| With the rapid development of China’s economy,the demand for non-ferrous metals is increasing.However,with the reduction of mining areas,the reserves and grade of nickel sulfide ore are gradually decreasing,while the demand for copper and nickel is increasing.Therefore,it is particularly important to select the appropriate mineral processing technology to improve the utilization rate of mineral resources.More than 90% of China’s nonferrous metals are purified by foam flotation.Foam flotation is a mineral separation and beneficiation method based on the difference of physical and chemical properties of ore surface.It separates different minerals by the different floatability of minerals produced by different physical and chemical properties of different minerals.This paper is based on the cooperation project between University of Electronic Science and technology and a company in Gansu.Aiming at the shortcomings of low level of automation,bad environment and the randomness of workers’ control in the nickel foam flotation process,data mining is carried out by using the huge production data in the database.There are five flotation columns for froth flotation of a company in Gansu.The process is cumbersome and the parameters are numerous.This makes mechanism based modeling almost impossible.With the development of computer computing technology and intelligent algorithm,non analytical modeling has gradually become the main way of data processing and modeling.In this paper,based on data-driven black box modeling method,using data mining related technology for analysis.In this paper,through the analysis of the data characteristics collected from the flotation field,a variety of data mining algorithms are used to establish the model between the input parameters and the output parameters.Finally,the model accuracy meets the requirements,and a set of flotation modeling software is developed and delivered to the factory.The work of this paper is as follows1.one Data feature analysis and data preprocessingIn this paper,SQL Server 2008 database is used as the data warehouse to receive the data from the field.There are many sources of data,including data from flow meters,barometer,concentration meter,grade analyzer,particle size analyzer,and image data from foam image acquisition system.These data are missing,repeated,normal condition data and abnormal condition data are mixed together,so the data characteristics of these different sources are analyzed respectively,and then the special data preprocessing method is developed to remove the abnormal data according to the data characteristics.2.Preliminary data processingThe normalization method is used to reduce the input data to the same dimension,so as to prevent some parameters with smaller values from being ignored.The principal component analysis algorithm is used to reduce the dimension of many parameters,which reduces the calculation of data mining,improves the operation speed of the follow-up process,and improves the accuracy of the model.In the clustering algorithm,K-means clustering algorithm is used to divide the data with small Euclidean distance into one group.The complex flotation condition is divided into four small conditions.By selecting stable condition data,the accuracy of subsequent modeling is improved.3.Data modelingData mining technology uses artificial neural network technology.In the artificial neural network algorithm,the nonlinear autoregressive neural network(NARX)is used according to the characteristics of nonlinearity,hysteresis and time series data of flotation process.Nonlinear autoregressive neural network has the function of memory and feedback.The data training results of the previous moment can affect the data input of the next moment.It makes the network dynamic and closer to the real system.Finally,root mean square error is used to evaluate the accuracy of the model.The root mean square error and the final stability are below 0.1,which meet the accuracy requirements of the model. |
PDF Full Text Request