Complex System Fault Detection Using Factor Analysis

Yilu Zhang
Submission Type: 
Full Paper
AttachmentSizeTimestamp
phmc_10_027.pdf599.72 KBAugust 23, 2010 - 12:23pm

The key to diagnostics and prognostics (D&P) technologies is the understanding of system failure modes, including the classification, the symptoms, the underlying causes, the progressive patterns, and the effects. However, the proliferating system complexity makes it difficult to exhaust every potential failure mode, especially under the mounting time-to-market pressure. Effective and efficient data analytical methodologies are being called for to compensate the popular DnP methods. In this paper, we propose to use a data modeling technology, Factor Analysis, to identify new/unknown failure modes and the signature of incipient failures, given limited prior knowledge of the system. Factor Analysis captures the dominant dependency underlying observable measurements of physical systems, and is sensitive to their changes, as demonstrated by the preliminary experimental results on two real-world datasets.

Publication Control Number: 
027
Submission Keywords: 
factor analysis
principal component analysis
K-means
fault detection
battery management systems
Submitted by: 
  
 
 
 

follow us

PHM Society on Facebook Follow PHM Society on Twitter PHM Society on LinkedIn PHM Society RSS News Feed