The insertion from the open clay face was necessary to avoid the disturbance of the clay in the upper part of the sample, along the trajectory of the pullout.
Experimental investigation into the influence of a keying flap on keying of plate anchors Christophe Gaudin, Mark Simkin, David J. White Centre for Offshore Foundation Systems, University of Western Australia Perth, WA, Australia
Conleth D. O’Loughlin Institute of Technology Sligo Sligo, Ireland
ABSTRACT Plate anchors are used in deep and ultra-deep waters to anchor floating offshore structures. Once installed vertically under the seabed with the aid of a suction follower they must be rotated or “keyed” in order to mobilize full anchor capacity. A keying flap attached to the top of the anchor is commonly used, aiming at reducing the vertical translation component of the installation path during keying, although its efficiency appears to be uncertain. To quantify the performance of this keying flap and to understand its mechanism, centrifuge tests were performed on an anchor model, using Particle Image Velocimetry (PIV) to monitor the trajectory of the anchor and the behaviour of the keying flap through examination of the soil failure mechanism upon keying. Results indicated that the keying flap did not affect the anchor trajectory, except by introducing an offset to the loading.
vertical load and geotechnical efficiency (Wilde et al. 2001, Ehlers et al. 2004). Prior to 2004, SEPLA field application was limited to temporary moorings for MODU platforms. The first permanent application occurred in 2006 for a Floating Production Unit in the GoM (Rigzone, 2006). Today the anchor is commonly used in deep waters. Similar to a VLA, the plate must be keyed (rotated) from its initial vertical position to an inclination normal to the load applied by the mooring line (Dove et al. 1998), hence exhibiting the maximum crosssectional area (normal to the loading direction) and optimizing the bearing capacity. This is achieved, with minimal delay, by tensioning the mooring line and loading the anchor to typically 20 to 30% of its maximum holding capacity. During the keying process, the anchor experiences vertical motion, resulting in a loss of embedment. For offshore marine deposits, which usually exhibit an increase of strength with depth, any loss of embedment results in an unrecoverable loss of capacity, which may be significant.
KEY WORDS: Geotechnical engineering, centrifuge modelling, soft soil, anchor, keying, failure mechanism.
Keying Flap
INTRODUCTION
In order to limit the vertical displacement of the anchor during keying and hence the potential loss of embedment, SEPLAs often feature a keying flap (Fig. 1).
SEPLA Concept The Suction Embedded PLate Anchor (SEPLA) concept borrows installation techniques from conventional suction caissons yet during its operation life it is effectively a Vertically Loaded Anchor (VLA) and as such is appropriate for soft cohesive soils. A modified suction caisson known as a follower houses a plate anchor inserted in a vertical slot at the base end. During installation, the follower and plate anchor are lowered onto the seabed and installed to the design depth in the same manner as a typical suction caisson. Water is then pumped into the top of the follower in order to retrieve it from the sea floor for reuse while the plate anchor remains in a vertical orientation at the design depth. The SEPLA incorporates the advantages of a conventional suction caisson, namely: proven installation methods along with known geographical position and penetration depth, together with the benefits of a VLA, namely: low cost, the ability to handle a high degree of
Fig. 1. Typical SEPLA with keying flap (Courtesy of ExxonMobil).
The keying flap is a solid piece of plate attached to the top of the anchor which is free to rotate a limited amount (usually about 20 degrees) away from the shank but is restrained in rotation towards the shank. Early field tests incorporated the flap with the aim of “[increasing] the vertical bearing area of the SEPLA and preventing it from moving back up its installation track when tensioned.” (Wilde et al. 2001,). This intended effect is illustrated in Fig. 2.
