Centrifuge model study on spudcan footprint interaction 4

3.3 Experimental setup 3.3.1 NUS geotechnical centrifuge – Full spudcan test The centrifuge model tests presented in this report were conducted on thebeam centrifuge at National Universi

3.2 Centrifuge modelling

In a keynote lecture by Schofield (1998), he pointed out that geotechnicalcentrifuge model testing is as valuable as the observational method, howevergeotechnical centrifuge model tests help solve problems where “theobservational method” cannot be used One of the conditions that cannot bereplicated for full scale test purposes is foundation behaviour of a mobileindependent leg jack-up rig deployed offshore (Schofield, 1998) Some studies

in relation to the behaviour of spudcan footings of jack-ups have beenundertaken using centrifuge modeling technique in the past few decades

includingMurff et al (1991),Dean et al (1993 and 1995), Tsukamoto (1994)

and more recent works by Stewart and Finnie (2001), Cassidy et al (2004),

Hossain et al (2004), Purwana (2007), Teh et al (2008)and etc.The offshoreindustry already has confidence in geotechnical centrifuge modeling (Murff,

1996) In addition, owing to complexity of soil behaviour and the limitations

in numerical modeling, physical modeling provides necessary data for thevalidation of numerical approaches and the refinement of the empiricalsolution for foundation response(Randolph and House, 2001)

The main focus of the present study is to investigate the footprintcharacteristics and the interaction between spudcan and footprint Both requirecreation of spudcan footprints by performing spudcan penetration andextraction The footprint characteristics required evaluation of soil shearstrength profiles beneath and around a footprint at various times Soilconsolidation condition is deemed to be important in defining the soilcondition beneath a footprint The time factor, Tv, for soil consolidation incentrifuge model and prototype can be expressed as follows:

2 2

p

p vp

where N is the ratio of centrifugal acceleration to the earth’s gravitationalacceleration cv is assumed to be the same for the same soil used in model andprototype Substitute Equation 3.2 into Equation (3.1) gives

Unlike surface footing, the spudcan test involves considerable penetrationinto the model ground of particularly in soft clay The soil failure mechanismwas found to change from shallow failure to deep failure over a range ofpenetration (Hossain et al., 2004 and Purwana, 2007) A correctly modeledover-burdened stress is required to trigger the soil back flow In the situationswhere considerable cavity formed above the infilled spudcan, the effect of theweight of the infilling (or the over-burdened stresses) on the net bearingresistance becomes significant Also, for spudcan extraction, the infilled soilprovides a seal against transient suction developed at the base of the spudcan(Hossain et al., 2004) It is readily seen that the correctly modelled soil over-burdened stress is essential to ensure that a more realistic soil failuremechanism is simulated Centrifuge model testing therefore allows the use ofsmall model structures to simulate a full size prototype stress field

3.2.1 Centrifuge scaling laws and model error

The scaling relationships between the model and its prototype can be derivedeither by dimensional analysis or by considering the governing equations and

system mechanics A standard basic scaling law is needed to ensure theconsistency response between the model and the prototype The centrifugescaling relations is shown inTable 3.1.

Unlike the earth’s gravity which is relatively uniform throughout thedepth, the acceleration field simulated in centrifuge varies in a model bothmagnitude and direction This is because the inertial acceleration field incentrifuge is given byR2 where is the angular rotational speed and R is theradius from the centre of rotation to any element in the soil model Hence, theartificial gravitational acceleration is expressed as

where N is the ratio of centrifugal acceleration to gravitational accelerationand g is the earth gravitational acceleration with a value of 9.81 ms-2 Astrategically selected nominal radii, Re can be adopted to minimize theshortcoming of g-level variation throughout the entire soil model Asillustrated by Taylor (1995), the stress variation can be minimized to lowerthan 5% if the Re is set to be the distance from the centre of rotary to one-third

of model height

3.3 Experimental setup

3.3.1 NUS geotechnical centrifuge – Full spudcan test

The centrifuge model tests presented in this report were conducted on thebeam centrifuge at National University of Singapore (NUS), see Figs 3.1(a)and (b) The working area of the platform is 750 mm × 700 mm and theheadroom available is about 1200 mm This 2 m radius centrifuge is designedfor a payload capacity of 40g-tonnes A stack of 100-tracks silver-graphite slip

rings is mounted on top of rotor shaft for power and signal transmissionbetween the centrifuge machine and the control room More detail informationabout NUS geotechnical centrifuge can be found in Lee et al (1991) and Lee(1992).

