Flow Measurement Using Flying ADV Probes -...

Dow

nloa

ded

from

asc

elib

rary

.org

by

UN

IV O

F A

UC

KL

AN

D o

n 08

/18/

14. C

opyr

ight

ASC

E. F

or p

erso

nal u

se o

nly;

all

righ

ts r

eser

ved.

Flow Measurement Using Flying ADV ProbesThomas MacDougal Clunie1; Vladimir I. Nikora2; Stephen E. Coleman3; Heide Friedrich4; and

Bruce W. Melville5

Abstract: An innovative measurement system of “flying” acoustic Doppler velocimeters was designed in order to allow rapid velocitymeasurement over a large flow field. Such measurements are necessary, for example, when measuring over a temporally varying andlocally nonuniform rough bed. The measurement technique was verified by comparison with measurements taken in the same flows usinga traditional stationary probe technique. Comparison showed that the flying-probe approach performs similarly to stationary measurementsin capturing the mean flow field and turbulent fluctuations. The data obtained from flying probe experiments can be used to describe theflow in terms of double-averaged hydrodynamic variables, obtained by averaging in time and spatial domains within a thin slab parallelto the mean bed. Examples are presented of flow measurements over a fixed flat bed, a fixed dune bed, and over mobile developing bedforms. It is shown that near-bed measurements suffer from boundary reflection interference, though affected data can be filtered out basedon the ADV-measured correlation coefficient. Measurement below roughness tops is possible, with in-bed records being detectable byspikes in measured signal-to-noise-ratio and by comparison with measured bed topography.

DOI: 10.1061/�ASCE�0733-9429�2007�133:12�1345�

CE Database subject headings: Flow measurement; Turbulence; Open channel flow; Sand; Bedforms.

Introduction

Natural open-channel flows typically move over irregular roughbeds. Although such flows can often be considered uniform in thebulk sense, the time-averaged flow may be spatially heteroge-neous, meaning measurements taken at a single vertical will notrepresent the whole flow, especially in the near-bed region. Todescribe the flow field as a whole, spatial averaging is necessary,with flow measurements taken at multiple points along a lengthlarger than local bed irregularities. In the case of a sediment-transporting flow, as well as spatial irregularity of the bed, the bedconfiguration may be changing with time. Here, the measurementdomain must be sampled over time. Previous studies often usedfixed beds to approximate a single point in development of amobile bed, allowing time for detailed flow measurement. Ex-amples include the fixed sediment–bed-form investigations of Lyn�1993�, McLean et al. �1994�, and Bennett and Best �1995�. To

1Graduate Student, Dept. of Civil and Environmental Engineering,The Univ. of Auckland, Private Bag 92019, Auckland, New Zealand.E-mail: [email protected]

2Professor, Dept. of Engineering, Univ. of Aberdeen, Aberdeen, AB243UE, U.K.

3Associate Professor, Dept. of Civil and Environmental Engineering,The Univ. of Auckland, Private Bag 92019, Auckland, New Zealand.

4Lecturer, Dept. of Civil and Environmental Engineering, The Univ.of Auckland, Private Bag 92019, Auckland, New Zealand.

5Professor, Dept. of Civil and Environmental Engineering, The Univ.of Auckland, Private Bag 92019, Auckland, New Zealand.

Note. Discussion open until May 1, 2008. Separate discussions mustbe submitted for individual papers. To extend the closing date by onemonth, a written request must be filed with the ASCE Managing Editor.The manuscript for this paper was submitted for review and possiblepublication on April 5, 2006; approved on June 15, 2007. This paper ispart of the Journal of Hydraulic Engineering, Vol. 133, No. 12,December 1, 2007. ©ASCE, ISSN 0733-9429/2007/12-1345–1355/
$25.00.
JOURNAL

J. Hydraul. Eng. 2007.1

overcome this restriction, an innovative system of “flying” acous-tic Doppler velocimeter �ADV� probes was devised for the task ofmeasuring a spatially variable and temporally varying flow. Theprobes measure continuously at a high frequency as they aremoved along the length of a laboratory flume, with such flightsoccurring at regular time intervals over the course of an experi-ment. The probes, attached to a moving carriage on rails abovethe flume, are intended to move fast enough with respect to anychanges in bed conditions that each “flight” can be assumed toapproximate a static snapshot of the flow over the dynamicallychanging bed.

A moving-probe approach to flow measurement known as the“flying hot wire anemometer” technique or FHA �Bruun 1995�has been used for some time, from the early study of Cantwell�1976� to more recent laboratory studies such as Hussein et al.�1994� and Cole and Glauser �1998�. FHA uses probes movingagainst the direction of mean flow, removing ambiguity as to thedirection of velocity in areas where flow separation and reversalcan occur. The technique also has the advantage of allowing con-venient probe calibration �Cole and Glauser 1998�. The presentsetup, similar in appearance though differing in purpose, adoptsthe “flying probes” moniker.

