Establishment And Application Of Mathematical Model Of Baosteel 's Iron Water Desulfurization Based On Data Mining Technology

In order to reduce the content of sulfur which is a harmful element in steel, hot metal desulfurization must be implemented before steel-making.Before modeling, a review of the algebraic model and ANN model was placed.The advantages and disadvantages of the existing model was also analyzed according to the using experience in Baosteel. A mathematic model which could predict the weight of desulfurizer was established base on data mining, according to the process of desulfurization using in Baosteel NO.2steel-making.During the phase of data pre-treatment, a certain amount of historical result was extracted from the database in online production system along with the rule of filter. As a common method reducing noise in dataset, ETL was also used before extracting. Besides, the technique of outlier detection made a great contribution to find the abnormal objects in sample set after extracting. In order to reduce the dimension of vector, the sample set was divided into small one. For the sake of making sure that the sub-set was mutual independence, the application of hypothesis test was essential.The main idea of model came from the knowledge of classifier in data mining. The k-nearest neighbor classifier would be the primary tool during finding the similar items with the current one in historical result. The amount of desulfurizer could be calculated using previous data. The internal structure of sample set may have an impact on the result of classification. Two algorithms which took internal structure into account were brought out as a complement of the traditional one. Furthermore, in order to find the optimal value of weight in calculating distance between objects, a method based on parameters experiment was presented.Finally, the thesis discussed the means of adding a model management sub-system in existing system in detail including system analysis and system design. The online function evaluation was also carried out. The results show that the hit rate of the hot metal desulfurization mathematic model based on data mining is87%and the model could be used in production field.