Liquor production is a traditional industry of china which has a long history and has a unique traditional process. But artificial taste is the main approach to liquor blending in the process. Taste varies from person to person, and is subject to many subjective and objective factors. Also physical and chemical indicators can not accurately controlled by taste, so the quality of liquor is instability, and it is difficult to reduce production costs. On the other hand, artificial taste need blending many attempts, so productivity is low. The valuable knowledge of experts can not form a long-term production which can service for the enterprise. With the improving of analytical tools and scientific and technology, the role of trace fragrance ingredients in liquor has been fully understood which creates an important condition for exploring the mysteries of liquor blending. On the basis of understanding of the relation between trace fragrance ingredients and the style of liquor, the application of modern computer technology to enhance the development of the traditional blending technology, is a undoubtedly driving force to promote the liquor technological level. In this paper, a number of studies are made to explore the intelligent model of liquor blending. A neural network model based on the relation between physic-chemical and sensory is built. First the lowest cost of blending is given through linear programming and then physic-chemical indicators are inputted into the model which predicts the sense indicators. This process is similar to blending through artificial taste. The sense indicators are given as an input into the expert system which produces the blending program. Also the sense indicators can be input into the model which produces the physic-chemical indicators, and then the combination program of lowest-cost is given through the linear programming.
修改评论