A HYBRID EXPERT SYSTEM FOR FORMULATED PRODUCT DESIGNS

摘要

The designs of formulated products, e.g. tobaccos and cigarettes, food products, paints and coatings, required an expert knowledge of the relationships between the recipe and processing of the ingredients and the resulting product properties achieved in the final product. This paper proposes an architecture for hybrid expert system for formulated product design that consists of a human knowledge base, a learning mechanism in the form of a series of neuro-fuzzy models, fuzzy set techniques to enhance capability of the user interface, and a cooperative mechanism between different parts. Main objective of the system is, taking the assessment on sensory quality of tobaccos for example, to enable the construction of flexible and extensible intelligent systems to partly replace panel tests of human experts, cut product development time, reduce overall developing expense, improve product designs and raise the level of quality management of enterprises. The assessment on sensory quality of tobaccos is a typical issue which is concerning with human sentience and subjectivity. Fusion of neural computing, fuzzy set theory and conventional programming techniques can provide an attractive opportunity to handle ambiguity in human mind, sensing with the real life uncertainty.