In this paper, we first describe the overall experimental set-upfor testing this new ADV application, and then provide back-ground information about the instruments used. This is followedby three examples where the flying-probe technique is tested. Themeasurements presented here were carried out under the NewZealand Programme of Sand Waves and Turbulence Research�SWAT.nz, e.g., Coleman et al. 2007�.

Experimental Setup

The flying ADV probes were set up in a glass-sided flume at theUniversity of Auckland, the flume measuring 440 mm wide,
380 mm deep, and 12 m long. A motorized carriage runs along
OF HYDRAULIC ENGINEERING © ASCE / DECEMBER 2007 / 1345

33:1345-1355.

Dow

nloa

ded

from

asc

elib

rary

.org

by

UN

IV O

F A

UC

KL

AN

D o

n 08

/18/

14. C

opyr

ight

ASC

E. F

or p

erso

nal u

se o

nly;

all

righ

ts r

eser

ved.

rails above the flume, chain-driven by a programmable variable-speed motor. Four ADV probes were mounted on the carriage,together with an array of 32 acoustic depth sounders to measurebed topography and water-surface profile. The ADVs, mounted atdifferent vertical levels as indicated in Fig. 1, simultaneouslymeasured velocity along the flume centerline at different levelsabove the bed. The longitudinal spacing of 120 mm shown in Fig.1 was determined to be the minimum spacing at which sonicinterference from adjacent probes had no noticeable effect onmeasurements. This minimum longitudinal spacing is dependenton probe geometry, vertical probe offset and probe acoustic oper-ating frequency, and thus needs to be reevaluated for a particularexperimental setup.

Measurements were taken continuously as the carriage movedat constant speed up and down the flume. When measuring whenmoving upstream, the relative velocities sampled are higher,which can lead to over-ranging of the ADVs. To eliminate this, ahigher velocity range setting is required, with associated in-creased measurement uncertainties. The depth-sounding probearray had further been found to influence a mobile bed whenmoving upstream, and so was lifted out of the flow for eachupstream traverse. The system was therefore designed to onlycapture flow measurements as the carriage moved downstream,with the carriage moving at a speed slower than the bulk flowvelocity.

The staggering of ADV probes and choice of carriage velocity,together with the fact that the ADVs have a remote samplingvolume 5 cm from the probe tip, all combined to minimize thepossibility of a wake effect from a probe influencing the measure-ments of downstream probes. The probes were positioned toallow measurements up to 5 cm below the highest roughnesstops, ensuring that the probes did not hit the bed during flight.

Carriage position over the measurement field was linked totiming in the continuously recorded ADV data file by microswitches located at either end of the measuring field. Theseswitches, tripped by the carriage as it passed, were wired into thekeyboard of the computer recording the ADV velocity data. Eachtrip of a switch was recorded as a flag at a particular time in thevelocity record. This allowed high accuracy �±0.01 s given a50 Hz sampling rate� recording of the time the carriage passed

Fig. 1. Flying-probe instrumentation comprising three-dimensionaldownward-looking ADVs measuring at multiple levels together withdepth sensors attached to a carriage on rails above a laboratory flume

given positions on the flume, simplifying later analyses. Using

1346 / JOURNAL OF HYDRAULIC ENGINEERING © ASCE / DECEMBER 20


this timing technique, the mean velocity of the carriage was ableto be checked to ensure it remained steady over the course of anexperiment.

The probes were aligned with the direction of carriage move-ment, meaning that the streamwise velocity u at each measure-ment point is simply a sum of the measured velocity um and thecarriage velocity uc, assuming this is steady. The lateral v andvertical w velocities are as measured, vm and wm, namely

u = um + uc, v = vm, w = wm �1�

Probe alignments were checked before measurements were taken,with recorded velocities for a downstream still-water carriageflight checked to see that the average lateral velocity tended tozero. Probe vertical alignment was checked with a spirit level.

The impact of carriage motion on measured velocity fluctua-tions was investigated by recording velocities in still water, thathad been earlier seeded and well-mixed, with the flying probesetup. Still-water velocity variances for a stationary ADV and forthe moving carriage are shown in Fig. 2. The velocity variance forthe stationary measurements, due to Doppler noise and possibleresidual motion in the water, can be seen to be essentially identi-cal to that when the carriage is moving. This suggests that anyvelocity fluctuations caused by the carriage motion were insignifi-cant compared with intrinsic Doppler noise for the present tests,even more so for the tests of flowing water, for which sources ofDoppler noise are more abundant �Voulgaris and Trowbridge1998; McLelland and Nicholas 2000�.

