IEEE Transactions on Nuclear Science, Vol. NS-29, No. I, February 1982
INSTRUMENTATION FOR REACTOR TWO PHASE NOISE DIAGNOSIS*M. K. De
Westinghouse Electric CorporationPost Office Box 355, Pittsburgh, Pennsylvania 15230
Summary
The aim of this work is directed towards devel-oping two particular applications of two phase noisediagnosis in nuclear power plants, cavitation damageprediction in circulating pumps, particularly forsodium cooled fast breeder reactor concepts and thedetection of incipient boiling and cavitation in pres-surized water or sodium cooled reactors. For theabove purpose a high frequency pressure-bar probe anda digital acquisition and processing system has beendesigned to find the amplitude and statistical charac-terics of pressure pulses from cavitation in a venturiwhich simulates conditions in hydraulic machinery andhigh speed flow channels. Significant results havebeen obtained about two phase inception phenomenon forthe development of instrument systems for the abovetwo applications.
Introduction
In the present state of art, it is still verydifficult to distinguish and analyze the acousticnoise from cavitation or boiling from other predomi-nant noise sources in a nuclear reactor1. It isdesirable, particularly for sodium cooled fast reac-tors, to be able to detect the onset of nucleate boil-ing, and determine the extent of cavitation in circu-lating pumps. The former is important in a fast reac-tor because of the positive void coefficient of reac-tivity in the core, particularly for localized boilingdue to insulating blockages, overenrichment errors andfission gas releasel. Such anomalies cannot bedetected by global parameters. The latter is impor-tant for any sodium circulating pumps due to a lack ofa priori knowledge of the extent of cavitation damageat a flow condition.
There is a lack of information at present aboutthe complex cavitation and boiling phenomena which isessential for accurate noise diagnosis. Attempts atdiagnosing the far field acoustic noise by measure-ments of power spectral density alone has preventedprevious researchersl12,3 from identifying some keyfactors in these processes. The present work is adetailed investigation of the phenomena. The discreteacoustic pulses from cavitation in a venturi which ispart of a high speed water tunnel facility have beeninvestigated by measurements of pressure pulse heightspectra, distribution of time interval between pulses,and statistics of pulses occurring in a time inter-val. This test facility simulates cavitation inhydraulic machinery and flashing in high speed flow asin reactor channels. A high frequency pressure-barprobe has been designed to measure the amplitudes ofthe discrete high frequency, and localized pressurepulses from cavitation. Presently there is no commer-cially available probe for such use. The high speedflow tests were only conducted for water due to avail-able facilities. However, pressure pulses from cavi-tation in sodium in an ultrasonic vibratory facilityhave been investigated and a comparison of cavitationin sodium and water is presently being conducted.
The present work, although limited in scope, wasconducted to increase an understanding of the funda-mentals of the above processes. The results obtainedapply to boiling inception due to a reduction in pres-sure (flashing), which is identical to cavitation.The next stage of this work will be directed towardsunderstanding the variation of the observations in
various geometries, liquids and flow conditions.Algorithms can then be written for optimal filtersbased on the signatures observed from cavitation andboiling.
A detailed description of the instrumentationdeveloped and highlights of the results are presentedhere. A full physical analysis of the resultsobtained shall be published in the future.
Pressure Transducer DesignAnd Observations
A Kistler 601A commercial transducer and a pres-sure-bar transducer developed by the author were usedto record the amplitudes of the pressure waves. Thecharacteristics of the two transducers are listed inTable 1. The size and frequency response of the com-mercial Kistler transducer make it less suitable forrecording the acoustic waves from cavitation becausethe typical maximum size reached by bubbles in thisventuri is `60 mils (1.52 mm)4, and risetimes ofthe emitted acoustic waves lie in the 1-10 lisrange. Also, the diameter of the liquid jets emitteddue to asymmetric bubble collapse is in the range1/10th to 1/100th the maximum bubble radius5.
Table 1. Characteristics of Pressure-bar andKistler-601A Transducers.
