lable at ScienceDirect

Progress in Nuclear Energy 53 (2011) 167e181

Contents lists avai

Progress in Nuclear Energy

journal homepage: www.elsevier .com/locate/pnucene

Review

On-line monitoring applications in nuclear power plants

H.M. HashemianAMS Corporation, AMS Technology Center, 9119 Cross Park Drive, Knoxville, TN 37923, USA

a r t i c l e i n f o

Article history:Received 24 February 2010Received in revised form29 July 2010Accepted 12 August 2010

Keywords:Nuclear power plantsNoise analysisSensor response-time testingSensing-line blockagesCalibration monitoringReactor diagnostics

E-mail address: hash@ams-corp.com.

0149-1970/$ e see front matter � 2010 Elsevier Ltd.doi:10.1016/j.pnucene.2010.08.003

a b s t r a c t

The nuclear power industry is working to reduce generation costs by adopting condition-based main-tenance strategies and automating testing activities. These developments have stimulated great interestin on-line monitoring (OLM) technologies and new diagnostic and prognostic methods to anticipate,identify, and resolve equipment and process problems and ensure plant safety, efficiency, and immunityto accidents. This paper provides examples of these technologies with particular emphasis on eight keyOLM applications: detecting sensing-line blockages, testing the response time of pressure transmitters,monitoring the calibration of pressure transmitters on-line, cross-calibrating temperature sensors in situ,assessing equipment condition, performing predictive maintenance of reactor internals, monitoring fluidflow, and extending the life of neutron detectors. These applications are discussed in the followingsections. Emphasis is placed on the principles of a core OLM method e noise analysis e and the technicalrequirements for an integrated OLM system are summarized.

� 2010 Elsevier Ltd. All rights reserved.

1. Introduction

The 2008 report of the U.S. Energy Information Administration(EIA) shows that the worldwide contribution of nuclear power isonly about 13 percent, fourth behind coal at 43 percent, natural gasat 22 percent, and renewables including hydro power at 18 percent.However, efficiency improvements, power uprates, new plantconstructions, and the gradual phasedown of fossil energy over thenext decade are all expected to increase the nuclear share ofworldwide electricity generation. Driving this trend is the growingperception that nuclear power is an environmentally friendly andlow-cost source of base-load electricity.

About 85 percent of the cost of electricity generation from fossilsources (excluding capital costs) is in the fuel (77 percent for coal,93 percent for natural gas), while operations and maintenance(O&M) costs account for only about 15 percent (23 percent for coal,7 percent for natural gas). In contrast, nuclear power generationcosts are only about 27 percent in fuel and 73 percent in O&M. Sincefuel costs are normally either constant or increase with time, thereis little room for fossil power plants to reduce generation costs,while nuclear power plants can easily reduce generation costs bysimply reducing O&M costs.

To reduce O&M costs, the nuclear industry has taken advantageof computer technologies to automate much of its testing and

All rights reserved.

maintenance activities. In particular, the industry has begun totransition from traditional time-directed, hands-on, and reactivemaintenance procedures to condition-based, risk-informed, andautomated maintenance strategies. This is partly because thecurrent generation of nuclear power plants has passed its mid-life,and increased monitoring of plant health is critical to theircontinued safe operation. This is especially true now that licenserenewal of nuclear power plants has accelerated, allowing someplants to operate up to 60 years or more. Furthermore, many util-ities are maximizing their power output through uprating projectsand retrofits. This puts additional demand and more stress on theplant equipment such as the instrumentation and control (I&C)systems and the reactor internal components making them morevulnerable to the effects of aging, degradation, and failure.

The on-line condition monitoring (OLM) technologies describedin this paper will meet several key needs of the nuclear industry,such as detecting sensing-line blockages, testing the response timeof pressure transmitters, monitoring the calibration of pressuretransmitters on-line, cross-calibrating temperature sensors in situ,assessing equipment condition, performing predictive mainte-nance of reactor internals, monitoring fluid flow, and extending thelife of neutron detectors. Continuous plant assessment throughOLM will enable plants to identify a degrading instrument orprocess in real time, rather than wait until the degradation causesa loss of function or proceeds to a failure. As a result, plants canfocus maintenance activities where they are most needed, planttrips can be reduced, maintenance schedules can be optimized, andplant outages can be shortened.

Table 1Estimated gains per operating cycle from implementation of an on-line monitoringsystem at the Sizewell B plant in the United Kingdom.

OLM

Applications Already Implemented Gains (U.S. Dollar)

Cross Calibration of Safety-Related Temperature Sensors $5,800,000.00On-Line Calibration Monitoring of Pressure,

Level, and Flow Transmitters$15,700,000.00

Total $21,500,000.00Planned OLMApplications (and MWe Gains)Reduction of Feed Water Flow Uncertainty (36 MWe) $24,000,000.00Improved Secondary Calorimetric Accuracy (6 MWe) $ 4,050,000.00Feed Water Temperature Accuracy

Improvements (5.25 MWe)$ 3,450,000.00

Cooling Water Optimization (4.5 MWe) $ 3,000,000.00Optimized Steam Drainage (3 MWe) $ 1,950,000.00Total $36,450,000.00Sum of Realized and Expected Gains $57,950,000.00

H.M. Hashemian / Progress in Nuclear Energy 53 (2011) 167e181168

Recently, British Energy (BE) performed a cost-benefit evalua-tion for an OLM system now implemented at the Sizewell B plant (a1200MWeWestinghouse PWR) in the United Kingdom. The resultsof BE’s calculation are presented in Table 1 in terms of areas ofpotential gains and the values of these gains in megawatt electricoutput (MWe) and/or the corresponding dollar amounts (Lillis,2008). According to BE calculations, the estimated total savingsper operating cycle would be about $58 million when the OLMtechnologies planned by BE are added to the implementationalready operating at the Sizewell B plant. This compares quitefavorably to the $5 million that has been estimated as the cost fora complete deployment of OLM at this plant.

Table 2 compares the benefits e direct and indirect (e.g.,reduced maintenance activity) e of maintenance performed withthe aid of OLM versus current I&C maintenance practices.

Table 2Benefits of on-line monitoring.