Successive series of centrifuge tests performed on reduced scaled rectangular anchors, without a keying flap, investigated more precisely, the effect of the load eccentricity (O’Loughlin et al., 2006), and of the weight and thickness of the anchor (Song et al., 2009). The results emphasized the dominant effects of the load eccentricity, the anchor thickness and of the submerged weight of the anchor. During keying, the applied pullout load, F, results in a vertical component Fv, which is resisted by the submerged weight W′ of the anchor and the vertical component of any soil resistance. The soil resistance rises with the anchor thickness t (as more bearing capacity is provided). The pullout load F also creates a moment, M, about the anchor plate, which increases with the load eccentricity e (Fig.3). The higher that M is relative to Fv, the higher is the rotation of the anchor in respect to its vertical translation and lower is the loss of embedment during keying. These different elements can be expressed in normalized ratios, leading to the following expression to assess the loss of embedment during keying (Song et al., 2009).
Fig. 2. Intended mechanism of the keying flap. As the anchor is moving upwards during keying, the shearing forces acting along the flap are supposed to create a moment around the hinge of the flap, due to the eccentricity eh of the hinge, resulting in a rotation of the flap away from the shank. Once rotated, the keying flap would provide extra bearing resistance against vertical displacements, consequently reducing the loss of embedment.
Loss of embedment The loss of anchor embedment during keying is a function of multiple parameters, including the anchor eccentricity, e, (defined as the normal distance from the padeye to the plate), the anchor thickness, t, the submerged weight, W′, and the soil properties.
B
Numerous studies, all experimental, have been performed in the past to investigate the respective influence of these parameters and to quantify the loss of embedment. From field tests using full scale and reduced scale rectangular anchors (1:4 and 1:5) featuring a keying flap, Wilde et al. (2001) reported a loss of embedment ranging from 0.5 to 1.7 times the anchor total height.
Fig. 3. Anchor equilibrium during keying (the weight of the shank and the horizontal component are omitted for simplification)
∆z 0.15 = B e t 0.3 M 0 B B ABs u
(1)
0.1
where Mo is the initial moment corresponding to zero net vertical load. For typical SEPLA anchor, with an eccentricity ratio e/B ranging from 0.4 to 1 this expression yields loss of embedment ranging from 0.2 to 0.5 times the anchor height B, which is in very good agreement with all the centrifuge results. Note that if the factor 0.15 is replaced with 0.2, an upper bound and hence conservative prediction of the experimental data is provided. This expression can be further refined by including the load inclination at the padeye, following results by Gaudin et al. (2009), as follows:
∆z 0.15 = B e t 0.3 M 0 B sin β B ABsu
(2)
0.1
π 2β
2
This empirical expression also accounts for (i) the load inclination and hence the reduced vertical component of the pullout force, (ii) the change in effective eccentricity from e to eβ (Fig. 3) and (iii) for the shorter rotation the anchor has to experience to complete keying (from π/2 for vertical loading to β for inclined loading). Comparisons of this expression with centrifuge results from Gaudin et al. (2009) and numerical results from Song et al., (2009) show good agreement as demonstrated in Fig. 4, though it is acknowledged that the load inclination varies during the keying process. All the studies performed so far focused on straight vertical or inclined pullout and ignored the behaviour of the chain during the keying process. As identified by Wang et al. (2010) the anchor keying is likely to include different stages in which the rotation of the plate is going to coincide with a change of load inclination at the padeye due to the motion of the chain through the soil from the recovery of the slack to an inverse catenary shape as the chain is tensioned. The effect of the chain behaviour on the keying of rectangular anchors (without a keying flap) in normally consolidated clays, using large deformation finite element approach has been investigated by Wang et al. (2010). Other parameters such as the anchor aspect ratio and the soil strength profile were varied. The results confirmed the experimental
findings from Gaudin et al. (2009), indicating that a lower load inclination (to the horizontal) results in a lower loss of embedment. The shorter rotation experienced by the anchor, to reach an inclination normal to the pullout load, is a more dominant effect compared to the reduction of the effective eccentricity resulting from the load inclination.