Figure 1 shows the pressure-bar transducerdesigned by the author. The probe was designed speci-fically for use in this system. Criteria for thedesign were minium overshoot, risetime and sensitivehead area. Various researchers697,8 have designedtransducers based on the pressure-bar principle foruse in shock tubes. Jones and Edwards9 studied it'sresponse to an underwater spark bubble. This is thefirst time a probe based on the pressure-bar principlehas been used to study natural cavitation in a flowingsystem.
PZT BA CERAMIC1/16" DIA.0.009" THICK
THINWIRE
SILICONERUBBEROC3110
THIN |TEFLONSPACERS |
I1
STAINLESSSTEEL TUBE BNC1/4" 0.D. CONNECTOR
COPPER 1/64" THICK
1O16"1DA. I SHIELOED l
II
Figure 1. Pressure-bar Transducer
As shown in Figure 1, the basic component is acylindrical copper pressure-bar which transmits anacoustic wave without longitudinal reflections until
*Work done for a PhD dissertation in nuclear engineering at the University of Michigan, Ann Arbor.
the original wave is reflected from the far end of thebar. This is true because the acoustic impedances ofthe copper rod (3.2 x 106 gm/cm2. sec) and thepiezoelectric ceramic (3.36 x 106 gm/cm2. sec) arevery close. For a one-dimensional case, the risetimeof a step pressure wave is the transit time of thewave front through the ceramic, in this case approxi-mately 0.05 ps. In reality, three-dimensionaleffects will limit the risetime of the observed wave-form and cause damped oscillations after the initialresponse. It has been shown theoretically10 thatthe risetime is proportional to (V)3/2(x/D)1/3(D/C), where V is rod Poisson's ratio, x the dis-tance of the ceramic from the front end of the rod, Dthe diameter and C the speed of sound in the rod. Thediameter of the copper bar is 2-4 times less than thatused by other researchers. This results in a smallerrisetime and a smaller surface area exposed to thewaves emitted by bubble collapses, as is desired.Edwards6 has discussed the effect of the parameterx/D. If the ceramic is placed at the front end of thebar, then the risetime would be limited by the transittime of the wave front through the ceramic. Largeamplitude oscillations occur in this case due to thereflection of stress waves from the radial boundary ofthe bar. Edwards has suggested an optimum value forx/D in the range 2-4. By constructing and testingprobes with different dimensions, a value of 8 wasfound to be optimum in the present study. The waveemitted by a collapsing bubble will propagate throughthe 0.5 inch (12.7 mm) bar section before arriving atthe ceramic. A thin wafer of pzt 5A ceramic was sol-dered to the copper bars with indium solder (50%indium - 50% tin) which has a low melting point of241°F (1160C). A low temperature solder was usedto avoid damaging the ceramic. The wafer was thenground to shape. The 7 inch (17.8cm) backing barresulted in a reflected wave arriving about 100 psafter the original wave. This allows the observationof reflected waves in the venturi, and waves frombubble surface oscillations, as discussed later.Silicone rubber was used to provide mechanical andelectrical insulation.
Optimum parameters were selected after testingthe transducer with a shock wave emitted by an under-water spark bubble. Condensers, charged from a high-voltage power supply, are triggered to discharge intowater between two electrodes, in a static spark cham-ber. Shock waves are emitted by the expansion andimplosion of the plasma. The transducers were placedflush with the wall of the chamber. Figures 2 and 3are oscillographs illustrating the responses of theKistler 601A and the pressure-bar probes in the sparkchamber. The Kistler rises in 2-3 ps and thienoscillates considerably (75%) whereas the pressure-barprobe rises in about 1 ps with minimal overshootoscillation (<25%). The actual risetime of the shockwave is evidently <1 ps as confirmed elsewhere11.