Conventional Maintenance Practice Maintenance by On-lineMonitoring

� Performed manually � Automated� Requires access to equipment � Performed remotely

and hands-off� Performed occasionally � Performed continuously� Detects problems after they have occurred � Detects problems

as they occur� Some maintenance can be

performed only when the plant is at shutdown� Most maintenance canbe performed duringplant operation

� Environmental and process conditioneffects are not typically included

� Environmental andprocess condition effectsare included

� Process monitoring is not possible � Process monitoringis possible

Indirect Benefits� Optimization of maintenance tasks� Reduced outage time� Labor cost savings (I&C technicians, operations,

utility support and supervisors, qualityassurance/quality control (QA/QC) personnel,health physics (HP) personnel,administrative personnel

� ALARA savings (as low asreasonably achievable)

� Trip reduction� Reduced potential for damage to equipment� Other benefits (e.g. cost and time of dress-out

to enter radiation controlled zone andlow-level waste (LLW) cost reduction

2. OLM fundamentals

The term on-line monitoring (OLM) describes methods, such asthe noise analysis technique, for evaluating the health and reli-ability of nuclear plant sensors, processes, and equipment fromdata acquired while the plant is operating. Although OLM tech-nologies typically apply to all types of nuclear power reactors, thispaper uses pressurized water reactors (PWRs) as the referenceplant since they are the type most commonly used in the Westernhemisphere. To control the PWR plant and protect its safety,several different kinds of sensors are employed to measure theprocess parameters (see Table 3). Fig. 1 shows a simplified sche-matic of the primary loop of a PWR plant and its importantsensors. Depending on the plant design and manufacturer, a PWRplant has 2e4 primary coolant loops; however, Russian PWRs(called VVERs orWWERs) have up to 6 loops. The normal output ofthese sensors can be used both to establish the health andcondition of the plant and sometimes to verify the performance ofthe sensors themselves.

Fig. 2 shows the output of a process sensor as a function of timeduring plant operation. Normally, while the plant is operating, theoutput of the sensor will have a steady-state value that correspondsto the process parameter indicated by the sensor. This steady-statevalue is often referred to as the static component or DC value. Fig. 2also shows a magnified portion of the sensor’s output signal toillustrate that, in addition to the static component, a small fluctu-ating signal is naturally present on the sensor output. The fluctu-ating signal, which is known as the signal’s dynamic or ACcomponent, derives from inherent fluctuations in the processparameter as a result of turbulence, random flux, random heattransfer, vibration, and other effects. It has long been known thatthe condition of a nuclear power plant can be effectively monitoredby analyzing these small fluctuations in the process variables, suchas reactivity coefficients, vibration amplitudes, and response times,around their stationary value. This technique, commonly known asnoise analysis, noise diagnostics, or reactor diagnostics, makes itpossible to discover the abnormal state of the system, whichregisters either as a shift of these parameters into non-permittedregions or the appearance of a changed structure of the noisesignatures, usually the frequency spectra. The advantage of thenoise analysis technique is that it non-intrusively measures processvariables under normal operation without requiring any externalperturbation.

One idiosyncrasy of the noise analysis technique is that a changein the measured signal characteristics may be caused either bya change in the transfer properties of the system or by a change inthe driving force, that is, the fluctuation of the parameter thatinduces the measured noise. Hence, performing a proper diagnosisrequires sufficient expert knowledge to choose the appropriatemodel on which the diagnostic algorithm is based. The noiseanalysis technique also involves an additional ambiguity: deterio-rating sensor characteristics can change the measured noise

Table 3Typical population of important sensors in pressurized water plants.

Sensor Measurement Typical Populationin a Reactor

RTDsa Temperature 16e60CETsb Temperature 50e100Pressure transmittersc Pressure, level, and flow 500e2500Neutron detectorsd Neutron flux 10e20

a Resistance temperature detectors.b Core-exit thermocouples.c Including differential-pressure transmitters.d Ex-core and some in-core neutron detectors.

Fig. 1. Primary loop of a Pressurized Water Reactor (PWR).

H.M. Hashemian / Progress in Nuclear Energy 53 (2011) 167e181 169

signature. In the Three Mile Island accident in 1979, for example,a role in the accident sequence was played by a failed sensor andthe control room personnel’s inability to realize its failure. Sensormalfunction, or just de-calibration, can also occur under much lessdramatic circumstances, through fouling, drift, response-timedegradation, and aging. As it turns out, noise analysis can be usedeven for sensor health analysis, by differentiating between sensordegradation/failure and system malfunction/anomaly.

Fig. 2. Normal output of a process sensor with illustra

Because the static (DC) and dynamic (AC) components of thesensor output each contain different information about the processbeing measured, they can be used for a wide range of monitoringapplications. For example, applications that monitor for gradualchanges in the process over the fuel cycle, such as sensor calibrationmonitoring, make use of the static component. In contrast, appli-cations that monitor fast-changing events, such as core barrelmotion, use the information in the dynamic component, which

tion of the DC and AC components of the output.

H.M. Hashemian / Progress in Nuclear Energy 53 (2011) 167e181170

provides signal bandwidth information. Fig. 3 illustrates howexisting data from process sensors is used for these applications.Note that in this figure the static data is analyzed using empiricaland physical modeling and averaging techniques involvingmultiplesignals, while dynamic data analysis entails time domain andfrequency domain analysis based on single signals or pairs ofsignals. For example, the dynamic response time of a nuclear plantpressure transmitter is identified by fast Fourier transform (FFT) ofthe noise signal. The FFT yields the auto power spectral density(APSD) of the noise data fromwhich the transmitter response timeis calculated. In applications where pairs of signals are used (e.g.,core barrel vibration measurements), the cross power spectraldensity (CPSD), the phase, and the coherence data are calculated todistinguish the vibration characteristics of various constituents ofthe reactor internal.

The types of OLM applications used in nuclear power plants arein large part determined by the sampling rates available for dataacquisition. Static OLM applications, such as resistance temperaturedetector (RTD) cross-calibration and on-line calibration monitoringof pressure transmitters, typically require sampling rates up to 1 Hz,while dynamic OLM applications such as sensor response-timetesting use data sampled in the 1 kHz range. Other, high-frequencyOLM applications, such as measuring the vibration of rotatingequipment andmonitoring loose parts, may use data sampled at upto 100 kHz. Fig. 4 shows examples of OLM applications that can beused in nuclear power plants, with their range of data samplingfrequency. For low-frequency (DC) data analysis, averaging tech-niques are typically used for redundant sensors, and empirical andphysical modeling techniques are used for non-redundant sensors.

Fig. 3. Online monitoring applications of static and

Because I&C sensors that measure temperature, pressure, level,flow, and neutron flux up to data sampling frequencies of around1 kHz represent the majority of measurement devices in nuclearpower plants, focusing this paper on the OLM applications thatmonitor these sensors will show to best advantage the potentialbenefits of OLM for nuclear plants. Other OLM applications, such asmeasuring the vibration of rotating equipment and monitoringloose parts, which primarily rely on high-frequency acquisition ofdata from accelerometers, are not discussed in this paper becausethey don’t acquire data from the existing process sensors of theplant.

