Health monitoring of hard disk drive based on Mahalanobis distance

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

47 Scopus citations

Abstract

A hard disk drive (HDD) is one of the core components of most computer systems. A failure of HDD may cause serious data loss and catastrophic consequences. Thus, health monitoring and anomaly prediction for HDD are critical to prevent data loss and make strategies for data backup. This paper analyzed the potential failure modes and failure mechanisms influencing on HDD reliability by FMMEA (Failure Modes, Mechanisms and Effects Analysis) method and performed the prioritization by estimating the risk priority numbers. The Head Disk Interface (HDI) and head stack assembly related failure and relevant performance parameters are identified as the dominant failure mode and health monitoring parameters. A novel strategy for anomaly prediction of hard disk based on Mahalanobis distance using SMART attributes is also suggested in this paper. Furthermore, a case study of HDD anomaly prediction based on the methodology presented in this paper is carried out. The experiment results showed that the proposed method is feasible.

Original languageEnglish
Title of host publication2011 Prognostics and System Health Management Conference, PHM-Shenzhen 2011
DOIs
StatePublished - 2011
Externally publishedYes
Event2011 Prognostics and System Health Management Conference, PHM-Shenzhen 2011 - Shenzhen, China
Duration: 24 May 201125 May 2011

Publication series

Name2011 Prognostics and System Health Management Conference, PHM-Shenzhen 2011

Conference

Conference2011 Prognostics and System Health Management Conference, PHM-Shenzhen 2011
Country/TerritoryChina
CityShenzhen
Period24/05/1125/05/11

Keywords

  • Mahalanobis distance
  • SMART
  • anomaly prediction
  • hard disk drive

Fingerprint

Dive into the research topics of 'Health monitoring of hard disk drive based on Mahalanobis distance'. Together they form a unique fingerprint.

Cite this