Knowledge is power
Nuclear plants use online monitoring for critical machinery, in-depth analysis
By Bruno Gauthier
For more than 30 years, Electricité De France (EDF) has been operating electrical power plants in France to increase the safety and availability of traditional and nuclear power plants. The company installed monitoring systems on primary circuits and turbo-generators of nuclear power plants to detect and investigate possible incidents, in addition to automated plant control.
It is now important to take into account how to accurately monitor the behavior of critical components. The company needed to archive events related to these components and their behavior to continuously evaluate their future ability to operate, and to schedule maintenance operations or replacements.
EDF has thus acquired a large knowledge base describing the behavior of the main elements of a nuclear power plant: reactor coolant pumps, turbines, generators, inlet valves, internal structures, and reactors.
Poste de Surveillance et d’Aide au Diagnostic (PSAD) is an online monitoring system designed to keep maintenance personnel informed of the working conditions of equipment through the in-depth analysis of data collected about its operating state.
The system’s main objectives are to monitor a power plant’s critical machinery and detect early any atypical operating conditions. It allows the operator to perform an in-depth analysis of detected problems and to help elaborate a precise diagnosis through the use of sophisticated graphing tools.
One specific advantage of the system is its capacity to monitor a fleet of nuclear power plants. It provides a maintenance knowledge database that can compare equipment behavior in different nuclear power plants and support feedback loop applied to the fleet. This is especially important for aging plants in managing equipment related to the extended lifetime of nuclear installations.
User-friendly analysis tools and user interface help facilitate the work of maintenance operators and remote experts. In addition to generic graphic tools adapted to vibration monitoring, the system provides dedicated tools for critical equipment monitoring in a nuclear power plant.
Condition-based maintenance stakes
The strategy of maintaining equipment for corrective maintenance when a failure occurs may result in plant unavailability, severe damages to the equipment itself, excessive maintenance costs, and safety and environmental risks.
Another concerning strategy, scheduled preventive maintenance, avoids concerns about corrective maintenance, but you can replace parts of equipment although they can be still available for some time. You must periodically inspect equipment, and these operations generate maintenance costs and loss of equipment availability.
Condition-based maintenance means monitoring equipment behavior, detecting anomalies or conditions of a future failure, and deciding the right time to intervene and maintain.
This strategy has economic and safety benefits:
It reduces the number of maintenance interventions and the number of inspections.
You can reduce radiation exposure of inspection personnel.
You can detect faults early and avoid damages.
You can better repair and maintain operations, provide spare components in due time, and then reduce spare parts inventory.
It is possible to build a knowledge database related to the working conditions of equipment, components, and equipment aging.
Finally, condition-based maintenance helps reduce maintenance costs, as well as increase equipment lifetime, equipment reliability, and equipment availability. Meanwhile it reduces the risk of failure and accidents.
Condition-based maintenance can be a solution for monitoring the evolutions of equipment’s working conditions, which can be related to future faults. The progress in data processing and computer technologies makes it possible. Methods and systems available today can see use for online condition monitoring with reasonable efforts and costs.
But technology alone is not the best response to implement this strategy. Condition-based maintenance requires highly skilled people in an organization to identify the cause of anomalies and to predict their evolution and modify working conditions of equipment or to manage maintenance operation.
Critical equipment monitoring
EDF uses the online monitoring system for critical equipment in nuclear power plants, which need continuous monitoring of working conditions. For other equipment, it is possible to use lighter solutions, such as portable data collectors and an off-line vibration analyzer.
The monitoring system actually allows operators to detect the following defects on critical equipment:
Primary coolant loop
Reactor coolant pumps—unbalance, seal degradation, bearing looseness, shaft cracks, frictions
Loose parts detection—localization, impacts intensity identification
Internal reactor structures—reactor vessel oscillation, core barrel vibrations, support rupture on thermal shield
Turbine—frictions, unbalance, thermal distortion, misalignment, cracks
Generator—rotor ventilation, vibrations, short-circuit, failure of the stator coolant system
Steam inlet valves on the turbine—valves jamming, opening and closing time
The first function of the monitoring system is to characterize working conditions and equipment behavior. Using data measured at a machine or process level, extracted features compare with reference features, which experts interpret. To get continuous measurement systems for critical equipment in nuclear power plants, depending on the types of equipment and the types of phenomena to interpret, we can measure vibration signals, acoustic signals, neutron flux, and process data such as temperature, pressure, flow, rotation speed, position, state, and delay.
Vibration signals acquisition measurements come from vibration sensors, such as accelerometers and velocity probes. With the monitoring system, they can see use for rotating machinery-like pumps, turbines, and generators. In nuclear power plants other than those from EDF, these kinds of sensors could also see use on the primary circuit for structure vibration measurement.
Spectral analysis, harmonics analysis, algorithms based on root mean square (RMS), or peak value of a vibration allow the identification of anomalies in the behavior of the machine related to a component, such as a shaft, bearing, or seal.
Acoustic signals acquisition measurements come from accelerometer sensors mounted on the coolant primary loop, especially on the reactor vessel and the steam generators. These sensors are working in the audible frequency range and help detect and locate loose parts in the primary circuits. The principle is to detect burst-type signals against background noise. For this function, PSAD manages alert thresholds. In case of threshold detection, the monitoring unit generates acoustic records sent to the diagnosis workstation for storing and expert analysis.
