Prognostics and Health Management of Electronics

7 Machine Learning: Diagnostics and Prognostics

Myeongsu Kang

University of Maryland, Center for Advanced Life Cycle Engineering, College Park, MD, USA

Prognostics and health management (PHM) has emerged as an essential approach for preventing catastrophic failure and increasing system availability by reducing downtime, extending maintenance cycles, executing time repair actions, and lowering life‐cycle costs. PHM helps companies to reduce inspection and repair costs and thus achieve competitive advantages in the global market by improving system reliability, maintainability, safety, and affordability [1]. Along with sensing and anomaly detection, diagnosis and prognosis are two of the four key elements of PHM; diagnosis is used to detect a fault/failure or degradation condition and identify the nature of the fault/failure (e.g. failure modes and/or mechanisms) or type of degradation, whereas prognosis is used to assess the future state of system health and integrate that picture of system health within a framework of available recourses and operation demand. This chapter provides a basic understanding of data‐driven diagnostics and prognostics, reviews recent advancements of diagnosis and prognosis techniques with a focus on their applications in practice, and discusses research opportunities that can lead to further improvement of PHM in both theory and practice.

7.1 Overview of Diagnosis and Prognosis

For those systems that are too complicated to have an explicit ...