1.0 V/div
20.0 psec/div
I Iv1.0 V/div
10.0 psec6div
TOP TRACE = W/O CHARGE AMPLIFIERS 20366 2
BOTTOM TRACE = W/CHARGE AMPLIFIERSAT 50 MV/PCB
Figure 2. Responses of Kistler (601 A) Probe to Underwater SparkBubble
10 Ps/div 5 ps/div0.5 V/div 0.2 V/div 2'
Figure 3. Response of Pressure-bar Probe to Underwater SparkBubble
Figure 4 shows the plexiglass venturi which ispart of a high speed water tunnel facility. Damage
DAMAGE SPECIMEN NO. 1
NOTE: ALL DIMENSIONS ARE INCHES
Figure 4. Venturi Flow Path
specimens and pressure probes can be mounted flush tothe wall of the venturi in the same axial plane. Theventuri throat velocity and the axial extent of thecavitation cloud is controlled by pump speed anddownstream pressure. Figures 5 and 6 show responsesof the two probes to bubble collapses and/or reboundsin the venturi. Risetime of the acoustic waves pre-sumably depends on their origin and intensity and isshown to range from r2 to ,1lO lis (Figures 5 and6). Large amplitude shock fronts would presumablysteepen at a faster rate. The slower risetime (10as) in Figure 5a may be due to a liquid jet impactas observed by Kling5 in a similar venturi, ratherthan shock waves. The Kistler probe does not respondaccurately to waves with risetimes <3 ps (Figure5b) due to its low resonant frequency. The secondpulse in Figure 5b arriving 40 us after the firstmay be due to a rebound of the bubble. Further evi-dence, and discussion of these phenomena, are providedlater. Figure 6 shows the minimal overshoot of thepressure-bar probe response. Reflected pulses in thepressure-bar at X1l00 ps intervals are shown inFigure 6a.
Reflections from the opposite venturi wall appearto be absent. Reflected waves would arrive '30 ps
5 after the initial wave and would be attenuated in thewater by a factor of RO/R, where Ro is the maximumbubble radius (r50 mils or 12.7 iam)4912 and Rs 2 cm. The reflection coefficient of water toplexiglass is 1r.2. The amplitude of the reflectedwave would then be P1% of original wave amplitude.
Lauterborn1l, usiaig Schlieren photography,found that the risetimes of such pressure waves is'P0.01 ups, corresponding to a cavitation shock wave
710
20366-3
.e
1. L l50 us/div 50 ps/div0 5 Vdiv 1.0 V/div
a b
Figure 5. Responses of Kistler (601 -A) Probe to Cavitation Bubblesin Venturi
100 ,us/div5.0 V/div
.1 I I20 ps/div2 V/div
20366-6
Figure 6. Responses of Pressure-bar Probe to Cavitation Bubbles inVenturi
thickness of P15 pm in water. No existing pres-sure-bar probe, including the one designed here, candetermine the duration of such steep-fronted shockwaves. The probe designed here does, however, measurethe approximately correct amplitude as the followingdiscussion will show.
The sensitivity (psi/volt) of the pressure-barprobe was calctulated in the following manner. Thetransmission coefficient from water to copper isP0.08. The spherical wave will be attenuated due todivergence4 by a factor of iPO in the 0.5 inch (Icm) front-end bar. By using the combined capacitanceof the ceramic, housing, cable and scope, and thepiezoelectric constant and surface area of the cera-mic, the pressure can be calculated using Coulomb'slaw, Q = CV. The details of the calculation can befound in Ref. 12. In the venturi the Kistler 601A(according to manufacturer's calibration) measuredpeak pressures up to 500 psi (34.5 bar). Using thecalculational method outlined above, the pressure-barprobe measured peak pressures up to P250,000 psi(1.7 x 104 bar) in the same venturi. The differenceis due to the different sensitive head areas (Table 1,factor of '10) and frequency responses of theprobes. The surface area impacted by a typical bubblecollapse liquid jet or shock wave would be '10 -1to p10 -2 of the maximum bubble radius5. Atypical maximum bubble radius observed4 in a venturisimilar to the one used for the present study is i50mils (12.7 pm). Therefore, the surfface diameterimpacted would be 0.5 to 5.0 mils. The ratio of theKistler probe sensitive head diameter (Table 1, 200mils) to a typical impacted diameter is in the range40-400. This explains the relatively low pressuresindicated by the Kistler (<500 psi., 24,5),whereas pressures needed to cause cavitation damageare no doubt much higher (range 104 - 105bar)12. The pressure-bar probe indicates pressuresthat can easily cause damage. This is the first
observation of the peak pressures of waves emitted bycollapsing bubbles in a flowing system. Fujikawa13,and others9 observed, also with a pressure-bar, thatthe peak pressures emitted by such bubbles in a sta-tionary system (cavitation bubbles created and col-lapsed by a shock wave) was in the range 104 - 105bar. A calibration of the present pressure-bar probewould be desirable. However, no suitable equipmentwas available, so that this could not be done as partof the present study. However, the results are in therange found by Fujikawal3 for a similar probe, whichwas calibrated in a shock tube.