3. Implementing OLM applications in nuclear power plants

In recent years, energy research funding has shifted towardsshort-termandappliedprojects (Montanari, 2008). Forexample, theElectric Power Research Institute (EPRI), the research arm of theEuropean Community (EC), and other organizations are promotingapplied research to globalize the energy sector. These effortscontrast with the classical practice of funding fundamental andlong-term research projects to promote innovations rather thanapplications. Reflecting this trend toward applied uses, the successof the noise analysis technique in nuclear power plant applicationsstimulated the industry to examine the feasibility of implementingan on-line monitoring (OLM) system that incorporates the noiseanalysis technique in both the current and next generation ofnuclear reactors for the purpose of dynamically testing sensors,measuring the vibration of reactor internals, and performinga variety of diagnostic applications. This OLM system will also give

dynamic data analysis described in this paper.

Fig. 4. Online monitoring applications versus sampling frequency.

H.M. Hashemian / Progress in Nuclear Energy 53 (2011) 167e181 171

plants the capability to verify the calibration of pressure, level, andflow transmitters as well as RTDs and thermocouples. The systemwill have built-in signal validation, noise analysis, and OLM algo-rithms that will enable nuclear power plants to check for: (1) cali-bration and response time of process instruments; (2) identifysensing-line blockages; (3)monitor the reactor coolant flow, and (4)alert the reactor operator of excessive vibration of reactor internals.

Such anOLM system canprovide plantswith the information theyneed to evaluate I&C sensors by providing applications that identifydrifting instruments, alert plant personnel of unusual process condi-tions, and predict impending failures of plant equipment. Moreover,operating nuclear power plants can use OLM technologies to improvetheir efficiency. For example, nuclear power plants are required tocalibrate important I&C instruments once every fuel cycle. Thisrequirement dates back 40 years to when commercial nuclear powerplantswerefirst licensed tobeginoperation.Basedoncalibrationdataaccumulated over these fourdecades, it has been determined that thecalibration of some instruments, such as pressure transmitters, doesnot drift enough to warrant calibrating all transmitters as often asonce every fuel cycle. OLM allows calibration efforts to be focused onthe instruments that have drifted out of tolerance, thereby savingplants a significant amount of the time and manpower.

The OLM system described in the following sections is based onexperimental and theoretical research activities performed overthe last 30 years in more than 100 nuclear power plants in theUnited States, Europe, and Asia. This system encompasses eight keyplant monitoring applications: detecting sensing-line blockages,testing the response time of pressure transmitters, monitoring thecalibration of pressure transmitters on-line, cross-calibratingtemperature sensors in situ, assessing equipment condition, per-forming predictive maintenance of reactor internals, monitoringfluid flow, and extending the life of neutron detectors. Theseapplications are discussed in the following sections.

3.1. On-line detection of sensing-line blockages

Chief among applications of noise analysis in nuclear powerplants is detecting sensing-line blockages. Sensing lines (also calledimpulse lines) are small diameter tubes that bring the pressure signalfrom the process to the pressure sensor. Depending on the appli-cation and the type of plant, pressure sensing lines can be as long as300moras short as 10m. The isolationvalves, root valves, andbendsalong their lengthmake themsusceptible toblockages fromresiduesin the reactor coolant, failure of the isolation valves, and otherproblems. Sensing-line blockages are a recurring problem in PWRs,

boiling water reactors (BWRs), and essentially all water-coolednuclear powerplants. Theyare an inherent phenomenon that causesthe sensing lines of nuclear plant pressure transmitters to clog upwith sludge, boron, magnetite, and other contaminants. Typically,nuclear plants purge the important sensing lines with nitrogen orbackfill the lines periodically to clear any blockages. This procedureis, of course, time consuming and radiation intensive, and moreimportantly, not always effective in eliminating blockages.Furthermore, with the exception of noise analysis, no way exists toknowaheadof timewhich sensing linesmaybeblocked. Also, unlessthe noise analysis technique is used, it’s not possible after purging orback filling a sensing line to verify that the line has been cleared.

Fig. 5 shows the cutaway of a partially blocked sensing line ofa nuclear power plant pressure transmitter. It’s clear from the figurethat this blockage can reduce the flow path in this sensing line byabout 40 percent. A blockage like this hampers the dynamic responseof the pressure sensor at the end of the sensing line. In particular,depending on the design characteristics of the pressure transmitter,a sensing-line blockage like this can cause the response time of theaffected pressure transmitter to increase by an order of magnitude.The degree of increase in the dynamic response depends on the“compliance” of the pressure transmitter. Compliance is a pressuretransmitter design parameter that relates to the physical displace-ment of the sensing element of the transmitter per unit of inputpressure. Some transmitters, such as those with sensing elementsmade of “bellows”, have a large compliance and are therefore affectedstrongly by sensing-line blockages. On the other hand, transmitterswith sensing elements made of stiff diaphragms have smallercompliances andare therefore less affectedbysensing-lineblockages.

The effect of compliance on the dynamic response of a pressuretransmitter was revealed in a research project performed by theauthor for theU.S.NuclearRegulatoryCommission (NRC) in theearly1990s (U.S. Nuclear Regulatory Commission, 1993). The goal of theproject was to characterize the effects of normal aging on theperformance of nuclear plant pressure transmitters by illustratingthe effect of compliance on the response time of representativenuclear-grade pressure transmitters from three manufacturers:Barton, Foxboro, andRosemount (see Fig. 6). Thedata in Fig. 6 is fromlaboratory tests measuring the response time of the transmittersusing a pressure ramp signal.

A significantly blocked sensing line can render thepressure sensoressentially useless or even dangerous. The danger here is that, due toa total blockage, the operating pressure may get locked in the trans-mitter and cause its indication to appear normal. Then, when thepressure changes, the transmitter will not respond andwill continue

Fig. 5. Photograph of a nuclear plant sensing line with a partial blockage.

H.M. Hashemian / Progress in Nuclear Energy 53 (2011) 167e181172

to show the locked-in pressure, which will confuse the reactoroperators and potentially pose a risk to the safety of the plant.

If a blocked pressure transmitter happens to be a part ofa redundant safety channel, it can trip the plant during a transient.More specifically, the indication of a blocked transmitter willobviously not match the other redundant channels, creatinga mismatch that could trigger a reactor trip. In fact, this problemhas occurred in France where partial blockages in flow transmitterscaused two French PWRs to trip during load flowing episodes in themid-1980s (Meuwisse and Puyal, 1987).

Fig. 6. Research results on the effect of sensing-line blockages on response time ofnuclear plant pressure transmitters.

