Isa Copper Smelting Ingredients Optimization And State Control Parameter Prediction Method Study

ISA furnace is a compact, fast reaction speed, high efficiency, strong adaptability, secondary energy use good, and comply with environmental requirements of the advanced smelting furnace in the world. ISA copper smelting process is very complicated, multi-phase, high temperature physical and chemical change process, with multi-variable, nonlinear, strong coupling, uncertain characteristics. The blending process is the preceding activities of ISA smelting. The optimal proportion of materials is a key of successful smelting, stable conditions of furnace and product quality. Those status parameters of melting bath and key equipment, such as matte grade, matte temperature, a ratio of Fe and SiO2of slag, fault condition of lance and brusque, are always coupled with blending process and difficult to detect and predict. In order to solve these problems, to achieve energy saving of the ISA copper smelting process, improve the utilization of resources and smelting equipment and maximize the production potential of the smelting process, promote the technical and economic index of the smelting process and the level of technology to achieve sustainable development of enterprises, the blending proportion optimization, the soft measurement of three main control parameters, the temperature of the matte, matte grade and ratio of Fe and SiO2of slag, and the prediction of the failure state of the key equipment linked up with the characteristics of the ISA furnace are researched in this thesis. The main research works are as follows:(1) To be aimed at blending proportion optimization for ISA furnace in copper smelting process, the intelligent optimization method based on adaptive ant colony algorithm is put forward. The first step, based on analyzes the blending process features, objective optimization function with the target at cost, considering the technology, quality, inventory constraints, is established. Then, the optimization problem is transferred into learning modeling in a variety of constraints by using using adaptive ant colony algorithm. At last, the blending proportion optimization is solved based on the modeling with the benefit of historical blending data. The simulating experimental results show that the optimizing method for blending can effectively reduce production costs and improve the efficiency of the blending system.(2) For ISA furnace copper smelting process, matte grade, the temperature of the matte, ratio of Fe and SiO2of slag are three main controlling parameters. According to high costs, large lag, difficult to measure, checking and controlling these three parameters is difficultl To be aimed at measuring these three key paramters, the soft measurement method is proposed based on generalized maximum entropy regression of the adaptive. At first, the input data of the smelting process are to reduce the dimensionality of pretreatment by locally linear embedding algorithm based on kernel clustering.Then, the operating conditions are detected by using hidden Markov model. Finally the generalized maximum entropy adaptive model is estabilished by combined with the operating condition detection model. The experiments show that the proposed method can significantly improve the error, improved measurement stability, and provide useful guidance for the factory production.(3) As a result of fluctuations of blending and furnace conditions irregularly, copper smelting is always accompanied by the fault of key equipment of ISA furnace. Due to melting bath stired strongly and invisibly, the faults are difficult to find and predict. To be aimed at fault monitoring and predicting, the method is proposed, which is based on kernel principal component analysis with a fusion of fuzzy C-means clustering feature samples and sparse least squares support vector machine (CSKPCA-SLSSVM). First, by using of Fuzzy C-Means algorithm to cluster the sampled data, the cluster center of the sample is formed as a base vector. Then based on the extracted characteristics of the sample data, the dimensionality of pretreatment is reduced by kernel principal component analysis. At last, based on T2and SPE statistics of ISA furnace failure to initial recognition, the final preliminary recognition results are classified and predicted accurately based on sparse least squares support vector machine fault diagnosis model. The experimental results show that this method can quickly show the change and the failure of the entire production process, contribute to monitoring of ISA furnace, and help to promote the monitoring level of similar industrial processes.