New Data Mining Methods in Modern Power Plants Optimizing Power Plant Operation through Analysis of Process Data stored in existing IC-System Archives

摘要

Thanks to modern measurement and IT technologies, it is now easy to measure almost any operating behavior in real time or process and archive measured data in contiguous storage. The collected data are only rarely analyzed and, in some cases, the stored data are never evaluated in any way. As a result, the unique opportunity to take full advantage of the information stored in a database relating to the condition of a measured object is wasted. With condition-oriented maintenance systems in particular, the large volume of information embedded, for example, in the vibration characteristic values is archived to virtually no useful effect. Vibration and process values are not merely indicators as to the current state of the plant or machine, but can, if analyzed correctly, be applied as a very helpful tool in optimizing the process in order to achieve a low wear regime, for example. However, it is virtually impossible to solve this task using conventional production management and maintenance tools. To be considered feasible, a tool must be capable of speedily processing very large data volumes of the MB or even GB order of magnitude. It must be powerful enough for the user to be able to analyze data in real time without inconvenient processing delays so that an overview of the collected material is immediately on hand. The fact that older data in particular are often inadequately documented is a further complicating factor and means that any tool must be tolerant of missing units or physical correlations.