Elaboration d'un système à base de connaissances hétérogènes pour la panification française | Development of a knowledge-based system based on heterogeneous knowledge for the French breadmaking process

In Food science, researchers have to deal with complex systems and with the fragmented state of knowledge. Formalisms of artificial intelligence (AI), by their ability to represent knowledge, allow the incorporation of know-how in models so to exploit the various types of knowledge. Our objective is to assemble the available knowledge on French breadmaking, the purpose is to simulate the prediction of states of dough and bread, like domain human experts can do. The object of study is the process of French breadmaking and the available knowledge is know-how and scientific knowledge. The study is focused on the first two operations of the breadmaking process. The conceptual framework of this work is that of qualitative reasoning and the research question is : "How to assemble knowledge about, on one hand, an operation of transformation based mostly on the domaine expertise and with a wide range of solutions, and, on the other hand, about an operation based primarly on scientific knowledge; and finally how to integrate them in the same process ?" We have shown: i/ how to complete the representation of the reference breadmaking process so that it may represent the processes used by bakers, ii/ how to limit the calculus imprecisions with the qualitative algebra and how using this formalism to prospect experts' reasoning, iii/ how to assemble the knowledge to develop phenomenological models so to simulate the dynamics of the fermentation process. Finally we have introduced theoretical tools for qualitative functions analysing.