All tests were conducted at 100g The photograph and schematic diagram

of the centrifuge model set-up for the present study are shown inFigs 3.2(a)

& (b) As the tests were to simulate foundation behaviour in offshoreenvironment, all tests were conducted underwater The model setup for fullspudcan tests comprises two main components, namely the model containerand the loading platform mounted on top of the container Details of eachcomponent are described in the following sections

3.3.1.1 Model container

The model containers are cylindrical tubs made by stainless steel Eachcontainer has a diameter of 550 mm diameter and height of 400 mm Watervalve is attached to the base of the container to facilitate water circulation Thewater valve is opened during soil sample consolidation and closed during thetest A complete test involves consolidation of soil sample and followed byperforming spudcan penetration without intervention of the centrifugal motion.Prior to spudcan penetration, the valve was closed mechanically in-flight usingdownward movement of the hydraulic piston

3.3.1.2 Loading platform and actuators

The loading frame is made by stainless steel with a plan area of 320 mm x 650

mm and height of 330 mm (seeFigs 3.3a and b) A 60 mm × 550 mm openingwas made at the middle of the loading frame This is to provide a continuous

passage for the vertical loading actuators to move at one axis Two verticalloading actuators were engaged, the bigger one (denotes VA1) was used toperform spudcan penetration and extraction, while the smaller one (VA2) wasemployed to perform T-bar tests The loading actuator VA1 is a hydrauliccylinder with a piston of 75 mm diameter and a rod of 37.5 mm diameter Atmaximum hydraulic pressure of about 60 bars, the big loading actuator is able

to provide a maximum compressive (downward) force of 2.65 tonnes andtensile (uplifting) force of 2 tonnes The difference between these two forces isdue to smaller effective area of the piston at the tensile side (to exclude the rodarea) The loading actuator VA2 is a hydraulic cylinder with a piston of 30

mm diameter and rod of 16 mm in diameter, and with capacity of 0.43 tonnes

in compression and 0.3 tonnes in tension under the maximum suppliedhydraulic pressure The maximum stroke is 300 mm for VA1 whereas 250 mmfor VA2 A flow divider was needed to divide the hydraulic flow into twoseparate passages into the two hydraulic control systems from a single outlet

of hydraulic supply When the loading actuator is engaged to perform thedownward or upward movements, a digital command is sent to a servoamplifier through a switch box to form a closed-loop circuit with feedbacksignals The actuator would then move whenever there is difference betweenthe command and feedback signals and stay stationary when these two signalsare identical

These two vertical loading actuators on top of a movable platformcomprising a pair of minirails, a stepper motor, a gearbox, a ball screw, acontrol device etc (see Fig 3.3b) Each minirail consists of a frictionlesslinear slider and six (6) compatible bearings The load carrying capacity per

each bearing and slider set is 350 kg The coefficient of friction of the linearslider and bearing sets is as small as 0.2 The moving platform facilitates themeasurements of soil shear strength at locations away from the centre of themodel container Potentiometers are used to monitor the elevation of thespudcan and the location of the moving platform The overall system allows amodel spudcan footprint to be created by performing a model spudcanpenetration and extraction at the centre of the model container The entireplatform is capable to move in one axis at 100g.

3.3.1.4 Model jack-up leg

As discussed in Session 1.1, there are two basic types of jack-up leg design;columnar and open-truss legs The open-truss legs are more common as theyare stronger in both bending and axial loadings In order to achieve the sameratio of bending and axial strengths in the proposed model, intricate modelbuilding technique to construct the truss-braces structure was involved Thisseems to be impractical as different rigs have different leg designs andinstrumentation on the model legs would be extremely difficult Since the soilreaction at the spudcan level is the main interest of this study, the spudcan leg

is modeled as a thin walled circular hollow section (columnar type) withflexural rigidity similarity.

