Back-propagation of imprecision in a cheese ripening fuzzy model based on human sensory evaluations

In the food processes, taking human sensory evaluations into account is relevant for the online control of the sensory quality of food products. However, the operator sensory measurements are subjected to higher imperfections than the conventional sensory measurements. Therefore, to determine the tolerated imprecision on the sensory evaluations according to the specified imprecision for the control variable is an important challenge for food processes. Thus, from a Takagi–Sugeno fuzzy model developed on an application dealing with the control of a cheese ripening process, the proposed approach allows to determine the maximum imprecision tolerated on the cheese coat and consistency input sensory measurements in order to satisfy a specified tolerance for the ripening degree inferred with the TS fuzzy model. Results show that for some ripening periods, the maximum imprecision tolerated on the input measurements is inferior to the human imprecision on these measurements, thus the model must be used carefully or measurements have to be repeated in these periods.