Study On The Method Of Fault Detection In The Smelting Process Of Electro-Fused Magnesium Oxide Based On Data

The electro-fused magnesium oxide smelting process involves a very complex series of physical and chemical reactions,the gas,liquid,solid and plasma states of MgO exist together,and various heat and energy transfer processes are coupled together at the same time,resulting in complex and changeable disturbances and frequent faults.However,the detailed working mechanism of the smelting process of electro-fused magnesium oxide is not easy to be analyzed,and it is difficult to establish an accurate mathematical model.In addition,a large amount of data generated in the smelting process can generally directly or indirectly reflect the overall operating state.therefore,this article adopts the method of based on the data to carry out research of fault detection.The image data of the melting process collected from a single Angle cannot be fully observed for the changes in the smelting of electro-fused magnesium oxide,and it is easy to miss the important characteristic information when the fault occurs.Therefore,this paper proposes a fault detection method based on graph regularized linear nonnegative matrix factorization(Multi-GLNMF)of multi-view.With multiple angles of image data to test for the variables,the high-dimensional space of sample data,through the Multi-GLNMF method is mapped to a low dimensional space,consistent low dimensional approximation to calculate the sample points,namely to extract the key characteristic information,and finally,the test statistic of sample points and the control limit,to judge whether the samples corresponding to the operation condition of the moment a fault.Current is important test variables in the process of fused magnesium melting,current data can indirectly reflect the whole running condition of the melting process,but the current data are easy to be interference,reflect the condition of hard to dig,used alone to fault detection limits is larger,the data and image data more intuitive reflect the operation state in the process of fused magnesia melting.However,the essential structure of image data and current data is different,and the operation of extracting the relevant features of normal state and fault condition cannot be carried out directly together.Therefore,this paper proposes a fault detection method based on collaborative learning of heterogeneous data.First,you need to use of two-way two-dimensional linear discriminant analysis to the characteristics of the image is divided into horizontal and vertical direction,then the characteristics and the current data in two directions study together,to extract the effective information,at the same time,the training of a linear classifier,it can promote the related in the process of training characteristics of learning,but also in the process of detection,directly determine the state of the samples some time,thereby,achieve fault detection of fused magnesia melting process.Based on the practical problems,fused magnesia melting process under the Windows environment,this paper designed the online monitoring system based on image data,using the network cameras from multiple angles collect the image data of fused magnesia furnace mouth work area,and smelting condition real-time display in the control room of the PC,at the same time,the application of fault detection method based on Multi-GLNMF algorithm for process monitoring.