In the selection of a more realistic leg flexural rigidity, a literature review

on the model legs used in other studies were carried out and summarized in

Tables 3.2 The actual field jack-up leg stiffness available published in thepublic domain are given in Table 3.3 The leg stiffness, EI generally rangesfrom 1.01×1012 Nm2 to 2.75×1012 Nm2 (except for Murff et al., 1991 where amuch lower stiffness was used) for spudcan diameter ranging from 7.6 m to 18

m Foo et al (2003) made a comparison of equivalent stiffness EI/L3 for fourdifferent jack-up rigs over the leg length L (Fig 2.7) They pointed out that theequivalent stiffness difference became less distinct when the jack-up operates

in deeper water or has longer leg length Owing to complexity of spudcan-soilinteraction, uniqueness of installation location and unknown lateral load, it isimpossible to determine the level of stiffening required for a jack-up leg (Foo

et al., 2003) It is apparent that there is no typical leg stiffness and theequivalent leg stiffness is dependent on the leg length (effective length fromspudcan to hull) In present study, a stiffness value of 1.18×1012 Nm2(the legstiffness of rig type 116C) was chosen for the first model leg (denotes Leg 1)fabricated Owing to limited headrooms in the NUS centrifuge, the maximumallowable leg length is 280 mm or 28 m in prototype This represents thespudcan installation in shallow water To account for the spudcan installation

in deeper water, the second leg (Leg 2) with much lower stiffness wasfabricated to simulate the equivalent stiffness in water depth of 70 m due tothe leg length limit By adopting the equivalent stiffness similitude shown

below, Leg 2 was designed to have EI magnitude of 15.6 times lower than that

of Leg 1

1 3 2

3

Leg Leg L

of the prototype

Both model legs were instrumented with 2 levels of full bridge axialgauges and 3 levels of full bridge bending gauges, as shown inFigs 3.6a & b.Full bridge strain gauges were selected as differential temperature can becompensated

To avoid the strain gauges being disturbed by soil back flow duringspudcan penetration, a steel tube was used as a shaft protector to separate thesoil and water from the model legs Consistent results obtained for twoidentical tests indicated no undesirable influences acting at the strain gauges.However, the limitation of engaging the shaft protector is that it substantiallyreduces the volume of back flow soil on top of the spudcan Comparisonbetween the thin model leg of 16 mm diameter and the Leg 1 with protected

shaft of 48 mm diameter were made as shown in Fig 3.7 The difference inaxial force during penetration in soft clay is insignificant up to a depth of 6 mand increasing gradually with depth as the soil back flow takes place.However, within the test range, the difference in axial force is less than 10%and even lower for penetration in firmer clay where the soil back flow isnegligible Thus, this suggests that the influence of the protected shaft on theaxial force can be accepted.

3.3.1.5 Calibration of the strain gauges of the model leg

Two levels of full bridge axial gauges namely A2 and A4 were installed atpositions shown in Fig 3.6a When the model leg experienced stress, thecorresponding strain would be captured by the axial gauges through the TMLstrainmeter that was installed on-board The axial force can be computed byconverting the strain to force by either adopting the theoretical approach ordirect load calibration Theoretically, the axial force can be calculated based

on the following equation (in elastic state):

a

EA force

wherea is the axial strain anda1 to 4 are the strain measured by each gauge atthe same level (in microstrain), given as follow

6 4

3 2 1

104

gauges (2 mm width x 5 mm length) are indeed attached on a curving surfacerather than a flat surface This may lead to some errors in the calculation of theaxial force based on Eq 3.6 The above shortcoming can be overcome byusing the direct load calibration method The model leg was placed vertically

in a triaxial frame When the base plate was lifted up, the model leg wascompressed between the base plate and the slip ring and the correspondingcompressive load was recorded as the strain gauges readings (captured by aTML strainmeter) By plotting the strainmeter reading (in micro-strain) versusapplied loads, the calibration factors can be deduced The direct calibrationmethod not only eliminates the potential errors due to curvature but also theprobable non-uniformity of the cross-sectional area and the existence of epoxylayer (used for waterproofing) A comparison between the theoreticalestimation and load calibration for A2 is presented inFig 3.7 For the reasonsdiscussed above, the calibration factor obtained from the load calibration isused to calculate the axial force

