Preprocessing Method Based on Pattern Recognition of Process Data and Its Applications
Xiang Li · Systems engineering and electronics · 2002
Based on the analysis of data measurement error sources in industrial process, this paper presents a data preprocessing method which is based on pattern recognition. This method only depends on the qualitative and partialy quantitative knowledge of the concerned domain and can deal with different errors.After preprocessing by this method the valid information is kept and the data with gross errors are eliminated, thus preventing any inconsistent data. The application of this robust method on data preprocessing in pulp cooking process has shown its effectiveness.