Data Acquisition System
Figure 7 is a block diagram of the digital systemused to record and process the amplitudes and statis-tical characteristics (eg. time interval distribution)of the acoustic waves sensed by the probes. The sig-nal from the Kistler transducer is initially amplifiedin a Kistler 556 charge-amplifier to give peak volt-ages to P5 volts in the venturi. The pressure-barprobe senses somewhat higher peaks (to P10 volts)due to the higher sensitivity of its ceramic (pzt 5Avs quartz), its lack of diaphragm, and its more accur-ate transient response. The signal may be shuntedwith a capacitor to reduce temperature effects on thecapacitance of the probe.
CONSOLE INTERFACE OSCILLOSCOPEDT = DATA TRANSLATION D TEKTRONIXDEC = DIGITAL EQUIPMENT AP 200-5 551 DUAL-BEAMCORPORATION MULTI-FUNCTIONCRDS = CHARLES RIVER FILTER
DATA SYSTEMS IPEAK DETECTOR
CALC I AND TIMING SIGNALSPRDGRAMMABLE I
CILOCK A/D CONVERTERDT 2769 DT 1762
_xMICROPROCESSORMODEM CON1TROLLER DEC -LSI -11/2 _ DACONVERTERDEC-DLV11-E FLOPPY DISK SYSTEM DT 2766DATA COUPLER _ CRDS -MF-1 1MULTI-TECH FM30
|CONSOLE INTERFACE IOSCILLOSCOPECALCOMP DEC-DLVII-F ITEKTRONIXI CALCOMP I TERSMINAL I5A 18N-AMFLIPLO_TTER DEC-LA-36 I I5B313N-TIME BASEI
Figure 7. Components Block Diagram
The cavitation pulses are in general much largerthan the background flow noise, which has frequencycomponents up to 'O KHz. An AP 220-5 multi-func-tion filter in high-pass mode is used to attenuate theflow noise, when flow system parameters result inreduced cavitation noise. The pulse amplitude isdetected by a peak detector designed here with slew-rate capability up to 10 volt/ps. It is digitizedby a DT-1762 analog-to-digital converter with timinglogic signals generated by a DT-2769 programmableclock, and timing circuitry designed here. Figure .8shows the sequence and operation of the pressurepulse, peak detector and timing signals.
A potentiometer setting in the programmable clockcauses its Schmitt Trigger I to fire if the inputsignal is greater than the set threshold (Figure 8).A 556 timer-integrated-circuit is used to produce adelayed trigger pulse to initiate conversion of thepeak detector signal to a digital number, which isthen sent directly to memory. The time taken forconversion and transfer to memory is 10 ps. A sec-ond 556 timer is used to produce a peak detector resetpulse. A dead time is created in which further fir-
ings of Schmitt Trigger I due to oscillations in theprobe output will not result in additional A/D triggerpulses.
The programmable clock is also used to record thetime interval between discrete pulses. The originalsignal may also be digitized directly by periodictrigger pulses out of Schmitt Trigger II of the pro-grammable clock. The pulse amplitudes and time inter-vals are processed with a DEC-LSI-11/2 microprocessorwith suitable software.
