ie anaysiso'resu:s 'rom

rou:ine oi e oac:es:sby BENGT H. FELLENIUS', PEng, DI. Tech, MEIC, MASCE

THIS PAPER deals with the analysis ofresults from axial testing of vertical singlepiles, i.e. the most common field test per-formed. Despite the numerous tests whichhave been carried out and the manyPapers which have reported on such testsand the analysis thereof, the understand-ing of pile test loading in current engi-neering practice leaves much to be de-sired. The reason for this is that the en-gineers have concerned themselves withmainly only one question: "Does the pilehave a certain least capacity?", findinglittle of practical value in analysing theactual capacity and the pile-soil interac-tion. This Paper aims to show that en-gineering value can be gained from ela-borating on a pile test —during the ac-tual testing in the field, as well as in theanalysis of the results.

The first portion of the notes make useof an earlier Paper by the author (Fellen-ius, 1975) However, additional views andrecent literature have been incorporated.

Testing methodsThe most common test procedure is

the slow maintained load method refer-red to as the "standard loading procedure"in the ASTM Designation D-1143 (ASTM1974) in which the pile is loaded in eightequal increments up to a maximum load,usually twice the predetermined allowableload. Each increment is maintained until

*University of Ottawa, Canada

zero settlement is reached, defined as0.01in/h ( = 0.002in/10 min.) The final

load, the 200% load, is maintained for 24hours The "standard method" is verytime consuming requiring from 30 to 70hours to complete. It should be realisedthat the words "zero settlement" are verymisleading, as the settlement rate of 0.01in(0.25mm)/hr is equal to a settlement of7ft (2.1m) /yr.

The "standard method" can be speededup by using the method of equilibriumproposed by Mohan et al (1967), wherethe load (jack pressure) is allowed todrop rather than being maintained bypumping. The equilibrium load value istaken as the load applied on the pile.

Housel (1966) proposes that each of theeight increments be maintained exactlyone hour regardless of having reached"zero" settlement or not. The Houselmethod of applying the load at equal timeintervals allows an analysis of movementwith time, which is not possible withthe "standard method". By plotting themagnitude of movement obtained duringthe last 30 minutes of each one-hour loadduration versus the applied load, twoapproximately straight lines are obtained.Provided the test has approached failure,that is. The intersection of the two linesis termed yield value.

A maintained-load test according toHousel's method takes a full day to per-form The points on the curve are stillvery few, but Housel's method is a defin-

ite improvement of the "standard method"and it has been incorporated as an op-tional method in the ASTM DesignationD-1143. However, it is the author's opin-ion that a test consisting of 16 equal in-crements of say 30 tons applied every30 minutes would provide a better testthan 8 increments of 60 tons applied every1.0 hour, because it would provide a bet-ter defined load-movement curve. Also,a similar yield value, and one not muchdifferent, can be evaluated from the move-ment during the last 15 minutes, providedthat readings are taken often enough andthat they are accurate. But why stop at16 increments, when 32 every 15 min-utes determine the load deformation curveeven better? The load is still applied ata constant rate in terms of tons per hourand no principal change is made.

Actually, the duration of each load isless important, be it 1.0 hour or 15 min-utes; it is the fact that the duration ofeach load is the same which is important.From this realisation, we can progressto the one that even shorter time inter-vals, and an increase of the rate of load-ing in tons/hour, are possible withoutimpairing the test. Actually, by using asshort time intervals as practically pos-sible, the time-dependent influences arereduced and a more truly undrained testis obtained. In those cases where a studyof the time-dependent, or drained condi-tions, creep, etc. is desirable, the testduration should be measured in weeks,

fn

0I-

00

300

200—

EXAMPLE N'

C. R. P. —TEST

DRIVEN JULY 9, 1975TESTED AUGUST 22, 1975DEPTH 76 I FT

PREBORED 30 FT

-0CLAY +gy

- 60

GLACIAL 'a oTILL o 6

500

200—f

I-0

CI

I I

DAV I S SON'S METHOD

EXAMPLE I

1I

I

I

I

I

I

I

I

I

I 00—

0 I I

0 I.O

MOVEMENT ( I NCHES)

Fig. 1. Load-movement diagram from CRP test

2.00 I I

0 I.O

MOVEMENT (INCHES)Fig. 2. Construction of Davisson's limit

2.0

September, 1980 19

IO IO 3

P( I N. /TONS)

CHIN S METHOD

EXAMPLE I

300-

200-

150-

DE BEER S METHOD

Iee

EXAMPLE I

V>

OI 00 ~

CI

0

5.10+

40-

300.05 0 IO O.I5 0.20 0.30 040 050 I 00 1.50 2.00

MOVEMENT (INCHES)

Fg. 4. Construction of De Beer's yield limit

C2

p = —Io = 235roasULT 4.25

I.O

MOVEMENT ( INCHES)

Fig. 3. Ultimate failure according to Chin

2.0

months or even years. A 48 or 72 hourtest is then vastly inadequate, and re-sults only in confusion

Tests which consist of load incrementsapplied at constant time intervals of 5,10 or 15 minutes are called Quick Main-tained-Load Tests (ML tests) and arefrom both technical, practical and econo-mical points of view superior to the slowML tests, They have been relatively re-cently introduced into North America, butare steadily gaining acceptance. The latestversion of the ASTM Designation has onequick ML method as an optional method.For instance, recently the Federal High-way Administration published an exten-sive users'anual for a Quick ML method(Butler & Hoy, 1977),

