The Integration Of Computational Intelligence And Application Research

The technique of computational intelligence has developed quickly, and the new intelligent algorithms have appeared continually. It is obvious that the computational intelligence methods are applied in many fields, and there are many research fruits. Meanwhile, the traditional mode of recipe design depends on vocational experts excessively. Its efficiency is very low, and it takes too much time and cost. The traditional mode has not satisfied the need of corporations.The thesis introduces the technique of computational intelligence to the process of product recipe design. Take the tobacco industry for an example, the paper analyzes firstly the design flow of cigarette product in detail, and take use of corresponding methods in the key point of flow. Then the whole methods system could assist the experts to accomplish the task of product design. The thesis divided the whole process into four important points, including for intelligent sensory evaluation, relation analysis, macro-distribution difference analysis and combination optimization. The data analysis runs through the whole process. Artificial neural network and support vector machine(SVM) have great ability of nonlinear approach, which are used in setting up learning model for forecasting the sensory quality. SVM is applicable to deal with high dimension and small sample problem. M5'model tree take use of subsection linear to approach the nonlinear relation. And the model which is built with this method is in the form of expression, which could be accepted easily by the common users. The paper construct the ladder-sample to extract knowledge from BP neural net in order to analyze the nonlinear relation between the output variant and input ones. And the method of fuzzy diffuse figure open out the tendency of two variants. Besides, the thesis pay attention to take use of classical multi-variant statistical analysis methods as complements of computational intelligence technique, which is used to analyze the relation among the variants form different angle. In the analysis of macro-distribution difference, the paper describes the data distribution with calculating statistic feature, which does not only reflect the difference among the samples form different collectivity, but also establish the foundation for later task of data analysis. Finally, the thesis takes advantage of genetic algorithms to bring out several recipes for experts to chose one as the final recipe.The sample data which is detected in the state of industry manufacture takes on the feature of information scarcity. Therefore, the thesis pays great attention to the task of data pretreatment before the application of computational intelligence methods. Meanwhile, the paper associates the knowledge of manufacture background in order that the result could be more accurate and interpretative.The thesis extends the method system of formula product intelligent assistant design to the beer and other industry. There are much commonness among these industries, so there should be great extension foreground of this methods system.