Similar to axial force, bending moment can be theoretically estimated bythe following equation (in elastic state):

y

EI moment

2 3

1

104

)(

table and the other end is cantilever Weights were hanged at the cantileverend and the moment acting at each bending gauge level was calculated bymultiplied the known force with the moment arm The corresponding bendingstrain was captured by the TML strainmeter The results are plotted togetherwith the theoretical estimation inFig 3.9 The calibration factor is the gradient

of the plot (in Nmm/micro-strain) As the bending gauges have similarlimitations as for the axial gauges, the calibration factors derived from thedirect calibration are adopted to calculate the bending moments

3.3.1.6 Derivation of the VHM acting at the spudcan

One of the main interests of the present study is to investigate the interactionbetween a spudcan and a footprint As the fully restrained connection ismodelled, the interaction is reflected in terms of vertical load (V), horizontalload (H) and moment (M) acting at the load reference point (L.R.P.) The V ismeasured by the axial gauges at A1 & A2, and either one can be used to find

V As the M and H at the L.R.P cannot be measured directly from theexperiments, they are evaluated by incorporating the bending moments usinglevels B1 to B3 into i) simple beam theory and ii) beam column theory.Detailed derivations based on these theories are presented inAppendix A

Figs 3.10a – dshows the H and M profiles at L.R.P for Legs 1 and 2computed using simple beam and beam column theories The computed H and

M for Leg 1 are identical using equations derived from both theories(Figs 3.10a and c) For Leg 2, the H computed using simple beam showsslightly larger magnitude than that of beam column (Fig 3.10b) The findingsuggests that if the the structure stiffness is sufficiently stiff within the loadrange (structure deflection is relatively small) which the P-∆ effect is

negligible (a situation where within the load range), the simple beam theoryprovides a reasonably good estimation When this is not the case, either theload is too large or the structure is slender, then beam-column theory should

be employed to compute the H and M For consistency, all results presented inthis thesis are computed based on beam column theory

3.3.1.7 Instruments and transducers

Difference stroke length Midori potentiometers were used to measure thelinear displacement The potentiometers give measurements with precision of

up to ±0.1% 100 mm and 50 mm potentiometers were used to measure thesoil settlement during the soil consolidation and test proper 300 mmpotentiometers mounted on the movable platform were used to monitor theelevations of loading actuators VA1 and VA2 The movements of these twoactuators were controlled via the differential between command and feedbackdisplacements from the respective potentiometer A 300 mm potentiometerwas used to monitor the position of the movable platform (Fig 3.3)

Miniature pore pressure transducers (PPT) were embedded in the modelground for the measurement of pore water pressure Two different capacities

of PPTs with maximum measurement of 3 bar and 7 bar were utilized in thetests The nominal sensitivity of the 3 bar PPT is 5.52 mV/V/bar, while for 7bar PPT is 2.4 mV/V/bar

Interface SML-series miniature load cell with capacity of 1000 lbs wasattached at the connection between the model leg and the vertical actuator tomeasure the total load acting at the spudcan This load cell has an accuracy of

±0.05% and a safe overload of 150% The load cell was only used for spudcaninitial penetration on undisturbed soil To avoid probable interference to the

load cell, the axial load was measured by the axial gauge instrumented on themodel leg (see Session 3.6) for spudcan penetration at certain offset distancefrom footprint centre.