The Quick ML method should aim for30 to 40 increments with the maximumload determined by the amount of reac-tion load available or the ultimate capa-city of the pile. For routine cases, it maybe diplomatically preferable to stay at amaximum load of 200% of the intendedallowable load, For ordinary test arran-gements, where only the load and thepile head movement are monitored, timeintervals of 5 minutes are suitable andallow for the taking of 2 to 4 readingsfor each increment (for instance, whenreaching the load, and at 2.5, 4.0 and 5.0minutes after starting to load) . Whentesting instrumented piles, where the in-struments take a while to read (scan), thetime interval may have to be increased.To go beyond 15 minutes, however,should not be necessary Nor is it advis-able, because of the potential risk of in-ffuence of time-dependent movementswhich may impair the test results. Us-ually, a Quick ML test is completed withintwo to three hours

20 Ground Engineering

A quick test which has gained muchuse in Europe is the Constant Rate ofPenetration test (CRP test), first pro-posed internationally for piles by Whitaker(1957 and 1963) and Whitaker & Cooke(1961). Manuals on the CRP test havebeen published by the Swedish Pile Com-mission (1970) and New York Depart-ment of Transportation (1974). In the CRPtest, the pile head is forced to settle ata predetermined rate, normally 0.02in/min (0,5mm/min), and the load to achievethe movement is recorded, Readings aretaken every two minutes and the test iscarried out to a total penetration (i.e.movement of the pile head) of 2-3in(50-75mm) or to the maximum capacityof the reaction arrangement, which meansthat the test is completed within two tothree hours.

The CRP test has the advantage overthe Quick Ml test in that it enables aneven better determination of the load-deformation curve, This is of particularvalue in testing friction piles, whensometimes the force needed to achievethe penetration gets smaller after a peakvalue has been reached. It also agreeswith the testing in most other engineeringfields, which regularly use CRP methodsto determine strength and stress-strainrelations.

To perform a CRP test, access is re-quired to a mechanical pump that canprovide a constant and non-pulsing flovvof oil. Ordinary pumps with a pressure-holding device, manual or mechanical, arenot suitable because of unavoidable loading variations. Also, the absolute require-ment of simultaneous reading of all loadand deformation gauges (changing contin-uously) could be difficult to achievewithout a trained staff, For these reasons,

the Quick ML method is preferable forinstrumented piles.

A fourth test method is cyclic testing.However, cyclic methods will not be des-cribed here; for details see Fellenius(1975), and references contained therein.In routine tests, cyclic loading or evensingle unloading and loading phases mustbe avoided. It is a common misconceptionthat unloading a pile every now and thenaccording to some more or less "logical"scheme will provide information on thetip movement. It will only result in adestruction of the chances to analyse thetest results and the pile load-deformationbehaviour, In non-routine tests and for aspecific purpose, cyclic testing can beused, but then after completion of aninitfal test and when the pile is instru-mented with at least a tell-tale to the piletip.

There is absolutely no logic in believingthat anything of value can be obtainedfrom cyclic testing, or occasional un-loadings, or one or a few resting periodsat certain load levels, when it is realisedthat we are testing a unit which is sub-jected to the influence of several soiltypes, is already under stress of unknownmagnitude, exhibits progressive failure,etc., and that all we know is what weapply and measure at the pile head, whilewe are really interested in what h'appensat the pile end.

Interpretation of failure loadFor a pile which is stronger than the

soil, the ultimate failure load is reachedwhen rapid settlements occur under sus-tained or slightly increased load —the pileplunges. However, this definition is in-adequate, because plunging requires verylarge movements and it is often less a

VlZ0

200—

BRI NOH HANSEN'S METHOD

( 90 96 - CRlTERlONI

205

909 205 = 185

7.0

BRINCH HANSEN'S METHOD

lao% cRITERIDN)P

Oct0

IOO-6.0— =C,B+C2

= 1.414. 10 l66 +3.956.10

CII-'l + C2

IO5096 I 67:066

MOVEMENT (INCHES)

20I 69

Fig. 5. Ultimate failure according to the90% criterion by Brinch Hansen

5.0+0.500

I

1.000I

I.500

MOVEMENT (INCHES)

/0.25&v BOTH ON

ES P„ = 211 TONS

I

2.000

Fig. 6. Ultimate failure according to the 80% criterion by Brinch Hansen

function of the capacity of the pile-soilsystem and more a function of the capac-ity of the man-pump system.

In the past, a common definition offailure load has been the load for whichthe pile head movement exceeds a cer-tain value, usually 10% of the diameterof the pile end. This definition does notconsider the elastic deformations of thepile, which can be substantial for longpiles, while it is negligible for short piles.In reality, a limit movement relates onlyto the allowable deformation limits of thesuperstructure to be supported by thepile, and not to the load test results.

Sometimes, the failure value is definedas the load value at the intersection oftwo lines, approximating an initial pseudo-elastic portion of the load-movement curveand a final pseudo-plastic portion, Thisdefinition results in interpreted failureloads, which depend greatly on judgementand, above all, on the scales of thegraph. Change the scales and the failurevalue changes also. A load test is influ-enced by many occurrences, but thedraughting manner should not be one ofthese.

