摘要 |
<p>Raw measured data is available as diverse quality- or process data obtained during manufacture or finishing. Laboratory quality parameters are also taken into account in raw data measurements. The raw data is assembled into data sets, each evaluated simultaneously to determine laboratory or quality measurements as target data from selection. A soft sensor algorithm running in the computer-based operation and correlation unit (12) carries out calculations from the input data. Each data set contains measured data which covers the same control volumes of web- or raw material, especially paper or its raw material. The raw measurement data is given a time stamp. The soft sensor algorithm is determined using mathematical processes, e.g. PCA, PLSR, SOM, fuzzy logic, neural networks or similar. The soft sensor algorithm runs automatically in the computer-based operation and correlation unit (12), as soon as new input data is available. The method is further elaborated and expanded, covering the following topics: automatic cycling and manual operation, display and print out, transport delay elimination, data filtering for accuracy, data averaging and windowing, target data, production rejection tools, adaptive learning and algorithmic coupling.</p> |