Some sensing-line blockages are so severe that the sensing linehas to be drilled to clear the blockage. This type of problem is thereason why measuring the response time of pressure transmittersis so important and why it is so surprising that even today, somenuclear power plants measure the response time of their safety-related pressure transmitters using conventional procedures thatexclude the sensing lines. These plants typically use a hydraulicpressure generator to input a pressure signal to the transmitter andmeasure its response time. In doing this, the sensor is isolated fromthe sensing lines. The research work documented in the U.S.Nuclear Regulatory Commission report NUREG/CR-5851 uncoveredthis flaw in the maintenance of nuclear plant pressure transmitters.As a result, many plants have recognized that they must measurethe response time of both their pressure transmitters and theirsensing lines. These plants have accordingly switched to the noiseanalysis procedure to verify the dynamic characteristics of theirpressure sensing systems.

3.2. Response-time testing of pressure transmitters

Pressure, level, and flow transmitters in nuclear power plantsbehave like filters on the natural plant fluctuations that are pre-sented to their inputs. That is, if one assumes that the input to thetransmitter exhibits wide-band frequency characteristics (which istypically the case for nuclear power plant fluctuations), informationabout the sensor itself can be inferred bymeasuring the transmitteroutput. This is the basis of the noise analysis technique that is usedto determine the dynamic response of pressure, level, and flowtransmitters in nuclear power plants (Thie, 1981).

The noise analysis technique is used to remotely measure sensorresponse time from the control room area while the plant is on-line.These measurements do not require the sensors to be disconnectedfrom the plant instrumentation or removed fromservice for the tests.That is, the tests are passive anddonot cause any disturbance to plantoperation. This reduces test time and helps to reduce radiationexposureof the testpersonnelwhowouldotherwisehave toenter thereactor containment to make the response-time measurements.

Dynamic response analysis is performed in the frequencydomain and/or time domain, and is based on the assumption that

Fig. 8. Examples of auto power spectral densities of nuclear plant pressuretransmitters.

H.M. Hashemian / Progress in Nuclear Energy 53 (2011) 167e181 173

the dynamic characteristics of the transmitters are linear and thatthe input noise signal (i.e., the process fluctuations) has properspectral characteristics. Frequency domain and time domain anal-yses are two different methods for determining the response timeof transmitters. It is usually helpful to analyze the data with bothmethods and average the results, excluding any outliers.

In frequency domain analysis, the APSD of the signal is gener-ated first, usually using an FFT algorithm. After the APSD isobtained, a mathematical function (model) that is appropriate forthe transmitter under test is fit to the APSD to yield the modelparameters. These parameters are then used to calculate thedynamic response of the transmitter. The dynamic response of thetransmitter can then be analyzed to determine the response time ofthe transmitter in situ. Fig. 7 shows an example of process noisethat enters a pressure transmitter and is subsequently filtered bythe transmitter. The response time of the transmitter can beinferred from the APSD with the proper analysis tools.

Under normal plant conditions, the APSDs of nuclear plantpressure transmitters have characteristic shapes that can be base-lined and compared with the APSDs of similar transmitters oper-ating under the same process conditions. Fig. 8 shows examples ofa few typical nuclear plant APSDs for steam generator level, reactorwater clean-up flow, and pressurizer pressure transmitters.

Through laboratory experiments, the noise analysis techniquewas validated for in-situ response-time testing of pressure trans-mitters. This validation work involved directly measuring responsetime and then using the noise analysis technique to compare theramp input signals with the response-time results. Table 4 showsrepresentative results of this validation work for seven differenttransmitters from various manufacturers of nuclear-grade pressuretransmitters. For each transmitter, the results of the directmeasurement of response time (ramp test) were compared withthe results of the noise analysis test; the difference between thetwo results is shown in Table 4. The details of the validation of thenoise analysis technique for response-time testing of nuclear plantpressure transmitters are documented in a comprehensive report,“Long Term Performance and Aging Characteristics of Nuclear PlantPressure Transmitters”, published by the NRC in March 1993 asNUREG/CR-5851 (U.S. Nuclear Regulatory Commission, 1993).

The research effort to validate the noise analysis techniqueincluded not only laboratory experiments but also in-plant trials inan operating nuclear power plant. In these trials, high-speed testtransmitters were temporarily installed upstream of plant trans-mitters, and noise data were collected from both the test trans-mitters and the plant transmitters. The data was then analyzed toobtain the mathematical transfer function of the plant transmitter;this function was used to calculate the transmitter’s response time.Subsequently, the response-time value obtained from the transferfunction calculations was compared with the results of both thenoise analysis technique and the direct measurement of thetransmitter response time. This exercise provided further evidence,based on in-plant measurements, that the noise analysis techniquecan yield the correct response time of nuclear power plant pressuretransmitters.

Fig. 7. Example of a pressure transm

3.3. On-line calibration monitoring of pressure transmitters

On-line calibration monitoring refers to monitoring the normaloutputof nuclear plant pressure transmitters duringplant operationand then comparing this data with an estimate of the processparameter that the transmitter is measuring. At most plants, theplant computer contains all the data that is needed to verify thecalibration of pressure transmitters, including data from plantstartup and shutdown periods used to verify the calibration of

itter filtering the process noise.

Table 4Representative results of laboratory validation of noise analysis technique forresponse time testing of nuclear-grade pressure transmitters.

Number Response Time (s)

Ramp Test Noise Analysis Difference

Barton1 0.05 0.09 0.042 0.17 0.20 0.033 0.17 0.25 0.084 0.12 0.15 0.035 0.12 0.20 0.086 0.11 0.15 0.047 0.12 0.18 0.06

Foxboro1 0.13 0.16 0.032 0.21 0.18 �0.033 0.16 0.13 �0.034 0.09 0.12 0.035 0.29 0.30 0.016 0.25 0.15 �0.107 0.28 0.25 �0.03

Rosemount1 0.05 0.06 0.012 0.32 0.28 �0.043 0.07 0.05 �0.024 0.10 0.07 �0.035 0.11 0.08 �0.036 0.09 0.08 �0.017 0.09 0.09 0.00

Other manufacturers1 0.15 0.15 0.002 0.21 0.18 �0.033 0.02 0.08 0.064 0.03 0.07 0.045 0.08 0.11 0.036 0.15 0.27 0.127 0.33 0.37 0.04

H.M. Hashemian / Progress in Nuclear Energy 53 (2011) 167e181174

instrumentsover theiroperating range.Using theon-line calibrationmethod, transmitter outputs are monitored during process opera-tion to identify drift. If drift is identified and is significant, thetransmitter is scheduled for a calibration during an ensuing outage.On the other hand, if the transmitter drift is insignificant, no cali-bration is performed for as long as 8 years typically. This 8-yearperiod is based on a 2-year operating cycle and a redundancy level offour transmitters. In this application, OLM is not a substitute fortraditional calibration of pressure transmitters; rather, it is a meansfor determining when to schedule a traditional calibration fora pressure transmitter.