Loss of embedment, ∆ ∆ze/B
1.6
Gaudin et al., 2009 Equation 2 Song et al (2009) - Transparent soil Song et al (2009) - FE analysis - t/B=0.02
1.4 1.2 1.0 0.8
an original design (Fig. 6). Fluke aspect ratios were identical having model lengths of 80 mm and widths of 20 mm (L/B = 4). The shank was not modelled to limit the investigation to the behaviour of the plate, but replaced by 5 mm in diameter cylindrical hollow tubes of varying interchangeable lengths, namely: 8 mm, 12.5 mm, and 20 mm giving eccentricity ratios (e/B) of 0.25, 0.4 and 0.625 respectively. Note that the eccentricity ratio is calculated in respect to the total height of the anchor (plate + flap). Two interchangeable keying flaps were attached to the anchors via robust hinges. Note that particular attention was paid to model correctly the hinge offset, in order to replicate the expected mechanism showed in Fig. 2. Two different flap heights x, namely: 7.5 mm and 10 mm giving flap/plate height (x/h) ratios of 0.375 and 0.5 respectively were also investigated. The addition of the flap also resulted in a negative offset, d, of the loading point compared to plate anchors studied so far, as shown in Fig. 6.
0.6 0.4 0.2 0.0 0.0
0.1
0.2
e B sin β
0.3
0.4
t B
0 .3
0.5
M ABs
0.6
0 u
0 .1
0.7
π 2β
0.8
2
Fig. 4. Loss of embedment during keying for varying pullout load inclination, anchor thickness and anchor weight. All these results provide insights into the keying mechanism allowing the various guidelines to be used more efficiently. The US Naval Civil Engineering Laboratory guidelines (NCEL, 1985) propose that this loss of embedment is twice the anchor height (B) in fine grained soils, whereas Det Norske Veritas guidelines (DNV, 2002) tentatively suggest a range of 0.4 – 0.8 times the anchor height for plate anchors incorporating a keying flap, increasing to 0.8 – 1.6 times the anchor height for plate anchors with no keying flap. These recommendations appear conservative compared to the numerical and experimental results.
Fig. 5. Testing chamber with flock and Perspex face.
Although further refinement can be achieved, the understanding of the behaviour of the keying mechanism is now relatively well established. This paper therefore focuses on investigating, through centrifuge tests, the mechanism associated with the keying flap and its efficiency in potentially reducing the loss of embedment.
CENTRIFUGE MODELING OF ANCHORS WITH KEYING FLAPS Testing apparatus and models The centrifuge tests were carried out using the UWA drum centrifuge (Stewart et al. 1998). It has a diameter of 1.2 m with a channel 0.2 m radial depth and 0.3 m high, for a maximum acceleration of 400 g. To facilitate optical measurement of the plate anchor keying process, tests were conducted in ‘beam centrifuge mode’ using six custom fabricated plane strain testing chambers with internal dimensions 258 mm long, 80 mm wide and 150 mm deep (Fig. 5). The chambers are modular, allowing either side of each chamber to be replaced with a transparent Perspex panel (see Fig. 5). Up to 6 boxes can be fitted inside the channel allowing for consolidation of several samples at the same time. Model anchors were designed based on Aker Marine SEPLA schematics similar to anchor prototypes previously tested in the field. Two 1:100 stainless steel, scale model anchors were built in house from
Fig.6. Model anchors and notations used The plate thickness for the fluke and flap was 3 mm. This is higher than the prototype thickness, but was necessary to ensure a good contact between the anchor and the Perspex window and hence an accurate track of the anchor trajectory for the subsequent PIV analysis. Keying flaps could rotate approximately 20° away from the shank but were restrained in motion toward it as for the prototype anchor. The anchor chain was plaited from three strands of thin stainless steel wire and was of sufficient length to ensure the anchors could be completely removed from the soil sample during testing. Once placed inside the sample box (Fig. 5), both edges of the anchor plate were in contact with the sides of the box resulting in plane strain
