Research On Monitoring And Diagnosis Of Process Minimal Fault Based On Multivariate Statistical Analysis

With the development of science and technology and the growth of social demand,process industry as an important basis for the development of national economy.Among them,the minor faults caused by equipment aging and environmental drift often cause major production accidents,resulting in serious personal injury,huge property losses and even irreparable environmental disasters.As the key technology to ensure its safe and stable operation,process monitoring and fault diagnosis have been widely concerned and developed rapidly.According to the production process of process industry and a large number of process data collected,the process monitoring model is established by multivariate statistical analysis method to realize the monitoring and diagnosis of minor faults.The main work is as follows:(1)A minimal method combining MEWMA and PC A is proposed for the minor faults in industrial processes.Firstly,the weighted sliding average of multivariate index is applied to filter the original sampled data to eliminate the noise.Secondly,two kinds of statistics T2 and SPE were extracted by principal component analysis(PCA).To simplify the monitoring process,the above statistics were fused into a single joint index.Then the whole process monitoring process is given.Finally,the rationality of the monitoring method is verified by case study.(2)PCA ignored the autocorrelation of data,so it proposed a method of micro-fault monitoring combining CVA and MEWMA.In this method,MEWMA is used to filter the original data,and then the state space model of CVA is used to maximize the correlation between the past data and the future data,to establish the process statistical indicators,and then to give the process monitoring process of MEWMA-CVA.Finally,the effectiveness of the method is verified through the TE process,and a systematic comparison analysis is made with the monitoring method proposed in the previous chapter.(3)The traditional diagnosis method is inaccurate because of the smudge effect,so the joint index gradient reconstruction contribution graph diagnosis method is proposed.Reconfiguration contribution graph is used to eliminate fault propagation.Then the change of fault at different sampling time is extracted by the combined index gradient method,and the fault origin is located accurately.The accuracy of the method is verified by the analysis of a fault case.(4)The system development of minimal fault monitoring on kingview lays a foundation for the above monitoring and diagnosis methods.The process industry is very complex,and the monitoring method for multivariate statistical analysis of minor faults should be further expanded in terms of non-linear variables and new data analysis methods to further improve the monitoring effect.