Reviews of the calibration histories of process instruments innuclear power plants have shown that high-quality instru-mentsdsuch as nuclear-grade pressure transmittersdtypicallymaintain their calibration for more than a fuel cycle of about 2 yearsand do not, therefore, need to be calibrated as often (U.S. NuclearRegulatory Commission, 1995, 1998). The validity of the OLMapproach in verifying the calibration of nuclear plant pressuretransmitterswas researched in themid-1990susingboth theoreticalwork and laboratory experiments and in-plant trials. The results aredocumented in a comprehensiveNRC report published inNovember1995 as NUREG/CR-6343: “On-Line Testing of Calibration of ProcessInstrumentation Channels in Nuclear Power Plants” (U.S. NuclearRegulatory Commission, 1995).

To perform on-line calibration monitoring, the output of redun-dant sensors is averaged. The average value, called the process esti-mate, is then used as a reference to determine the deviation of eachsensor from the average of the redundant sensors and to identify theoutliers. For non-redundant sensors, averaging obviously cannot beused todeterminea reference value. Therefore, if there is insufficientinstrument redundancy, the process estimate for calibration moni-toring is determined by analytical modeling of the process. Both

empirical and physical modeling techniques are used in this appli-cation, although empiricalmodels are preferred because theycan beadapted to various processes and different operational envelopes.

In particular, empirical modeling techniques involving neuralnetworks have been researched for on-line calibration monitoringapplications in nuclear power plants (as documented in NUREG/CR-6343). Although neural networks are effective, the nuclear industrydoes not currently favor this application because of difficulties indetermining theuncertaintyof their results. As such, othermethods,such as averaging or analytical modeling techniques, have beendeveloped for monitoring the calibration of pressure transmitters.

3.4. In-situ cross calibration of temperature sensors

PWR plants often employ 20e40 RTDs to measure the fluidtemperature in the reactor coolant system. The temperaturesmeasured by the RTDs are used by the plant operators for processcontrol and to assess the operational status and safety of the plant.As such, the calibration of the RTDs is normally evaluated at leastonce every refueling cycle. Each RTD must meet specific accuracyrequirements in order for the plant to continue to produce poweraccording to its design specifications. There are also about 50 core-exit thermocouples (CETs) in PWRs to provide an additional way tomonitor reactor coolant temperature. Accuracy for CETs is not asimportant as for RTDs because CETs are used mostly for tempera-ture monitoring. Nevertheless, CETs are sometimes cross calibratedagainst RTDs to ensure that their output is reliable.

In each loop of a PWR plant and for each core quadrant,redundant RTDs and CETs are used to minimize the probability offailure of any one RTD or CET affecting the safety of the plant. Thisredundancy of temperature sensors is the basis for a method ofevaluating the calibration of RTDs and CETs called cross calibration.In cross calibration, redundant temperature measurements areaveraged to produce an estimate of the true process temperature.The result of the averaging is referred to as the process estimate. Themeasurements of each individual RTD and CET are then subtractedfrom the process estimate to produce the cross-calibration resultsin terms of the deviation of each RTD from the average of allredundant RTDs (less any outliers). If the deviations from theprocess estimate of an RTD or CET are within acceptable limits, thesensor is considered in calibration. However, if the deviationexceeds the acceptance limits, the sensor is considered out ofcalibration and its use for plant operation is evaluated.

Traditionally, cross-calibration data has been acquired usingdata acquisition equipment that is temporarily connected to testpoints in the plant instrumentation cabinets. The traditional cross-calibration method, while highly accurate, causes the plant to loseindicationwhen the data is being acquired, and costs the plant timeduring shutdown and/or startup to restore the temperature indi-cations. Now, with new and more advanced plant computers, RTDand CET measurements can be collected in the plant computer,which provides a centralized location for monitoring and storingthe measurements. Using on-line data from the plant computer forcross calibration can save plants startup and shutdown time, whileproducing results that are comparable to the traditional method.

3.5. Equipment condition assessment

Static analysis methods may be used for other purposes besidesevaluating the health of individual sensors as in on-line cross-calibration and transmitter calibration monitoring. Equipmentcondition assessment (ECA) applications take the idea of on-linecalibration monitoring a step further by monitoring for abnormalbehavior in a group of sensors. An example of ECA is illustrated inFig. 9, which shows a simplified diagram of a typical chemical and

Fig. 9. Simplified diagram of chemical and volume control system components.

Fig. 10. Normal operation of chemical and volume control system flow parameters.

H.M. Hashemian / Progress in Nuclear Energy 53 (2011) 167e181 175

volume control system (CVCS) in a PWR. The primary functions ofa typical CVCS in a PWR are:

1. Controlling the volume of primary coolant in the reactorcoolant system (RCS)

2. Controlling chemistry and boron concentration in the RCS3. Supplying seal water to the reactor coolant pumps (RCPs)

Several transmitters are typically used to monitor variousparameters related to the operation of the CVCS. Fig. 9 highlightsthe normal operation of a few of these parameters:

1. Charging Flow e measures the flow rate of the coolant beingprovided from the volume control tank (VCT) to the RCS andRCP seals

2. Reactor Coolant Pump Seal Injection Flow emeasures the flowrate of the coolant provided to the RCP seals

3. Seal Return Flow e measures the flow rate of the coolantreturned to the VCT from the RCP seal injection

4. Letdown Flow e measures the flow rate of the reactor coolantas it leaves the RCS and enters the VCT

During normal operation, the measurements of these parame-terswill fluctuate slightly, but should remain at a consistent relativelevel. However, in abnormal conditions such as an RCP seal leak,some parameters may exhibit upward or downward trends, indi-cating a problem in the plant. For example, Fig. 10 shows the fourflow signals mentioned above during normal operation of a PWRplant (the actual flow rates are scaled to simplify this example). AsFig. 10 shows, the flows remain at relatively constant rates relativeto one another.

Fig. 11 shows how these flow signals may appear at the onset ofan RCP seal leak in this PWR plant. In this example, the onset of theRCP seal leak is first indicated by a downward trend in the sealreturn flow measured at time T1. This is followed by an increase incharging pump flow at time T2 as the charging pump compensatesfor the loss of coolant due to the RCP seal leak.

Of course, an abnormal trend in an individual parameter such asseal return flow could mean that the sensor is degrading; however,

abnormalities in related parameters that occur close together intime are more likely to indicate the onset of a system or equipmentproblem. Early warning of these types of failures is thus the keybenefit of ECA. More specifically, early warning of impendingequipment failures can provide nuclear plants with increased plantsafety through early recognition of equipment failures and reduceddowntime stemming from timely repair of affected equipment.