behaviour. O-rings were fitted on both sides of the anchor plate to ensure that no clay would squeeze through the interface between the Perspex window and the anchor, compromising the quality of the observations. Conducting the model tests adjacent to the transparent Perspex panel facilitates optical measurement of the displacement and rotation of the plate anchor, which was conducted by a digital camera placed within a custom made cradle that supports the camera lens at high acceleration levels. Of note is the slightly shorter length of the flap compared to the fluke on the non-viewing side, and the absence of the O-ring. This was done to allow the flap to rotate freely rather than contacting the back of the testing chamber. A Canon S50 camera with a 5 Mega Pixel resolution (2592 x 1944 pixels) was used for digital image capture. The camera was set to continuous shooting mode which resulted in a full-resolution capture frequency of about 1 Hz. Remote triggering of the camera was achieved through a miniature actuator fixed to the camera cradle and activated remotely from the drum control room. In order to facilitate conversion from image measurement scale (i.e. pixels) to model measurement scale (i.e. millimetres), a grid of reference markers was embedded into the Perspex panel along the soil-Perspex interface. Synchronisation of the logged data and the captured images was achieved by placing a liquid crystal display timer whereby data appeared on each image and were recorded in the data file along with the load applied during pullout and the displacement of the cable. The general testing arrangement is presented in Fig. 7.
for two-way drainage during consolidation. To infer the soil strength profile, T-bar tests were conducted in a dedicated soil sample which had undergone the same preparation and consolidation as the test samples. A dedicated sample was necessary due to anchor trajectory resulting in a lack of virgin soil after the test. Six T-bar tests were conducted at velocities ranging from 0.1 to 3 mm/s. A velocity of 2 mm/s was deemed necessary to ensure drainage conditions similar to those mobilised during anchor pullout. A soil profile featuring zero strength at the soil surface and a strength gradient a prototype scale of 2.72 kPa/m was measured.
Test setup and testing programme Once consolidation was achieved, the strongboxes were removed from the centrifuge, sealed and stored in the laboratory. Testing was conducted on one sample at a time. For each test, the strongbox face located away from the digital camera was removed and the anchor inserted carefully from the side, ensuring the fluke and flap remained vertical. The chain was installed vertically, even for inclined loading, so the testing sequence would replicate the in situ process. The depth that the anchor was inserted resulted in an embedment ratio of 3.3, ensuring deep behaviour of the anchor as demonstrated by Merifield et al., 2003. The insertion from the open clay face was necessary to avoid the disturbance of the clay in the upper part of the sample, along the trajectory of the pullout. The back face was then replaced and the face adjacent to the digital camera was removed. The O-ring on the anchor side was made visible before flock was sprinkled on the clay surface in order to maximise the contrast for subsequent PIV analysis (White et al. 2003). The clear Perspex viewing pane was then lightly greased on the clay side in an attempt to reduce friction with the anchor during pullout, before being screwed back in place. Distances from the anchor to the edges and the bottom of the sample box were at least two times the anchor height to ensure that no border effects would influence the anchor behaviour. An example of the test setup once the anchor was installed is presented in Fig.8.
Fig. 7. Testing arrangement in the drum centrifuge channel (after White et al. 2005).
