Multiple correspondence analysis to study failures in a diverse population of a cable

Swati Sachan, Chengke Zhou, Rui Wen, Wubin Su, Chenjie Song

Research output: Contribution to journalArticlepeer-review

5 Citations (Scopus)
126 Downloads (Pure)


The study of failure behavior of a diverse population of cables is challenging. Previous attempts have failed to capture the complexity of cable system failures due to an independent analysis of multiple failure causes or influential factors. In this paper, the multiple correspondence analysis (MCA) is proposed for simultaneous analyses of multiple variables responsible for the cable failures and classification of cables into homogeneous groups in terms of past failure behavior. The proposed classification method is less subjective as it gives equal consideration to all the cable features. The methodology has been applied to the main cable section and cable joint failure data of a diverse population of cables obtained from a Chinese utility company. The failure data have six categorical variables related to cable features and failure characteristics. The application of MCA provided an enriched view and understanding of failure behavior by allowing visual exploration of the failure patterns and associations. Based on the past failure, the cable sections and joints were classified into three and four groups, respectively. The failure trend of each classified group is evaluated separately. Results show that failure history and trend of each classified group is different. Thus, they must be analyzed and treated differently in the forecasting or maintenance planning procedures.
Original languageEnglish
Pages (from-to)1696-1704
Number of pages9
JournalIEEE Transactions on Power Delivery
Issue number4
Early online date5 Oct 2016
Publication statusPublished - Aug 2017


  • cable system failures
  • Multiple Correspondence Analysis


Dive into the research topics of 'Multiple correspondence analysis to study failures in a diverse population of a cable'. Together they form a unique fingerprint.

Cite this