Research On The Optimized Desulfuration Static Model Based On Generalized Genetic Algorithm

With our joining WTO, "Informationalization improves Industrialization" becomes a key strategy for every industrial enterprise. It is a very important problem for steel-iron-making enterprise to remove the sulfur efficiently, improve the quality of steel and design the new products, however, the traditional iron water pre-desulfuration method based on manual control causes the instability in the process, which harms the quality of steel and increases the production cost. As a project of one big national steel-iron corporation, a desulfuration static model is designed in this dissertation. The modeling process is based on Data Mining technology, which can automatically discover the rules and knowledge in desulfuration process, and then provides the decision support, improves the desulfuration effect and decreases the cost. The successful use of this static model will be a solid prerequisite for full automatic dusulfuration in the future.The Radial Based Function (RBF) neural network is adopted as the modeling algorithm. To overcome the difficulty in determining the RBF center numbers and spread, a kind of Generalized Genetic Algorithm is introduced, which follows the analysis of the basic rules of genetic algorithm. The new hybrid algorithm determines the center numbers and spread adaptively to reach the optimal tradeoff between the training accuracy and the generalization, so it increases the prediction accuracy of the model.The dissertation explains the ideas and characters of Generalized Genetic Algorithm. The evolution method that the directional evolution is combined with the directional transfer of local optimal status to reach the global optimization rapidly is explicated from three aspects: syllogism, mathematic deduction and biologic evolution. A kind of improved real-coded Generalized Genetic Algorithm based on "group in group" and multi-population generation method is carefully introduced and successfully applied. The experiment results prove that the proposed algorithm is obviously better than standard genetic algorithm.After analyzing the desulfuration techniques, this dissertation optimizes the desulfuration static model through the effective data preprocessing and Generalized Genetic Algorithm. The performance of the improved model proves better than traditional RBF model by the experiments.Finally, a kind of Locally Weighted Linear Regression model is introduced. After comparing the model performance between active learning and lazy learning, this dissertation briefly analyzes the data distribution in dusulfuration.