3.6. Predictive maintenance of reactor internals

A research project published in NUREG/CR-5501 (June 1998),“Advanced Instrumentation and Maintenance Technologies forNuclear Power Plants,” investigated such OLM applications as noiseanalysis for measuring the vibration of reactor internals and othercomponents such as RCPs (U.S. Nuclear Regulatory Commission,1998). Typically, vibration sensors (e.g., accelerometers) arelocatedon the topandbottomof the reactor vessel to soundanalarmin case the main components of the reactor system vibrate

Table 5Typical frequencies of motion of reactor internals at pressurized water reactorplants.

Reactor Component Average ResonantFrequency (Hz)

Fuel Assembly 3.0Core Barrel Beam Mode 9.7Core Barrel Shell Mode 23.1Thermal Shield 12.5Reactor Vessel 18.5Reactor Coolant Pump 25.0

Source: NUREG/CR-5501.

Fig. 11. Chemical and volume control system flow parameters at the onset of a reactorcoolant pump seal leak.

H.M. Hashemian / Progress in Nuclear Energy 53 (2011) 167e181176

excessively. However, neutron detectors have proved to be moresensitive than accelerometers in measuring the vibration of thereactor vessel and its internals. This is because the frequency ofvibration of reactor internals is normally below 30 Hz, which iseasier to resolve using neutron detectors than accelerometers.Accelerometers are more suited for monitoring higher-frequencyvibrations.

Fig. 12 shows the APSD of the neutron signal from an ex-coreneutron detector (NI-42) in a PWR plant. This APSD contains thevibration signatures (i.e., amplitude and frequency) of the reactorcomponents, including the reactor vessel, core barrel, fuel assem-blies, thermal shield, and so on. It even contains, at 25 Hz, thesignature of the RCP rotating at 1500 revolutions per minute, whichcorresponds to25Hz. These signatures canbe trended to identify theonset of aging degradation, which can damage the reactor internals.The neutron noise approach has been recognized as a predictivemaintenance tool that can help plants guard against vibration-induced mishaps that may be encountered as plants age, makingthem more vulnerable to challenges to their structural integrity.

Over the last 10 years, numerous plants have begun programs tomeasure reactor internal vibration using neutron noise analysis andthen trend the results so as to identify changes and signs ofdegradation. Table 5 shows average values for the resonant

Fig. 12. Auto power spectral density containing vibration signatures of reactorinternals.

frequency of vibration of reactor internals of PWR plants. Theresonant frequency of the RCP also shows up on the neutron noisesignal, as shown in Table 5, at 25 Hz corresponding to 1500 rpm(revolutions per minute).

The results in Table 5 are the average of neutron noisemeasurements made by the authors and others in 15 PWR plantsaround the world representing ABB, Westinghouse, Babcock andWilcox (B&W), Areva (i.e., Framatome and Siemens), and Mitsu-bishi Heavy Industries (MHI) plants. The details are presented inNUREG/CR-5501 (U.S. Nuclear Regulatory Commission, 1998).

Neutron noise analysis has repeatedly demonstrated its effec-tiveness in nuclear power plants. For example, in 1991 the DiabloCanyon, California, PWR plant began to experience an increasingnumber of neutron signal alarms at the end of one of its operatingcycles. Regulators ordered the plant to identify the source of thealarms and shut the plant down if the alarms revealed a significantsafety issue. However, they gave the plant permission to continueto operate while investigating the cause of the alarms. It waspostulated that the alarms could be due to excessive vibration ofcore barrel or other reactor internal components. Therefore, usingthe output of the ex-core neutron detectors, the vibration of reactorinternals was measured. The results showed normal vibrationlevels for the reactor internals. Next, neutron signals were crosscorrelated with core-exit thermocouples (CETs) and other sensors.This helped narrow the problem down to core flow anomaliescaused by new fuel assemblies installed in the plant in a previousrefueling outage. Further analysis of the noise data demonstratedthat the flow anomalies were not significant, and the plant was thusallowed to continue to operate. In the meantime, the neutron alarmsetpoints were raised to limit the frequency of the alarms.

In another application of neutron noise analysis, the cause ofa rod stepping problem in a PWR plant was identified usingsignals from existing ex-core neutron detectors. In this plant,neutron signals are used in the rod control system. Due to neutronflux spikes, the reactor regulation system would move the controlrods in and out of the plant. The first step in resolving the problemwas to measure the neutron noise and verify that the neutron fluxspikes were not caused by any abnormal vibration of reactorinternals. Once the vibration of the reactor internals wasconfirmed to be normal and no significant flow anomalies werediscovered, the plant was notified that this problem did notthreaten the safety of the plant and could be resolved by (1)raising the setpoint that triggers the rods to move, or (2) placinga low-pass electronic filter at the output of the neutron detectorsto dampen the spikes.

3.7. Fluid flow monitoring

In a PWR plant, 50 thermocouples are located on the top of thecore. These thermocouples are normally used tomonitor the reactorcoolant’s temperature at theoutputof the core. Theycanalsobeused

H.M. Hashemian / Progress in Nuclear Energy 53 (2011) 167e181 177

in conjunction with the ex-core neutron detectors to monitor flowthrough the reactor system. More specifically, by cross correlatingsignals from the ex-core neutron detectors and CETs, it is possible toidentify the time required for the reactor coolant to travel betweenthe physical location of the neutron detectors and the thermocouple(Fig. 13). The result, referred to as transit time (s), can be used withcore geometric data to evaluate the reactor coolant’s flow throughthe system, identify flow anomalies, detect flow blockages, andperform a variety of other diagnostics. The same concept has beenused tomeasure primary coolantflow in PWRplants. The procedureis referred to as transit time flow measurement (TTFM).

In a PWR plant, Nitrogen-16 (N-16) is produced by fast neutron-induced reaction with oxygen in the primary coolant water. The N-16 is a radioisotope of nitrogen and has a half-life of 7.35 s. In itsdecay back to oxygen, it emits high-energy gamma rays. As theN-16 is transported by the primary coolant, these gamma rays canbe detected by radiation monitors (N-16 detectors) installed on theprimary loop hot leg piping (Fig. 14a). The coolant flow can then bedetermined by measuring fluctuations in the intensity of the N-16gamma radiation and analyzing them with the cross-correlationmethod (Fig. 14b).

An alternative to the cross-correlation technique for deter-mining transit time is to analyze the data in the frequency domain.

