By Richard Rusaw, Electric Power Research Institute, & Mohammed Yousuf, Exelon Nuclear
Nuclear plants are increasingly turning to on-line monitoring to improve plant productivity, reliability, and asset management. Exelon has been an early adopter, and is applying fleetwide online monitoring technology developed by the Electric Power Research Institute (EPRI) across its 17 nuclear units to obtain real-time information on component condition, evaluate and manage risk, optimize maintenance, and reduce costs. The program continuously monitors plant equipment and identifies low-level anomalies at a very early stage.
Monitoring equipment health provides a wealth of information to support business decisions. Equipment condition data can help avoid unplanned or corrective maintenance of power plant components, which is one of the most significant expenses for power plants in terms of parts and labor and replacement electricity costs. Online equipment condition monitoring can provide early warning of potential failure by detecting incipient indicators of equipment degradation via advanced pattern recognition and signal processing technologies. Online monitoring (OLM) also can help engineers estimate the remaining useful life of critical plant components that are expensive and difficult to replace.
EPRI has been developing online monitoring technology since the late 1980s. In particular, EPRI adapted advanced pattern recognition technology that was in development at national laboratories for use in nuclear power plants. In addition, EPRI supported the development of the technical infrastructure for online monitoring and worked with early adopter utilities including Exelon to apply the infrastructure and technology in a series of pilot projects.
The advanced on-line monitoring technology developed by EPRI incorporates pattern recognition software, digital measurement and signal processing, and advanced modeling and analytical methods. The numerical models employed in pattern recognition software can automatically detect much smaller changes than possible with data trending, and are amenable to higher levels of automation and integration into advanced information management systems, thereby supporting modern business management programs.
Advanced digital measurement and signal processing techniques are important because the large amounts of data from plant equipment do not necessarily provide useful actionable information. Preprocessing technologies, for example, can condense unfiltered data into organized data sets for efficient use in analytical programs. And because monitoring programs do not have inherent intelligence about the behavior of monitored equipment or systems, advanced analytical methods help to optimize data requirements and supporting measurements, and to interface with the operating plant and plant personnel.
Although online monitoring is used extensively and effectively in other industries, it is still in the early implementation stage in the nuclear power sector. To support broader application, EPRI has published guideline reports incorporating operating experience and lessons learned from success¬ful implementations over the past 20 years. For example, Guideline for On-Line Monitoring of Nuclear Power Plant Instrument Channel Performance (1022988) and Requirements for On-Line Moni¬toring in Nuclear Power Plants (1016725) provide the necessary guidance and requirements for plant implementation of online monitor¬ing in both the existing fleet and the next generation of nuclear power plants.
Using these guidelines and building on EPRI’s technology development efforts, Exelon implemented a decentralized fleetwide online monitoring system. The system monitors equipment health via an array of instrumentation and control sensors that measure process parameters including temperature, pressure, level, flow, vibration, and neutron flux to verify proper sensor operation or to identify differences from expected behavior that may indicate sensor degradation, process anomalies, or equipment problems. A server runs the advanced pattern recognition and other software tools to analyze the sensor data and evaluate equipment health and plant performance. Results are provided to respective systems engineers at each plant for assessment, including optional email or pager notifications when monitored components deviate from specified normal operating conditions.
Exelon’s monitoring program supports the company’s “Fundamentals of Intolerance for Unexpected Equipment Failure.” The OLM system identifies anomalies at low levels and helps plant staff either mitigate the deviating condition in timely manner, or, if the deviating condition cannot be corrected immediately, closely monitor the trend and more effectively plan the corrective work. Exelon decided to automate the OLM system to achieve a higher level of performance.
Exelon has developed a unique notifications framework linked to the monitoring scheme that automatically informs responsible plant staff of deviating plant conditions. A tiered alarming scheme is built into the program to track the degradation of real-time values. Each trend in the program is equipped with “Hi Residual,” “Hi Warning,” “Hi Alarm,” “Hi Out of Range,” and “Low Warning” alarms. The program improves productivity of plant staff and supports systems engineering, operations, and maintenance by focusing attention on areas of greatest concern.