Typical Processed Data
Figures 5 and 6 show that pressure pulses fromventuri cavitation bubbles occur discretely in time.To verify this observation an A/D converter was usedto digitize the original wave form from the Kistler601A transducer to study the pressure pulses over aperiod of 250 ms. Since the sampling rate was limitedto 70 KHz, a low-pass filter set at 50 KHz was used toattenuate the 130 KHz resonant frequency of the trans-ducer, also resulting in the attenuation of the peakpressures of the cavitation pulses by 24 db/octavefrom the cut-off point (50 KHz). Figure 9 shows a
typical segment of the data, running from 0 to 4.2msecs. It should be noted that the pulse amplitudes
are in reality much larger than the background flownoise, and are indeed discrete. The repeated signalat about 10 KHz (Figure 9) confirms that some of thebubbles do indead oscillate, as indicated earlier.
This phenomenon has been photographically observedhere by Ivany and Hammitt4 in a similar venturi.The oscillation frequencies (5.0 to 20.0 KHz, Fig. 9)correspond to the natural frequencies of bubblesobserved by Ivany and Hammitt, which ranged in dia-meter from 10 to 50 mils (0.254 to 1.27 mm). Theoscillations may be excited by turbulent pressurefluctuation in the liquid. The pulse from bubblerebound4'14 is sometimes larger than that frombubble collapse as shown in Figure 9.
Further work is planned to obtain statistics ofthe occurrences of collapses, rebounds, oscillations,liquid jet impacts, and the spatial origin of thebubbles. This can be achieved by modifying the hard-ware and software for a dual-channel system. Outputsof two probes 1800 apart, in the same axial plane,can then be simultaneously recorded.
Figure 10 shows a typical pulse height spectrummeasured by the Kistler transducer. The vertical
4000
3600
3200
2800
2400
e 2000
0U 1600
1200
800
400
4 6
AMPLITUDE OF PULSE (VOLTS)
lX104
0
1 X 102
1 X 101
10
Figure 10. Typical Pulse Height Spectrum
lines (Figure 10) indicate the variation in thethreshold setting shown in Figure 8. In Figure 10 anexponential decay of the spectrum is shown followed bya power law variation. The decay constant of thespectrum acquired with the Kistler was found to beabout a factor of two smaller than that acquired withthe pressure-bar probe at the same flow conditions.Some difference would be expected since the Kistlerprobe will record less correctly a large pressurepulse, if the bubble collapses close to its relativelylarge diaphragm. The decay constant represents infor-mation about the probability distribution of thepulses in the spectrum, but does not indicate anythingabout their amplitudes. The Kistler probe functionsfairly well in acquiring the relevant shape of thespectra, but fails to record the correct absoluteamplitudes of the pressure waves. It has been foundthat the area under a pulse height spectrum curve can
be correlated to cavitation damage and frequencyshifts in the power spectrum, thus possibly allowingdamage predictions in pumps from a priori correlations.
Figures 11 and 12 show a typical time intervaldistribution between pulses, and a histogram of thenumber of observations of recording N pulses (x-axis)in a fixed time interval, in this case 1 second.Briefly, it has been concluded from these observationsthat two phase inception due to pressure perturbationsmay be modeled as random inception of pulse trains,with bubble clusters incepting periodically within a
712
20366 9
1600
1400
1200
Z 1000
400
z
C
Bouj
E 600
400
200
0
6 12 18 24 30
TIME INTERVAL (MSECS)
Figure 1 1. Typical Time Interval Distribution
a00
600 -
0
0c
co
LUm
z
'I,
0*
400 0 0
0
200 -
0
0
0 \
0 20 40 60 80
NUMBER OF COUNTS IN THE TIME INTERVAL
Figure 12. Typical Counting Statistics
train. This observation may be significant, but yetneeds to be theoretically formulated and measured forvarious flow geometries. This information can then beused in designing optimum filters for extracting inci-pient two phase noise for reactor surveillance.