Fig. 13. Illustration of cross-correlation principle involving a neutron d

The Fourier transform of the detector data can be used to determinethe phase spectrum of the data in the frequency domain. The phasespectrum can then be used to determine the transit time (Fig. 15).

3.8. Life extension of neutron detectors

Effective managing of the aging of neutron detectors depends tosome degree on the detector manufacturer and the strategy of thenuclear plant for verifying the performance of nuclear instrumen-tation systems. Some manufacturers recommend that detectors bereplaced as often as once every 5 years; other manufacturers statethat their neutron detectors can be used for as long as 40 years ifthey are in good working condition. In the latter case, manufac-turers sometimes recommend cable testing and static and/ordynamic performance monitoring as a way to verify that theneutron detectors are in good working condition.

The dynamic response of neutron detectors can be monitoredusing the noise analysis technique. Results of such tests in a U.S.nuclear power plant are shown in Table 6 for four ex-core neutrondetectors, each with an upper and a lower sensor. As these resultsdemonstrate, the response time of the detectors increases duringthe first two decades and then stabilizes. This is expected ofneutron detectors as well as other sensors.

etector and a Core-exit thermocouple to determine transit time.

Fig. 14. a. Detector placement. b. Typical detector noise data record.

Table 6Results of trending the dynamic response of neutron detectors in a nuclear powerplant.

Item Tag Number Response Time vs. Years in Service

10 Years 20 Years 24 Years 30 Years

1 NI 41 Upper 0.61 0.77 0.60 0.722 NI 42 Lower 0.61 0.79 0.64 0.713 NI 42 Upper 0.62 1.00 1.11 0.904 NI 42 Lower 0.57 1.03 1.25 0.935 NI 43 Upper 0.64 0.97 1.04 0.826 NI 43 Lower 0.54 1.06 1.10 0.787 NI 44 Upper 0.73 0.88 0.67 0.858 NI 44 Lower 0.59 0.86 0.68 0.72Average Response

Time (s)0.61 0.92 0.89 0.80

H.M. Hashemian / Progress in Nuclear Energy 53 (2011) 167e181178

In addition to trending response times, the noise output ofneutron detectors can be examined for signs of other problems inthe nuclear instrumentation circuit, such as cable and connectoranomalies. Fig. 16 shows the APSD of a neutron detector before andafter the onset of a cable degradation problem. In this case,analyzing the APSD reveals a difference in the neutron detectordynamic response resulting from an increase in cable capacitance(Arndt and Miller, 1991). Continuous monitoring of neutron

Fig. 15. Frequency domain phase analysis to determine transit time.

detectors can reveal problems in the neutron detector circuit,enabling plant personnel to schedule maintenance accordingly.

4. Integrated OLM system

The research efforts documented in NUREG/CR-6343, NUREG/CR-5851, and NUREG/CR-5501 have helped provide the foundationfor developing an integrated OLM system to verify the static anddynamic performance of nuclear plant I&C systems (U.S. NuclearRegulatory Commission, 1995, 1993, 1998). Table 7 shows severalexamples of OLM applications in PWR and BWR applications. It alsoindicates whether the particular applications require AC signal, DCsignal, or both. Note that some of the techniques listed in Table 7 arefor sensor condition monitoring and diagnostics, and the others arefor process condition monitoring and diagnostics.

An integrated OLM system would consist of a data acquisitionmodule (hardware and software) and a software-based data pro-cessing or analysis module implemented on a fast computer. Thedata acquisition module should include signal isolation devices aswell as fast sampling capabilities (e.g., 1000 Hz). The system can bebuilt into the design of new plants or deployed as an add-on feature

Fig. 16. Auto power spectral density of a neutron detector showing degradation due toan increase in cable capacitance.

Table 7Examples of nuclear power plant applications of on-linemonitoring using signals forexisting sensors.

Application Signal Plant Types

DC AC PWR BWR

In-Situ Response Time Testing of ProcessInstrumentation

O O O

Instrument Calibration Monitoring O O OCross Correlation Flow Monitoring O O OOnline Detection of Venturi Fouling O O OOnline Detection of Sensing Line Blockages,

Voids, and LeaksO O O

Fluid and Gas Leak Detection O O O OEquipment and Process Condition Monitoring O O O OCore Barrel Vibration Measurement O OOnline Measurement of Temperature

Coefficient of ReactivityO O

Aging Management of Neutron Detectors,CETs, and Other Sensors

O O O O

Measurement of Vibration of In-Core Flux Monitors O OCore Flow Monitoring O ON-16 Flow Measurement O O

Checkmark (O) means that the application shown is based on AC and/or DC signalanalysis (whichever is identified by the checkmark) and that the application isuseful in PWRs and/or BWRs (whichever is identified by the checkmark).

Fig. 17. Block diagram of on-line monitoring system.

H.M. Hashemian / Progress in Nuclear Energy 53 (2011) 167e181 179

to the existing generation of plants. Fig.17 shows a block diagram ofthis system.

It’s essential that the data acquisition module used in the OLMsystem be fast. In the current generation of nuclear power plants,data from process sensors is normally sampled by the plantcomputer at rates of one sample per second or slower. This isadequate for applications such as calibration monitoring but not fordynamic analysis. To analyze dynamic signals, at least 100e1000samples per second are normally required. In the newgeneration ofnuclear power plants, this requirement can be accommodatedsimply by bringing the data into the plant computer through a fastanalog-to-digital (A/D) converter and providing adequate storageto save the data for subsequent retrieval and analysis. However, inthe existing generation of reactors, retrofitting the plant with a fastdata acquisition system and new storage provisions can becomplicated. In fact, even recent digital retrofits in nuclear powerplants have not provided the necessary means for fast datacollection and storage. As such, in the current generation of nuclearpower plants, a separate data acquisition system must be installedto collect dynamic data.

Fig. 18 shows a block diagram of a dedicated data acquisitionsystem for OLM. The system not only acquires data for OLMapplications but also takes calibration signals and simulated faultsto verify its own operation and calibration. In its signal conditioningmodule, it contains a gain amplifier, DC offset circuits, and an anti-aliasing filter to separate the AC component of a sensor output fromits DC component (see Fig. 19).

The data acquisition system shown in Fig. 18 was designed anddeveloped by the author and installed for about 3 years in the early1990s at the McGuire nuclear power station in the United States.The work was done under an R&D contract funded by the NRC (U.S.Nuclear Regulatory Commission, 1995). This system was used toprovide data for developing and validating the first versions of OLMalgorithms and software packages. In designing a data acquisitionsystem for OLM implementation in nuclear power plants, whetherit is for the new generation of plants or the old generation, the dataacquisition system must have ample electrical isolation and a veryhigh input impedance. The input impedance must remain higheven if the power to the equipment is lost. Table 8 lists some of thekey requirements for an OLM data acquisition system.