Soil preparation and characterization Normally consolidated kaolin clay was used to reconstitute the soil samples. It has been extensively characterized in previous projects undertaken at UWA (Lehane et al. 2009). Samples were prepared by mixing commercially available kaolin clay powder with water to form a slurry at 120% water content (twice the liquid limit) and subsequently de-aired under a vacuum close to 100 kPa. The slurry was subsequently placed in the testing chambers in flight at 20g before self weight consolidation at 100g for approximately 4 days, resulting in a normally consolidated sample. During this time additional slurry was added to the sample to maintain the required sample height of ~120 mm and a 510 mm layer of water was maintained at the sample surface to ensure saturation. Drainage blankets placed at the base of the boxes allowed
Fig. 8 Example of test setup After anchor installation, the supporting beam and the pulley system were installed and aligned in order to achieve the required loading inclination. The sample box was then placed into the centrifuge with the second testing box separated by 180 degrees to balance the centrifuge. The pullout cable was attached to the load cell and to the actuator after which the sample was reconsolidated at an acceleration of 100 g for a day. After the reconsolidation period, the digital camera
was activated and vertical loading commenced at a displacement rate v of 0.1 mm/s, giving a dimensionless velocity of vB/cv higher than 30 (where cv is the coefficient of consolidation of 2 m2/y), thus ensuring undrained behaviour (Randolph et al., 2005) A total of 4 tests were performed examining three different eccentricity ratios, two different flaps to plate height ratios and 2 different load inclinations β. The experimental programme is summarized in Table 1. Note that the 90 degrees inclination corresponds to vertical loading in respect to the attachment point.
Flap – height ratio
Load inclination
x/h 0.5 0.5 0.375 0.375
β (°) 90 90 90 60
CENTRIFUGE RESULTS AND DISCUSSIONS Plate motion during keying and the resulting loss of embedment was determined from Particle Image Velocimetry (PIV) analysis of the images taken during testing (White et al., 2003). Each digital image was divided into ∼8,000 interrogation patches, each covering a zone of soil approximately 1 mm square (Fig. 6). Each of these patches was tracked using a correlation algorithm, to identify the movement of that patch of soil between a pair of images, with a measurement precision of 0.004 mm for the field of view used during these experiments. Close range photogrammetry was then used to provide a rigorous basis for converting from pixel measurements to object space (mm) data. This applies the necessary corrections for image distortion arising from the non-coplanarity of the image and object planes, and non-linearity within the camera due to fish-eye and barrelling.
Fig. 9. Anchor motion at four different pullout stages. Test 1 (e/B=0.625, x/H=0.5, β=90°). 90
Plate inclination to horizontal (°)
Table 1. Experimental programme Test Eccentricity Eccentricity based on fluke relative to fluke width + flap width e/h e/B 1 1 0.625 2 0.625 0.4 3 1 0.625 4 1 0.625
Test 1 - e/B=0.625 Test 2 - e/B=0.4
80 70 60 50 40 30 20 10 0 0
5
10
15
20
25
30
Vertical displacements (mm)
Behaviour of the keying flap
Fig.10. Anchor rotation during vertical pullout for test1 and test 2.
Fig. 9 present images at different stages of the rotation for test 1, featuring an eccentricity ratio e/B of 0.625 and a vertical pull out. It is evident from the observation that (i) the keying flap did not activate (rotate) during the keying process, (ii) the plate experienced some vertical displacements in addition to the rotation and (iii) the flap rotated once the keying was complete, resulting in a plate inclination not normal to the pullout direction. Indeed, the high eccentricity ratio of the anchor in test 1 results in significant rotation of the anchor in respect to the vertical motion as shown in Fig. 10 which is consistent with the observations of O’Loughlin et al. (2006) and Gaudin et al. (2009), but the same behaviour is observed for Test 2 (see Fig. 11), featuring a lower eccentricity ratio and experiencing more vertical displacement in respect to the rotation of the anchor (Fig. 10). It is then evident (Fig. 9) that the mechanism expected to activate the keying flap does not occur, even when the anchor is experiencing significant vertical displacements, i.e. when the shearing forces along the flap are significant.
Fig. 11. Anchor motion at four different pullout stages. Test 2 (e/B=0.4, x/H=0.5, β=90°).