Instrumentation

ADVs

Flow velocities were measured using four SonTek 10 MHzADVs, measuring synchronously at different vertical levels. TheADVs measure three-dimensional point velocities at a remotesampling volume by measuring the Doppler shift of acousticwaves in backscatter from particles in the flow. The samplingvolume is approximately cylindrical in shape, with a diameter of6 mm, corresponding to the diameter of the transmitting trans-ducer, and a height of 7.2 mm using the regular pulse-receivewindow �SonTek 2006�. The ADV samples velocities of particlesin the flow relative to the probe, which are assumed to be equiva-lent to flow velocities. In the present experiments, these seedingparticles were the fines found naturally in the bed sediments used,or artificially-seeded hollow glass spheres �of diameter 8 �m, asprovided by SonTek�.

The ADV uses pulse-to-pulse coherent Doppler backscatter todetermine velocities. Transmitted acoustic pulses are reflected offsmall particles in the water, their phase Doppler shifted. The dif-ference in phase between a pair of received reflections is used tocalculate a velocity along the receiver axis. The three axial ve-locities, measured sequentially, are transformed to give a velocityin three-dimensional Cartesian coordinates �Kraus et al. 1994�.The ADV uses two pulse pairs for each velocity measurement or“ping,” allowing for calculation of a correlation coefficient foreach receiver every ping. The correlation coefficient gives a mea-sure of data quality �SonTek 1997�. The use of different timedelays between pulses in each pair enables separation of the con-tributions of signal and noise in the sampled reflection �Lhermitte
and Serafin 1984�. More details on ADV operational principles
07

33:1345-1355.

Dow

nloa

ded

from

asc

elib

rary

.org

by

UN

IV O

F A

UC

KL

AN

D o

n 08

/18/

14. C

opyr

ight

ASC

E. F

or p

erso

nal u

se o

nly;

all

righ

ts r

eser

ved.

and performance can be found in Nikora and Goring �1998�,Voulgaris and Trowbridge �1998�, McLelland and Nicholas�2000�, and García et al. �2005�.

ADVs encounter problems when measuring at certain dis-tances close to a boundary �Table 1� that may affect near-bedmeasurements. The problem occurs when the first set of acousticpulses reflect off the boundary and are received by the probe atthe same time as the second set of pulses in the pulse-pair isreflected from particles in the sampling volume and received. Thetwo different time lags �1 and �2 between pulses in successivepulse pairs mean that there are two distinct distances zbi at whichthis boundary interference occurs, given by

zbi = �icwater/2 �2�

where cwater�speed of sound in water.These distances, calculated from Eq. �2� with default time lags

and confirmed by correspondence with Sontek, are shown inTable 1. The time lags are roughly user set by adjusting the ve-locity range in the ADV software �as lag times also control thelargest velocities measurable�. When starting measurements, theADV software measures the distance to the boundary, and adjuststhe lag times to minimize potential interference problems

Fig. 2. Carriage motion as

Table 1. Indicative Weak Spot Distances for User-Set Velocity Ranges

Velocity range�cm/s�

Boundary distance �mm�a

Early model probes Current Field ADV probes

3 300, 600 480, 960

10 180, 360 240, 480

30 84, 180 120, 240

100 36, 96 36, 84

250 21, 36 30, 78aDistances from experimental trials, and confirmed by correspondence
with SonTek.
JOURNAL


�SonTek 1997�. Avoidance of reflection interference may not bepractically possible when measuring over a varying bed levelhowever. In the flying-probe approach, the distance between thesampling volume and boundary is constantly changing and devi-ating from the initial distance the pulse timing was optimized for.This effect should be kept in mind when interpreting the datafrom flying ADV probes.

Measuring near to an irregular bed will mean that sometimesthe sampling volume is immersed within the bed, yielding invalidvelocity measurements which must be removed from the record.Such segments of records can be easily identified by a significantrise in the signal-to-noise ratio �SNR�, as the probe receives avery strong reflection from the boundary. In the current setup,in-bed records were roughly identified by comparison of the sam-pling volume elevation with the measured bed topography. Pre-cise filtering out of in-bed measurements was then based onexceedance of a maximum threshold for SNR.

Depth Sounders

Attached to the carriage and measuring concurrently with theADV probes was an array of 32 ultrasonic depth-sensing probes�Fig. 1�, developed by Seatek Inc. One of these probes was up-ward looking, giving a measure of water surface slope over thetest length. The remaining probes were focused on the sand bed,measuring changes in bed topography �Friedrich et al. 2005�. Tosynchronize the ADV and Seatek measurements at the com-mencement of an experimental run, a fixed object �measured bythe depth sensors and identified by a spike in SNR for the ADVmeasurements as the sampling volume passed through the object�was traversed and then removed from the flume.

Acoustic interference between ADV and depth-sensing probeswas not observed for the present experimental results, due to thephysical separation of the probes, and the fact that the instrumentsoperate at different acoustic frequencies, namely—10 MHz for

red by ADVs in still water
measu
the ADVs and 5 MHz for the depth sensors.