To be useful, a failure definition mustbe based on some mathematical rule andgenerate a repeatable value that is in-

dependent of scale relations and the opin-ions of the individual interpreter. In someway, it has to consider the shape of theload-movement curve or, if not, it mustconsider the length of the pile (whichthe shape of the curve indirectly does) .Without such proper definition, every in-

terpretation becomes meaningless.The test results given as a load-move-

ment curve in Fig. 1 will be used to pre-sent nine different definitions of failure.The example pile is a 12in (305mm) con-crete pile installed through 60ft (18.3m) ofsensitive clay, 10ft (3.0m) of clayey siltand 6ft (1.8m) of silt. The pile wastested six weeks after driving. Methodof testing was the CRP method. The pilestarted to plunge when the test loadreached 200 tons, but at the maximumload of 206 tons the load necessary toachieve the movement was still incroac-ing,

In Fig, 2 is applied a method proposedby Davisson (1972), also referenced byPeck et al (1974). Davisson's limit valueis defined as the load corresponding tothe movement which exceeds the elastic

compression of the pile by 0 value of0,15in (4mm) plus a factor equal to thediameter of the pile divided by 120. For the12in dia. example pile, the value is 0.25in(6mm). The Davisson limit was developedin conjunction with the wave equationanalysis of driven piles and has gainedwidespread use in phase with the in-creasing popularity of this method ofanalysis. It is primarily intended for testresults from driven piles tested in ac-cordance with quick methods.

Fig. 3 gives the method proposed byChin (1970 and 1971) for piles in applyingthe general work by Kondner (1963). Themethod assumes that the load-movementcurve when the load approaches the failureload is of hyperbolic shape. By the Chinmethod, each load value is divided withits corresponding movement value andthe resulting value is plotted against themovement. As shown in Fig 3, after someinitial variation, the plotted values fall ona straight line. The inverse slope of thisline is the Chin failure load,

Generally speaking, two points will de-termine a line and a third point on thesame line confirms the line. However, be-ware of this statement when using Chin'smethod. It is very easy to arrive at afalse Chin value if applied too early in

the test. Normally, the correct straightline does not start to materialise untilthe test load has passed the Davissonlimit. As a rule, the Chin failure load isabout 20% to 40% greater than the Davis-son limit. When this is not the case, it isadvisable to take a closer look at all thetest data.

The Chin method is applicable to bothquick and slow tests, provided constanttime increments are used. The ASTM"standard method" is therefore usually notapplicable. Also, the number of monitoredvalues are too few in the "standard test";the interesting development could wellappear between load increment numberseven and eight and be lost.

Fig. 4 presents a method proposed byDe Beer (1967) and De Beer fk Wallays(1972), where the load movement valuescre plotted in a double logarithmic dia-gram. When the values fall on two ap-proximately straight lines, the intersec-tion of these defines the failure value.De Beer's method was originally proposedfor slow tests.

Fig. 5 illustrates a method proposed by

Brinch Hansen (1963), who defines fail-ure as the load that gives twice themovement of the pile head as obtainedfor 90% of that load. This method, alsocalled the 90% criterion, has gained wide-spread use in Scandinavia (Swedish PileCommission, 1970), Brinch Hansen (1963)also proposes an 80% criterion definingthe ultimate load as the load that givesfour times the movement of the pile headas obtained for 80% of that load. The80% criterion failure load can be estima-ted by extrapolation from the curve tobe about 210 tons. (Some references haveconfused the 80% and 90% criteria, anduse, erroneously, for the 80% criterion themovement of the 90% criterion),

In Fig. 6, Brinch Hansen's 80% criterionis shown in a plot —which is very simi-

ilar to that of Chin —V p/ II plotted against~). The ultimate failure value is deter-mined from the criterion that a point, co-ordinates P6 3„, on the curve is the pointof ultimate failure when the point, co-ordinates 0.80P6 0.253„, also lies on theload-movement curve. The criterion givesthe following simple relationships to usein calculating the ultimate failure, P„:

1

P2VC, C„

C2

C,where C, is the slope of the straight line

and C2 is the y-intercept in the V p/A„plot,Fig. 6.

When using the Brinch Hansen 80% cri-terion, it is important to check that thepoint 0.80P6/0.253„ indeed lies on themeasured load-movement curve.

In the example case, P„ is 211 tons,which agrees well with the value extra-polated from the load-movement curve,directly.

Brinch Hansen's 80% criterion postulatesthat the load movement curve is approxi-mately parabolic. Chin postulates that itis approximately hyperbolic. The shapeof the actual curve is obviously closeenough to both mathematical curves toallow both approximations. BrinchHansen's 80% criterion results generallyin a failure value about 10% lower thanChin's value, No'te that both methods al-low the latter part of the curve to be

September, 1980 21

EXAMPLE i EXAMPLEI

250250- FULLER AND HOY S METHODMAZURKIEWICZ'S METHOD

BUTLER AND HOY S METHOD

ZD3

208200 /

/X 200—

I/3

150I-

En

0 150- ON

O

CI IOO Ci

0100-

I NE50

50-I

1.000.50 1.50 2.00

MOVEMENT (INCHES)

Fig. 7. Ultimate failure load according to Mazurkiewicz0

I

0.50I

1.00I

I.50 2.00plotted according to a matnernatical rela-tionship, and —which is cften very tempt-ing —they make an "exact" extrapolationof the curve possible. That is, it is easyto fool oneself into believing that theextrapolated part of the curve is as trueas the measured.

In Fig. 7, the method put forward byMazurkiewicz (1972) is illustrated, A ser-ies of equal pile head movement linesare arbitrarily chosen and the correspond-ing load lines are constructed from theintersections of the movement lines withthe load-movement curve, From the in-tersection of each load line with the loadaxis, a 45" line is drawn to intersect withthe next load line These intersectionsfall, approvimately, on a straight line, theintersection of which with the load axisdefines the failure load. Also, this methodis based on the assumption that the load-movement curve is approximately para-bolic. Consequently, the interpreted fail-ure load of Mazurkiewicz's method is closeto that of Brinch Hansen's 80% criterion.However, when drawing the line through

MOVEMENT (INCHES)

the intersections according to Mazurkie-wicz, some disturbing freedom of choiceis usually found.

