基于多证据融合决策的间歇过程测量数据异常检测方法 - 化工学报.pdfVIP

基于多证据融合决策的间歇过程测量数据异常检测方法 刘伟 ，王建林，邱科鹏，于 涛，赵利强 旻 （北京化工大学信息科学与技术学院，北京 100029 ） 摘要：间歇过程测量数据的高维、非线性、非高斯分布特征直接影响过程测量数据异常检测的准确性，为了融 合多源数据异常检测信息，提升间歇过程测量数据异常检测精度，提出了一种基于多证据融合决策的间歇过程 测量数据异常检测方法，该方法通过引入证据理论（Dempster-Shafer ，D-S ），采用主焦元判别伪证据和重新计算 证据权重改进冲突证据处理方法，减小了冲突证据对多证据融合决策结果的影响，提高了间歇过程测量数据异 常检测的准确率。构建了基于多证据融合的测量数据异常检测模型并将其应用到间歇过程测量数据异常检测决 策判决中。实验结果表明，该方法能够融合多证据信息，有效地处理冲突证据，实现了间歇过程测量数据异常 检测，降低了误检和漏检率。 关键词：间歇过程；D-S 证据理论；冲突证据；多证据决策；测量数据异常检测 DOI: 中图分类号： TQ 277 文献标志码：A 文章编号： Detection method of abnormal measured data for batch processes based on multi-evidence fusion decision LIU Weimin, WANG Jianlin, QIU Kepeng, YU Tao, Zhao Liqiang （College of Information Science and Technology，Beijing University of Chemical Technology ，Beijing 100029，China） Abstract: The high-dimensional, non-linear, non-Gaussian distribution of measured data in batch processes directly influence the detection accuracy of the abnormal measured data. In order to fuse multi-source abnormal detection information and increase the detection accuracy of the abnormal measured data for batch processes, a detection method of the abnormal measured data based on multi-evidence fusion decision was proposed. By introducing the Dempster-Shafter evidence theory, the proposed method used the main focal element to identify fake evidence, and then recomputed the weight of evidences. Through the reduction of the influence caused by the conflicting evidences, the detection accuracy of the abnormal measured data was improved. Furthermore, the detection model of the abnormal measured data was constructed with the multi-evidence fusion decision, and it was applied to the decision-making of the abnormal measured data