110007466329

The Japanese Geotechnical Society

NII-Electronic Library Service

TheJapaneseGeotechnical Society

SOILS AND FOUNDATIONSJapanese Geetechnical SocietyVol.

49, No. 5. 699-709, Oct. 2e09

EVALUATION OF SASW TEST CONFIGURATIONS ANDASSOCIATED DATA UNCERTAINTIES IN GENERATING

SITE SPECIFIC DISPERSION CURVES

MousuMI MuKHERJEEi) and Apagl'r PRAsHANTti)

ABSTRACT

Spectral analysis of surfaee waves (SASW) technique is an in-situ nondestructive method used for near surface soilsite profiling. In the present study experimental uncertainties related to this mcthod are assessed in connection with testconfiguration specification and data quality. Field experimentations have been conducted for this purpose with variousspatial configurations of source to near receiver distance and inter-receiver distance under various impact source mag-nitudes and heights of fall, Based on the experimental findings, a method has been proposed to identify the frequencyrange ofinterest based on a threshold minimum cross power spectrurn value. Further, it is proposed to consider subsetstacking combination with good data quality which can be estimated based on weighted evaluation of the coherence ofeach frequency and importance of different frequency ranges. A procedure of obtaining the site specific dispersioncurve is illustrated with due consideration to the issues associated with data quality and power spectra,

Key words: coherence, dispersion,power spectra, rayleigh wave,SASW Test (IGC: C21C81D7)

INTRODUCTION

Near-surface soil site characterization is an importantissue related to various types of geotechnical constructione,g., highway or railway track foundation and earth-quake related problem. In-situ tests are widely used fordetermination of soil stifiness and material property forthis purpose, Since its introduction in the mid-80s, theSASW (Spectral Analysis of Surface Waves) method hasgained a large role into in-situ testing fer determinationof soil stifftiess at small strains. This method can be usedfor testing a large area of soil deposits economically.SASW test is performed by generating waves of a widerange of frequency under a vertical impact loading,Waves in the form of signals are then captured by sensorsand then soil profiling is done by generating experimentaldispersion curve and appiying subsequent inversionmethod. However, this method is alTected by a largedegree of uncertainties due to various factors such as farfield and near field effects arising from attenuation of sar-face wave and body wave interference, mechanical andelectrical noise at low frequencies, and specifieations ofdata acquisition add to such uncertainties. As a result,the construction of the experimental dispersion curve isstrongly affected by the operator's experience and deci-sion making for significant and corrected data actually re-quired for interpretations. It leads to the analysis being aconsiderably time consuming process and involves sig-

nificant human intervention and manipulation during theprocess,

Heisy et al. (1982a) first mentioned the concept ofSASW. Nazarian and Stokoe (1986, 1987) developed theexperimental and theoretical aspects ef the SASWmethod as applied to geotechnical and pavement en-

gineering. They developed a suitable method for conduct-ing the in-situ tests. Sayyedsadr and Dmevich (1989) de-veloped a software code SASWOPR to operate on SASWdata for generating a combined dispersion curve. As co-herence and phase values were directly used as input, dataquality assessment depending on energy content andrelevance of frequency range could not be considered,and coherence values were the sole signal quality decidingfactor, Some literature is also available addressing the is-sues related to the source and receiver spacing and arran-

gement. Heisy et al. illustrated (1982b) the factors thataffect the source and receiver geometry. Based on the ex-perimental studies, it was suggested that the source tonear receiver distance (S) might bc equal to intcr-receiverdistance (d), provided that wavelengths Z 3d are eliminated from the data. Nazarian and Stokoe(1983) recommended that Common Receiver Midpoint(CRMP) geometry was the most suitable for SASW testin case of soil site due to less scatter in dispersion curvearising due to non-homogeneity. In CRMP geometryreceivers are placed at equal distance apart from an ima-ginary centerline and they are shifted by equal distancc

/)mFormerly Research Eellow, Dept. of Civil Enginccring, Indiai Institute of Techno!egy, Ka"pur, lndia.

Assistant Professor, ditto ([email protected]}.The manuscript for thLs paper was received for revievif on August 1I, 2008; approved on June 8, 2e09,Written discussions on this papei should be submitted before May 1, 201O to the Japancse Gcotechnicat Sociely, 4-3S-2, Seneoku, Bunkyo-ku,Tokyo 112-OOI1, Japan, Upon request tbe closing date may be extended one month.