As shown in Fig. 12 the moment τ × s created around the flap hinge by the shearing forces acting along the flap is balanced by the one, R × a created by the bearing pressure at the back of the flap resulting from the rotation of the anchor. As soon as the rotation of the anchor is initiated, the moment, R × a, which can be much larger than the one created by the shearing forces due to both a higher magnitude of the forces and a higher level of arm around the flap hinge, appears to hold the flap in place, preventing any rotation. Therefore, unless the plate experience only vertical translation, i.e. without any rotation, the keying flap will never activate during keying. This point is further developed by the identification of the failure mechanism around the anchor during the keying. Fig. 13 presents the instantaneous velocity field for tests 1, 2 and 5. M h = Ra − τs = 0
F
a
R Mh s
(a) (b) (c) Fig.13. Soil failure mechanism during keying for test 1 (a), test 2(b) and test 4(c) (axes show vertical position relative to soil surface and horizontal position relative to end of strongbox, both in mm). In all three failure mechanisms shown in Fig. 13, the flap is pushing soil in the clockwise direction. The net soil reaction force on the key is therefore in the anticlockwise direction (consistent with the force R in Fig. 12), tending to hold the flap against the stops, rather than allow it to be activated.
Loss of embedment The total loss of embedment has been determined for each test as the embedment at the end of the plate rotation, and is plotted against the
Note that the soil velocities have been normalised by the vertical velocity of the anchor and subsequently scaled by a factor of 10 to improve the appearance of the plot. For test 1 (Fig 13a), the failure mechanism corresponds to a pure ellipsoidal rotational mechanism, with a large diameter equal to the anchor height, typical of fast rotation with limited vertical displacements. Once the mechanism is initiated, the shear resistance and hence the pullout load remains constant until the mechanism develops to a larger one. Accordingly the loading curve exhibits a plateau during this stage (results not presented here). The same mechanism was observed by Gaudin et al. (2009) for an anchor with identical eccentricity ratio, but without keying flap, highlighting the inefficiency of the flap. The failure mechanism of test 2, where more vertical displacements occurred in respect to rotation is presented in Fig.12b. The failure mechanism adopts a different pattern that test 1 with a shearing plane localized at the anchor soil interface on the front side of the anchor and a localised circular shearing plane at the back side of the anchor, But as for test 1, the mechanism is identical to mechanism observed by Gaudin et al (2009) for plate anchor without keying flap and with slightly smaller eccentricity ratio, highlighting again the absence of effect of the flap during the keying process. Note that the load inclination does not affect the behaviour of the keying flap during keying. Similar observations are made for tests 4 featuring a load inclination of 60 degrees to the horizontal. The failure mechanism in test 5, with a high eccentricity ratio is a pure rotational mechanism (Fig 12c), similar to the one observed by Gaudin et al. (2009) for identical load inclination.
0.1
2
coefficient e t M 0 π on Fig. 14, along with B sin β B ABs 2β u Eq. 2 and the data from Gaudin et al. (2009) featuring anchor rectangular anchor without keying flap. 0.3
1.6
Loss of embedment, ∆ ∆ze/B
Fig. 12. Idealised forces acting on the flap hinge during keying.
1.4
Gaudin et al., 2009 Equation 2 Anchor with keying flap
1.2 1.0 0.8 0.6 0.4 0.2 0.0 0.0
0.1
0.2
0.3
e t B sin β B
0.4 0 .3
M ABs
0.5 0 u
0.6 0 .1
π 2 β
0.7
0.8
2
Fig. 14. Loss of embedment of anchors featuring keying flap. It is evident from Fig. 14 that the data fit very well with Eq. 2 and the data from anchors without a keying flap, validating the conclusions made from the analysis of the anchor motion and failure mechanism. As the keying flap does not rotate during keying, the flap does not create any reduction of embedment. Surprisingly, the padeye offset does not seem to influence the loss of embedment either although it results in a higher rotational compoment, compare to a centred padeye. The magnitude of the padeye offset in the
current study may not be high enough to generate any significant effect and more work is required on this aspect.
