Prognostics in Battery Health Management

Prognostics and Health Management (PHM) has seen a resurgence recently with new service offerings in industry for guaranteed uptime and with military requirements asking for cost-containing condition-based maintenance (CBM) implementations. A chief component of PHM is prognostics, which is also its least mature element. Prognostics attempts to estimate remaining component life, given that an abnormal condition has been detected. Key to useful prognostics is not only an accurate remaining life estimate, but also an assessment of the estimate’s confidence. The latter is often times expressed through a probability density function that envelopes the prediction, by allowing the computation of confidence bounds around it. It is the uncertainty estimate that poses particular challenges to the prediction since it must account for various sources stemming from measurements, state estimation, model inaccuracies, future load uncertainty, etc. In this article, we examine these issues using battery health management as a test case. Batteries form a core component of many machines and are often times critical to the well being and functional capabilities of the overall system. Failure of a battery could lead to reduced performance, operational impairment and even catastrophic failure, especially in aerospace systems. A case in point is NASA’s Mars Global Surveyor which stopped operating in November 2006. Preliminary investigations revealed that the spacecraft was commanded to go into a safe mode, after which the radiator for the batteries was oriented towards the sun. This increased the