33:1345-1355.

Dow

nloa

ded

from

asc

elib

rary

.org

by

UN

IV O

F A

UC

KL

AN

D o

n 08

/18/

14. C

opyr

ight

ASC

E. F

or p

erso

nal u

se o

nly;

all

righ

ts r

eser

ved.

Data Analysis

The velocity data record for an experiment, contained in a singlefile, was easily separated into “flights” with the inclusion in thesaved file of the flags from microswitch passing. Each flight con-tained a series of single-point velocity measurements at a longi-tudinal spacing s dependent upon carriage speed uc and ADVsampling frequency fs, specifically:

s = uc/fs �3�

The double-averaging approach for rough-bed flows �Nikora et al.2001; 2007a,b� provides a framework within which to analyze themeasured velocity data. Similar to Reynolds’ decomposition, aninstantaneous velocity u at a level z above the mean bed level canbe thought of as a double-averaged �in time and space� velocity�u� combined with a spatial fluctuation u �the difference betweenlocal time-averaged velocity and the double-averaged velocity�and a temporal fluctuation u�, i.e.,

u�x,y,z,t� = �u��z� + u�x,y,z� + u��x,y,z,t� �4�

The limitation of the flying-probe data record compared withmore traditional stationary ADV sampling techniques is that therecan be no time-averaging of flow at a certain position. This limits

Table 2. Experimental Properties

Experimentalrun

Averagevelocity�m/s�

Shearvelocity�m/s�

Flowdepth�mm� Bed sediment

Flat bed 0.56 0.0235 150 Sand,d50=0.8 mm,

fixed

Fixed dunes 0.50 0.0459 120 Sand,d50=0.8 mm,

fixed

Mobile bed�run ndf14a�

0.38 0.0220 240 Sand,d50=0.24 mm

Fig. 3. Comparison of mean velocity profile for flyin



the decomposition of data into a double-averaged component anda fluctuating component arising from variations in both time andspace, i.e.,

u�x,y,z,t� = �u��z� + u��x,y,z,t� �5�

where u� cannot be further decomposed into u and u�. For thepresent application of describing the evolution of flow over amobile boundary, the change with time of these reduced compo-nents �u� and u� may be investigated. Over the present fixed beds,however, multiple flights provide an ensemble of points at eachspatial location allowing for separate determination of the fluctu-ating components u and u�.

Verification of Flying-Probe System for Fixed Beds

Tests over a Flat Bed

The system was tested over a flat bed to verify its basic operationin a typical rough-bed boundary-layer flow. Flow conditions forthis flow and later tests are provided in Table 2. The bed was a flatsurface covered with glued 0.8 mm sand. Measurements along theflume centerline were made with the flying-probe system, withthe “flight length” extending 3.75 m from the starting point5.75 m downstream of the flume entrance. The flight took 15 s ata carriage speed of 0.250 m/s, with an extra 2.5 s of carriagemovement either side of the measured flight length allowing forcarriage acceleration and deceleration. Measurements were alsotaken using an ADV stationary in the flow, measuring for a pro-longed period at a number of elevations. For both approaches, theprobes sampled at 50 Hz with a user-set velocity range of100 cm/s. A mean velocity profile from the two measurementtechniques was compared in terms of u�z� for the stationary probeand �u��z� for the flying probes. The results, shown in Fig. 3, werefound to match closely, where dimensionless elevation z+

=zu* /�, where u*�bedshear velocity and ��fluid kinematic vis-cosity.

e and stationary-probe measurements over a flat bed
g-prob
07

33:1345-1355.

Dow

nloa

ded

from

asc

elib

rary

.org

by

UN

IV O

F A

UC

KL

AN

D o

n 08

/18/

14. C

opyr

ight

ASC

E. F

or p

erso

nal u

se o

nly;

all

righ

ts r

eser

ved.

For the fixed-bed runs, flume slope was adjusted to give uni-form flow conditions. Shear velocities were calculated in thereach-averaged sense, from the product of the energy slope Se andthe hydraulic radius of the flow R, namely

u* = �gRSe �6�

where g�gravitational acceleration.

