Multivariate monitoring of a carbon dioxide removal process

Abstract

Process hardware improvements have significantly increased the amount of information collected in industrial facilities, which allows the use of tools such as multivariable statistical analysis for process monitoring. Nevertheless, such statistical models tend to be static and extremely general when implemented in process facilities with several modes of operation, as is the case of carbon dioxide removal processes. This work demonstrates the use of multivariable statistical analysis for process transitions between different modes of operation. Continuous process analytics are used to define key variables, named “state variables”, to determine the current mode of operation. This work also makes use of parallel coordinates to illustrate the simultaneous visualization of several transition paths and statistical tests. Such tests apply confidence limits that are appropriate to the current mode of operation. This methodology is successfully tested in the CO2 capture plant at the University of Texas in Austin. The results show the effectiveness of using this type of application to detect abnormal operating conditions at different levels of operations.