Layered Classification for Improved Diagnostic Isolation in Drivetrain Components

Smith, Mike; Byington, Carl S.; Kalgren, P.W.; Parulekar, Ashish; DeChrisopher, M.

doi:10.1109/aero.2006.1656097

Cited by 4 publications

(1 citation statement)

References 1 publication

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The solution they devised is by data fusion to amplify the S/N ratio. Smith & Byington (2006) applied principal component analysis (PCA) to obtain a layered classification of the condition indicators (CI) with a condensed feature set.…”

Section: A Brief Literature Reviewmentioning

confidence: 99%

A new life system approach to the Prognostic and Health Management (PHM) with survival analysis, dynamic hybrid fault models, evolutionary game theory, and three-layer survivability analysis

2009

2009 IEEE Aerospace Conference

View full text Add to dashboard Cite

In this paper, I propose a new architecture for PHM, which is characterized by life-system approachtreating PHM as a hierarchical system with fundamental properties similar to those of life systems. Conceptually, besides drawing on the important concepts from existing PHM theory and practice such as life cycle, condition-based maintenance (CBM), remaining useful lifetime (RUL), I draw on the dynamic hybrid fault models (DHF) from fault tolerance theory and agreement algorithms, three-layer survivability analysis from survivable network systems (SNS), population dynamics from population ecology, and survival analysis from biostatistics and biomedicine. Methodologically, three main mathematical tools: survival analysis (including competing risks analysis and multivariate survival analysis), dynamic hybrid fault models and evolutionary game theory (EGT) are applied for PHM modeling and analysis. Operationally, the three-layer survivability analysis is applied to deal with the so-called UUUR(Unpredictable, latent, Unobserved or Unobservable Risks) events and to achieve sound decision-making. Overall, the advantages of the new architecture include: (1) Offer a flexible architecture that is not only compatible with existing components/approaches of PHM, such as lifetime, reliability, maintainability, safety, data-driven prognostics, and model-based prognostics, but also readily extendable to incorporate fault tolerance, survivability, and security. (2) Utilize survival analysis, competing risks analysis, and multivariate survival analysis for better modeling of lifetime and reliability at both individual and population (group of components) levels.(3) Approach the fault tolerance and reliability of the monitoring sensor network in PHM and those of the underlying physical system with the DHF models. (4) Analyze the system survivability (sustainability) with the three-layer survivability analysis approaches. (5) Capture security events with UUUR events and incorporate security policies into PHM.

show abstract

Section: A Brief Literature Reviewmentioning

confidence: 99%