. These findings show that FT-NIR spectroscopy analysis of rapidly-cooled PC samples show potential to be used for predicting fat droplet size, PC sauce viscosity, and oven melt diameter of samples later in shelf life. Further calibrations of molten PC to final properties would show the feasibility of using a rapid, in-line FT-NIR method for process and quality control purposes.
  • Thumbnail Image
    Item
    Development Of Fourier Transform Near Infrared Spectroscopy Methods For The Rapid Quantification Of Starch And Cellulose In Mozzarella And Other Italian-Type Cheeses
    (2019-09) Vazquez, Leilany
    Flow-aids consisting of starch and cellulose are added to prevent caking and sticking in grated and shredded cheese and are used as carriers for added antimycotics to prevent mold growth. The accurate quantification of these flow-aids involves difficult wet-chemistry methods. When too little antimycotic is added, quality issues can occur. Conversely, when too much flow-aid is added in order to dilute the cheese for economic gain, the reputation of the dairy industry is damaged. As a way to prevent the over or underuse of these ingredients, Fourier Transform Near-Infrared Spectroscopy (NIR) could be useful. The goal of this research was to investigate whether calibrations could be made to quantify starch and cellulose in Asiago, Parmesan, and Romano cheeses, alone and separately. Samples of Asiago, Parmesan, and Romano loafs were shredded, and 0 – 5.66% of a starch/cellulose flow-aid was added. Treatments were ground, weighed, formed into a ball and pressed in the middle of the glass petri dish to create a homogenous scanning surface. Samples were scanned using a BUCHI NIRFlex N-500 FT-NIR spectrometer (BUCHI Labortechnik AG, CH). NIRCal 5.2 Chemometric Software (BUCHI Labortechnik) was used to analyze the spectra after first dividing the spectra of the 2,367 samples into 1,578 calibration and 789 validation samples. The spectra were treated with standard normal variate to minimize variations and optimize the calibration. The calibrations obtained had an r2 greater than 0.94 and worked better when they were specific to one specific cheese (ie. Asiago) instead of one kind of cheese (ie. Hard-grated cheese). Future research will determine if cellulose and starch can be identified and quantified separately in the same sample, and the effect of different starch and cellulose types on quantification.