3.3.2 NUS centrifuge – Half spudcan test

3.3.2.1 Half-spudcan test setup

Half-spudcan tests were performed to investigate the soil failure mechanismsduring initial spudcan penetration and extraction and the spudcan re-penetration at certain offset distance from the initial penetration site.Fig 3.11

shows a close view of the half-spudcan soil sample Fig 3.12 shows the fullview of the half-spudcan test setup A rectangular container with dimensions

of 530 mm (length) × 400 mm (width) × 550 mm (height) was used The frontface of the container is made of transparent Perspex that allows theobservation of soil movement The image processing technique evaluates thesoil movements by tracking the patches with identical pixels between twoimages The pixels are recognised by the unique texture of each patch Sincethe kaolin clay used is in plain white colour, it does not provide sufficienttextures for the PIV program to track the soil movements To overcome this,light “flock” powders and beads that are black in colour were sprinkledrandomly onto the frontal surface of the clay block The front wall of thecontainer is detachable which can be disassembled for sprinkling the flocksand bits on the white clay surface When it is assembled, the connectionbetween the front wall to the two side walls is water-tight

The half-spudcan was fabricated to have the dimension of exactly half

of the model spudcan with a diameter of 100 mm (Fig 3.4(a)) To prevent soil

or water ingression through the contact face, a 4 mm thick highly compressible

and extensible rubber was attached on the half-spudcan’s face A rod (tosimulate the model leg) was used to connect the half-spudcan to the verticalactuator (VA1) The rod was positioned at the centric of the half-spudcanwhich was slightly away from the centre of the full spudcan This allows thehalf-spudcan to be placed closely against the wall A layer of transparentgrease was applied on the inner face of the transparent Perspex to reduce thefriction between the spudcan face and the wall.

The image capturing system includes a JAI CV-A2 progressive scancamera, which was mounted on a frame extended from the centrifugeplatform This provided sufficient distance between the camera and the frontalface of the soil sample Images were directly stored in a on-board personalcomputer (PC) that was installed on the centrifuge arm The PC was remotelycontrolled by a computer in the control room where the image capturingprocess could be easily monitored To enhance the accuracy of the experimentyielding high resolution images, only the soil movements within the zone ofinterest was captured The captured images had a dimension of 1600 pixels ×

1200 pixel (horizontal × vertical) The image capturing rate was set at 1frame/second

3.3.2.2 Image processing technique

Particle image velocimetry (PIV) technique was used to quantify the soildisplacement by comparing a pair of images The concept of applying PIVtechnique in geotechnical problem was first implemented by White et al.(2003) As this technique enhances the precision and accuracy compared toprevious image-based methods, it has been widely used in qualifying andquantifying soil movements in many geotechnical problems including spudcan

foundation behaviour, tunnel-pile interaction, pipe-soil interaction and subseaanchor.

In the present study, the GeoPIV8 program developed by White &Take (2002) was used to analyse the soil movement The first step of theimage analysis involved meshing the first image As the clay surface wasartificially randomly textured, each patch had a unique identity that wasrecognised in pixel The search for the same patch in the subsequent imagewithin the pre-defined search zone was done using a set of matchingalgorithms The distance between the initial patch (the first image) and thesearched patch (the subsequent image) that gave the peak correlation wasdenoted as the soil displacement in image-space The conversion from theimage-space to object-space required consideration of image distortion thatwas spatially varied throughout the image (Taylor et al., 1998; White et al.,2003) The control markers with known distances were fixed on the inner face

of the transparent wall and served as calibration tools (Fig 3.11) They wereused to derive the calibration factors to convert the soil movements fromimage-space to object-space (in mm) Study on the patch sizes was performedand the pixel size of 20 × 20 was found to be efficient in computing timewithout much compromise the accuracy

3.3.3 UWA drum centrifuge

The drum centrifuge at the University of Western Australia (UWA) has adiameter of 1.2 m with twin centric shafts that coupled with a precision servomotor.Figs 3.13 and 3.14 show a photograph and a schematic cross-sectionaldiagram of the drum centrifuge A detailed description of the drum centrifugefacility at UWA was reported by Stewart et al (1998) The outer sample

containment channel is 300 mm in height and 200 mm in radial depth Themaximum rotational speed of the channel is 850 rpm which is equivalent to anacceleration of 485g at the base of the channel and 364g at the top of a 150

mm thick sample The central tool table is allowed to rotate differentiallyrelative to the channel rotation to the desired positions The tool table is lockedwith the rotating channel at synchronous rotation during testing The tool tablehas an actuator that can be controlled to move vertically and horizontally,including the circumferential movement, the testing tool has three axesmovements during the in-flight testing One of the important features of thecentre tool table is that it can be stopped independently without affecting thechannel’s motion The actuator can then be lifted up to the top of the drumcentrifuge for changing the testing tool such as spudcan, T-bar, laser scanningdevice etc