Normalized turbulence intensity u� /u* was compared for thestationary-ADV and flying-probe techniques, assuming that u=0for the flying case �appropriate for a flat bed�, with good agree-ment between the results �Fig. 4�. Also presented in Fig. 4 is theempirically derived relationship for turbulence intensity sug-gested by Nezu and Nakagawa �1993� of

u�/u* = 2.30e�−z/H� �7�

where H�flow depth.The measurements compare reasonably well with the relation-

ship of Eq. �7�, once they have been corrected for Doppler noise�e.g. by identifying the noise floor from autospectral analyses,�Nikora and Goring 1998; Voulgaris and Trowbridge 1998�. How-ever, one may also note that our data are systematically higherthan that predicted by Eq. �7�. This difference is probably due tothe combined effect of remnants of measurement noise, secondarycurrents, and errors in shear velocities used for normalization �i.e.the bulk estimate, Eq. �6�, may be different from the local center-line value of shear velocity�. The uncorrected data highlight theeffect that changing lag time has on ADV noise levels andsubsequently velocity variance. Between z+=700 and 850�z=30–35 mm�, where lag times �i are automatically reduced toavoid boundary interference �akin to increasing the velocityrange�, Fig. 4 shows a rise in measured velocity variance for boththe stationary-ADV and flying-probe techniques.

A comparison of Reynolds stress �u�w�� profiles for the twotechniques �Fig. 5�, shows that both measurement techniques per-

Fig. 4. Comparison of normalized turbulence intensity profile forsymbols represent data corrected for Doppler noise, open symbols re

form similarly, providing Reynolds stress estimates to within

JOURNAL


±25% of theoretical expectations for the two-dimensional uni-form flow. However, some deviations from the theoretical lineardistribution could be expected due to the effects of secondarycurrents. The relative difference between values of Reynoldsstresses from flying and stationary probes are within measurementerrors related to the limited sample lengths.

Tests over Idealized Dunes

To test the ability of the flying-probe setup to accurately measurevelocities in a more complex flow field, flying-probe runs weremade for a flow over a bed of fixed idealized dune shapes, withresults compared with stationary ADV measurements for the sameflow �Coleman et al. 2006�. These dunes have a two-dimensionalshape, spanning the width of the flume, with a wavelength of750 mm repeated along the flume. The dunes have a curvingcosine-shaped upstream slope of length 680 mm and height40 mm, with a straight downstream face at an angle of approxi-mately 30° from horizontal. The dune shapes, coated with a0.8 mm sand, reflect the shape and size of bed forms that can beexpected to form in equivalent mobile-bed tests.

The varying boundary introduced new complexities for testingthe flying-probe system. Measurements when the sampling vol-ume was within the bed needed to be removed from the recordeddata below roughness tops. Erroneous measurements caused byboundary reflection interference also needed to be correctly dealtwith.

Fig. 6 shows that in-bed ADV measurements can be detectedby a rise in SNR. The heightened level of SNR, however, de-creases as the sampling volume is immersed deeper within thebed. Testing of the trend in SNR with sampling volume immer-sion showed that when the centre of the sampling volume wasgreater than 10 mm within the bed, SNR levels were no longerelevated above normal measuring levels. Filtering of in-bed mea-surements were therefore accomplished first by comparison with

-probe and stationary-probe measurements over a flat bed. Closedt uncorrected data.

flyingpresen

the measured bed profile, removing velocity measurements that


33:1345-1355.

Dow

nloa

ded

from

asc

elib

rary

.org

by

UN

IV O

F A

UC

KL

AN

D o

n 08

/18/

14. C

opyr

ight

ASC

E. F

or p

erso

nal u

se o

nly;

all

righ

ts r

eser

ved.

were well within ��10 mm below� the bed, and second by usingthe SNR spikes as precise markers of when the sampling volumeentered or exited the bed.

The correlation coefficient was not found to be useful in indi-cating boundary position. Correlation coefficients were found tovary greatly over the length of a dune, decreasing in areas of highturbulence and in areas affected by bed reflection. Although cor-relations did decrease near to the boundary, as noted by Martin etal. �2002�, they then increased with the sampling volume partiallyimmersed within the bed.

Fig. 7 shows the detection of bad data caused by bed reflectioninterference, as identified by a drop in the correlation coefficient.For the measurements shown of a velocity range setting of

Fig. 5. Comparison of Reynolds stress profile for flyi

Fig. 6. Average SNR recorded by an ADV for flying



±100 cm/s, this occurred when the sampling volume was be-tween 30 and 40 mm from the bed. This matches the value ofzbi=36 mm from Table 1, based on Eq. �2� and confirmed bySonTek, the observed vertical extent of this “weak spot” beingdue to the size of the sampling volume. SonTek �1997� suggestthat the correlation coefficient should ideally be above 70% foracceptable data, while McLelland and Nicholas �2000� suggestthat the correlation for any single receiver must be above 60% toobtain valid readings. Removal of data with low correlation val-ues is routine in analysis of ADV data. The present data werefiltered for a minimum correlation of 60% for any receiver and70% for the average correlation, which was found to acceptablyremove any bad data caused by bed reflection interference.

be and stationary-probe measurements over a flat bed

rements taken at an elevation below roughness crests

ng-pro

measu

07

33:1345-1355.

Dow

nloa

ded

from

asc

elib

rary

.org

by

UN

IV O

F A

UC

KL

AN

D o

n 08

/18/

14. C

opyr

ight

ASC

E. F

or p

erso

nal u

se o

nly;

all

righ

ts r

eser

ved.

