The Methods Of Fused Magnesia Industrial Process Fault Monitoring Based On Manifold Learning

With the development of modern technology,the procedure of industrial process is becoming more and more complex,the requirement of continuous,high efficiency and safety for process control becomes increasingly higher.The abnormal situation in the production being found timely and effective becomes more and more important.The method of process monitoring is an effective way to solve the above-mentioned problem.And the technology of data-driven process monitoring technology has been widely applied to the field for fault monitoring.In this paper,on the basis of previous work,we carried out the research and application about a fault monitoring method of semi-supervised discriminant embedding based on pseudo-label penalty,a fault monitoring method of multi-view potential common subspace heterogeneous industrial data collaborative modeling based on discriminant information and a fault monitoring method of multi-view potential common subspace heterogeneous industrial data collaborative modeling based on mixed information.(1)In order to solve the shortcomings and defects of the semi-supervised discriminant embedding algorithm,this article is improved on the basis of the above algorithm.First,the online feature extraction is performed by means of regression,and the expression of the extracted feature are enhanced.Second,the pseudo-label of unmarked data is utilized,and the class discrimination information of the unmarked data and the local geometric structure are fully utilized to enhance the clustering of similar samples and the discreteness of heterogeneous samples and improve the accuracy of fault monitoring.(2)Considering the existing data-driven fault diagnosis method,the traditional data and multimedia stream heterogeneous data are separately modeled for fault diagnosis and monitoring.And the inevitable connection of the big data is ignored and the inherent characteristics of big data is lost.So a fault monitoring method of multi-view potential common subspace heterogeneous industrial data collaborative modeling based on discriminant information is proposed.The traditional physical and chemical variables and the data of images and videos are collaboratively modeled,and the internal relationship between their discriminating information is fully explored,and the accuracy of fault monitoring is improved.(3)Considering the fault monitoring method of multi-view potential common subspace heterogeneous data collaborative modeling based on discriminant information,the discriminant information learning is performed on the data features of all perspectives.The learning information may be relatively simple,and the complementary data features of all perspectives may be weak.This part proposes a fault monitoring method of multi-view potential common subspace heterogeneous industrial data collaborative modeling based on mixed information.Different perspectives adopt different feature learning algorithms,including discriminant information,principal element information,and positive and negative constraint information,enhancing complementarity learning among various perspectives data,and provide more clear positioning of industrial production objects to improve fault monitoring accuracy.Finally,we use the proposed algorithms in this paper for the process of fused magnesia furnace to simulate and analysis.Test their performance of fault detection and diagnosis respectively.Compared with some traditional methods,we verify the feasibility and effectiveness of the proposed method in this thesis.