699

NII-Electionic



TheJapaneseGeotechnicalSociety

700 MUKHERJEE AND PRASHANT

from that line after each test set up during field test, as il-lustrated in Fig. 1. Hence, the soil site under testingremains the same after the whole test assuring relativelyhomogeneous soil site condition. Sanchez-Salinero et al. (1987) analytically studied themost feasible source-receiver configuration. They indicat-ed that a desirable distance between the sQurce and thenear receiver is equal to the distance between the adjacentreceivers and for that set-up the wavelengths (A) consi-dered during analysis of the field data were suggested tobe equal to or less than one half of the distance betweenthe receivers (A




EVALUATION OF SASW TEST CONFIGURATIONS 701

:zo,e 'ws

Guide -Rod

BasePIateX

'ssl

,Heightof

fall

EVALUATION OF DATA QUALITY VVITHSTACKING OF SIGNALS

One of the main objectives of the present study was tofind out quality data subset from the stacking done foreach configuration. Considering various factors inducingthe uncertainty of data, the procedure of evaluating thedata quality for SASW test interpretation was divided infour major steps discussed subsequently.

tp

Fig, 3, Schematic diagram ef the londing arrangement

100axE)=glogo

1

../.1

l

Shuntdamping34%

::4.tt t-//1 ... t/t tttt tttttttt t /t t

t t/ tt II / /ttt /tttt/t ttt ttt ,1-,:...:1:1/F.1, 11/'i.'ili/,..111i'i.. .,//111.11i

t./ .t/tl. 1..

ttttt/tt /ttt t -. 11/ ../.. ./. . .t:.i,. t. .l. t. 1111' //. / //

1 10Frequency(Hz) 100

Fig. 4. 0utput response curve for thc geophone used during test

data stacks were recorded for each setup configuration.The SASW control unit was used for data recording dur-ing the test. Two geophones with 4.5 Hz natural frequency wereused as sensors. Figure 4 gives the response curve of thegeophones used in the present study with shunt resistanceof 34%. The shunt resistance controls electrical dampingof the geophone. A large value of shunt resistancedecreases the sensitivity of sensor to mechanical vibrationand at the same time reduces amplification in responsedue to natural frequency of geophone, Though the linearoutput is obtained from 20 Hz, the minimum frequencyrange considered for the present analysis is twice thenatural frequency, i.e., 9 Hz to avoid loss of lower fre-quencies which are important for deep soil site profiling.The nonlinearity due to electrical resonance of the geo-phone for the frequency range 9-20 Hz may not sig-nificantly affect the obtained results. Below 9 Hz, outputvalue does not truly represent the variation due to inputsignal and original signal along with the ambient noise getamplified creating swamp in the output signal. The maxi-mum frequency required for analysis depends upon thedepth of interest and in case of near surface soil sitecharacterization the minimum depth of interest is ap-proximately O.5m, It is observed that 150-200 Hz fre-quencies are able to yield waves of such depths (Nazarian,1984; Foti, 2ooO; Ganji et al., 1998). For the presentstudy maximurn frequency considered is 200 Hz.

intet]pretation of Coherence Fleinction Generally coherence function is used to assess dataquaLity at any fyequency. It is a measure of the degree bywhich two signals are linearly correlated. Coherencefunction value close to unity is considered as a good qual-ity data. Nazarian and Stokoe (1986) considered a mini-mum acceptable value of coherence function O.9 i) theiranalysis for pavement system, Sayyedsadr and Drnevieh(1989) used a default value of O.98 as a minimum coher-ence rnagnitude. Al-Hundai (1992) used a minimumfiltering criterion of O,9 for coherence funetion. For thepresent analysis threshold minimum coherence value wasconsidered as O.98 to ensure good data quality. The phasedifference between two signals may be more than once cy-cle (2n); however, the phase angle is calculated to have itsvalues between -z to +n. These values, which arereferred to as wrapped phase angle, are corrected to in-clude the number of missing cycles based on an expectedresponse of phase distribution. Such process of the phasecorrection is called unwrapping in which the phase mag-nitude is shifted by 2n for each missing cycle, Figure 5presents coherence and wrapped magnitudes of phase an-

gle, each varying over the frequency range 9-200 Hz, forgood and bad quality data. It can be observed that fre-quencies with low coherence (Fig. 5) lead to irregularjumps in the wrapped phase which can lead to misinter-pretations in phase velocity calculations during furtheranalysis. However, such irregular jumps were sometimespresent in the wrapped phase in spite of high coherencevalues corresponding to those frequencies. These may bedue to the low power content of both the signals resultingin a good coherence value, though the signal is affeeted bythe background noise. Such irregular jumps can also benoted at soil layer interfaces due to sharp change of medi-um properties. These issues are further discussed later un-der the section on Signal 9uatity Based on Power Spec-trum.

Dij7lerent Stacking Combinations for Enhancing Data9uality The coherence values were calculated by averaging thespectral quantities of signal over the number of datastacked. Ifit is assumed that the background noise is ran-dom in nature, by averaging the surn of the backgroundnoise in the signals, it can be reduced by a large extent(Nazarian, 1984). Theoretically, the higher the number ofsignals averaged the more enhanced the fina} results willbe. Practically, the number of averages should be op-timized. Law of statistics suggests that the reliability of

NII-Electionic





1

o.v

o,s

O.7Y

e.G:/

o.s8-

o.4

O.3

O.2

".1

o

3.S

2.5E

,.,-;"

e.sE-;e

-o.sk:

-d.sl-2.,'.3.5

Highcohe'enceregio]i

/ (goodql]ali"'datn)

o 50 100Fitq"ency(H-)150 200

1,o"'cnhetence:egion

(ilrc"u]a]jUMPS)

y,./Sk.it..

