Joint treatment of imprecision and variability in food engineering: Application to cheese mass loss during ripening

Food processes are complex systems because of the complexity of microbiological and/or physicochemical activities and the combinations of these activities are responsible for the physical, chemical, biological and structural changes in food properties. As a result of time limits, financial constraints and scientific and technological obstacles, available knowledge about food processes is often vague, imprecise and incomplete. These considerations, plus the random nature of knowledge, lead to uncertainty that must be taken into account in the decision-making process. It may occur in practice that some model input variables and parameters can be represented by probability distributions (due to observed variability and sufficient statistics), while others are better represented by possibility distributions (due to imprecision), or by the Dempster–Shafer belief functions (due to partial observed variability and partial ignorance). This paper applies recent methods in order to represent and propagate uncertainties relative to the input variables and parameters of a cheese ripening mass loss model in the presence of imprecise and incomplete knowledge. The joint propagation of variability and imprecision through the model combines interval analysis with Monte-Carlo simulations and provides lower and upper probability bounds (referred to as Belief and Plausibility, respectively) of exceeding a certain value of cheese mass during the ripening process.