| Basic oxygen furnace (BOF) steelmaking is one of the most important steelmaking method in the steel industry. It is also the most important steelmaking method in China. In order to improve the productivity and save energy, the researches of modeling and control control technology in BOF steelmaking process are of great significance. But for the reason that the processes of BOF steelmaking are complex, the mechanisms are still obscure and hard to described. The control of steelmaking processes are mainly depended on experiences, so efficiency is not high and resources are wasted easily. Along with the fast-developing of automation techniques and computer technology, the level of BOF steelmaking automatization has remarkably improved. One of the most important techniques is the case-based reasoning (CBR) method. Its features apply to the modeling problem of BOF steelmaking. But in the practical problem of BOF steelmaking the data is so large and has a lot of uncertainties. Thus, the proposal of an improve CBR method which considering the demand of practical BOF steelmaking is important to improve the productivity of BOF steelmaking in our country.Soft computing refers to those methods which pursue approximate solution effectively. It helps deal with the problems in BOF steelmaking. In this thesis, soft computing technology is introduced in the process of case description, case retrieval and case reuse to increase the accuracy and efficiency of the model. The proposed methods are used to calculate the required oxygen blowing amount in BOF steelmaking process, predict the carbon content and temperature. The main contents of this thesis are as follow:(1) This thesis designed an improved case description method based on mutual information, fuzzy and rough set algorithm. At first, this thesis uses mutual information and rough set technique to realize condition attributes reduction, reducing the number of attributes at the content of valuable information; Fuzzy description is introduced in condition attributes after reduction. Finally, the cases are described in the form of fuzzy rules and fuzzy membership degree values. The effectiveness of the proposed method is validated by simulation on UCI data, and the proposed method is applied to calculate the required oxygen blowing amount in BOF steelmaking, fairly good results have been obtained.(2) Based on fuzzy C-Means algorithm and neural network, this thesis developed an improved case retrieval method. When the case-base is large, it will lead to some uncertain effect so that reducing accuracy and efficiency in case retrieval. In this thesis, fuzzy c-means clustering method is used to divide the case base into several sub space firstly, and each case was assigned with a category label. Finally, new case is classified by the reserve pool based model, and similar cases are selected with case density algorithm within the certain classification. Besides, the gradient descent method and neural net algorithm are used to obtain the weighting coefficients of attributes. Simulation results are used to verify the effectiveness of the method. The simulation results on UCI data and the prediction results of carbon content in BOF steelmaking indicate that our proposed methods are effective.(3) An improved case revise method based on support vector machine (SVM) and case reuse method based on hybrid expert system is given in this thesis. For case revise, it mainly considers the differences between condition attributes and solution attributes. When establishing the relationship between them, it uses SVM to calculate the solution of current problems making use of the nonlinear learning ability of SVM. Further more, it proposes a model for meature the direct relationship between condition attribute and solution based on hybrid expert system. Simulation results of the calculation of the required oxygen blowing amount in BOF steelmaking process show that the proposed methods are effective. |
PDF Full Text Request