' .,to '10ri i15.0lttt.t.,t.t..t'Htt..ttttt

1

1'reqi]encl'(IIz}

o

Fig. 5. Coherence amd wrapped phase angle of cross povver spectrH for

test configuration: S=d=4 m, mnss == 10 kg, nnd ht=O.9 m

getting a value closer to the real value by averaging thesignals is inversely proportional to the square root of thenumber of experiments. Based on their experimentalstudy Nazarian (1984) proposed 5 stacking for averagingthe signals.

In the present study experimental data were stacked six

times and the spectral quantities were calculated for vari-ous stack combinations, considering three to six stacksout of six stacks obtained for each setup configuration.Such analysis helped in idcntifying the combination forwhich the noise effects were less in comparison to theother combinations. The good quality data was estimatedthrough increase in magnitude of coherence function.Coherence values for different frequencies are presentedin Fig. 6 for both 3 stack and 5 stack combinations. It isevident that 3 stack results yielded good result than 5stacks. Though the interpreted data quality may increaseby decrease in the number of seiected stacks, it is impera-

tive to maintain a minimum number of stacking to ensureconfidence in representatiyeness of data. It was observedfrom the prescnt study that four stacks combination maybe taken to satisfy both coherence and representativityconditions.

Iilnding the Best Stacking Combination Based on Cu-mutative VVeight

lt was required to establish a qualitative index to getthe best stacking combination out of all combinations un-der consideration for a setup configuration. Such indexwas calculated over the whole frequency range (f,) of in-

1

G.9 1

es

O.T:

O.6;

O.5t

O.4

Q.3

O.2

O.1

o

\,,g,,'11h XS: i,Ii For3 st"ckillgs

l/i'/SN'-- Fovs stnckings

e so

f"

ii

",}

l.[EI

----T-= tOO IS()]'tvqucnny'(llz)

100

Fig, 6. Different stacking combinatiolls for e"haiicing data qualit}' (Test Configurntio": S== d=4 m, mass = 10 kg, ht=O,9 m}

terest with emphasis on the data quality of each fre-quency and frequency range of importance for everycombination of the stacks. In case of soil profile investi-gation, the lower range frequencies yielding longerwavelengths were of greater interest which could allowsampling of deeper soil stratum. Hence, the frequencyrange was first diyided into two equal segments and aweight (wi, i) was assigned for each of these frequency seg-ments. The first segment containing lower range of fre-quencies was given higher weight than the second one. Acurnulative weight (CW) was then calculated where fur-ther weights (w2,i) were given based on the coherencevalue (CH) at the corresponding frequency. n

Cur= E) w,,i w,,] (1) i=1

wi,i=2 Asmean value of the considered frequency range

wi,i== 1 f>mean value of the considered frequency range

w,,,:O CH



The JapaneseGeotechnical Society


10UOO

sooott'

fioo"EurS'

4000zuuo

[)

1o.sU.

o,StL:8

n.4n.21)

o20 40 60 80 1[ra 11" I41) IGO IW) 200 Frcquencycai)

U2o 4o eo so 1oo no 1- tso lso 2oo Freq"ellny'(Hz)

Fig. 7. Cross power spectrum (CPS) and Coherence for testconfiguration: S!d=2 m, mass = 10 kg, ht =O.9 m

4

3E,:tli'ogn.-1$.

-2

.3-4

2an

os..-

.2tsE

-4=Z

-6ft;

-8S' -10-12

o 50 100 15e 200

FTeq"encyCHi)

Fig. 8, Wrapped aficl U"wrapped Phase of cross power spectra for test coiifiguration: S=. d=2 m, mass=10 kg, ht =09 m

cies. The auto-power spectra give an estimate of the fre-

quency distribution of energy for each signal. The cross-power spectrum helps in identifying the predominant fre-quencies that are present in both the signals. Figure 7 il-Iustrates the cross power spectrum curve and coherencevalue distribution for test configuTation d =: S=2 m withimpact mass o'f 1O kg and height of fall O.9 m. It was ob-servedthat the power content was low up to 50 Hz and in-creases significantly for 60 to 100 Hz, Figure 8 shows thewrapped and unwrapped phase lag of second receiverwith respect to the first one. Though coherence values

(Fig. 7) were quite good for the region O to 50 Hz, thephase velocities were found to be negative due to negativephase lag (Fig. 8) obtained at frequencies lower than 20Hz. On the contrary, unwrapping of such negative phasesyields very Iow velocities at considered frequencies. Suchnegative phase lag values may be due to the presence ofother background noise or signal with direction from sec-ond sensor to the first one, Hence, masking was requiredfor obtaining feasible phase velocity for these frequen-cies.

