Reconciliation de donnees en ligne (French text).

摘要

Pulp and paper mills are starting to use, more frequently, information management systems that are coupled with data management systems. However, until now, those data have been used through impromptu analysis, which does not allowed the full exploitation of the advantages related to the implementation of information systems. Consequently, the implementation of an on-line methodology allowing the correction of measurements would ameliorate system efficiencies.; A methodology allowing the utilization, by applying an adequate data treatment strategy, of reliable on-line data in the automation of high-level tasks is elaborated in this study. This methodology requires to be robust and to allow the improvement of real-time data accuracy and precision. This general objective is partially achieve by the investigation of these sub-objectives (1) To compare the efficiency of on-line abnormalities detection techniques. (2) To compare the performance of a new on-line steady state identification method, performed in 3 steps, with existing methods. (3) To demonstrate the capacity of implementation of the methodology on industrial data. (Abstract shortened by UMI.)