Use of SPC in non-discrete processes


Thread Starter

Andrew Ng

Does anybody have any experience with using SPC (statistical process control) software for a non-discrete process? I am wondering if it can help in improving a typical chemical process e.g. distillation etc.
Statistics are by nature discrete. So, your view is incorrect ! Is it ? If you have a continuous model of your process, yes statistical methods will help you in many ways. But, for this, you must have math tools. For example: Mathcad, Matlab, Origin, Minitab...There are many more specialised softwares.

The idea that SPC only works for discrete processes is one of the old but persistant wives tales about SPC. The diference is that the dimension/variable to be measured is not so obvious in a continuous process as it is in a discrete manufacturing operation. You have to identify the possible assignable causes and carefully choose from your process variables what to track to reveal process troubles. For example, in a heating application it may do little good to track the temperature as your variable in SPC, because it is an actively controlled element, but you might track the controller's response rate to a specific disturbance, or you might track changes in heater power consumption. An increasing or decreasing trend in your controllers output will be revealed by SPC methods and would reveal a process change. These are just examples and may not mean anything to sone particular process. Someone with extensive knowledge of the process in question should be involved in the choice of what to track. And you may not get it right the first time. In a particular vacuum process it took us a long time and several tries before we found a meaningful measurement to track, now it reveals alot about the health of the process. A pretty good book on it is "Understanding Statistical Process Control" by Donald Wheeler and David Chambers. Be patient, creative, and don't give up because the benifits are tremendous.