An expert system for automated flavour matching – Prioritizer

Abstract

Flavour matching can be viewed as trying to reproduce a specific flavour. This is a time consuming task and may lead to flavour mixtures that are too complex or too expensive to be commercialized. In order to facilitate the matching, we have developed a new mathematical model, called Prioritizer. Based on the chemical composition of a mixture of volatiles (a target flavour profile), from a set of different ingredients, the model can suggest an initial selection of ingredients needed to reproduce the flavour. The model also allows the user to prioritize properties of the flavour, given the cost of the different raw materials, the allowed number of ingredients and relative importance of flavour components (represented e.g. by perception threshold values). The limit of detection of the equipment used to measure the target flavour profile puts a limit to what flavours can be matched, but it will be shown that the model offers a good initial estimate for reproducing flavour.