Role of Data Reconciliation in Process Plants, 1992 Engineering Conference Proceedings

Data reconciliation is a topical subject as more mills become automated and the average manager or process engineer is threatened with the vast landfill of paper reports with varying and inconsistent information. In Europe, data reconciliation is called the single number system. Sometimes reconciliation of the data is for operations decisions and sometimes it is required by contract or law. Variation and uncertainty in these data causes, at best, confusion and, at worst, obfuscation of serious problems.

Data reconciliation is the technique of modifying raw data until they form an unique, internally consistent picture. Some would claim that this makes the data more accurate but this is not necessarily so. It makes the data more reliable by demanding that known relationships are forced true and generating additional information from non process measurement sources such as shipments and receipts.

This paper will briefly describe the three primary methods used for data reconciliation; stream selection, regression, and SLP optimization. The use of reconciliation for data set preparation, model matching, business data, problem detection, and legal validity will then be examined individually. Each of these takes advantage of certain characteristics of data reconciliation to improve an application.