On the other side of frequency range, from 100 Hz to150 Hz the power value decreases again and after 160 Hzbecomes nearly constant with a very low magnitude,From Fig. 7, it can be observed that the coherence valuedecreases to a large extent at these low power contentregions and spurious jumps are also present in wrappedphase after 150 Hz (Fig. 8). In the present study, suchlow power regions were masked out to avoid these spuri-ous jumps and for identifying the frequency range withgood data quality, The regions of masking for Iow powerhave been indicated in unwrapped phase distribution

75ftu5ttkb

1,5

22,5

33.5

44.5

55.S

o

PhaseVle]ocity{mis)

10fi 200 3eo 400

F;g. 9. Experimentnl dispersion curve for d=S==2m, mass=:10 kg, ht=:O,9m

shown in Fig. 8, The final dispersion curve after applica-tion of masking has been shown in Fig, 9. From the ex-perimental results, it was observed that the threshold

power magnitude value was 500 (mV)2 for the instrumentused for profiling of the soil site.

Each instrument carries specific sensitivity of measure-ment, which may depend on the specifications of bothsensors and data acquisition system, The measurementswith low power do not necessarily represent the trueresponse, albeit the coherence value of such measure-

ments may be high for few associated frequencies. Theapproximations in transformation of signal to frequencydemain and interference of noise make it complex to ac-curately determine the threshold power value for eachmeasurement. As a pragmatic approach, an instrument

NII-Electionic




704. MUKHERJEE AND PRASHANT

can be calibrated for its response by performing a numberof measurements in different configurations of testing,

and then the threshold power value can be taken as an

average magnitude of power below which the associatedfrequency range mostly shows: (i) considerably low co-herence values for a number of frequencies, (ii) negativevalue of phase lag, (iii) irregular jumps in phase distribu-tion, and (iv) unacceptable phase velocities. The sameprocedure was followed in this study (e.g,, Figs. 5-8),which yielded the threshold power value of the instru-ment to be 500 (mV)2.

700006onoo5.0000tt

4oonogut:

3000020000loooeoo

so 100Frequclc}'(Hz}ISO 100

NEAR FIELD AND FAR FIELD EFFECTS ANDPOWER SPECTRUM

Near field and far field effects were analyzed in thepresent study based on cross power spectrum values toavoid body wave effects and poor signal quality due to at-tenuation of surface waves respectively, The filteringcriterion proposed by Heisey (1982) was considered inthis study. According to that filtering criterion the mini-mum wavelength should be greater than half of the inter-receiver distance and rnaximurn wavelength should beconsidered as thrice the inter-receiver distance. In thepresent analysis power spectrum curves were studied to

predict the combined efiect of source impact and spatial

configuration of source and inter-receiver distance forsatisfying the above mentioned filtering criterion, Farfield effect could be eliminated by selecting proper upperfrequency cut off from the threshold cross power spec-trum value, Through masking of higher frequencies withlower power content, adulteration of signal quality due toattenuation of surface waves could be avoided. From Fig. 9, it can be observed that the minimumwavelength yielded after masking satisfies Heisey's filter-ing criterion; i.e., minimum wavelength should be greaterthan half of the inter-geophone distance. Near fieldeffects are related to the increase of phase velocity valuesdue to presence of body wave in the wave generated bythe impact load during SASW test. A detailed discussionon the spectral quantities has been presented in the fol-lowing sections threugh a comparative study, whichwould guide to select the proper source type and its spa-tial arrangement to eliminate near field effect by satis-fying Heisey's filtering criterion. The focus of discussionis the combined effect of source type, height of fall of im-pact load and S/d arrangernent on the dispersion curveand power spectrum values,

EFFECT OF SOURCE TYPE

Figure 10 presents the variation in power spectrummagnitudes, Auto Power Spectrum at first station (APS1)and Cross Power Spectrum (CPS), under increasingsource impact and different height of fall for a test setupwith d= 2 m and S= 1 m. For the same S and d configura-tion, increase in either the magnitude of load (wt) or theheight of fall (ht) of impact load increased the powerspectrum values due to increase in input energy. A con-

3Uoeo!:ow!onoo'sE

isolluzu10000fiowou

s" IUOFrtgvent}'OII)ISO !Uu

Fig, 10, Power Spectru ford=: 2 m and S= 1 m under different $ource

impact conditions, (a) Auto po-,cr spectrum (APSI) for near receiyer (b) CToss power spectrum (CPS) ror two receiyers

