NASA is developing increasingly autonomous systems that can perform missions with a high degree of certainty with minimal human intervention. Examples of such mission include rovers operating in Mars, where the missions are extremely long, and therefore multiple components and subsystems will degrade and fail over the duration of the mission. However, due to the long communication delays between Mars and Earth, these systems cannot be monitored and diagnosed by mission control like any other near-earth mission. The proposed capability will be invaluable to NASA for such operations by (a) Predicting failures before they disrupt the mission, (b) Reducing false positives of such prediction with the proposed diagnosis-driven prognosis, and (c) identifying the remaining useful capability of the system. This will enable NASA to focus on the mission planning and recovery aspects, and manage the health of the system, rather than being blindsided by unexpected failures.
The potential applications for DoD and Commercial users are even larger. This is because they are likely to operate multiple systems, a fleet of vehicles for example, that have the opportunity of periodic preemptive maintenance. To address these customers' needs, we will develop a decision-support module on top of the proposed capability here that will allow the customer to define his own business rules. Such business rules will help the customer answer questions like "if the system has a scheduled downtime window of 2 hours tomorrow, what pre-emptive repairs should I perform within that maintenance window so as to minimize the chance of unscheduled downtime (due to failure) in the next 10 days". For enterprise-wide logistic planning, this decision-making capability will also help optimize the cost of additional opportunistic maintenance versus the cost of additional downtime if such maintenance were not performed. The capability developed here is key to proving the business case for prognostics in commercial and military applications.