In Fig. 8 a simple definition proposedby Fuller 82 Hoy (1970) is shown. Thefailure load is equal to the test load forwhere the load movement curve is sloping0.05in/ton (0.14mm/kN),

Fig. 8 also shows a development of theabove definition proposed by Butler & Hoy(1977) defining the failure load as the loadat the intersection of the tangent sloping0.05in/ton, and the tangent to the initialstraight portion of the curve, or to a linethat is parallel to the rebound portion ofthe curve. As the latter portion is moreor less parallel to the elastic line (seeFig, 2), the author suggests that the in-tersection be that of a tangent parallelto the elastic line, instead.

The Fuller 82 Hoy method penalises thelong pile, because the larger elastic move-ments occurring for a long pile, as op-posed to a short pile, cause the slopeof 0.05in/ton to occur sooner, The But-ler 82 Hoy development takes the elasticdeformations into account, substantiallyoffsetting the length effect.

Fig 9 shows the construction of thefailure load as proposed by Vander Veen(1953), A value of the failure load, PH„,is chosen and values calculated from1n (1 —P/PDI,) are plotted against themovement. When the plot becomes astraight line, the correct PH„has beenchosen, The Vander Veen method was pro-posed long before programmable pocketcalculators were available. Without those,however, the application is very time-consuming.

500i

190EXAMPLE I195

VANDER VEEN'S METHODCOMPARISON OF FAILURE CRITERIA200

-40-

CHIN 235

8RINCH HAN5EN 80 / ZH MAZURHIEWICZ 208FULLER AND HOY 203

90 /o 205'~VANDER VEEN 205200—-30- DE REER I86

IHOY 185

60XOI-

2IO

C3

O220

-2.0-~ 230

100—

-I 006 13

I1207 f408 09 IO

MOVEMENT DNCHES)

Fig. 9. Ultimate failure according to Vender Veen 0I I

0 I.OMOVEMENT (INCHES)

F'ig >0 Comparison of nine failure criteria

2.0

22 Ground Engineering

Fig. 8. Ultimate failure according to Fuller & Hoy and Butler & Hoy

50 200

ih

O

40-

30—

20—

//O)O /

e

I) No

/''Ir

<u f—1 'XAMPLE N~l

jV~J

l/

EXAMPLE N'I A

1'!/WL''''i'' '' i'''' i'IOO 200

LOAD CELL (TONS)300

150

O

o IOO

Is INCHES DIAMETER FLAT JACK LOAD CELE

—I 0-

—20-

—30-50 100 150

—40—READING

Fig. 12. Variationsin load cell calibration due to eccentric loadapplication, temperature change, and reduced loading area

In Fig, 10, the nine values determinedabove are plotted together, As shown, theDavisson limit of 181 tons is lower thanall the others and the Chin value of 235tons is the highest, The other sevenvalues are grouped more or less togetheraround an average of 200 tons,

It is difficult to make a rational choiceof the best criterion to use, because theone preferred is heavily dependent onone's past experience. One of the mainreasons for having a strict criterion is,after all, to enable a set of compatiblereference cases to be established. Theauthor prefers to use not one but three orfour of the criteria given in Fig. 10, Thepreferred criteria are the Davisson limitload, the Chin failure load, the BrinchHansen 80% criterion, and the Butler& Hoy failure load. In the case of an en-gineering report, the preference and ex-perience of the receiver of the report mayresult in the use of one of the othercriteria, in addition.

The Davisson limit is chosen becauseit has the tremendous merit of allowingthe engineer, when proof testing a pilefor a certain allowable load, to determinein advance the maximum allowable move-ment for this load with consideration ofthe length and size of the pile. Thus, asproposed by Fellenius (1975), contractspecifications can be drawn up includingan acceptance criterion for piles proof-tested according to quick testing methods.The specifications can simply call for atest to at least twice the design load, asusual, and declare that at a test loadequal to a factor, F, times the designload, the movement shall be less than the

elastic column compression of the pile,plus 0.15in, plus a value equal to thediameter divided by 120. The factor Fis a safety factor and should be chosento a value of 1.5 to 1.8 depending on cir-cumstances,

The Chin method is chosen because itallows a continuous check on the test,if a plot is made as the test proceeds, anda prediction of the maximum load thatwill be applied during the test. Suddenkinks or slope changes in the Chin lineindicate that something is amiss witheither the pile or with the test arrange-ment, The Chin value has the additionaladvantage of being less sensitive to im-precisions of the load and movementvalues,

The Brinch Hansen 80% criterion ischosen because it usually gives a P„valuewhich is close to what one subjectivelyaccepts as the true ultimate failure value.The value is smaller than the Chin value.However, the criterion is more sensitiveto inaccuracies of the test data than isthe Chin criterion

The Butler & Hoy method is chosenprimarily because of its resemblance tothe Davisson method In some cases aDavisson limit load can be obtained with-out the interpreter being willing to ac-cept intuitively that the pile has reachedfailure. (In such cases, the Chin valuewill be much higher than the Davissonlimit), However, the Butler & Hoy slopeof 0.05in/ton is not approached unlessfailure is imminent, and absence of aButler & Hoy failure indicates —in addi-tion to a high Chin value —that theDavisson value is imprecise. The reasons

—50-Fig. 1 f. Errorin jack-load determined with manometer vs. correctload determined with load cell

for the latter can be wrongly chosenvalues of pile elastic modulus or pilelength, or imprecise or erroneous load ormovement values, Also the Butler & Hoymethod permits an acceptance criterion forproof-tested piles to be formulated andincluded in the specifications However,the Butler & Hoy method requires thepile head movement to be large enough toreach the Fuller & Hoy point, whichrestricts the use of the definition in thiscontext,

Influence of errorsThe test results shown in Fig. 1 and

used in the preceding discussions arefrom a test where an electrical straingauge load cell was used to determinethe load applied on the pile. In the test,the pressure in the jack was monitoredby means of a manometer, which had beencalibrated together with the jack. Yet theload determined from the manometer read-ings was inaccurate. Fig. 11 shows thedifference between the load determinedfrom the jack pressure and the load deter-mined by the load cell, as plotted againstthe load cell load.