Complete results obtained by this instrumentationsystem, theoretical forniulations, and explanations ofthe observed results can be found elsewhere12415'16.
Conclusion
A pressure-bar probe, and a digital data acquisi-tion and processing system has been developed, andsuccessfully used to find the amplitude and statisti-cal characteristics of pressure pulses from flow cavi-tation. The observations presented here in detailshow that cavitation bubbles incept discretely, andsome of the bubbles oscillate at frequencies in therange 5 to 20 KHz. These results, and other observa-tions briefly presented here provide knowledge aboutcavitation inception and growth that will be usefulfor developing two phase noise diagnosis systems.
Acknowledgement
Financial support was provided by the Office ofNaval Research and the National Science Foundation.
References
1. Carey, W. M., et al; "Detection of Sodium VaporBubble Collapse in a Liquid Metal Fast BreederReactor", SMORN-II, 1977, Argonne NationalLaboratory.
2. Lush, P. A. and Hutton, S. P., "The RelationBetween Cavitation Intensity and Noise in a Ven-turi-Type Section", Proc. Int'l Cof. on Pump andTurbine Design, Glasgow, Scotland, Sept. 1976.
3. Ramamurthy, A. S. and Bhaskaran, P., "VelocityExponent for Erosion and Noise due to Cavita-tion", Proc. ASME/CSME Appl. Mech., Fluids Engr.and Bioengr. Conf., Niagara Falls, N.Y., June,1979.
4. Ivany, R. D., "Collapse of a Cavitation Bubble inViscous, Compressible Liquid-Numerical and Expe-rimental Analyses", Ph.D. thesis, Nucl. Engr.Dept., Univ. Mich., 1965.
5. Kling, C. L., "A High-Speed Photographic Study ofCavitation Bubble Collapse", Ph. D. thesis, Nucl.Engr. Dept., Univ. Mich., 1969.
6. Edwards, D. H., "A Piezo-electric Pressure-BarGauge," J. of Sci. Instruments, v. 35, Sept. 1958.
7. Jones, I. R., "Beryllium Pressure Bar HavingSubmicrosecond Risetime", Review of Sci. Instru-ment, vol. 37, no. 8, Aug. 1966.
8. Ragland, K. W. and Cullen, R. E., "PiezoelectricPressure Transducer with Acoustic Absorbing Rod",Rev. Sci. Inst., v. 38, n. 6, 740-742, June 1967.
9. Jones, I. R. and Edwards, D. H., "An ExperimentalStudy of the Forces Generated by the Collapse ofTransient Cavities in Water", J. of Fluid Mech.7, 596 (1960).
10. Baganoff, D., "Pressure Gauge with One-TenthMicrosecond Risetime for Shock Reflection Stu-dies", Rev. of Sci. Inst., v. 35, n. 3, p. 288,March 1964.
11. Lauterborn, W. and Ebeling, K. J., "High SpeedHolography of Laser-Induced Cavitation Bubbles inLiquids", Proc. 7th Int'l Symp. Non-linear Acous-tic, Aug. 19-21, 1976, Blacksburg, Va.
12. De, M. K., "Acoustic Waves from HydrodynamicCavitation", Ph. D. thesis, Nucl. Engr. Dept.,Univ. Mich., Ann Arbor, Mich., 1980.
13. Fujikawa, S. and Akamatsu, T., "ExperimentalInvestigations of Cavitation Bubble Collapse by aWater Shock Tube", Bull. JASME, v. 21, Feb. 1978.
14. Hickling, R. and Plesset, M. S., "Collapse andRebound of a Spherical Bubble in Water", Physicsof Fluids, 7, 7-14, 1964.
15. De, M. K. and Hammitt, F. G., "New Method forMonitoring and Correlating Cavitation Noise toErosion Capability", submitted Trans. ASME, J.Fluids Engr., 1981.
16. De, M. K., "Statistical Characteristics of Inci-pient Two Phase Flow for Reactor Noise Diag-nosis", to be submitted.