After removing in-bed records, and filtering bad data, flying-probe profiles of mean velocity �Fig. 8� and Reynolds stress �Fig.9� were again compared with those from stationary-probe mea-surements �Coleman et al. 2007�, where bulk flow properties aregiven in Table 2. In Figs. 8 and 9, h is the dune height and theorigin of the vertical coordinate z is at the mean bed level, mean-ing data from z /h=−0.5 to 0.5 are measured beneath roughnesstops. The flying-probe centerline measurements spanned 6 dunes,the 8th to 13th in a series of identical dunes, with the carriagetraveling at 150 mm/s. The results presented in Figs. 8 and 9 arethe ensemble averages of 10 flights or 60 dune lengths. Thestationary-ADV measurements were taken over the length of the12th dune in the series, at 20–40 mm spacings �5 mm verticalspacings� along the centerline for a duration of 2 mins at eachpoint. A stationarity analysis confirmed that the 2 min sampling

Fig. 7. Average ADV correlation coefficient for flying measuremeninterference

Fig. 8. Comparison of mean velocity profile for flying-probe and statments spanning from dune troughs to one dune height above crest le

JOURNAL


durations were sufficient sampling time for mean velocity, veloc-ity variance, and Reynolds stress values to converge. Both mea-surement techniques used a sampling frequency of 50 Hz.

The comparisons of Figs. 8 and 9 show that the flying-probesystem provides a good representation of the flow field over thefixed dunes. The advantages of the flying-probe system are theease and speed of measurements, providing a spatial resolution ofmeasuring points that would have been extremely time intensiveto achieve with a stationary probe. The drawback is the preva-lence of weak spots from boundary interference occurring withflying probes. Weak spots are avoided for stationary probesthrough automatic adjustment of lag times by the ADV software.With flying probes over a varying bed level, bad data from bound-ary reflection interference are inevitable, however, and filteredresults subsequently contain gaps from removed data. Leaving

n at an elevation above roughness crests, identifying bed reflection

-probe measurements over a fixed idealized dune bed, with measure-

ts take

ionaryvels


33:1345-1355.

Dow

nloa

ded

from

asc

elib

rary

.org

by

UN

IV O

F A

UC

KL

AN

D o

n 08

/18/

14. C

opyr

ight

ASC

E. F

or p

erso

nal u

se o

nly;

all

righ

ts r

eser

ved.

such gaps in the record will bias double-averaged results, as thegaps are correlated with bed geometry. For the present compari-sons, these gaps were filled with linearly interpolated artificialdata.

Flying-Probe Measurements over a Mobile Bed

Mobile-bed tests were carried out to capture information on theflow field over a sand bed under bed-form-generating flows�SWAT.nz, Coleman et al. 2007�, with bed development fromplane-bed conditions being measured in the present 440 mm wideflume for a range of flow rates, flow depths and sediment types.For such dynamic-bed phenomena, irregular local variations inflow due to the presence of moving, growing bed forms necessi-tate measurement over a large area in a short time. The flying-probe system is ideal for this application of measuring andparameterizing a developing flow field.

The carriage carrying the flying probes moved downstreamover a measurement length of approximately 4.5 m in 18 s. Withthe rate of change of bed forms small in comparison with eachflight time, averaged flow quantities from each flight can be takento describe the flow at a particular stage of bed development.Changes with bed development in the distributions of velocitymoments with depth, or the profile of average fluid stresses withdepth, can thereby be captured with this flying-probe setup, offer-ing significant insight into the development of flow over a mobilesandy bed.

As bed forms developed for the mobile-bed tests, flow resis-tance, and thus the water surface slope, increased. The flumeslope and boundary condition for flow depth were adjusted togive uniform flow for the fully developed bed forms, and thusflow was nonuniform early on in the development period. Water-surface slope measurements were used to monitor the develop-ment of flow resistance, as indicated by the reach-averaged shear

Fig. 9. Comparison of Reynolds stress profile for flying-probe anmeasurements spanning from dune troughs to one dune height above

velocity, expressed in Eq. �6�.



Indicative results for test ndf14a of the “SWAT.nz” experimen-tal runs are shown in Fig. 10, where the test name expresses runproperties �narrow flume, deep flow, fine sediment� that are pre-sented in Table 2. The double-averaged velocity at elevation z=46 mm above the mean bed level is shown over time as bedforms develop in Fig. 10�a�. Fig. 10�b� reflects the developmentof the bed configuration, from a flat bed to equilibrium-sized bedforms, showing the change in standard deviation of bed elevationsover time. Also presented is the change in time of spatially aver-aged velocity fluctuations �u�

2 �0.5= ��u+u��2�0.5 and �w�2 �0.5

= ��w+w��2�0.5 at z=46 mm �Fig. 10�c��. Fig. 10�c� shows that theoverall fluctuations increase as bed forms grow, due to the emer-gence of form-induced fluctuations as well as increases in turbu-lence. The development of shear velocity for the run is presentedin Fig. 10�d�. This series of plots shows that for experimentndf14a, the major hydrodynamic changes occurred in the first40–50 mins, with the slower-developing bed forms having onlyreached perhaps 50% of their final height over this period.

