Lab-data Based Process Control—Comparing SPC and Automatic Control

Abstract: In many process industries samples are taken and analyzed in laboratories to monitor and ensure proper production conditions. If the analysis result is not within specification limits measures are taken to correct errors. It has proved to be hard to take the appropriate measures. The problem description fits into the confines of control theory and the aim is to make a fairly simple solution to a very complex problem. For example, a standard proportional-integral-controller block in the control system can not be used. As a case study, this project tries to stabilize and control the content of residual formaldehyde in a Neo production plant situated in Perstorp. The content is affecting the yield of the two main reactants in the reaction and therefore important in an economical aspect. The thesis answers the following questions: 14 What is statistical process control and is it the solution to the case study? 14 What are the similarities and parallels between classic automatic control and statistical process control? The theory of statistical process control is based on defining processes to be in statistical control or out of statistical control. The founder of the theory, Dr Shewhart, puts it into words like this: While every process displays variation, some processes display controlled variation, while others display uncontrolled variation. Many tools are developed to signal when a process changes from displaying in-control properties to out-of-control properties. Quite a few of those tools are described in the statistical process control chapter. A chapter recalling some useful parts of the automatic control theory is followed by a chapter describing tools and theories in the borderland between statistical process control and classic automatic control. The two theories contradict each other in some aspects. While statistical process control always try to identify the actual cause of production malfunction, automatic control can in some cases hide the actual cause by manipulating the control parameter. Tools coping with situations like that have been developed and are described in chapter 7: Comparing and Combining SPC and Automatic Control. Analyzing the whole process in depth is too complicated and had to be limited into focusing on the parameters dominating the variations in the residual formaldehyde content. Extensive data analysis pointed out the formaldehyde loading not to be sufficiently accurate. A control system consisting of both feedforward and feedback terms is designed. Only the feed forward term could be tested in reality. The semi-manual Excel based solution showed good results in time periods when used correctly. The conclusion is that statistical process control is not the solution to the case study problem but nevertheless interesting, especially for its diagnostic tools. The theory should not be considered an alternative to automatic control but as a good complement. The project shows that it is definitely possible to stabilize and control the residual formaldehyde content. In order to do this an automatic implementation is needed.