基于SVDD的多时段间歇过程故障检测

发布时间：2018-11-09 20:00

【摘要】：支持向量数据描述(SVDD)不要求过程数据满足正态分布,已应用于间歇过程故障检测。现有的SVDD间歇过程故障检测方法采用聚类分析和模型识别划分间歇过程时段,多时段划分的准确性较低,制约了多时段间歇过程故障检测精度的提高。针对上述问题提出了一种基于SVDD的多时段间歇过程故障检测方法,利用SVDD的超球体半径值与支持向量个数的变化划分间歇过程的多时段,并建立了不同时段的SVDD故障检测模型,使用待检测样本点的球心距与对应时段的超球体半径之差检测过程故障,实现了多时段间歇过程的时段划分与过程故障检测。发酵过程仿真实验和实际生产实验结果表明,该方法能够准确地划分间歇过程的多时段,并且能够针对不同时段进行故障检测,具有较高的检测率。
[Abstract]:Support vector data description (SVDD) does not require process data to satisfy normal distribution and has been applied to batch process fault detection. The existing SVDD batch process fault detection methods use clustering analysis and model identification to divide the batch process period, and the accuracy of multi-period division is low, which restricts the improvement of fault detection accuracy of multi-period batch process. In order to solve the above problems, a fault detection method based on SVDD for multi-time intermittent process is proposed. The variation of the radius of hypersphere and the number of support vectors of SVDD is used to divide the multi-period of intermittent process, and the model of SVDD fault detection in different periods is established. Using the difference between the spherical center distance of the sample point to be detected and the hypersphere radius of the corresponding period to detect the process fault, the interval division and the process fault detection of the multi-time interval intermittent process are realized. The results of fermentation process simulation and actual production experiments show that the method can accurately divide the multi-period of batch process, and can detect faults in different periods, and has a high detection rate.
【作者单位】：北京化工大学信息科学与技术学院;
【基金】：国家自然科学基金(61240047) 北京市自然科学基金(4152041)项目资助
【分类号】：TP277

【相似文献】