The error (overestimation) in the jackpressure load is substantial and variesbetween 10 and 25%, being mostly 15-20%. In unloading the pile, the error wasmuch smaller. This is not the worst, nor thebest, example the author has met, but isa typical case for the equipment usedin the industry of today,

Fig. 11 also shows similar results fromanother test, called Example IA, whenin loading the error was less than 5%.On the other hand, the error in unloadingwas large, This seems to have involvedjacking equipment of a much better qual-ity than that used in Example 1. However,Example 1A is from an identical pile lo-cated about 20ft (6m) away at the

September, 1980 23

44fI

~ I K'N

J ~m ~ ~

~l R

I I I I

,'.-;-~ e(Left), A 650 ton kentledge arrangement(lor testing 200lt long 16.5in concretepiles). Note that the measuring beam isshielded from sunshine and wind

ltg

same site and tested two days later us-ing the same equipment and method.

Based on the above and many similarmeasurement results, the author con-cludes that if one wants to ensure animprecision smaller than about 20%, aload cell must be used, The jack and jackpressure are too erratic to be reliable Acalibration of the jack and manometer forone pile is not relevant to even a neigh-bouring test pile. The reason for the un-reliability is that the system is beingrequired to do two things at the sametime; both provide the load and measureit, and load cells with moving parts areconsiderably less reliable than thosewithout, Calibrating testing equipment in

the laboratory ensures that no eccentricloadings, bending moments, or tempera-ture variations inffuence the calibration.However, in the field, all these factors arepresent to influence the test results to anunknown extent, unless a load cell is used.

Naturally, many structures are safelysupported on piles which have been testedwith erroneous loads, and as long as weare content to stay with the old rules,loads and piling systems, we do not needto improve the precision. The error isincluded in the safety factor. That is whyfactors as large as 2.0 and 2.5 are appliedand such numbers are really more ignor-ance factors than safety factors. However,if we want to economise and continue toincrease allowable loads as geotechnicalknowledge increases, we cannot acceptpotential errors as large as 20 to 25%. In

the author's opinion, we cannot accepterrors exceeding 10%, and this require-ment necessitates the use of load cells.

However, the fact that a load cell isused is no guarantee for precise loads.Fig. 12 shows calibrations performed ona flat-jack load cell under varying condi-tions. The heavy centre line is a regularcalibration curve obtained when using 3in(76mm) thick full-width steel plates onboth sides of the load cell and applyingthe load through a spherical bearing(swivel plate). This curve is readily re-peatable. However, by moving the loadonly 2in (51mm) off-centre, a differentcalibration was obtained. By letting thetemperature drop, a third line was ob-tained. The greatest influence was obtain-ed by removing the steel plates and load-

24 Ground Engineering

ing only a centre area of the load cell.Of course, the load cell of Fig. 12 is

unsuitable for use in the field, wheretemperature variations and eccentric load-ing cannot be avoided. In a load test,the geometric centre does not necessarilycoincide with the load centre. Therefore,it is necessary to check the calibration ofthe cell and its sensitivity to eccentric loadapplication.

The foregoing discussion has dealt withthe imprecision of the load value. But theprecision of the movement values canalso be critical, If the "failure" criterionis a maximum settlement of 1.75in(44mm), an error of 0.25in (6.5mm) isof no consequence when the maximummovement recorded is 1.5in (38mm) orless, which it is on most proof testingoccasions. However, errors of this de-gree of magnitude greatly influence theshape of the curve and the various met-hods of interpretation of failure loads. Inparticular, Davisson's limit is sensitive tothese errors,

It must be remembered that the mini-mum distances from the supports of mea-suring beam to the pile and the platformetc., as recommended in the ASTM Desig-nation, are really minimum values; gen-erally they give rise to errors of little con-cern for ordinary testing, but they are tooclose for research or investigative testingpurposes.

One of the greatest villains for spoilinga load test is the sun The measuringbeam must be shielded from sunshine atall times,

Analysis of results using tell-taledata

Fig. 13 shows the results from a QuickML test on a 130ft long (40m) 12in(300mm) precast concrete pile. The pilehad a total cross-section of

124in'800cm'),the area of steel reinforcementwas 1.9in'12cm'), and the pile circum-ference was 41in (107cm). The pile wasloaded in steps of 22.4 tons, and a loadcell was used to determine the test load.The failure loads evaluated in accordancewith the nine methods are given in thegraph. Scatter of the values is similar tothat shown in Fig. 10.

In the test, a centre pipe had beencast in the pile allowing a tell-tale to be

(Above). Load cell and swI vel plate(spherical bearing) on a hydraulic jack

inserted down to the pile tip to monitorthe compression of the pile and the piletip movement, As will be shown, thisrelatively simple and cheap addition tothe test arrangement greatly enhanced thevalue of the test results.