Data acquisition consists of two onboard computers, one serving thetool table and the other one serving the channel Communication between thecentrifuge testing tools and computers in the control room takes place seriallyvia sliprings For additional instrumentations installed inside the sample such

as pore pressure transducers, wireless connection is used to transfer themeasurements to the data processor in the control room

3.3.3.1 Model spudcan and leg

The model spudcan with largest diameter of 14.55 m in prototype was used inthis study A schematic cross-section of the model spudcan is shown in Fig.3.15(a) The model leg is a thin walled circular hollow section with prototypeflexural stiffness value of 3.2×1012 Nm2 with the prototype length of 38 m

shown inFig 3.15(b).Fig 3.16 shows top view of the model spudcan and legattached on the centre tool table prior to the centrifuge spinning.

3.4 Sample preparation

3.4.1 NUS beam centrifuge

Malaysia kaolin clay was chosen in this study instead of marine clay, mainlydue to its relatively high permeability which could reduce the time forconsolidation significantly In addition, the physical properties are moreconsistent which is capable giving fairer comparison amongst tests Itsphysical properties were described by Goh (2003) and shown inTable 3.5.The sample was prepared using a similar method as those used by Goh(2003) and Purwana et al (2005) Firstly, dry kaolin powder was mixed withwater at a water content of 1.5 times its liquid limit, which is 120% of thepowder weight They were mixed together in a 550 mm circular mixer invacuum condition (where a constant suction was applied) for 4 hours and fullysaturated clay slurry was produced at end of the 4 hours mixing Prior to theclay slurry pouring, a 30 mm thick sand layer that served as drainage layer wasplaced at the base of the model container A layer of silicon grease was applied

on the inner wall of the model container to reduce the friction between thesidewall and the soil The clay slurry was then carefully poured into the modelcontainer The trapping of air pockets were minimized as the clay slurry wasplaced under water

After the clay pouring process was completed, the sample was subjected

to a gradually increase in loading pressure up to 20 kPa at 1g At the end of 1gloading, the soil sample achieved an effective vertical stress of 20 kPathroughout the depth The soil sample was then moved onto the centrifuge and

subjected to self-weight consolidation at 100g for at least 7 hours Thisresulted in a thin layer of over-consolidated clay profile for the top 30 mm (or

3 m in prototype) and normally consolidated profile below this depth Thisrelatively stiffer layer provided a sufficient bearing resistance to support thepotentiometers used to measure surface settlement

To prepare for a full over-consolidated clay sample, the preparationprocedure is the same except the sample was subjected to a much higher butgradually increased pre-consolidation pressures at 1g before subject to self-weight consolidation at 100g The magnitude of final pre-consolidationpressure is dependent on the desired clay shear strength profile The higher thefinal pre-consolidation pressure, the stronger the clay is The shear strengthprofile of the undisturbed clay sample was evaluated using T-bar and ballpenetrometers

3.4.2 UWA drum centrifuge

In this thesis, 10 tests performed in a full drum sample at 200g will bepresented The drum sample was first divided into two test sites where onehalf was normally consolidated clay and the other half was over-consolidatedclay Five tests were conducted in normally consolidated (NC) clay and fivewere in over-consolidated (OC) clay Australian kaolin clay was used and itskey properties are shown inTable 3.6

To prepare NC clay, a 12 mm thick sand layer was first placed at thebase of the drum channel to provide a drainage path for the clay sample Theclay slurry was then sprayed into the channel while the centrifuge wasspinning at 20g The slurry was then allowed to consolidate overnight at 200gbefore the second top-up of slurry took place The complete clay sample was

subsequently consolidated at 200g for two days when at least 90%consolidation was achieved The consolidation of clay was monitored by thepre-installed pore pressure transducers At this stage, the sample wasapproximately 160 mm thick with a normally consolidated strength profile.The undrained shear strength profile of the sample was evaluated byperforming several T-bar tests and ball penetrometer tests.