Today, even some of the best data acquisition systems on themarket can load down the plant circuits and cause plant trips. Assuch, nuclear power plants are very cautious about impedance andisolation requirements. Theminimum input impedance for the dataacquisition system allowed in nuclear power plants is typicallyabout 1 MU, and the minimum electrical isolation is 500 V. A moredesirable systemwill have an input impedance of 1000MU and willprovide 1000 V of electric isolation.

OLM data, whether it is acquired using a dedicated data acqui-sition system or from a plant computer, must first be qualified or

Fig. 18. Dedicated data acquisition system for on-line monitoring.

H.M. Hashemian / Progress in Nuclear Energy 53 (2011) 167e181180

validated. This can be accomplished automatically using algorithmsembedded in software packages that perform statistical calcula-tions on the raw data. More specifically, the mean value, variance,skewness, and kurtosis of the raw data are calculated and trended.The results of these calculations are then used to identify andremove anomalous data records. For example, the skewness fora normal data record is expected to be near zero, and its kurtosisshould be equal to 1. This and similar other criteria are used todistinguish between normal data records and outliers.

4.1. Challenges to OLM implementation

To achievewidespread adoption, OLMmust overcome a range ofobstacles, from convincing utilities to trust its performance androbustness and believe in its economic benefits to reassuring themthat it will neither produce costly false alarms nor interfere with

Fig. 19. Data flow in separating the AC componen

the safety and process control and data acquisition system of theplant. But the primary impediment to widespread adoption isregulatory constraint.

To make it worthwhile for utilities to retrofit their plants withOLM technologies, nuclear industry regulators must allow OLM toreplace the conventional techniques for maintaining safety-relatedequipment. Toward this end, the U.S. Nuclear Regulatory Commis-sion (NRC Project, 2000) issued a Safety Evaluation Report (SER) in2000 that accepted the OLM concept for condition-based calibra-tion of safety-related pressure transmitters in nuclear powerplants. However, according to the SER, each plant must still apply tothe NRC and receive approval for OLM implementation if it is to beused in lieu of traditional calibration of safety-related equipment.This has hindered the widespread use of OLM in the nuclearindustry. For that reason, in fall 2008, the author began workingwith representatives of the U.S. nuclear industry and nuclear power

t of a sensor output from its DC component.

Table 8Key requirements for on-line monitoring data acquisition hardware.

OLM Data Acquisition Requirements

Minimum Optimum

Sampling rates 100 Hz 1000 HzInput impedance 1 MU 100 MU

Electrical isolation 500 V 1000 VInput channels 16 32A/D resolution 12 bit 24 bitSampling time 1 h Continuous

H.M. Hashemian / Progress in Nuclear Energy 53 (2011) 167e181 181

plant vendors to help initiate an effort to obtain generic NRClicensing for the use of OLM in nuclear power plants. If approved,such generic licensing will allow nuclear power plants to imple-ment OLM without having to apply for an individual license foreach plant. This will doubtless incentivize the industry to imple-ment OLM at an accelerated pace.

5. Summary and conclusions

Over the past 40 years, an array of techniques has been devel-oped for equipment and process condition monitoring. Because ofregulatory constraints, cost of implementation, and other factors,these techniques have been used in nuclear power plants mostly onan “as-needed” basis rather than for routine condition monitoringapplications. Now, with the advent of fast data acquisition tech-nologies and proliferation of computers and advanced data pro-cessing algorithms and software packages, condition monitoringcan be performed routinely and efficiently using dedicated equip-ment installed at the plants.

This paper reviewed a class of condition monitoring technolo-gies that depend on data from existing process sensors during allmodes of plant operation including startup, normal operatingperiods, and shutdown conditions. The data may be sampledcontinuously or periodically depending on the application. Thesteady-state (DC) component of the data is analyzed to identifyslowly developing anomalies such as calibration changes in processsensors. The fluctuating (AC) component of the data is analyzed to

determine such parameters as the response time of pressuresensors or to measure the vibrational characteristics of reactorinternals, check for blockages within the reactor coolant system,identify flow anomalies, and provide other diagnostics.

The AC and DC data acquisition and signal processing techniquesdescribed in this paper can be integrated together to provide an on-line monitoring (OLM) system for nuclear power plants. This paperintroduced eight key applications of this system together with therequirements for implementing it in nuclear power plants. SuchOLM systems should be built into the design of the next generationof reactors to contribute to optimized plant maintenance byproviding automated measurements, condition monitoring, anddiagnostics. As for the current generation of reactors, they shouldbe retrofitted with OLM systems as utilities begin to appreciatetheir benefits and as regulators realize the added benefits of OLM tonuclear reactor safety.

References

Arndt, S.A., and D.W. Miller, 1991. The use of radiation detection noise for signalvalidation and sensor degradation monitoring. In: Proceedings of the Sympo-sium on Nuclear Reactor Surveillance and Diagnostics (Gatlinburg, Tennessee,USA), SMORN VI 2, 66.01e66.12.

Lillis, D., 2008. Oral Presentation at AMS, November 2008. Tennessee, USA, Knoxville.Meuwisse, C., Puyal, C., 1987. Surveillance of neutron and thermodynamic sensors

in EDF nuclear power plants. In: Paper Presented at the CORECH Meeting,Chatou, France.

Montanari, G.C., 2008. Envisaging links between fundamental research in electricalinsulation and electrical asset management. IEEE Electrical Insulation Magazine24 (6).

NRC Project No. 669, July 2000. Safety Evaluation by the Office of Nuclear ReactorRegulation: Application of On-line Performance Monitoring to Extend Calibra-tion Intervals of Instrument Channel Calibrations Required by the TechnicalSpecifications e EPRI Topical Report (TR) 104965 On-Line Monitoring ofInstrument Channel Performance. U.S. Nuclear Regulatory Commission:Washington, D.C. Nuclear Energy Institute, 2008, NEI, Washington, DC.Affordable Energy to Support Economic Growth.

Thie, J.A., 1981. Power Reactor Noise. American Nuclear Society, La Grange Park,Illinois, USA.

U.S. Nuclear Regulatory Commission, 1993. Long Term Performance and AgingCharacteristics of Nuclear Plant Pressure Transmitters. NUREG/CR-5851.

U.S. Nuclear Regulatory Commission, 1995. On-line Testing of Calibration of ProcessInstrumentation Channels in Nuclear Power Plants. NUREG/CR-6343.

U.S. Nuclear Regulatory Commission, 1998. Advanced Instrumentation and Main-tenance Technologies for Nuclear Power Plants. NUREG/CR-5501.