Projecte llegit

Títol: Prediccion de proteína y grasa en leche mediante dispersión de luz y fluorescencia del triptófano

---

Estudiant que ha llegit aquest projecte:

MARÍN ANGLADA, XAVIER (data lectura: 15-09-2023)
   Cerca aquest projecte a Bibliotècnica

---

Tutor/a o Cotutor/a: SEPULCRE SANCHEZ, FRANCISCO LUIS

Departament: DEAB

Títol: Prediccion de proteína y grasa en leche mediante dispersión de luz y fluorescencia del triptófano

Data inici oferta: 06-02-2023      Data finalització oferta: 06-10-2023

---

Estudis d'assignació del projecte:

MU TECH4AGRI+FOOD

Lloc de realització:


En empresa (cal signar un conveni de cooperació)

        Tutor/a Extern: Manuel Castillo
        Institució/Empresa: Universitat Autònoma de Barcelona

Paraules clau:
Predicción, Grasa, Proteína, Leche, Retrodispersión de luz, Fluorescencia de triptófano, Óptica, Fibra, Cremado, Espectro, Espectral, VIS, VIS-NIR, NIR, Modelos, Regresión, PLS.

Descripció del contingut i pla d'activitats:
L'objectiu és la detecció de proteïna en la llet mitjançant espectroscòpia de fluorescència.

Preparació de patrons i calibració de mostres

Estudi de mostres

Anàlisi de resultats

Overview (resum en anglès): Backscattered light at 880 nm and tryptophan fluorescence (excitation and emission at 280 and 350 nm, respectively) were evaluated to predict milk fat and protein content (0.50 to 5.00% and 2.50 at 3.90%, respectively) with FluorLite CoAguLab, a coagulation sensor to predict curd cutting time to maximize cheese yield, which contains an optical probe at the bottom of the measuring cell. Simultaneously, creaming of fat globules was measured with Turbiscan LAB to see how it affected light scattering. Alternatively, VIS-NIR and NIR light backscattered spectra were obtained to investigate at which wavelengths could assist in improving the prediction models. Results were statistically analysed by analysis of variance, linear and non-linear predictive models, and PLS regression. The initial light backscatter signal at 880 nm is linearly correlated with fat content and improve in non-linear models with R^2 of 0.9798 and 0.9816, respectively. Protein content can be predicted with an initial light backscatter signal at 880 nm for different fat levels at 0.50, 2.75, 5.00% and fat-corrected non-linear models with R2 of 0.9894, 0.8548 and 0.9861, respectively. Tryptophan fluorescence is uncorrelated due to an unknown interactions. As time passes, the backscattered light signal and tryptophan fluorescence are diminished by creaming. The wavelengths of 700, 1250 and 1850 nm and the ratios of 750/500 and 1675/1475 show correlation with the fat content. Still, it is necessary to increase the number of observations to avoid overfitting in predictive models.

---