8stseU

8EbgU

s:liiU

tO,9oso.7O.6o.s

1o.posO.7O.6o.s

1u.yo.sO.7O,6O.5o

so 10UFrequcncy{Hz)150 200

Fig. 11. Coherenceys.)'reqllencyford=2mafidS=1munderdiffer-

ent source impact conditions (a) mass=5 kg and ht=025 m, (b) mass ==5 kg alld ht=O,5 m nnd (c) mass=18 kg and ht=O.5m

tinuous increase in coherence function (Fig. 11) indicatedimprovement in the data quality, with increase in impactloads due to increased energy values for the same spatialconfiguration of experimental set-up. Figure 10 indicatesthat the distribution of energy over the frequency rangewas independent of the height of fall and it only de-pended on the mass used for generating the impaet. For




13

tL's:f-7;t91113

ISO]oo

EVALUATION

PhascVeluc".)'(m/s)

2so 3oe 3so400

OF SASVgi TEST CONFIGURATIONS

4SO

Fig.12. Experimental dispersion curve for d=2m, S=1m under differcn't impac"oad

lower mass (5 kg) the energy was distributed over a widerange of frequency (40-170Hz) with peak value at110-120 Hz. Whereas, in case of higher mass (18 kg) theenergy distribution was concentrated over a lower rangeof frequency (35--140 Hz) and peak values at 100-1 10 Hz.Figure 12 depicts the experimental dispersion curve for 5kg and 18 kg load with the same height of fall of O.9mspatial configuration of d=2 m and S= 1 m . For the low-er mass (5 kg), the upper frequency cut off extended to ahigher frequency and smaller wavelengths were present inthe experimental dispersion curve. Similarly, in case of 18kg higher wavelengths were present in the dispersioncurve due to dominant lower frequencies.

EFFEC[I] OF SPATJAL CONFIGURATION OFSENSOR AND SOURCE

Auto power spectrum and cross power spectrum curveswere observed for different configurations of S and d.Subsequently, the change of coherence function valuesand the quality of dispersion curve were a)so analyzed forsuch test configurations.

EfiZiet of Source to AJear Receiver Distance The experiments were carried out at four ditferent Sand keeping dthe same. Figure 13 depicts the variation ofpower spectrums (APSI and CPS) with different S/d ra-tio (for d=4m) for the same impact load (18kg) andhejght of fall O,9 m. It was observed that with increase inSfd ratio pewer spectrum magnitude decreased sig-nificantly due to attenuation of energy for travellinggreater source to receiver distance. The rate of decreaseof power spectrums were higher when S/d>1. In suchcases, the power spectrum value went below 500 (mV)2which was below the threshold value. So, no dispersioncurves could be obtained in such conditions, The coher-ence function and the dispersion curve (for S/ds1) arepresented in Figs. I4 and 15 respectively. The experimen-tal dispersion curve was plotted after masking the zonewith cross power spectrum values belew the threshQldvalue. The frequency range of analysis was chosen basedon power spectrum (Signat 2uatity Based on Power spec-trum), and the considered frequency was 30-110 Hz in

;gl7

lsoeo

20000

ISOOO

10DOO

-;gEu

sooo

o

7UOO6000sooo4eooscoo1000100Dioo

o

o so 100FrequeneyMz)15.0 200

o 50 looFrcqlimic.yCll7.}150 200

705

Fig. 13. Power Spectru ford=4m, mass=18 kg, ht=O.9 ill, a"d um-

der different Sld sourcc impact conditions, (E} A"to power spec- tvum (APSI) for fiear rcceiver (b) Cross pos'er spectrum (CPS) for trv'o recelvers

ustsEv

ssbEU

to.posO.7O.6o,s

1o.vo.sO,7O.Go.s

d)J" 406080 100 120 140 160 ISO 200Freque:]cy(Hz)

U 50 100Freque]cy(lh)ISO 200

Fig. I4. Coherence value distribution for Sld[= O.5 and 1.0 Cl'est configuration: d=: 4 m, mass == 18 kg, ht == O.9 m)

the case of Sfd=1 and 20-100 Hz in the case of S/d=O.S. Coherenee values (Fig, 14) of these regions weregreater than the acceptable value (O.98) indieating a gooddata quality, Small variation was observed between themagnitudes of phase velocities (Fig. 15) at a particularwavelength, though the pessible depth of exploration incase of S/d=O.5 was more with maximum wavelength,J,.. == 21 m than that in case of S/d = 1 with Aniax=7 m.





1

3 5a7ggg'

iir/

i3f 15

17

19

21

15e200Phase lleloc;t}. (mls)

2SO 300350400 27000

2400"

21000ny.

ISooo:t

ISOOOt,

l!VVO

gooo

GUUO

30oo

ooso lno lso

Fvcqucncv{Hl)

200

Fig, 16, Cross power spectrum (CPS) bet}yecn two geopholles for S=dconfigurutlon(muss=18kg,ht=iO,9m)

Fig. 15, Experimental dispersion curve fer Sfd=O.5 and 1.0 (Test configuration: d=4 m, muss = 18 kg, ht=O.9 m>

At low Sfd ratio, it is diencult to ascertain that the wavetravelling to the receivers are surface waves and not bodywaves. It is quite Iikely that the interpretation of velocity

at high A (Iow frequency) had some infiuence of bodywaves, which could mislead the interpretations, Heisey'scriterion for such near field effect suggests the maximumwavelength acceptable for a4 m inter-receiver distance tobe 12m. Hence, the data obtained from Sfd=O,5 condi-tion can be considered in the analysis provided maximumwavelength considered as thrice the inter-receiver dis-tance.