Post keying behaviour Examination of the images post-keying gave insights of the influence of both the keying flap and the loading point offset on the anchor behaviour and trajectory. Fig. 15a shows the anchor after keying for test 4, featuring a load inclination of 60 degrees. It is evident that (i) the plate is not aligned normal to the pullout direction (60 degrees from the horizontal in this case) and (ii) the keying flap has indeed now rotated (after the keying process). As the anchor starts to experience translation, the soil pressure acts only on the front side of the anchor (as opposed to the pressure acting during rotation, see Fig. 13) pushing the flap backwards, activating it. It appears therefore that although the keying flap does not influence the behaviour of the anchor during keying, it does affect the post keying behaviour. In conjunction with the loading point offset (relative to the full anchor width, B = x+h), it results in an unsymmetrical distribution of the soil pressure in the front face of the anchor leading to an inclination of the anchor to the loading direction (as the offset creates a moment around the centre of the anchor). It also results in an asymmetrical flow round failure mechanism as illustrated in Fig. 15b. This mechanism is different to the one identified by Gaudin et al. (2009) for plate anchors loaded at an angle in their centre which takes the shape of an ellipsoid centred at the top edge of the anchor.
Fig. 16. Reduction of vertical bearing capacity due to the loading point offset Applying load with an offset also results in a change of the anchor trajectory. The incremental displacement vector (using conjugate kinematic axes) is normal to the yield envelope (according to normality, which is appropriate for undrained conditions). This shows that the anchor no longer translates steadily in the direction of the load, V. In all tests the anchor experienced significant horizontal displacement, increasing with the load inclination, which can be quantified using a yield envelope defined in the full (V,H,M) space.
CONCLUSIONS Centrifuge tests have been performed on rectangular plate anchors with a keying flap, modelling typical SEPLAs, in order to investigate the influence of the flap on both the loss of embedment and the post keying anchor holding capacity. (a) (b) Fig. 15. Post-keying condition from test 4: (a) anchor orientation (b) instantaneous failure mechanism (e/B=0.625, x/H=0.5, β=60°). The change of mechanism results in a change of anchor trajectory and most importantly a reduction in the anchor bearing capacity. Due to the friction along the edges of the box, it is not possible to derive the precise anchor capacity from the experimental results. More work is required to identify clearly compatible kinematic mechanism from Fig. 15b and compute an upper bound solution to derive an appropriate bearing capacity factor. However, the simple yield envelope for a plate anchor, provided by O’Neill et al. (2003), provides an indication of the behaviour of the anchor. The yield envelope is presented in Fig. 16. For a vertical (or inclined) pullout aligned through the centre of the anchor, no moment is generated on the anchor and the vertical capacity reaches the maximal theoretical one (V/Vmax=1) corresponding to pure uniaxial loading. If the anchor experiences pure moment loading (V=0), the moment capacity reaches the theoretical maximum one (M/Mmax=1). As the anchor is subjected to an eccentric vertical load, a moment is created which has a magnitude in proportion to the offset d. The point associated with this moment on the yield envelope results in a maximum vertical load V which is lower than the maximal bearing capacity for pure vertical loading. The amplitude of the reduction of the bearing capacity is a function of the offset d and the shape of the yield envelope.
