Research in the past few years has been advocating the use of an on-board health management system in engineering systems used for time-critical, safety-critical, and cost-critical missions. An accurate health management system constantly monitors the performance of the engineering system, and directly aids decision-making regarding the operation of the engineering system. First, it is necessary to perform diagnosis and check whether there are faults in the system. Fault diagnosis typically consists of three steps: fault detection, fault isolation, and fault estimation. After diagnosis, it is necessary to perform prognostics and thereby, predict the future behavior of the system by analyzing possible failure models and degradation models. The most important goal in prognostics is the continuous, online prediction of remaining useful life, and this facilitates decision-making activities such as mission re-planning, fault mitigation, etc.
A major challenge in prognostics and health monitoring is that there are several sources of uncertainty that affect the performance of both the engineering system and the health management system. For example, the loading conditions and operating conditions of the engineering system may be uncertain. Mathematical models are built to emulate the time-dependent behavior of such engineering systems, and these models are used by the health management system for state estimation and prediction. It may not be possible to accurately model practical engineering applications, and therefore, such models may impart additional uncertainty to diagnosis and prognosis. The sensors, which are part of the health management system, may not be accurate due to measurement errors, and this may prevent accurate estimation of the system health state. As the result of the presence of such sources of uncertainty, it is important to account for these sources of uncertainty during diagnosis, prognosis, and decision-making. While the topic of uncertainty quantification in diagnosis has gained attention in literature, the importance of uncertainty significantly increases in the context of prognosis, since the focus is on predicting future behavior, which is far more challenging and uncertain than fault diagnosis.
The overall objective of this project is to develop a computational framework for uncertainty quantification and management in prognostics, that can help (1) in identifying the various sources of uncertainty that affect prognostics; (2) systematically quantify the combined of the different sources of uncertainty on prognostics and estimate the uncertainty in prognostics and Remaining Useful Life (RUL) prediction; and (3) provide important and useful information for decision-making activities in the context of prognostics and health management.