Efilect of bi ter-Receiver Distance The depth up to which soil profile can be determineddepends on the chosen inter-receiver distance and the ap-

plied impact load, Three inter-re eiver distances wereconsidered (2 m, 4 m, and 6 m) for S/d= l condition and18kg impact source and O.9m height of fall. Figure 16shows the cross power spectrum and dispersion curves areplotted in Fig. 17, In the case of d=2m and 4m, theCPS values (Fig. 16) were greater than the thresholdpower and their masking was performed based on thecriteria described in Signat euatity Based on Power Spec-trum. The dispersion curves were plotted after applica-tion of masking (Fig. 17) and it was observed that themaximum wavelength (19 m) obtained from d=2m wasgreater than that (8 m) of d=4m for 18 kg mass withheight of fall O.9 m. However, for satisfying the filteringcriterion of near field effect, the maximum wavelengthshould be taken as 6 m for the inter-receiver distance of 2m. Further, the minimum wavelength was nearly thesame for both inter-receiver distances (2m and 4m).Hence, it is advisable to take lower masses (5-10 kg) forsmaller inter-receiver distance to get a better representa-tive frequency range (EFFECT OF SOURCE TYPE). Incase of d=6m, the CPS magnitude went below thethreshold magnitude and a scattered dispersion curve isobtained (Fig. 17), For this condition, to get a better dis-persion curve, it is suggested to use either a heavy source,or generate a greater irnpact under enhanced height offall. In either case, greater input energy is required to betransferred to the soil. Input energy can also be increased

1

3

gsgg7Tls9

11

13

o 100Phase Velecit), (mls)

200 300400soe

Fig. 17. Dispersien curve for S= d co"figuration (d=2, 4, a"d 6m, mass!=18kg,ht=O.9m)

by reducing the S/d factor. However, in such case bodywave interference may be present in the acquired signaland hence analysis results can be misleading.

COMBINED EFFECT OF SOURCE TYPE ANDSPATIAL CONFIGURATION OF SOURCE ANDRECEIVER ON POWER SPECTRUM

From EFFECT OF SOURCE TYPE and EFFECT OFSPATIAL CONFIGURATION OF SENSOR ANDSOURCE it can be observed that the power spectrumvalues get largely affected by the source type and spatialconfiguration of source and receiver. For smaller receiverdistance (2 m) higher frequencies are of moTe importanceto generate low wavelengths for sarn pling shallow depths,Lower masses (5-10 kg) can be used for such cases forgenerating higher frequency range (Fig. 10), However, insuch a case it is possible that such a mass may lead to asignal with low power and hence height of fall should beincreased





configurations contributed moTe energy with Iittle varia-tion in wavelength and phase velocity (Effect ofSource toNbar Receiver Distance and Fig, 15), In the presentstudy, it was observed that the etfect of body wave notonly depends on source configuration but also on its mag-nitude.

GENERATION OF DISPERSION CURVE

For each test set up, the waves were generated by im-pact load and captured at some distance with the help of

two geophones. The process was repeated for six stacks.A procedure of obtaining the dispersion curve for the soilstrata was developed with due consideration to the issuesassociated with data quality and power spectra. This

procedure can as well be used for automated generationof dispersion curve from time domain data so as to be

used further during inversion!forward analysis to even-tually determine the shear wave velocity profile of thestrata, The steps involved in this procedure are discussedbelow.

1. 7'kansprmation of TVme Domain Data to ne- quency Domain: The two tirne domain signals (y] (t) and y2(t)) acquiTed through the geophones during each stack were transformed into frequency

domain using Fast Fourier Transform to obtain the

corresponding Iinear spectra (Yi(ca) and }3.(w)). 2. Masking of R'equencies with Low Power Spectra Values: lnitially, the frequency range of analysis was eonsidered as 9-200 Hz based on the sensor

specifications (EXPERIMENTAL PROGRAM), and the spectral quantities (APSI, APS2 and CPS) were calculated for each of the data stack, Then

these quantities were averaged in various cornbina- tions of five, four and three stacks out of the six stacks. Through observation over power spectra,

the final frequency range of analysis was identified. The maximum frequency range among all the stack combinations where CPS values were obtained more than the threshold power value was taken as the frequency range of importance (Fig. 10). The minimurn power magnitude was considered to be 500 (mV)2. 0ne should note that the threshold power magnitude could be instrument dependent. 3. Masking of F)'eguencies with Low Coherence and finding the Best 9uality Data Set: In the next step, coherence values, cumulative weights based on frequency and coherence CIiinding the Best Stacking Combination Based on Cumutative PVeight) and count of deleted points were estimated for each of

the combinations, The quality of each combination

set was defined based on the cumulative weights as

described in Finding the Best Stacking Combination Based on Cumulaiive PVeight. The best data set was se!ected out of these combination sets based on the

quality ot' data and degree of representativeness, Tbe quality of data was assured by cumulative weight and at the same time degree of representativeness of data was retained by considering a

minimum four stacks in combination set of analysis.

