Identifying Suitable Degradation Parameters for Individual-Based Prognostics

Introduction

Unforeseen equipment failure is costly, both in terms of equipment repair costs and lost revenue. Discovery of unanticipated pressure vessel head degradation at the Davis-Besse nuclear plant led to a 25-month outage and estimated repair costs exceeding $600 million (Union of Concerned Scientists, 2009). In September, 2008, a turbine generator malfunction at the D.C. Cook nuclear plant resulted in a fire which led to eventual manual plant shutdown. Turbine repairs totaled $332 million, in addition to lost revenue during the one-year outage (World Nuclear News, 2008). Enterprise server downtime can be even more costly, resulting in a possible loss of $6.4 million per hour for brokerage operations and $2.6 million per hour for credit card authorization services (Feng, 2003). Traditional maintenance strategies fall into one of four categories: corrective, preventive, proactive, and condition-based. Corrective maintenance occurs only when equipment fails or malfunctions, such as replacing car tires only when they will no longer hold air. While this maintenance strategy avoids any unnecessary maintenance by only repairing components or systems which have already failed, it also reduces equipment uptime, resulting in lost revenue. Alternatively, preventive, or periodic, maintenance occurs on a time-based schedule, such as replacing car tires after 30,000 miles of use. It is completed every cycle, regardless of need, and is intended to occur frequently enough to preclude any failures from occurring in service. Clearly, this maintenance strategy can result in a significant amount of unnecessary maintenance but reduces the occurrence of failure. Often, preventive maintenance regimes are less costly than corrective due to the added cost of failure during operation and associated damage to other components; a tire blow out while driving may require towing or replacement of damaged wheels. A third maintenance strategy, proactive maintenance, attempts to remove failure modes from a system by performing ongoing maintenance to reduce the probability of the fault or redesigning the system to remove the fault. Maintaining proper alignment in a car to reduce the probability of uneven tire wear is a form of proactive maintenance intended to increase the lifetime of the tires. However, sometimes this is not feasible due to design or cost limitations. Condition-based maintenance can provide a more elegant and cost-effective maintenance strategy. Condition-based maintenance involves monitoring equipment health and performing maintenance actions on an as needed basis. Condition-based maintenance can be facilitated by a health monitoring system. Health monitoring systems commonly employ several modules that perform specific functions, including but not limited to: system monitoring, fault detection, fault diagnostics, prognostics, and operations and maintenance planning. System monitoring and fault detection modules are used to determine if a component or system is operating in a nominal and expected way. If a fault or anomaly is detected by the monitoring system, the diagnostic system determines the type and, in some cases, the severity of the fault. The prognostics module uses all the available information—including sensed system measurements, monitoring system residuals, and fault detection and diagnostic results—to estimate the remaining useful life (RUL) of the system or component along with associated confidence bounds. With this information in hand, a system could be shut down for maintenance or system operation may be adjusted to mitigate the effects of failure or to slow the progression of failure, thereby extending the equipment life until it is more convenient to perform maintenance.