The graph in Fig. 13 also shows themovement of the pile tip and the mea-sured compression of the pile. After aload of 70-90 tons, the measured compres-sion plots in a straight line, indicatingthat the part of the added load used forovercoming shaft resistance is constant.It would be highly improbable that theconstant value is other than zero. There-fore, the applied additional load remainsunreduced by shaft friction straight downto the pile tip, and the slope of thecompression line is equal to the slopeof the elastic line. The combined elasticmodulus of the pile determined from thisslope is 5,1 x 10" psi (35000 Mpa).

According to a method of analysis pro-posed by Trow (1967), the pile tip startsto move when the elastic line becomestangential to the load movement curveof the pile head, and the load appliedthereafter goes straight to the piletip. The analysis by Trow is valid for alinear, i.e. triangular or rectangular, dis-tribution of shaft resistance. The testresults presented in Fig. 13 show that ata load of about 70-90 tons, the elasticline, established from the measured com-pression, becomes parallel to the load-movement curve. Consequently, accordingto Trow's method of analysis, the shaftfriction must be approximately linearlydistributed, and the shaft friction valuecannot be greater than about 70-90 tons.

When assuming constant unit shaft fric-tion, i.e. rectangular shaft resistance,the distribution of load in the pile be-comes linear and from knowledge of thecompression of the pile, Fellenius (1969)has shown that simple relations can beestablished for the load at the pile endand the total shaft resistance, as shownin Fig. 14. At pile head loads of 224, 246and 280 tons, the measured compressionswere 096, 1.07 and 1.24in, respectively.The values result in calculated pile tiploads of 160, 182 and 216 tons, respec-tively. The corresponding calculated pileshaft resistance was 64 tons for all threepile loads.

300—

O

UJx 200—UJ

a.

EXAMPLE N'2

QUICK M.L. TEST

255

JJ VANDER VEEN

250MA 2 U R K I EW I C 2

P 270IFULLER HOY

DE BEERL250

BUTLER-HOY

240DAVISSON

PILE HEAD—

PILELENGTH=L

AREA -"A L/p

Pskin

Rip Pave PheadLOAD

ASSUMED LOAD

DISTRIBUTION IN THEPILE (CON "TANT UNIT

SHAFT FRICTION ALONG

THE PILE I

CJKE0

100—

P =A E BLave L

tIP ave head

skin head tID

P I LE T IP—tip

DEPTH

I

I

3.00 I 4 00

0 25

MOVEMENT ( INCHES)

I'

100 2 00 500

increments of 22 tons are too large tojustify the refinement. An increment of10 tons, instead, would have shown muchmore precisely the load-movement de-velopment during the first 100 tons ofapplied load.

Lacking adequate soil data, the truetip and shaft loads cannot be determined.However, they lie somewhere in betweenthe mentioned figures. The results of thecomplete analysis have been plotted in

Fig. 18 showing the load-movement cur-ves for the tip and the shaft (vs headmovement) for the two extreme distri-butions of the shaft resistance, Detailedknowledge of the soil profile could nar-row the ranges. However, for most prac-tical purposes, determining the shaft re-sistance to be somewhere in between 64and 96 tons, as in the subject case, isgood enough

The C'alue has additional analyticalsignificance. The ratio C's plotted in

Fig 19, both as a function of the loadat the pile head, and as a function of theinverse of the load. According to thederivation by Leonards fk Lovell (1978),the plot of C's the inverse of P is astraight line, if the change of compression,dII, for a change of load, dp, is a constantvalue, This is the case when the maximumshaft resistance is reached and surpassedby the applied load.

The equation for the line is:

Fig. 13. Load-movement diagram from Quick M.L. test withmeasurement of pile tip movement. Example 2

The value of 64 tons is less than thepreviously established maximum possibleof about 70-90 tons. For many reasons, itis prob'able that the unit shaft resistanceis not constant. Recently, Leonards 82

Lovell (1978) proposed a method of ana-lysis using measured pile compression,which allows a variety of distributionsof shaft resistance to be tried in theanalysing of the test data.

Leonards 8I Lovell established the fol-lowing refations:

C is 0.5. The previously mentioned threeloads and average loads give values ofx of 0.714, 0.740 and 0,772, respectively.Insertion in the Leonards 8I Lovell rela-tion gives values of the tip loads, whichare equal to the ones calculated pre-viously, i.e. 160, 182 and 216 tons.

The nomogram of Fig. 16 is applicablewhen assuming a triangular distributionof shaft resistance. In this case, the ratioC becomes 0.667 and the calculatedvalues of x are 0.571, 0.610 and 0.658respectively, resulting in the correspond-ing pile tip loads of 128, 150 and 184tons, and a shaft resistance of 96 tonsfor all three loads.

In these days of the pocket calculator,it is easier to work with the equationsdirectly, as opposed to using nomograms.In Fig. 17, the equations behind the no-mograms in Figs. 2 and 3 are presented.A third pattern of shaft resistance isadded, which is useful for piles in homo-geneous clay. The reduction of shaft resis-tance at depth ~ is intended for use whenanalysing a progressive mobilisation ofshaft resistance.