4, Unrvrapping of PV)'apped Phase and Construction ofDispeflsionCurve: Thecalculationofphasean- gle and its unwrapping was done for the best stack-

ing combination. Using this phase information,

Rayleigh wave phase velocity was calculated for

each frequency from the known receiver distance. User interactive interfaces were used in the devel-

oped code during automated unwrapping of the phase data to avoid problem arising from fietitious

jumps due to low coherence and real jumps with phase jump less than 2n due to change of stratification. To consider real jumps with phase jump less than 2n due to change of stratification, the jump tolerance was assumed to be 3n/2. For the phase jumps in between z to 3n12, the unwrapping was first done both by considering and without consider-

ing the phase jurnp and dispersion curves were plotted, The decision was then made heuristically

whether to consider the jump or not. The accuracy of interpretations is governed by therepresentativeness of the data, and the output of

proposed method promises more reliable data in com-

parison to the conventional procedures. Besides selecting

good coherence data from multiple stacks, it is effective inminimizing ambiguities of unwrapping procedure andmasking of the unrepresentative data based en simplerules. Hence, the strength of proposed method is its fastand efrective automated procedure, Figure IS ilLustratesan example of the advantages of the proposed method,Figure 18(a) shows thTee dispersion curves of different dvalues based on the conventional empirical approachconsidering the coherence )O.98 for full stack combina-tion and rimits of included wavelengths as d/2}A23d.The dispersion curves in Fig. 18(b) are the results ofproposed method, which eliminated the case of d=6mdue to low power content. The large scatter of data in thecase of conventional approach indicates uncertainty ofdata, and the interpretations based on any of the test con-figurations will have a large error bar associated with it.The proposed approach provides only the representative

11

150Phflse

Ve]ocits.' (mis) 200 ZSO300 ISO

3:t4f:.o5:.6S7

s

9

lo

d;2]n "i"-

d-:ilmiE,

'

/t,spa-wt:liR'

ig'

-.

o- -o'''1?-''"''"''"''--''''"'''

,:L:t . -e-

oa g

o. `.

"" -O-T-- .r-...-.

iol {a}

.PhHseNlePocit],(]nls) 200 ISO300

Fig. 18. Comparison of conve"tional filtcring criteria and proposed mcthod of annl},sis for dispersion curves (S==d=2, 4, and 6m, mass=18kg,ht=O,9rn)

NII-Electionic





data based on the quality of collected multiple stacks. Itencourages variation of test configurations for obtaining

the data for wider wavelength range instead of using theunrepresentative data of a setup producing misleadingresults. As another example, in Fig, 8 the unwrapping atlow frequencies to correct for the positive phase valuescouid be avoided based on the assumption of low powerdata being unrepresentative. Such unwrapping wouldhave produced interpretation of velocity ranging from150 m/s to 50 m/s for the wavelength between 1 m to 2.3m, which was considerably lower than the velocities sug-gested by the other test configurations.

CONCLUSION

It is imperative to understand the amount and thecause of uncertainty induced through the instrument limi-tations and test configuration while interpreting theresults from an SASW test. Several issues associated withloading arrangements, spatial configuration, sensorspecifications, and analysis scheme were discussed in thepaper. Based on the observations, a simple methodologywas proposed eventually to obtain a representative dis-

persion curve for the measurement domain. Key findingsof the present study are giyen below,

1, The power spectrum values could be used to determine the frequency range for which the response would be interpreted based on propagation of induced mechanical waves and not due to surround- ing noise or sensitivity of sensors. It was proposed to consider a threshold minirnum cross power spectrum va!ue for selecting the frequency range. The magnitude of the value could be site and instrument dependent. 2. Different stacking combinations of the same test configuration were considered during analysis to obtain quality data. The best data stack was identified based on cumulative weight calculated over the frequency range of interest. The weights were as-

signed based on the magnitude of coherence of each frequency and importance of different frequency ranges for every combination of the stacks,

3. A comparative study was presented to show the im-

portance of source impact magnilude to obtain a

good quality data satisfying the coherence require-

ment and also its effect on frequency range of in-

terest, power spectrum values and dispersion curve. The dorninant frequency content produced by a source could be a key factor to be considered while performing a test for certain ground depth of interest. The falling mass source was used during this

study and the results indicated that the interpretations could be infiuenced not only by the energy induced by the source but also by the mass of the falling object.

