Intelligent Network System Of Wine Blending Based On Neural Network

The wine blending of the finished product wine in traditional wine enterprise production management adopts the combination of sense organ analysis with experience scope. It is easily influenced by subjective and objective environment factors with human taste. It causes bigger fluctuation and difference in different batch of wine product. This situation leads to unsteady product quality and lower productive efficiency, and can't optimize product cost. The precious experience coming from domain expert's can't transform into knowledge to serve the enterprise for long time.In recent years, neural network found it's extensive application in various engineering fields, because of its ability of large-scale parallel processing, fault tolerance, self-organization and self-adaptability and strong association ability. But it had not been seen to solve the wine blending problem using computer. The craft technique of the wine blending in fact is a combinatorial optimization problem, since 1990's, with the application of high-precision analysis instrument and electronic technology, it became possible to solve the combinatorial optimization problem in wine blending procedure using computer. As a result, the classic programming and target programming can't still satisfy the requirement of tasting quantization. Therefore, how to construct knowledge-based expert system, and construct neural network machine learning model, how to relate the optimized solution of practical problem to the stable states of neural network, how to map the optimization procedure of practical problem to evolution procedure of neural network system, and improve the work of computer-aided wine blending are the intention of this prosect.According to the above analysis and problem, this thesis finished the following 5 main works:1. Collect data inside wine at chemistry workstation using gas chromatography analysis technology, construct mathematical model using analysis means, draw the product quality characteristics curve, and find out the mirco-cosmic appearance of wine.2. Introducing the liner programming, target programming foundation theories principle, and according to actual craft technique of the wine blending, set up the combinatorialoptimization mathematical model of original wine.3. Making use of the production rule model to construct the wine taste knowledge-based expert system, at the same time introducing neural network Backpropagation algorithm to construct neural network case learning model, create tasting quantization system with machine learning.4. Integrate various systems model, and introduce multi-threading parallel computing method.5. Set up the computer management network system of the wine blending.