To discuss the results of the analysisof the load test presented above, the as-sumption of constant unit shaft frictionalong the entire length of the pile re-sulting in a shaft load of 64 tons is pro-bably incorrect, However, the shaft loadof 96 tons calculated on the assumptionof triangular distribution of shaft resis-tance is greater than the maximum pos-sible shaft load. To arrive at a shaft loadin between 64 and 96 tons, the analysiscould be repeated with a C ratio between0.500 and 0.667, chosen either from Fig.15 with two rectangular shaft frictionpatterns or from Fig. 16 with an uppertriangular and a lower rectangular pattern.For instance, C = 0.58 determines theshaft load to be 76 tons. However, nojustification is available for further re-finement. Such justification would havebeen, for instance, a definite change ofsoil profile at some depth, Also, the load

C' C

1 —Cwhere

x = ratio between the pile tipload and the load appliedtO the pile head (P„P= x X P),

C' ratio of measured compres-sion to column compression,the latter being the com-pression of a free columnsubjected to the same loadas the pile,

C = ratio of elastic compressionof the pile at a load P sup-ported totally by shaft fric-tion to the column compres-sion for tho same load.

1C' n —K-

PThe ratio C's known from the mea-

sured data, The purpose of the analysisis, either from knowledge of x, i.e. thetip load, to determine C, i.e. the relativedistribution of shaft resistance, or, in-

versely, from knowledge of relative dis-tribution of shaft resistance determinethe tip load.

It is not necessary to know the factualshaft resistance in order to establish theratio C. Leonards & Lovell (1978) havedetermined C for two principal patternsof shaft friction, and these are presentedin the nomograms in Figs 15 and 16. Thecase of constant unit shaft resistance isa special case of the nomogram of Fig.15; the two friction values are equal and

whereAE da

n = —x—L dp

The factor n is equal to 1 only if all shaftfriction has been mobilised, as in thiscase, which, as pointed out by LeonarasIk Lovell, is not necessarily always thecase.

In Fig. 20, an additional example isgiven. Results are shown from a QuickML test on an 84ft long, 12in precastconcrete pile driven through clay and intoa very competent glacial till, The ditgramshows the measured pile head movement,pile compression, and the head load versus

September, 1980 27

Fig. 14. Shaft and tip load calculation from measurement ofpile compression (tip movement) when assuming constant unitshaft friction

I.O

0.9

SHAFT FRICTION PAT

fS fsGROUND

SURFACE

I

I

I

P I LET I P

l ~ I.O I I I K

SHAFT FRICTION

fs fsGROUND

SURFACE

0.9 I

I

iI

PILE I

I

PATTERN

0 8 'll

C=I—

UJ

O

U

UJOo C7

I ),'(I ((I I SI2 I- K II K) )

O.&e-s-RATIOS OF —= kfs,

0.8

)—ZUJ

O

UUJO

0.7

0.60.6

0.55- )- Xk)- )(k .

0.50

0 05l I

I.O05

0 05I

I

I.O

RATIO 7/L = RATIO Jf/I

Fig, 15. Coefficient C for various distributions of unit shaft friction Fig. 16. Coefficient C for various distributions of unit shaft friction

measured tip movement, plus the calcu-lated shaft resistance and tip load. Thecalculations are performed assuming thatthe distribution of shaft resistance fol-lows the third pattern in Fig. 17 withl = L, and k equal to the same ratioof shear resistance as found by vane sheartesting. (For details on the soil profile,and an older, much more time-consumingand arbitrary, method of analysis of thetest results, see Fellenius & Samson,1976).

The pile test is not carried to ultimatefailure, However, the Leonards-I ovellmethod of analysis of the simple tell-tale measurements of the tip movementmakes it possible to establish that themaximum shaft resistance acting alongthe pile is 240 tons and the maximum tipload mobilised is 185 tons. Indeed, thisis a result well worth the expenditure of abit of time and money,

As in the previous example, additionalsupportinq information is gained from aplot of C'ersus 1/P for the results, andalso C'ersus P. As seen in Fig 21, whenqoing from riqht to left in the diagram,the C's 1/P is at first a curved linelater becominq a straight line pointing toC' 1.00, 1/P = 0. The P value for whenthe curve becomes a straight line deter-mines the point (load) where all shaftfriction is mobilised, in this case for P =300 tons.

The results of the second test are al-most certainly affected by a residual loadcaused by the reconsolidation of the clayafter driving and estimated to be about

28 Ground Engineering

20 tons. This load acts on the pile be-fore the start of the test loading and, as aresult, the compression of the pile doesnot start from zero, but has an initialvalue of about 0.08in. Adjusting the

C'aluesaccordingly, and recalculating theresults show that the adjusted maximumshaft resistance is 210 tons and the maxi-mum tip load is 235 tons. The differencein calculated tip load is 50 tons, 2.5 timesthe estimated residual load.

The above shows how sensitive themethod of analysis is to residual loads,and, therefore, also to inaccuracies of themeasurements. However, the old subjec-tive methods are actually even more sen-sitive, but because of their arbitrary nature,this is not always evident. In contrast,the Leonards & Lovell method allows adetermination of the extent of uncertain-ties influencing a test It provides, there-fore, the engineer interpreting the datawith a zone of reliability and a confi-dence he would otherwise have felt onlybecause of his ignorance of the uncer-tainties involved and of their effects,

To take full advantage of the Leonards& Lovell method of analysis, the testingmethod should not be the "standardloading procedure", which for all techni-cal and economical purposes is the worstmethod to use, but the "Quick load testmethod". The load increments must beapplied at constant intervals (for practicalreasons usually 5 minutes, to allow for atleast three readings per increment). Theincrements should be small enough to allowfor at least 30 or 40 increments before

kr

2

()-kl- ~ k2 2

2-2 A i)-k)- x k

)

L

2)+k ~ z)I- 8s k +

Fig. 17. Mathematical expressions for co-efficient C for various distributions of unitshaft friction

reaching the maximum test load, Naturally,a reliable load cell should be used tosupplement the jack manometer, and everyeffort must be made to ensure reliablemovement values.

