基于模式聚类的过程数据预处理方法及应用

Preprocessing Method Based on Pattern Recognition of Process Data and Its Applications

在分析工业过程数据测量误差源的基础上 ,提出了一种基于模式聚类的数据预处理方法。该方法只依赖于领域的定性知识和半定量知识 ,能处理不同的误差干扰 ,预处理后的数据保留了原始测量数据的有用信息 ,剔除了过失数据 ,从而防止了矛盾数据的出现。将该方法用于制浆蒸煮过程的数据预处理中 。

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.