Research On Process Monitoring Of Process Industry Based On Multivariable Statistic Analysis

Nowadays, the research of process monitoring has been a key role in the field ofindustry process researching. As a modern production system, which has been gettingincreasingly complicated and Intelligent, the process industry is be supposed to be a moresecure and more efficient system. And a reliable and advanced process monitoring system isunquestionable playing a important role on that. As a consequence, the research which isbased on some theories such as principal component analysis、 kernel principal componentanalysis and some other methods is carried out in this paper.First of all, this paper analyzes the process monitoring method and its development trend.And the paper makes analysis and description on the methods of fault detection and faultidentification based on multivariate statistical analysis. In addition, some common statisticalindicators such as T2and SPE and the relation of each other in the fault detection of oneindustry is analyzed. The advantage of multivariate statistical analysis is given by thecomparisonby the univariate statistical analysis methed. What is more, the application of kernelmethod in process monitoring is given by a series of general formulations are developedsecondly, the Tennessee Eastman (TE) chemical process simulation in themanufacturing process is analyzed in this paper, and then, the data of TE is acquired for theresearch of process monitoring. Moreover, a hardware in the loop simulation (HILS) is biultfor the futher research. On that basis, the monitoring by principal component analysis andkernel principal component analysis is simulated and verificatied, and the result indicates theadvantage of KPCA methed in a more complex process.At last, theory of sparse representation is proposed and applied to the TE and HILSprocesses to reduce to the modeling data and its operation. The result shows it can remarkableincrease the efficiency of the monitoring of KPCA..