Monitoring And Realization Of Electrofused Magnesium Oxide Melting Process Based On Multi-angle Cooperation

Increasingly complex processes are a major feature of modern industrial processes.Once complex industrial processes break down,the consequences are unimaginable.Therefore,it is necessary to monitor and analyze the faults of complex industrial processes.Data-driven process modeling and monitoring is a common technique.Based on previous work,this paper carries out the following work:A multi-perspective manifold analysis fault monitoring method based on image feature regression is proposed,a heterogeneous data fusion fault monitoring method based on cluster manifold is proposed,and a remote video online monitoring and alarm platform is developed.The main research contents are as follows:(1)In order to solve the previous problems of missing data from a single perspective,unable to accurately reflect the smelting status,cumbersome multi-perspective analysis and difficult real-time analysis,in this paper,a fault diagnosis method for manifold analysis based on image feature regression is proposed.The first part of this paper is to collect video data of smelting from multiple perspectives to make up for the lack of data from a single perspective.Second,the idea of feature regression and the method of manifold analysis are adopted to simplify the calculation while preserving the maximum information amount of the picture,which has a strong real-time analysis capability and high accuracy of fault detection.(2)In order to solve the problems that single type of data modeling leads to insufficient utilization of industrial data,the establishment of a single model,and the failure to fully explore the characteristics of industrial big data,this paper proposes a cluster manifold based heterogeneous data fusion fault monitoring method.In the smelting process of electro fused magnesium oxide,the data that can be collected include:video data from three perspectives,and current from three electrodes.The dimension of current data is small,the data sample is small,the modeling is simple,and the robustness of the model is poor.The dimension of video data is large,the data sample is large,the modeling process is complex,and the model effect is good.The algorithm proposed in this paper neutralizes the current model and the video data model,which can fully mine the internal connection between video data and current data,while retaining the integrity of big data and fully mining the characteristics of industrial big data.(3)According to the algorithm proposed in this paper,a remote video online monitoring and alarm platform was built to analyze the real-time smelting data of the factory online.This platform has verified the fault monitoring method of multi-perspective manifold analysis based on image feature regression,and achieved good monitoring effect while monitoring the smelting process in real time.