Sachin Kumar
Prognostics and Health Management Group
Center for Advanced Life Cycle Engineering (CALCE)
University of Maryland
College Park, MD 20742, USA
and
The DEI Group
Millersville, MD 21108, USA
Michael Pecht
Prognostics and Health Management Group
Center for Advanced Life Cycle Engineering (CALCE)
University of Maryland
College Park, MD 20742, USA
and
Prognostics Health Management Center
City University of Hong Kong
Abstract:
Prognostics and Health Management is an enabling technology with the potential to solve complex reliability problems that are due to complexity in design, manufacturing, and maintenance. There are several different mathematical techniques that can assist in performing prognostics and health management of electronic systems. These techniques can be categorized into statistical reliability, life cycle loads, state estimation, and feature-extraction based models. The selection of the appropriate model depends on the application environment. This paper presents a methodology for selecting the correct model to perform diagnostics and prognostics in electronic systems based on a user’s application environment. The model selection method is based on five properties, including usability, accuracy, performance, applicability at the system level, and flexibility of the model. Based on all this information, a comparison is made between the five prognostic models to show the advantages and disadvantages of each. Finally, recommendations are given for selecting the most appropriate model for system fault diagnostics and prognostics of electronics. While this methodology used in this study for analysis of electronic systems, it can be extended to other applications as well.
Keywords: Electronics, health monitoring, diagnostics, prognostics, data-driven, physicsof- failure (PoF)
Click here for complete article.
This article was originally published in the International Journal of Performability Engineering (IJPE) (www.ijpe-online.com)