CIESC Journal ›› 2013, Vol. 64 ›› Issue (5): 1674-1682.DOI: 10.3969/j.issn.0438-1157.2013.05.024

Previous Articles     Next Articles

Fault detection of complex chemical processes using Mahalanobis distance-based local outlier factor

MA Hehe, HU Yi, SHI Hongbo   

  1. Key Laboratory of Advanced Control and Optimization for Chemical Processes of Ministry of Education, East China University of Science and Technology, Shanghai 200237, China
  • Received:2012-08-17 Revised:2012-12-12 Online:2013-05-05 Published:2013-05-05
  • Supported by:

    supported by the National Natural Science Foundation of China(61074079)and Shanghai Leading Academic Discipline Project(B504).

基于马氏距离局部离群因子方法的复杂化工过程故障检测

马贺贺, 胡益, 侍洪波   

  1. 华东理工大学化工过程先进控制和优化技术教育部重点实验室, 上海 200237
  • 通讯作者: 侍洪波
  • 作者简介:马贺贺(1985-),男,博士研究生。
  • 基金资助:

    国家自然科学基金项目(61074079);上海市重点学科项目(B504)。

Abstract: Complex chemical processes often have multiple operating modes to meet the changes of production conditions.Meanwhile, process complexity makes the within-mode data usually follow an uncertain combination of Gaussian and non-Gaussian distributions.The multimode characteristics and the uncertainty of data distribution within one single mode make the conventional multivariate statistical process monitoring(MSPM)methods unsuitable for fault detection.In this paper, a novel Mahalanobis distance-based local outlier factor(MDLOF)method was proposed.The local structure of variables was taken into account by employing Mahalanobis distances, and the local density of the surrounding neighborhoods was also considered.A density-based monitoring statistics which was robust regardless of data distribution was obtained by this way.Finally, the validity and effectiveness of MDLOF were illustrated through a numerical example and the Tennessee Eastman process.

Key words: multimode process monitoring, fault detection, local outlier factor, Mahalanobis distance

摘要: 为了满足实际的生产需要,复杂化工过程往往包含多个运行模态。同时过程的复杂性使得同一模态下的数据分布是一种高斯分布和非高斯分布混合存在的不确定情况。数据的多模态分布特性以及同一模态下数据分布的不确定性使得传统多元统计监控(MSPM)方法很难给出令人满意的结果。针对这一问题,本文提出一种新的马氏距离局部离群因子(MDLOF)方法进行故障检测。通过利用马氏距离挖掘变量局部结构中包含的有用信息,并对样本的邻域密度加以考虑,形成对数据分布具有鲁棒性的基于密度的监控指标。最后通过数值仿真例子及Tennessee Eastman过程验证其有效性。

关键词: 多模态过程监控, 故障检测, 局部离群因子, 马氏距离

CLC Number: