| In this dissertation, the subject root in the cross-cutting projects in Zhang Jia Gang LianFeng Iron and Steel Company─"YongGang the Integrated Management System production running ". For the LianFeng steel company seven bell-less top blast furnace of 350-580m3 existing not high ironmaking degree of automation, and temperature controling relying entirely on the human experience of operational, this paper researched such as issues mainly. In the analysis of on-site smelting process and a large number of production data, basing on the use of modern intelligent algorithms to establish predictive controlling model of blast furnace, arms to guide operations and improving quality and output of hot metal.On the basis of reading plenty of domestic and overseas literature, the dissertation introduces the domestic and foreign present status and development tendency of blast furnace prediction model.Through analysizing and researching the production status, production processes, and existing problems, aiming at the LianFeng Steel Company blast furnace production, proposes the establishment of furnace temperature prediction model for the program.Based on the existing data acquisition system,integrating of ironmaking production process data and various information subsystems data; Through the analysis and processing of data on production, using neural networks predicts model blast furnace temperature. This dissertation describes the data collection point for the specific blast furnace acquisition process, data processing and furnace temperature prediction model building process.In this dissertation, analysis the data needed to predict furnace from the perspective of iron-making mechanism and process in-depth. Through analysizing the field devices of blast furnace, the production of data storage status in detail,determined the source of data requiredand the corresponding interface standard. Introduction of architecture and functionality the existing data acquisition system briefly. And then discussed the blast furnace integration approach of heterogeneous data source data in detail.Finally focused on representing the concrete realization of data acquisition process of the blast furnace automation system. Ultimately to realizing data collection for the forecast model of requirements data.At the same time, taking into account the online collectig data has many noises in the blast furnace and impacts modeling accuracy.Therefore, on-line collected data have been pre-blast furnace.Including first and second data preprocessing and data of normalization and choice.Providing a solid data base for building predicting model.On the base of introducting the basic idea of neural networks briefly, focusing on inroducing the process of established blast furnace temperature prediction model based on artificial neural network. Firstly,the use of correlation analysis methods, analysized of the correlation during process parameters,explored the lag time of process parameters impacting on the blast furnace hot metal silicon content [Si]Then among the parameters affecting furnace filtered out the correlation coefficient larger parameters as model input parameters, eventually determined the neural network input and output. Finally the use of processed data of production process as a sample to train neural network.The network output value compared with the actual values.Experimental results show that the improved BP network prediction of blast furnace silicon content in hot metal hit rate was significantly higher than the standard BP network.It meets the the actual production needs. So the models can help blast furnace foreman controlling blast furnace parameters to achieve predictive control of the blast furnace. Achieveing raising hot metal production and economic efficiency.Comeing up to modeling the desired results.The dissertation finally summarizes the work which did to this dissertation topic, points out deficiency,simultaneously forecasts the work which dissertation topic next stage has to do.
|
PDF Full Text Request