Research On Fault Detection Of Fused Magnesium Oxide Based On Manifold Study

In the actual production of fused magnesium oxide,the production process is complicated and the control is difficult.Due to the continuous melting of the charge during the smelting process,there are many forms of solid,gaseous and molten states in the molten pool.At the same time,the magnesium oxide powder generates a large amount of gas when it is melted,which easily causes a "eruption" phenomenon,which requires timely detection of abnormalities and malfunctions in the process.Based on the previous work,this paper proposes a semi-supervised cost sensitive fault detection method based on graph and a heterogeneous data collaborative modeling fault diagnosis method based on two-dimensional neighborhood preservation projection:(1)In practical engineering applications,the fault information of a large number of data samples is unknown,and calibration of unlabeled fault samples is a time consuming and laborious task.In order to solve this problem,a semi-supervised cost sensitive fault detection method based on graph is proposed.According to the manifold hypothesis of the semi-supervised method,the method uses the tag information of the tag data and the manifold structure information of the unlabeled data to mine the missing categories of tag information and delineate the early fault classes,and is sensitive to the early failure category.The method of diagnosis.Finally,the data of the fusedmagnesium oxide production process were simulated and analyzed,and the feasibility and effectiveness of the proposed method were verified.(2)Due to the limitation of the production environment,the number of physicochemical variables that can be measured in some production processes is small,which results in an incomplete amount of information and cannot accurately reflect its production status.For this problem,this paper proposes a heterogeneous data collaborative modeling fault detection method based on two-dimensional neighborhood preservation projection.On the basis of monitoring the video data,the physical data is added for collaborative modeling,and the extracted low-dimensional vector contains more comprehensive and accurate state information by adding the tag local information of the manifold and the data.Finally,the method was applied to the fault detection of fused magnesium oxide production process,and the feasibility and effectiveness of the proposed method were verified.