4, The effect of spatial configuration of source and inter-receiver distance was studied based on the cross power spectTum magnitude. Heisey's filtering

criterion was satisfied by choosing proper source ar-

rangement and type depending on the power spec-

trum values for eliminating near field and far field issues. The depth of exploration can be inereased by having larger inter-receiver spacing and suMcient impact energy with low frequency content propagating to both the receivers from the source.5. The proposed method of analysis has the potential to automate the generation of experimental dispersion curve with reduced uncertainty. Such develop- ment will eliminate some decision making during data processing and will reduce the time of calcula- tion of experimental dispersion curve by a great amount, This experimental dispersion curve then

again can be used as input of the inversion procedure for profiling the soil site,

NOTATION

APSI

APS2

CPSdfAsWl,i

W],i

Yi(t), ),2(t)Yi(to),k(co)

Auto power spectrum at 1st geophone posi-tionAuto

power spectrum at 2nd geophone po-sltlonCross-power

spectrum of two geophonesInter-receiver distanceFrequencyWavelength of Rayleigh wave

Source to near-receiver distaneeWeight assigned for each of these fre-

quency segrnentsWeight based on the coherence value ateach frequencySignals in time domainLinear Spectra obtained by transformingthe time-domain signals into frequencydomain signals using Fast Fourier Trans-formation (FFT)

REFERENCES

1) Al-Hundai, M. O. (1992): Dirnculties with phasc specLrum unwrapping in spcctral analysis of surface waves nondestructive tcsting or pavements, Can. Geotech. J.,29, S06-511.2} Foti, G. {2000): Multistation mcthods for geotcchnical characteTi- zation using surface waves, PhD Desffertation, Politecnico di Mila-

no, Italy.3) Ganji, V., Gukunski, N. and Na7arian, S. (1998): Aulomated inversion procedure for spectral analysis of surface waves, J, Geo- tech, and Geoenv. Engg., ASCE, 124(8), 757-77e.4} Heisey, J. S., Stokee, K, H,, II and Meyer, A, H. (1982): Moduli or pavement systems froin Speetral analysis of surface waves, T)ans-

portation Research Record, (852), 22-31.5} Heisey, J. S., Stokoe, K. H., IT, Hudson, W. R. and Meyer, A. H, (1982): Determination ofin situ shear wave velocities frorn Spectral anaiysis of surface waves, Researrch Report No. 256-2, Center for

Transportation Research, The University at Texas at Austin.6) Hiltunen, D, R, and Woods, R. D. (199e): Variables affecting Lhe testing of pavements by the surface wave rmcthod, T)'ansportation

Research Record, (1260), 42-52,7) Marosi, K. T. and Hi]tuncn, D. R. (2004): Characterization of spectral analysis of surface waves shear wavc vclocity measurcment un-

certainty, J. Geoteeh. and Geoenv. Engg., ASCE, 130(10), 1034LI041.

NII-Electionic





8) Nazarian, S. and Stokoe. K. H., ]I (1983): Evaluation of moduli andthicknessofthepavemenLsystemsbySpectral-Analysis-or-Sur-

face-Waves method, Researeh Roport Nb. 256-4, Center for Trans-

portation Research, The University at Texas at Austjn.9) Nazarian, S. (1984): In situ deterinination el' elastic moduli of soil

deposits and pavement systerns by spectral-analysis-of-surfacc- waves method, Ph,D. Dissertation, University of Texas at Austin,

Austin, Texas, U.S.A.10) Nazarian, S, and Stokoc, K. H., II (1986): Use of surfacc waves in pavernent evaluation, 7)ansportation Research Record, (1070), 132-144.11) Nazarian, S, and Stokoe, K. H., Il (1987): Nondestructive delineat- ing changes in modulus profiles of secondary roads, TYcrnsportation Research Record, 1136 TRB, National Research Council, Washin-

gton D.C,12) Nazarian, S. and Stokoe, K. H., II (1989): Nondestructive evalua- tien of pavements by surface wave method, Jsi I}it. Symp. Nbndes- tructive Testing ofPavements and Backcalcuiation ofMbduli, STP 1026, ASTM, 119-137.13) Sanchez-Salinere, 1., Roesset, J. M., Shao, K. Y., Stokoe, K. H.

and Rix, G. J, (1987)/ Analytical evaluation of variables affecting surface wave testing of pavements, T)ansportation Researeh

Record, (1136), g6-95,14) Sayyedsadr, M. and Drnevieh, V. P. (1989): SASWOPR: A program to operate on spectral analysis of surface wave data, lst inter.

Symp. Nondestructive Tlesting Qf Pavements and Backcalculation ofModuti, STP 1026, ASTM, 670-6S2.

NII-Electionic

110007466329

Documents

soil profiling

corrected data

good data quality

situ nondestructive

sasw test igc

near receiver distance

experimental dispersion

power spectra