The double-averaged velocity time series from all four probesmeasuring synchronously at different levels for run ndf14a arecombined in Fig. 11 to give a contour plot of velocity-profiledevelopment with bed form growth. The probes were located atlevels of 5, 32, 59, and 86 mm above the initial mean bed level,with the contours being obtained by interpolation. In particular,this plot shows the slowing of near-bed flow as the bed formsgrow into the measurement domain.

Conclusions

A flying-probe system using ADV probes attached to a steadilymoving carriage was devised to measure velocities rapidly over alarge measurement area, the system using a set of probes to mea-sure velocities simultaneously at multiple vertical levels, includ-

ionary-probe measurements over a fixed idealized dune bed, withlevels

d statcrest

ing levels beneath roughness tops. Means of correcting measured

07

33:1345-1355.

Dow

nloa

ded

from

asc

elib

rary

.org

by

UN

IV O

F A

UC

KL

AN

D o

n 08

/18/

14. C

opyr

ight

ASC

E. F

or p

erso

nal u

se o

nly;

all

righ

ts r

eser

ved.

data for in-bed measurements and boundary-reflection interfer-ence are identified. Measurement records may be resolved intodouble-averaged �in time and space� and fluctuating components,

Fig. 10. Indicative results over a mobile bed showing: �a� time ddevelopment through standard deviation of bed elevations; �c� deve�w�

2 �0.5 �in the streamwise and vertical directions, respectively� at an e�determined from measured water surface slope�

which can be used to describe the average flow. Examples are

JOURNAL


presented of flow measurements over a fixed flat bed, a fixed dunebed, and developing bed forms in a mobile sediment bed. Thesystem was found to perform as accurately as stationary probes in

ment of �u� at an elevation 46 mm above the mean bed; �b� bednt of spatially-averaged fluctuating components of flow �u�

2 �0.5 andn 46 mm above the mean bed; and �d� development of shear velocity

eveloplopmelevatio

measuring mean flow and fluctuating flow components.


33:1345-1355.

Dow

nloa

ded

from

asc

elib

rary

.org

by

UN

IV O

F A

UC

KL

AN

D o

n 08

/18/

14. C

opyr

ight

ASC

E. F

or p

erso

nal u

se o

nly;

all

righ

ts r

eser

ved.

Acknowledgments

This research was funded by the Marsden Fund �UOA220� ad-ministered by the Royal Society of New Zealand. The writerswould like to thank Stuart Cameron for his help in preparingfigures, and his helpful review of the paper. The helpful sugges-tions of anonymous reviewers are gratefully acknowledged.

Notation

The following symbols are used in this paper:cwater � speed of sound in water;

d50 � median sediment size;fs � ADV sampling frequency;g � gravitational acceleration;H � flow depth;h � bed-form height;R � hydraulic radius;Se � energy slope;s � longitudinal spatial resolution of flying

measurements;t � time;

u ,v ,w � instantaneous velocity streamwise, spanwise,vertical;

u � spatial fluctuation in time-averaged velocity;u� � temporal fluctuation in velocity;u* � bed-shear velocity;uc � streamwise carriage velocity;

um, vm, wm � point velocity measured by ADV;u�, w� � combined spatial and temporal velocity

fluctuation, i.e., u�= u+u�;x ,y ,z � longitudinal, cross-stream, and vertical

coordinates;z+ � dimensionless elevation;zbi � elevation at which boundary reflection

interference occurs;�1, �2 � time lags between pulses in an ADV pulse

Fig. 11. Development of double-avera

pair;



� � fluid kinematic viscosity; and�u� � double-averaged velocity;

References

Bennett, S. J., and Best, J. L. �1995�. “Mean flow and turbulence structureover fixed, two-dimensional dunes: Implications for sediment trans-port and bedform stability.” Sedimentology, 42�3�, 491–513.

Bruun, H. H. �1995�. Hot-wire anemometry. Principles and signal analy-sis, Oxford University Press, Oxford, U.K.

Cantwell, B. J. �1976�. “A flying hot-wire study of the turbulent nearwake of a circular cylinder at a Reynolds number of 140 000.” Ph.D.dissertation, California Institute of Technology, Pasadena, Calif.

Cole, D. R., and Glauser, M. N. �1998�. “Flying hot-wire measurementsin an axisymmetric sudden expansion.” Exp. Therm. Fluid Sci., 18�2�,150–167.

