化工学报 ›› 2009, Vol. 60 ›› Issue (1): 168-171.

• 过程系统工程 • 上一篇    下一篇

基于支持向量数据描述的非高斯过程故障重构与诊断

张建明,葛志强,谢磊,宋执环,王树青   

  1. 浙江大学智能系统与控制研究所,工业控制技术国家重点实验室; 浙江大学工业控制技术研究所,工业控制技术国家重点实验室
  • 出版日期:2009-01-05 发布日期:2009-01-05

Non-Gaussian process monitoring and fault reconstruction and diagnosis based on SVDD

ZHANG Jianming,GE Zhiqiang,XIE Lei,SONG Zhihuan,WANG Shuqing   

  • Online:2009-01-05 Published:2009-01-05

摘要:

提出一种基于支持向量描述(SVDD)的统计过程监控与故障重构及诊断算法,避免了PCA、PLS等传统统计过程监控方法假设过程数据服从高斯分布的不足。鲁棒故障重构算法通过迭代保证重构后的数据对应的SVDD监控统计量最小化。诊断算法根据故障集中的不同故障重构后监控统计量是否恢复正常,确定实际发生的过程故障。CSTR过程的仿真研究表明了所提出方法的有效性。

关键词:

支持向量数据描述, 故障诊断, 统计监控

Abstract:

A novel support vector data description(SVDD)statistical process monitoring and fault reconstruction & isolation approach is proposed to address non-Gaussian multivariate process systems and overcome the deficiencies of traditional PCA and PLS methods.A robust iterative approach is involved to reconstruct the fault signals and minimize the SVDD monitoring statistics.By comparing the reconstructed statistics with different fault candidates,a fault diagnosis strategy is proposed to isolate the actual process fault.A study of the application in CSTR simulation demonstrates the efficiency of proposed method.

Key words:

支持向量数据描述, 故障诊断, 统计监控