The OLM system at Exelon is a fleetwide application, currently monitoring more than 1,400 assets. The monitoring encompasses all major safety-related components and power production assets across the fleet. The next phase of the project will cover all non-safety related components supporting safety functions. There are approximately 15,000 sensors programmed into the OLM application.
The OLM program trends equipment condition by monitoring the rate of change of degradation through a unique algorithm incorporated into the software. The algorithm determines the “residual” value for a particular piece of equipment based on real-time equipment performance and a reference value built into the program. The residual is the difference between the expected values generated by the software and the measured values from the plant. In normal operation, the residual should be a near-zero, stable signal. Changes to the statistical behavior of this normally well-behaved signal are easily detected and can indicate an unexpected or fault condition that warrants investigation [Clarkson, S. A, and R .L. Bickford, Path Classification and Remaining Life Estimation for Systems having Complex Modes of Failure, MFPT 2013, Cleveland, OH, May 2013.] Figure 1 illustrates the residual concept.
Figure 1. Residual-based monitoring enables improved fault detection and prognostics. The residual is the difference between the observed signal and the predicted signal.
Early Detection Examples
The OLM program has successfully identified several plant anomalies at an incipient stage and helped Exelon staff take timely mitigating actions to ensure safe plant operations. Several examples are described below.
Steam Generator Level Control Card Failure
The OLM program identified a failed circuit card in the feedwater regulating valve controls while the valve was controlling the steam generator level in automatic mode. When the circuit card failed, the steam generator started to rise. Although the rate of change in level was small, the OLM system was sensitive enough to detect that the residual exceeded the pre-set alarm set point of Hi Warning. In response, the OLM system informed plant staff of a changing plant condition.
The early detection helped the plant staff to switch steam generator level control from automatic to manual until the card was replaced. Notably, the anomaly was detected at an incipient stage that could not have been detected by human eyes until at reaching a more mature stage. The OLM program, therefore, prevented a plant transient and helped plant staff to take timely corrective actions.
Figure 2. This graphic shows how the OLM system detected an out of range condition in the steam generator level control assembly. The residual value is shown in green, the Hi-Warning setpoint in yellow, the Hi-Alarm setpoint in red, and the Hi-Out of Range setpoint in purple.
Condensate Pump Bearing Temperature
The OLM program identified a rising pump bearing temperature at an incipient stage. The early detection alarm was initiated automatically by the program when the bearing real-time value (green trend line) exceeded the Hi-Warning (yellow line) setpoint. The early detection enabled plant staff to perform a flow balance of the turbine building closed cooling water system. The bearing temperature dropped back to pre-anomaly conditions as soon as the bearing cooling flow was adjusted. The early detection helped prevent degradation of the pump bearing, which could have resulted in a plant transient.
Figure 3. This graphic shows how the OLM system detected a high alarm condition in the condensate pump bearing temperature. The residual value is shown in green, the Hi-Warning setpoint in yellow, the Hi-Alarm setpoint in red, and the Hi-Out of Range setpoint in purple.
Circulating Water Motor Stator Temp
The OLM program successfully identified temperature degradation in the circulating water motor stator winding. The automatic alarm was initiated when the rate of change of winding temperature caused the residual (green trend line) to exceed the preset Hi-Warning setpoint. The degradation was identified at an incipient stage, helping plant staff take timely corrective actions and plan the motor replacement work according to plant schedule. The OLM helped prevent a plant derate and protect the work planning cycle.
Figure 4. This graphic shows how the OLM system detected an out of range condition in the temperature of a circulating water motor stator winding. The residual value is shown in green, the Hi-Warning setpoint in yellow, the Hi-Alarm setpoint in red, and the Hi-Out of Range setpoint in purple.
Nuclear Power Internaional Issue Archives
View Power Generation Articles on PennEnergy.com