Coleman, S. E., Nikora, V. I., McLean, S. R., Clunie, T. M., Schlicke, E.,and Melville, B. W. �2006�. “Equilibrium hydrodynamics concept fordeveloping dunes.” Phys. Fluids, 18�10�, 105104-1–12

Coleman, S. E., Nikora, V. I., Melville, B. W., Goring, D. G., Friedrich,H., and Clunie, T. M. �2007�. “New Zealand programme of sandwaves and turbulence research: SWAT.nz.” 6th Int. Symp. on Ecohy-draulics, Christchurch, New Zealand �poster on CD�.

Friedrich, H., Melville, B. W., Coleman, S. E., Nikora, V. I., and Clunie,T. M. �2005�. “Three-dimensional measurement of laboratory sub-merged bed forms using moving probes.” Proc., 31st IAHR Congress,Seoul, Korea, 396–404.

García, C. M., Cantero, M. I., Niño, Y., and García, M. H. �2005�. “Tur-bulence measurements with acoustic Doppler velocimeters.” J. Hy-draul. Eng., 131�12�, 1062–1073.

Hussein, H. J., Capp, S. P., and George, W. K. �1994�. “Velocity mea-surements in a high-Reynolds-number, momentum-conserving, axi-symmetric, turbulent jet.” J. Fluid Mech., 258, 31–75.

Kraus, N. C., Lohrmann, A., and Cabrera, R. �1994�. “New acousticmeter for measuring 3D laboratory flows.” J. Hydraul. Eng., 120�3�,406–412.

Lhermitte, R., and Serafin, R. �1984�. “Pulse to pulse Doppler sonarsignal processing techniques.” J. Atmos. Ocean. Technol., 1�4�, 293–308.

locity profile over growing bed forms
ged ve
Lyn, D. A. �1993�. “Turbulence measurements in open-channel flows

07

33:1345-1355.

Dow

nloa

ded

from

asc

elib

rary

.org

by

UN

IV O

F A

UC

KL

AN

D o

n 08

/18/

14. C

opyr

ight

ASC

E. F

or p

erso

nal u

se o

nly;

all

righ

ts r

eser

ved.

over artificial bed forms.” J. Hydraul. Eng., 119�3�, 306–326.Martin, V., Fisher, T. S. R., Millar, R. G., and Quick, M. C. �2002�.

“ADV data analysis for turbulent flows: Low correlation problem.”Proc., Conf. on Hydraulic Measurements and Experimental Methods�CD-ROM�, Estes Park, Colo.

McLean, S. R., Nelson, J. M., and Wolfe, S. R. �1994�. “Turbulencestructure over two-dimensional bed forms: Implications for sedimenttransport.” J. Geophys. Res., [Oceans], 99�C6�, 12729–12747.

McLelland, S. and Nicholas, A. �2000�. “A new method for evaluatingerrors in high frequency ADV measurements.” Hydrolog. Process.,14�2�, 351–366.

Nezu, I., and Nakagawa, H. �1993�. Turbulence in open-channel flow,Balkema, Rotterdam, The Netherlands.

Nikora, V. I., and Goring, D. G. �1998�. “ADV measurements of turbu-lence: Can we improve their interpretation?” J. Hydraul. Eng.,124�6�, 630–634.

Nikora, V. I., Goring, D. G., McEwan, I., and Griffiths, G. �2001�. “Spa-tially averaged open-channel flow over rough bed.” J. Hydraul. Eng.,

JOURNAL


127�2�, 123–133.Nikora, V. I., McEwan, I. K., McLean, S. R., Coleman, S. E., Pokrajac,

D., and Walters, R. �2007a�. “Double-averaging concept for rough-bed open-channel and overland flows: Theoretical background.” J.Hydraul. Eng., 133�8�, 873–883.

Nikora, V. I., McLean, S. R., Coleman, S. E., Pokrajac, D., McEwan, I.K., Campbell, L., Aberle, J., Clunie, T. M., and Koll, K. �2007b�.“Double-averaging concept for rough-bed open-channel and overlandflows: Applications.” J. Hydraul. Eng., 133�8�, 884–895.

SonTek. �1997�. Acoustic Doppler velocimeter (ADV) principles of op-

eration, San Diego.SonTek. �2006�. “Setting the record straight: ADV data acquisition rates

and sampling volume size.” �http://www.sontek.com/apps/lab/advdarsv/advdarsv.htm� �Mar. 29, 2006�.

Voulgaris, G., and Trowbridge, J. H. �1998�. “Evaluation of the acousticDoppler velocimeter �ADV� for turbulence measurements.” J. Atmos.Ocean. Technol., 15�1�, 272–289.


33:1345-1355.

Flow Measurement Using Flying ADV Probes -...

Documents

Transcript of Flow Measurement Using Flying ADV Probes -...