Industrial Process Monitoring And Fault Diagnosis Based On Multivariate Statistical Analysis

With the rapid development of mass production and new technology, automatic control systems for modern industrial processes have been expanding to increasing complexity and larger scale, in the new era with highly progressed production ability, science and technology. However, the more complicated system, on the other hand, brings that in case of unexpected fault it normally takes longer time and expense to figure out and eliminate, or even causes human injuries or environmental problems. Therefore, research on integrated process control system, which includes such roles as control, monitoring, diagnosis, proves to have important theoretical and practical value. Fault diagnosis, provides scientific methodology to detect and diagnose system exception, and find out fault source, frequency, severity, tendency etc., so as to take timely and effective solution.Multivariate statistical technique is an important branch in the research of fault diagnosis. The paper systematically elaborates this technique from many aspects, discussing the existing methods applied in two typical processes: continuous process and batch process, as well as their weakness. And the respective solutions are introduced. Contents in this paper are as following.1. The paper summarizes the definition, contents, catalog and developing tendency of fault diagnosis, especially provides complete and historical introduction on multivariate statistical technique.2. Briefly introduces Principal Component Analysis (PCA) and its application in fault diagnosis, and brings several computational methods of PCA, based on which Principal Component Recursion (PCR) and Partial Least Squares (PLS).3. Summarizes the design and develop ways of monitoring and fault diagnosis systems based on PCA for multivariable control system. Introduces the main statistical variables, multivariate control charts and the analysis ways of statistical variables. Then this paper takes a model of continuous process for example, to show and analysis the result of fault diagnosis based on PCA.4. Describes the modeling, process performance monitoring, and fault detection for batch processes using Multiway Principal Component Analysis (MPCA) method. And the Dynamic Time Warping (DTW), which used to synchronize the time length of data trajectories observed is introduced. Based on these techniques two improved MPCA,Multi-model MPCAand Slide-model MPCA, are proposed, and takes a model of batch process for example to introduce the two MPCA.In the end, contrast and analysis on most research methods on fault diagnosis are made, in order to figure out issues that deserve more attention and deeper understanding for future research in multivariate statistical analysis.