Suivi en temps réel des attributs de qualité de la purée de pomme durant la transformation : mécanismes physicochimiques et modélisation statistique

This thesis aimed to identify cooking markers of apples during their transformation into puree in order to understand and model in real time the physicochemical changes occurring in the product. Quality criteria—viscosity, particle size, volatile organic compounds (VOCs), color, and biochemical parameters—were characterized at four key stages of the cooking process, while VISNIR spectra were continuously recorded every 10 seconds. A full two-factor, two-level factorial design (temperature and grinding speed) was implemented to study the effects of these parameters on the composition and structure of purees made from Golden Delicious apples. Fifteen VOCs belonging to the families of alcohols, aldehydes, esters, and ketones were identified and quantified. Major changes occurred between 45 °C and 95 °C, with a significant loss of compounds at 95 °C. Regarding structure, particle size and viscosity also proved to be relevant markers of product evolution. Two modeling approaches were used: a phenomenological approach, based on Henry’s law and the ideal gas law, was applied to the VOCs and showed thatchemical reactions predominated over volatilization. A predictive modeling approach of quality criteria from VIS-NIR spectral data using the SO-PLS method enabled the prediction of four parameters: viscosity, particle size and acetaldehyde concentration. Then, using all data organized into six blocks, the SO-PLS method successfully predicted viscosity (RPD = 4.00; R² = 0.89). This thesis constitutes a preliminary study paving the way for online and continuous control of food processing. The acquisition of additional data on apples and other plant-based products over several years will contribute to developing new perspectives for real-time quality monitoring of food products.