Preservation of juices is essential to obtain microbial safe products. There are various established methods as pasteurization. Heretofore, only the kinetic figures of microbial inactivation were considered but not those of reaction impairing the chemical quality. For a gentler processing, knowledge of the kinetics of relevant chemical conversion reactions is necessary. 5-(Hydroxymethyl)-furfural (HMF) formation and the color change of juices are important attributes. The non-isothermal Rhim method was used to determine the activation energy and pre-exponential factor for HMF formation in different juices and an isothermal method for the reaction order. Values for the activation energy from 133 to 303 kJ/mol were obtained with a zeroth reaction order. A correlation between HMF and the color change could be found. Based on the kinetic figures, lines with equal effects for the chemical changes and for the lethal effect on microorganisms were calculated. Time-temperature settings for the gentlest treatment could be found.

Thermal preservation is widely used for the stabilization of fruit and vegetable juices. However, this method can potentially impair product quality and it is necessary to categorize and quantify these changes in order to optimize the treatment process. Here, carrot juice was treated with different time temperature combinations selected to achieve equivalent microbiological lethality. Different temperatures were selected - low temperature long time (LTLT) with 90 and 95 °C and high temperature short time (HTST) with 120 and 125 °C. The thermally treated juice exhibited significant differences in physical, chemical and sensory parameters in contrast to untreated juice. HTST treatment had less of a negative effect on the juice. A significant deterioration in odor, appearance and taste was observed after the LTLT treatment compared to an untreated reference juice. The juice exhibited elevated acid taste, a shift in olfactory profile from white to red vegetables and diminished homogeneity. Significant changes in sugar composition, pH, carotenoids and color were also observed during this treatment. The juice that had been treated with LTLT exhibited a higher amount of monosaccharides, a lower pH value, a reduced quantity of carotenoids and a perceptible difference in color in comparison to the untreated and fresh juice. Therefore, it can be concluded that varying temperatures show different effects on juice quality, despite the same microbiological lethal effect. This must be taken into account when designing the pasteurization process.

With the implementation of EU Directive 2019/904, which mandates a minimum of 25% recycled polyethylene terephthalate (rPET) in single-use beverage bottles by 2025, there is an urgent need for reliable analytical methods to verify regulatory compliance. This study evaluates a data-driven approach to quantify and classify recycled PET based on near infrared (NIR) and mid infrared (MIR) spectral data.Investigations were conducted on two sample systems: industrial preforms containing various recyclate types and electrospun nanofiber tissues with controlled rPET concentrations (0–100%). Data analysis utilized Random Forest models for classification and Lasso regression for quantitative prediction.The results demonstrate that NIRS data allows for robust classification of recyclate types in preforms, achieving precision and recall values exceeding 0.8. For quantitative determination in the fiber samples, the multimodal data fusion of NIRS and MIRS spectra yielded the highest accuracy, with a coefficient of determination of R2 = 0.93 and a Root Mean Square Error of Prediction (RMSEP) of 8.24%. However, leave-one-out validation indicated a decrease in performance for unknown material compositions, where the average prediction error rose to 21.97%. Overall, these findings highlight the potential of spectroscopic techniques for monitoring recycled content within a data-driven framework. To ensure practical application in heterogeneous material streams, future research must focus on improving model generalization through larger, more diverse calibration datasets and advanced transfer learning to bridge the gap between controlled laboratory settings and variable industrial waste.