系统仿真学报 ›› 2021, Vol. 33 ›› Issue (9): 2066-2073.doi: 10.16182/j.issn1004731x.joss.20-0362

• 仿真建模理论与方法 • 上一篇    下一篇

基于LLE与K均值聚类算法的工业过程故障诊断

李元, 耿泽伟   

  1. 沈阳化工大学 信息工程学院,辽宁 沈阳 110142
  • 收稿日期:2020-06-16 修回日期:2020-07-10 出版日期:2021-09-18 发布日期:2021-09-17
  • 作者简介:李元(1964-),女,博士,教授,研究方向为基于数据驱动的工业过程故障诊断。E-mail:li-yuan@mail.tsinghua.edu.cn
  • 基金资助:
    国家自然科学基金重大项目(61490701); 国家自然科学基金(61673279)

Fault Diagnosis of Industrial Process Based on LLE and K-means Clustering Algorithm

Li Yuan, Geng Zewei   

  1. Department of Information Engineering, Shenyang University of Chemical Technology, Shenyang 110142, China
  • Received:2020-06-16 Revised:2020-07-10 Online:2021-09-18 Published:2021-09-17

摘要: 工业过程中各类数据间具有一定的相似性,单纯利用K均值算法对其进行故障诊断时,存在很大的错误率。提出一种基于局部线性嵌入(Locally Linear Embedding, LLE)的K均值聚类算法,将正常数据运用LLE算法降维并求出投影矩阵,利用投影矩阵将原始故障数据映射到低维空间,再利用K均值算法对其聚类,建立检测与诊断模型。将此方法应用于田纳西-伊斯曼(Tennessee-Eastman, TE)过程中进行故障检测与诊断,并同传统K均值算法及LLE算法对比,结果表明:提出的新方法具有更高的正确率,同时可以有效地对未知类型的故障数据进行判别。

关键词: K均值聚类, 局部线性嵌入, 田纳西-伊斯曼(Tennessee-Eastman,TE)过程, 故障诊断

Abstract: Because of the similarity of various types of data in the industrial process. The fault diagnosis using the K-means algorithm has a large error rate. A K-means clustering algorithm based on Locally Linear Embedding (LLE) is proposed. the normal data is reduced by the LLE algorithm and the projection matrix is obtained. The projection matrix is used to map the original fault data to the low-dimensional space and the K-means algorithm clusters is used to carry out the data to establish a detection and diagnosis model. The method is applied to the fault detection and diagnosis in the TE (Tennessee-Eastman) process and is compared with the traditional K-means algorithm and LLE algorithm. The results show that the proposed new method has a higher accuracy rate, and could effectively identify the unknown types of fault data.

Key words: K-means clustering, locally linear embedding,TE process, fault diagnosis

中图分类号: