PROGRESS in ACOUSTIC EMISSION XII
The Japanese Society for NDI, 2004
ACOUSTIC EMISSION LOOSE PARTS MONITORING
SOTIRIOS VAHAVIOLOS, MARK CARLOS* & ATHANASIOS ANASTASOPOULOS**
MISTRAS Holdings Group, 195 Clarksville Road, Lawrenceville, NJ 08550, USA,
* Physical Acoustics Corporation, 195 Clarksville Road, Lawrenceville, NJ 08550, USA
** Envirocoustics S.A., El. Venizelou 7 & Delfon, 144 52 Athens, GREECE
ABSTRACT
The purpose of the Loose Parts Monitoring Equipment (LPMS) is to monitor for, detect and
evaluate large, metallic loose parts, in the primary coolant system of Nuclear power plants. Loose
parts can come from deteriorated components or from items inadvertently left in the primary
coolant system during construction. It is important to detect these loose parts before they can
create a partial flow blockage and could increase the potential for control rod jamming, either of
which could trigger a major problem in the nuclear system operation.
The purpose of this presentation is to discuss the application and operational aspects of a Loose
Parts monitoring system for Nuclear power plants as well as to demonstrate how present day digital
AE systems and processing strategies can improve LPMS operation and uniquely detect and
evaluate, metallic loose parts, in the presence of severe environmental noise and vibration.
Emphasis is given to one of the most critical aspects of a Loose Parts Monitoring system, i.e.
minimizing the false detection of loose parts events, a major problem existing in older analog loose
parts system designs. Loose part alarm definition strategies are presented together with some
unique signal processing functions and the Loose Parts detection algorithm which is a multiparameter, multiple level, time and frequency based analysis aiming in eliminating false calls.
KEY WORDS
Loose Parts, Acoustic Emission, Condition Monitoring, Nuclear, Steam Generators, Light-Water
Reactor Power Plants.
INTRODUCTION
The purpose of the Loose Parts Monitoring Equipment (LPMS) is to monitor for, detect and
evaluate metallic loose parts, using transient acoustic signal analysis. A loose part, whether it be
from a deteriorated component or from an item inadvertently left in the primary system during
construction, refueling, or maintenance, can contribute to component damage and material wear by
frequent impacting with other parts in the system. In addition to that a loose part can pose a serious
threat of partial flow blockage and could increase the potential for control rod jamming, either of
which could trigger a major problem in the nuclear system operation.
- 271 -
For example [1], loose parts object 3.18cm long and 1.91cm in diameter, identified in steam
generator, resulted in tube wear 38% through wall. As stated in [1], the indication of tube wear was
present in prior cycles but was not reported because the indication was distorted due to its location.
In another example [1], unit shut down performed when the steam generator primary to secondary
leak rate increased to unacceptable levels. Follow up inspection identified a leaking tube and given
the nature (volumetric) and location of associated flaws, were attributed to wear from loose parts.
United States Nuclear Regulatory Commission (NCR), recognizing the importance of early
detection of loose parts, requires from applicants for construction permits and operating licenses to
commit to a loose-part detection program. The NRC has developed hardware criteria and
programmatic (operational) criteria for loose-parts detection programs [2]. Loose-parts detection
programs are also in effect at most PWR operating facilities. In addition to requirements stated by
NCR in [2], loose parts systems design and function is described in relevant ASME code [3]. Both
NCR and ASME documents are used around the world.
In compliance with requirements of NCR and/or ASME, low frequency Acoustic Emission
(AE) sensors (also known as accelerometers) are located at strategic positions along the reactor
vessel (in direct contact with pipes and the vessel) where loose parts are most likely to travel. The
accelerometers detect the impact of these loose parts as they come in contact with the pipe or vessel
wall. This impact usually occurs with high force due to the speed at which these large loose parts
are traveling. The accelerometers are able to detect the mechanical vibration propagating from the
point of impact and convert this mechanical vibration into an electrical signal which is amplified in
the system preamplifier, processed by the Loose Parts data acquisition system and recorded by the
data acquisition computer. The software is the key thread needed by the data acquisition computer
to provide a friendly user interface for the operator, analyze the acoustic emission transients,
differentiate between noise and impact signals and alert the operator and control room when a loose
part is detected.
Existing analog based Loose Parts systems used a simple threshold detection strategy to detect
large impact signals, which would trigger a loose parts alarm and trigger an analog tape recorder,
which would record the waveforms on all active channels for up to 2 minutes. This then required
the use of an “expert” to replay these signals on a multi-channel oscilloscope to determine if the
signals received were from noise or an actual loose parts event. Present day digital system have the
capability of not only automating the entire process but at the same time offer, more stringent,
multi-level, alarm detection criteria, eliminating false alarms due to noise and the need for an
“expert” operator analysis.
The following describes each important part of a Loose Parts monitoring system followed by a
description of the advanced software and signal processing and alarm detection.
AE SENSORS AND CHARGE PREAMPLIFIERS FOR LOOSE PARTS MONITORING
Special AE sensors are required for loose parts monitoring. First, since the sensors need to
operate on a very hot nuclear reactor pipe or vessel wall, the sensor must be capable of
withstanding very high temperatures and nuclear radiation exposure. The sensor must withstand
continuous temperatures as high as 325 degrees C (or higher) and gamma radiation as high as 1000
MRads, over its 40 year rated life. In addition, the sensor must be sensitive to mechanical
vibrations up to 20 kHz while at the same time immune to electrical and electromagnetic
disturbances.
- 272 -
Due to the very high temperatures in the nuclear containment area and the high radiation
environment, special considerations are required when it comes to installing and using AE
preamplifiers. Traditional voltage preamplifiers cannot be used as the preamplifier must be located
a long distance (sometimes up to 100 meters) away from the system to keep within the temperature
and the radiation limitations of today’s electronics. In these cases a charge preamplifier is used.
The specific advantage of a charge preamplifier, over a voltage one, is that the preamplifier can be
located long distances away from the sensor without degradation in signal amplitude.
Other preamplifier considerations that have been taken into
account in the specific application is in the ability to remotely
change the gain of the preamplifier as well as the ability to run
diagnostics remotely including an on-board “preamplifier signal
input” which injects a known amplitude and frequency sine wave
signal into the front of the preamplifier, and an “Auto sensor Test”
function to excite the sensors on the structure to determine
operability. These remote control gain and diagnostic functions are
very important especially when the preamplifiers and sensors are
inside containment and it is necessary to determine system integrity
at specific test intervals.
DATA ACQUISITION SYSTEM CONSOLE
A data acquisition system console as show in Figure 1, contains
the digital data acquisition system utilizing multiple Physical
Acoustics, PCI-8’s, 8 channel Acoustic Emission system on a single
PCI card, the industrial chassis and PC compatible computer, a user
friendly, intuitive, graphical user interactive interface, an alarm and
control room interface and a color printer. The data acquisition
processes all AE signals from the sensors and performs real time
feature extraction and processing to monitor the health of the system
and detect suspect loose parts. The system also provides automated
and user demand frequency analysis to help distinguish between
noise and loose parts. The console has been thoroughly tested and
certified to operate in the presence of EMI and has also been
seismically tested to assure its operation, even in earthquake
conditions.
Figure 1: LPMS Cabinet
LOOSE PARTS MONITORING STRATEGIES and NON CRITICAL ALARMS
The Loose Parts Monitor software has been designed with multiple data presentation screens,
specifically for easy understanding and to offer a complete graphical and icon related analysis for
fast problem identification. The software operates in two main modes: Manual mode allows the
user to view the system operation, set adjustable test and operational parameters, set discrimination
levels, alert levels, and carry out extra diagnostic tests for system performance verification.
Automatic mode continuously monitors for loose parts events and displays system status. Upon
detection of a loose parts event, local alarms are sounded, signals are sent to the control room and
local video monitor displays an analysis mode to provide further information regarding the loose
parts event to the operator. In automatic mode, all AE activity is logged for future replay.
Supervisor mode protects the system from unauthorized tampering with critical system settings.
- 273 -
There are 6 basic AE monitoring software screens, presenting either time driven data or hit driven
data. The main time driven data screen, shown in Figure 2, presents monitoring of average AE
signal level (ASL) versus channel for background noise monitoring, leak or other problem
detection. As far as the ASL versus Channel setup data is concerned, it can be seen for this function
that the setup items in Figure 3a are related to the viewing screen. The first two setup items, on the
left, are the alarm level settings for the Low and High ASL alarms. On the right are the ASL High
and Low Alarm Time Constants i.e. the amount of time that the ASL must remain out of limits
(High and Low, respectively), per channel, for an ASL alarm to come on. The next item is the ASL
Time constant which allows the user to select an averaging period to assure that the ASL versus
channel display is reacting to the background noise changes and not short time bursts which will be
occasionally picked up by the AE transient analysis anyway. In addition to the ASL Low and High
alarms, Cabinet Door Open and Low disk space are also being continuously monitored as part of
the Non-Critical Automatic Diagnostic tests, as presented in Figure 3b.
Figure 2: Average Signal Level and associated alarm levels & status
(a)
(b)
Figure 3: Non-Critical alarms set-up and Non-Critical Alarm Diagnostics
- 274 -
Other monitoring screens includes, AE signal level versus time where a user can utilize trending
principles to determine channels of concern, the AE monitoring screen which as shown in Figure 4
contains multiple key AE graphs including Hits versus channel, Hits versus time, Hits versus
amplitude and Amplitude versus Energy correlation graph. In addition, there is a real time
waveform analysis screen, Figure 5, which allows a timed or on-demand waveform capture and
frequency analysis and Correlation analysis to aid in location of sources.
Figure 4: AE monitoring screen – summary of Hit Driven Data
Finally there is a test log screen and a Sensor Layout screen. The Test log screen presents the
measured Time and Hit Driven data values as well as the system diagnostics, critical and noncritical alarms history. Sensor Location Screen shows the relative placement of each sensor,
overlaid (and labeled) on a graphic depiction of the plant components that are being monitored.
The sensor color indicates the status. Colors vary to show status. For example, green indicates all
things are normal, yellow indicates a Non Critical Failure associated with that channel, while Red
indicates a loose parts alarm on that channel. At this point it is worth mentioning that the software
is customized for each specific installation. More specifically, depending on the monitoring
components and the respective sensors installed on each component, provisions are made for
viewing only part of the data e.g. the Steam Generator A and the associated Coolant Pump A.
LOOSE PARTS ALARM DETECTION
The most important part of the new digital Loose Parts Monitoring system is the critical alarm
detection mode. As mentioned earlier, it is extremely important to eliminate false calls. Any alarm
in a nuclear plant needs to be taken very seriously and because of that, each alarm must be genuine
and important. Loose parts systems have suffered a poor reputation over the years due to the older
analog systems lack of good alarm discrimination and need to be followed up with a manual,
human assisted (via monitoring replayed waveforms on an oscilloscope) decision. It was for this
- 275 -
reason that a lot of time was spent to come up with an automated, on-line, multiple level alarm
criteria in the Loose Parts Monitoring system.
In order to overcome limitations posed by the amplitude based alarms, we have implemented a
three level alarm criteria including both time and frequency analysis, as well as incorporating
pattern recognition techniques in the decision making. The following is the alarm detection
process that has been employed in the Loose Parts Monitoring system. First, a “suspect” loose
parts alarm is generated when AE activity meets a defined 2 dimensional alarm based on a
correlation graph of two AE features, as can be seen in the example of Figure 4. Once detected,
this starts an alarm determination process that includes performing a pattern recognition analysis
using AE feature analysis from time and frequency domain. If the criterion for a loose parts alarm
is met, a frequency analysis is performed comparing the high frequency/low frequency content
against the alarm criteria for a loose part. Only after all three conditions are met, will a loose part
be triggered.
Upon detecting a loose part, a 2 minute waveform capture mode is initiated including pre-trigger
waveform information. In addition, an alarm output signal is generated to the control room to
indicate a loose parts detection. Users can view, manipulate and analyze waveforms from the “first
arrival” sensor and up to three neighboring sensors, as presented in the example of Figure 5, or to
analyze any user selected waveforms. Waveform data is automatically stored for future analysis
and information is available for post processing waveform review. Also, the location determination
can be analyzed using waveform correlation techniques built into the system, Figure 6.
Figure 5: Waveforms processing screen
In addition to the three levels loose part alarm criteria, Control Rod Drive Movements (CRDM),
can be identified and processed as separate case in order to minimize false call due to CRDM
operations.
- 276 -
Figure 6: Correlation based Location Display
DISCUSSION & CONCLUSIONS
An advanced, digital, AE Loose Parts Monitoring system has been developed that is capable of
operating in a nuclear environment, monitoring the primary coolant system of a nuclear reactor.
The system replaces old analog signal processing techniques that were very prone to false alarms.
The benefit of this new system includes the advanced diagnostics to minimize test time as well as
the application of advanced, critical alarm detection criteria that has been developed to eliminate
false calls, and a very user friendly graphical user interface.
Loose parts detection and alarm criteria fine tuning is performed during system calibration by
means of base line impact testing. The equipment needed for carrying out the testing procedure to
determine the loose parts settings, includes an Electronic Loose Parts Simulator, which will
simulate a loose part, hammers of different masses, metallic balls of different masses as well as any
other device that can create a non loose part noise or false type of signal. This might include an air
jet, a scraping source, inducing flow noise and other process noises that can be induced. For each
specific installation, base line impacts are performed on metallic plates/walls and pipes simulating
the reactor vessel and piping respectively in order to set up the initial conditions of alarms. The
system is then fine tuned upon final installation.
A feasibility study for loose part energy/mass estimation was performed using data from simulated
impact events. The AE data and the respective waveforms were analyzed by means of NOESIS [5]
signal processing and pattern recognition software. Features extracted from the time and frequency
domain were transformed using principal component analysis. Representative part of the data was
used to train a Back Propagation Neural Network (NN), while the remaining part of the data was
used for testing purposes. The preliminary results, proved the feasibility of energy/mass estimation
using NN. Experimental and research work in this field is in progress, aiming to classifier
validation first in laboratory condition and later in field conditions.
REFERENCES
1) NCR Information Notice 2004-10, “Loose Parts in Steam Generators”, USA Nuclear Regulatory
Commission Office of Nuclear Reactor Regulation, (2004).
2) U.S. Nuclear Regulatory Guide #1.133, (1981)
3) ASME OM-S/G-1997, part 12, (1997), 67.
4) LPMS Operating & Technical Manual, Physical Acoustics Corp, (2002)
5) NOESIS V. 4.0, Reference Manual, Envirocoustics S.A., (2004)
- 277 -