Fig 22 shows another method of pre-

300

~(~-TIP VS BUTT LOAD

INVERSE OF LOAD AT PILE HEAD ( I /P)3 -3 .3

0 I)0 510 IDIO1.00 'XAMPLE 2

MEASURED COMPRESSION

COLUMN COMPRESSION

0 90-

200-f.-TIP

0 80-

VIXO

C3ctO

0.70-

/P

IOO- ———p ———--p--p- -p .p3

—SHAFT

----0-----O--O--O-—

0.60.

0.50-

;7

0-,0 1.00 2.00 3.00

MOVEMENT (INCHES)

Fig, 18. Example 2. Load-movement diagram for shaft and tip loads

0.400 IOO 200

LOAD AT

PILE

HEAD�(TONS)Fig. 19 Example 2. Ratio C'lotted vs load at the pile headand vs the inverse load

300

500

HEAD LOADYS TIP MOVEMENT COMPRESSION INVERSE LOAD (TONS )

400—HEAD

01.00—

-35 IO

-310.10 15 10

-3

V3ZO

CI

O200—

SHAFT LOAD-VS

HEAD MOVE MENT

TIP LOADVS

HEAD MOVEMENT

C

0.50—

~

~

.r~

.r wJI

~rVS P

100—

0.10—

00 2.00

I01 I

0 I 00 200 300 400 500 600

MOVEMENT ( INCHES) LOAD ( TONS 'I

Fig. 20. Example 3. Load-movement diagram from Quick M.L. test Fig. 21. Example 3. Ratio C'lotted vs load at the pile headwith measurement of pile tip movement and vs the inverse load

senting the results of an analysis of a piletest (Example 2). The load distributionin the pile is shown for three differentloads at the pile head. The straight line

represents the load distribution for con-

stant unit shaft resistance (rectangulardistribution) and the curved line that fora triangular distribution of shaft resistance.The interesting point in this graph is thatto fulfill the condition that both load dis-

tributions give the same average load in

the pile, the two areas A'nd A" mustbe equal.

The condition indicated in Fig, 12 canbe developed to determine non-linear load

September, 1980 29

00

100

LDAD (Toifs)200

PILE HEAD300

B(ITT LOAD

50—

TELLTALESE

zi-CLiuto

TiP TELLTALE

d distribution from tell-tale measure-Fig. 23. Determination of loa is ri

ments

- ~—'0

1.0-

)Z2.0-X

Iu

3.0-

i-

the i le during2. V tical load distributions in p'ig.22. Example . ericthe test

100—

E. Ik Wa//ays, M (1972): "Frankipi

'd bases", La Techniquepiles with overexpande as

des Travaux, No 333, 48 pp."Static measurements o f

I tli t f P'Iur". Design and ns a aCellular Structures, Envo Publ.Foundations and Ce! u ar

Co., Edited by H-Y Fang, pp,1972: "High capacity piles",Davisson, M. T.

ASC i oi SLecture Series,

Foundation Construction,52 pp,

H. (1969): "Bearing capacity off!I I t t", Rofnction piles Results s

24 pp (I 5 edish)Research, Report No.1975): "Test loading of piles.

d oof testingMethods, interpretation an newprocedure", Proc. ASCE, Vo

Samson, L. (1976): "Testingof dnveability of concrete piles anto sensitive clay", Canadian o,

~ pp.. 139-160.

ick-load test method conven-t of ti movemen tional methods and interpre a i, as roposed by va lue that measurements o i

. 78-86,wo- can provide in the anay'd'

that is strongly recom-,... e

also for routine tests. For clos - sM4 ppave been In m~~d~d asan,d t It b il

'23, the diagram en

COnCrete pi eS respo, ASCE,response cohesive soils", ASCE,precast prestressires a little bit of advance planning

be can be cast in thetip and a second one ocate

f the SO that a Centre tube Can e ts on higile. H-piles could necessitate som, osium on e awo

din and a few hours of pre L d, . 388-g ph " gd paration. Cast-in-place pi es a

u h eluded.of the simplicity and low re a-

2 from the pi e eds, lotted at In view o

tive cost coupled with t e argt e w

ell-tale locations ivined, by no means "

n w tea~a, eo emi

II of extra information gai, w

ale Is itt 6an te

hausted in this Paper,considering ex ain ti meaSure- N y DOT sSo

POSSIe true loa ex ec...,, . E. Ei Tho nbu, T.

c dure SCP4/74, 35 pp,i 'fill the conditions m n

"Foundation Engineering", Secon

distribution line must fu i'

m n

, A'nd A" and B'nh athema- 8 8rBnCE)S

I d" A I 8 k 1 weisOn aS firat been ASTM Standards, Part

p we oathe "true" load distribution has irs

I 88

nsan J. (1963); Discussion Hyper- o massumed to cons'

IcSV SM4 pp 241-242 1

r to he 48th nnuaHo, H. E. (1977): "Users manual o a

'ck-load method for founda-load. It also mean

I Highway Adminis-tion losad testin " F dere igDe op

Whr kChin, F. K. (1970): "Estima onsistance resu s i

"'gh oad of Piles not earned, . 'ou s, eo ec n

Southe t A i Co f, o Soi ngn

Discussion, "Pile tests. Whiraker,....: newIn the ana ys'rkansas River Project", AS

PP0-932. h. an ounuse was made o

Whiraker, T. (1963): "The constan rvan het grensdraag vermo- i amovement tr pi

ingen op staal" tration tes oqen van zend onder fundeTijdshrift der openbar Werken van ', . 'a acity o a

Se tember, 1980 31has been to demonstrate t e trem

ep