Visual observations and PIV analysis showed that the keying flap does not activate (i.e. rotate relative to the anchor fluke) during the keying process, due to the level of rotation experienced by the anchor in respect to the level of vertical displacement and the soil resistance forces imposed due to the overall failure mechanism of the anchor. Once keying is completed, the flap does rotate. In combination with the padeye offset resulting from the presence of the flap, this leads to an asymmetric failure mechanism. This change in mechanism results in a lower pullout bearing capacity compared to an anchor of same area with a centred padeye. The amplitude of the reduction in holding capacity depends mainly on the offset of the padeye. The keying flap therefore appears to be inefficient during keying and has a detrimental effect on the final holding capacity. It is therefore recommended that keying flaps should not be used. The padeye offset, which should supposedly result in a lower loss of embedment maybe kept, providing that the gain in strength from the reduction of the loss of embedment offsets the reduction in holding capacity resulting from the change in mechanism post-keying. Further work is therefore required to capture the full influence of the padeye offset to (i) understand its influence on the keying behaviour and (ii) quantify the resulting reduction in holding capacity. ACKNOWLEDGEMENTS This work is supported by the Australian Research Council through the
ARC discovery scheme DP0771348. This support is gratefully acknowledged. The experimental study and the PIV analysis have been performed while the second author was a final year student at UWA. The contribution of centrifuge operator Bart Thompson is also gratefully acknowledged. The third author is supported by an Australian Research Council Future Fellowship (FT0991816). REFERENCES Gaudin C., Tham K.H., Ouahsine S. (2009). "Keying of plate anchors in NC clay under inclined loading". J. of Offshore and Polar Eng., IJOPE, Vol. 19, No 2, pp. 1-8. Lehane, B.M., O’Loughlin, C.D., Gaudin, C. & Randolph, M.F. (2009). "Rate effects on penetrometer resistance in kaolin". Géotechnique, Vol. 59, No. 1, 41–52. Merifield, R.S., Lyamin, A.V., Sloan, S.W., Yu, H.S. (2003). "Threedimensional lower bound solutions for stability of plate anchors in clay". J. Geotech. and Geoenvir. Eng., Vol. 129, No 3, pp. 243-253. O’Loughlin C.D., Lowmass A.C., Gaudin C., Randolph M.F. (2006). "Physical modelling to assess keying characteristics of plate anchors" Proc. Of the 6th Int. Conf. on Physical Modelling in Geotechnics, Hong Kong, Vol. 1, pp. 659-666. O’Neill, M.P., Bransby, M.F., Randolph, M.F. (2003) "Drag anchor fluke-soil interaction in clays". Canadian Geotechnical Journal, Vol. 40, No. 1, pp. 78-94. Randolph, M.F. (2004). "Characterisation of soft sediments for offshore applications" Keynote lecture, Proc. of the 2nd Inter. Conf. on Site Characterisation, Porto, Portugal, Vol. 1, pp. 209-231. Rigzone. (2006). "Permanent SEPLA Mooring Unit Installed in Gulf". Rigzone Magazine, www. rigzone.com/news/article.asp?a_id=30962. April 4, 2006. Song, Z., Hu, Y, O’Loughlin, C.D., Randolph, M.F. (2009). "Loss in Anchor Embedment during Plate Anchor Keying in Clay". J. of Geot. And Geoenv. Eng., ASCE, Vol. 135, No 10, pp. 1475-1485. Stewart, D.P. & Randolph, M.F. (1994). "T-bar penetration testing in soft clay". J Geotech. Engr. Div., ASCE, Vol. 120, No 12, pp. 2230-2235. Stewart, D.P., Boyle, R.S. and Randolph, M.F. (1998). "Experience with a new drum centrifuge" Proc. Int. Conf. Centrifuge ’98, Tokyo, Japan, Vol. 1, pp. 35-40. Wang, D., Hu, Y., Randolph, M.F. (2010). "Keying of rectangular plate anchors in normally consolidated clays". J. of Geot. And Geoenv. Eng., ASCE (submitted in Sept. 2009). White, D.J., Take, W.A. & Bolton, M.D. (2003). "Soil deformation measurement using particle image velocimetry (PIV) and photogrammetry". Géotechnique, Vol. 53, No 7, pp. 619-631. White D.J., Randolph M.F. & Thompson B. 2005. "An image-based deformation measurement system for the geotechnical centrifuge". Int. J. Physical Modelling in Geotechnics 5(3):1-12 Wilde, B., Treu, H. and Fulton, T. (2001). "Field testing of Suction Embedded Plate Anchors" Proc. 11th Int. Offshore and Polar Engr. Conf., pp. 544-551.