Upon completion of the test program for NC clay, a 30 mm thick sandlayer was sprayed on top of the sample The drum was then ramped up to 300gfor 2 days Once the pore-pressures stabilized, the water was drained off andthe drum centrifuge was stopped The sand layer was then carefully scrapedaway The sample was then ramped to 200g and left to re-consolidate foranother two days After the pore pressures stabilized, an OC sample ofapproximately 150 mm thick was obtained with over-consolidation ratio, OCR

of 12 at 1 m and 2 at 20 m below the mudline based on an assumption of aconstant vertical effective stress, v’, of 6.5 kN/m3 The undisturbed shearstrength profile of the sample was evaluated by ball penetrometer tests

3.5.1 Tests done in NUS

The ball penetrometer was used to evaluate the shear strength profile bymeasuring the penetration resistance of the ball while penetrating into the soil.Plastic solution for the flow around a sphere (Randolph et al., 2000) providedthe basis for obtaining estimation of shear strength directly from the measuredpenetration resistance

where qball is the penetration resistance of the ball (that can be obtained bydividing the load cell reading with the projected area of the ball), and Nball isthe bearing capacity factor for the ball For measurements of shear strengthprofile of a footprint, the ball penetrometer was found to give a comparablebut more stable reading than the T-bar This is because the soil profile within afootprint is highly variable which may cause bending on the T-bar due toasymmetrical resistance pressure and subsequently affects the measurements.

In this particular situation, the ball penetrometer has an advantage over T-bar

as it experiences less bending problem

The ball penetrometer used in the tests conducted in NUS centrifugehas a diameter of 11.9 mm, as shown in Fig 3.17(a) The ball penetrometercomprises a spherical ball that is attached perpendicularly to a rod Axialstrain-gauges were instrumented on the rod right above the ball that served asload cell The ball and is changeable with a T-bar (5 mm in diameter × 25 mm

in length) using the same load cell, as shown inFig 3.17(b) They were bothmade from aluminium alloy and fabricated in UWA To minimise the errordue to the difference in over-burden stress, the rod is to be designed as small

as possible However, owing to the smallest size of strain gauges available isaround 1 mm in width and also the strain gauges need to be attached on arelatively flat surface, the rod diameter is limited not to be smaller than 3 mm.With a layer of protective (or waterproofing) sleeve, the smallest shaftdiameter becomes around 5 mm With the ball diameter of 11.9 mm, the areaaspect ratio is around 1:2.4 Following Watson et al (1998), the bearingcapacity for the ball penetrometer, Nball, is taken as 10.5 The ball was installed

at a penetration rate of 3 mm/s It is established that this rate is sufficiently fast

to pertain the process of penetration in undrained condition based on thevelocity group parameter proposed by Finnie (1993).

The resistance during extraction provides a measure of the remouldedstrength of the soil The residual strength of the soil can be obtained byexercising the penetrometer penetration and extraction for a few cycles untilthere is no more change in the resistance

3.5.2 Tests done in UWA

In this study, one of the main interests is to evaluate the soil condition of aspudcan footprint In order to obtain a more representative soil strengthcondition, the area ratio of the ball to the spudcan should be maintained as low

as possible Hence, it is desirable to use miniature penetrometers with thesmallest possible diameter To overcome the constraint of the rod size asdiscussed in previous section, a new fabrication technique was developed andreported by Lee (2009), which the entire ball, shaft and embedded straingauges were cast in epoxy Fig 3.18(a) and (b) show the photograph anddrawing of the 5 mm diameter ball penetrometer Details of this ultra-smallball fabrication and the material properties were reported by Lee (2009) In thepresent study, a miniature ball penetrometer of diameter 5 mm with shaftdiameter of 2.5 mm was used to evaluate the undrained shear strength for soilsample prior to spudcan penetration and also after the footprint was formed.The area ratio of the ball to the spudcan is rather small, approximately 0.005.The ball was installed at a penetration rate of 3 mm/s It is established that thisrate is sufficiently fast to pertain the process of penetration in undrainedcondition based on the velocity group parameter proposed by Finnie (1993).The undisturbed su profile of the soil sample measured by the ball