Neutron flux measurements concern the variation of neutron flux issued from reactor vessel at specific points. They allow you to detect reactor vessel oscillations, barrel vibrations, and fuel assembly vibrations.
Process data acquisition
Some process data, such as speed rotation for pumps or turbines, see use in correlation with vibration measurements in digital signal algorithms. The online monitoring system manages other process measurements to detect some equipment failures that are not correlated to vibration, such as steam inlet valves on the turbine. We can compare position and delay measurements with reference values and allow the detection of jamming defects or abnormally long stroke times. Take generators as another example—temperature and pressure measurements allow you to monitor thermal conditions of a generator.
Finally, a function called perturbography samples process data after some predefined conditions. You can define these conditions through one operating state or a combination of several operating states of a machine. When the condition occurs, the monitoring unit samples one or several process measures at a configurable frequency and duration. Records go to a diagnosis workstation for storage and analysis. EDF uses this function for monitoring some components related to the working conditions of the turbine and the generator during transition phases such as run-up and run-down.
Data calculation, monitoring functions
A descriptor defines all process data in the monitoring system. It is an abstract entity that represents one piece of information related to the monitoring task, usually some measurement of machine behavior.
The monitoring system can measure or compute descriptors at three different levels:
Level 1: Descriptors received directly from the acquisition boards; the measurement and its format are determined by the type of board that generates the descriptor.
Flow rate, temperature, pressure, etc., from an analog-to-digital conversion board
RMS or peak value from a dynamic input board
Harmonic amplitudes and tachometer from a synchronous analysis board
Spectra from a spectral analysis board
Level 2: Descriptors as real-time results of level 2 data processing functions the monitoring unit evaluates using level 1 and/or level 2 descriptors in its real-time database:
Arithmetic function of level 1 descriptors
Sliding average of a level 1 descriptor
Mean and sigma value of level 1 descriptors
Level 3: Descriptors resulting from level 3 data processing functions and computed from descriptors of any level and/or external state variables. Unlike the other two types of real-time descriptors, these are computed off-line by the diagnosis workstation using measurements stored in the local database. Level 3 descriptors describe long-term behavior such as:
Resonance frequency damping
The progression of an average vibration vector
Long-term average and standard deviation of a harmonic
Warning functions can generate alert and warning messages about operating faults at two levels:
The monitoring unit in real time (A temperature descriptor, level 1, exceeds some upper bound. A shock factor, level 2, reaches its threshold.)
The diagnosis workstation, using archived data from the local database, such as long-term progression of a harmonic descriptor, or level 3.
A monitoring system configuration is based on the definition of these descriptors and the data processing functions selected from the system’s library. The configuration of a monitoring function consists of the following operations:
Selecting the type of each sensor and its parameters (sensitivity, offset, physical units)
Selecting the type of signal processing for each sensor input (level 1 descriptor) and the processing parameters (gain, pass-band, harmonic, filter)
Configuring additional data processing (level 2 or 3 descriptors) with algorithm function library
Configuring automatic fault detection (from the warning function library) and associated parameters (thresholds, time spans)
Analysis, diagnosis tools
An important requirement for this kind of system is the user interface ergonomic characteristics. The monitoring system allows facilities to efficiently access information. A warning frame located on the bottom of all screens of a diagnosis workstation gives operators a general overview of an installation. This frame gives a concise evaluation of the state of all monitored equipment based on the existence and severity of any warning messages. In case of a defect on equipment, alarm logbook indicates the details and the chronology of events.
The working space and context definition facility allows operators to rapidly restore a previous analysis work. This facility is based on the working-set concept. A working set can define a collection of sensors, data, and time periods to examine. Then various users can save work they define and recall it again later whenever they need to repeat the same analysis for a same or different time frame.
All data declared in the system has a specification sheet that describes the functions and data used to process it, such as sensor parameters, acquisition board configuration, level 2 data processing functions, and warning functions. You can display this information in a frame by clicking on the descriptor’s identifier, wherever it appears on screen—in a warning message, on a graph display, or elsewhere.
Graphic tools can help users plot any variable on different scales, either linear or logarithmic, to display information as Nyquist diagram, 3-D perspective, and waterfall diagram. There are also specific tools dedicated to nuclear power plant domain, facilitating the analysis of some equipment working conditions:
Perturbography data display
Valves operations monitoring
Machine slow-down analysis
Natural frequencies identification
Machine behavior report
Acoustic signal listening for loose parts detection
Following a detected fault or for a specific case study, operators and experts use data graphing tools and dedicated diagnosis tools to diagnose the exact physical condition of the machines concerned. You can process analysis on site or remotely. This remote analysis capability helps users compare different sites equipped with the same component and helps them gather the best specialists of a technical domain. Experts at EDF can remotely access information of a defined power plant while keeping an eye on the average case, collaborating to the best practices and increasing their knowledge.
Monitoring units (MU) provide data acquisition, digital signal processing, first level of data processing, anomalies detection, and data compression.
Local station (LS) provides data processing, local storage (short-term database), and data export to the public server.
Publication server provides long-term data archiving and secured access to database-to-remote expert users.
Remote station and site station are diagnosis workstations dedicated to experts—in-depth and post-incident analysis and assistance to onsite operators.
ABOUT THE AUTHOR
Bruno Gauthier is I&C project manager at the Energy & Utilities center of Atos Origin, an international information technology services company specializing in systems integration and managed operations out of Meylan, France.
Return to Previous Page