Suryasentana, S. K. & Lehane, B. M. (2014). Géotechnique 64, No. 3, 186-194 {http:/idx.doi.org/10.1680/geot.13.P026] Numerical derivation of CPT-based p—y curves for piles in sand S. K. SURYASENTANA® and B, M, LEHANE? The formulations for the lateral load—displacement (p-y) springs conventionally used for the analysis, of laterally loaded piles have been based largely on the back-analysis of the performance of small- scale instrumented piles subjected to lateral load. Although such formulations have been employed with much success in industry, their applicability to large-diameter piles, such as those used to support offshore wind turbines, is uncertain and has necessitated further research in this area. Moreover, with the growth in popularity of in-situ cone penetration tests (CPTS), there are demands for a theoretically supported direct method that can enable the derivation of p-y curves from the CPT end resistance (qe). In this paper, a numerical derivation of CPT-based p-y curves applicable to both small- and large-diameter laterally loaded single piles in sand is presented. Three-dimensional finite-element analyses are performed using a non-linear elasto-plastic soil model to predict the response of single piles in sand subjected to lateral loads. The corresponding CPT q. profile is derived using the same soil constitutive model by way of the cavity expansion analogue. An extensive series of computations of the lateral pile response and CPT qe values is then employed to formulate a direct method of constructing p-y curves from CPT qe values. The proposed method is shown to be generally consistent with existing empirical correlations and to provide good predictions in relation to the measurements obtained during lateral load tests on instrumented piles in an independent case study. KEYWORDS: finite-element modelling; in situ testing; piles; sands INTRODUCTION ‘The p-y method is the most commonly employed approach to analyse the behaviour of piles subjected to lateral loads. This method idealises the soil as a series of independent springs distributed along the pile length. The springs are assumed to be approximately independent as the width of a pile is generally much smaller than its length (Vesic, 1961). Each spring has its own load-transfer relationship known as the p-y curve, where p is the soil resistance force per unit pile length and y’ is the local pile deflection ‘The most widely used design method for deriving p-y curves in sand is the API method (API, 2011). This method evolved from the seminal paper of Reese et al. (1974) who compiled results from lateral tests on relatively small piles. ‘The most recent version of the recommendations is based on the interpretation of O'Neill & Murchison (1983). These recommendations adopt a hyperbolic tangent form for the p-y curve and the only input parameter required is the ssand’s peak friction angle ((').The high sensitivity of the formulation to the selected ¢" value is, however, problematic given the difficulties faced by geotechnical practitioners who need to employ an in-situ test such as the standard penetra- tion test (SPT) or cone penetration test (CPT) to assess the sand density and hence friction angle. The difficulties in selecting appropriate stiffness and strength parameters for the prediction of lateral pile response hhas prompted some research into the development of a p-y formulation involving direct input of an in-situ test param- eter. Houlsby & Hitchman (1988) have confirmed through experiments that the CPT end resistance (qe) is affected Manuscript received 7 February 2013; revised manuscript accepted 30 October 2015, Published online ahead of print IS January 2014. Discussion on this paper closes on | August 2014, for further details see pi * Rio Tinto Northparkes, Parkes, NSW, Australia; formerly. The University of Western Australia, Crawley, WA, Australia School of Civil and Resource Engineering, The University of Wester Ausialia, Crawley, WA, Australia 186 significantly by the in-situ horizontal stress (0fo), while studies of Salgado & Randolph (2001), among others, show that numerical predictions for qe value depend on Oia, @’and the sands stiffness characteristics. It is therefore plausible that the response of sand adjacent to a laterally Joaded pile correlates, at least in an approximate way, with the CPT ge value. Lee et al. (2010) explore the dependence on qe of ultimate resistance in a numerical study entailing the evaluation of ge for sands with a variety of relative densities and in-situ stress states. These ge values were then compared with the expression for limiting stress (Py) pro- posed by Broms (1964) to deduce the following relationship between P, and qe values in silica sands (where stresses are expressed in kPa, p’ is the mean effective stress and p, is a reference stress = 100 kPa) P, = 0196042 pp," wo Novello (1999) and Dyson & Randolph (2001) have proposed empirical CPT:based p-y formulations for ealear- cous sand. These formulations, which will be discussed later in the paper, were derived by double integration and difer- entiation of bending moment profiles recorded! in lateral test results on centrifuge model piles and on a 368 mm dia. pile installed in a test pit backfilled with sand. from the Bass Strait, Australia. While these formulations provided a rea- sonable match fo the experimental observations in the re- spective studies, their general applicability to the full range of sand types and pile diameters is unknown, This paper addresses such uncertainty by using the finite-element (FE) method to develop CPi:based py rela tionships for piles in sand. The sand is assumed to be undisturbed by the process of pile installation and is mod elled as a non-linear elasto-plastic material. The interface between the pile and sand is assumed fully rough through- out. The responses of laterally loaded piles with a wide range of diameters in a variety of different sands are predicted in full three-dimensional (3D) FE analyses. The corresponding CPT 4. profile for each sand is derivedNUMERICAL DERIVATION OF CPT-BASED P-¥ CURVES FOR PILES IN SAND 187 numerically using the same soil constitutive model by way of the cavity expansion analogue in a two-limensional (2D) axisymmetric FE analysis. An extensive series of computa- tions predicting pile lateral response and CPT q. values was performed for sands with different strengths and non-linear ss characteristics, These results are then used to sti formulate a direct method of constructing p-y curves from CPT ge values. The proposed method is ‘shown to be consistent with the “power law” empirical relationship of Novello (1999) and to provide good predictions. for the response recorded in lateral tests on field-scale piles. in an independent case study. NUMERICAL APPROACH Constituive soil model A non-linear elasto-plastic constitutive model was em- ployed to model the sand in the analyses, This model, which is referred to as the hardening soil (HS) model, is described by Schanz er al. (1999). The model parameters are intro- duced in the following and their full descriptions, as well as detailed explanations of the model, including verification are available online (see hitp:/wwwplaxis.n) (a) Stiftness varies with stress level according to a power law, ‘with an exponent (m) of 0-5 assumed for sand. (). Plastic straining due to shear loading occurs such that the Young's modulus (Z) in a drained triaxial testis given by the following hyperbolic relationship festeonf) we (2) ® where ¢ isthe deviator stress, gy isthe maximum deviator stress, jo isthe reference confining pressure, Reis the failure ratio (set to 0-9 for this study) and E%! is the E value at pr (which can be determined when 4 is 50% of 4) (6) Plastic straining due to primary compression is controlled by the one-dimensional (1D), stress-normalised, ocdo- metric stiffness (E%,), which is usually similar in itude to ES, (see hup:/www pais. loading and unloading within the current yield surface is defined by a separate, stress-normalised, modulus (E"), which is typically about thre times ES! (see inp /wwwplaxis). (e) Failure occurs in accordance with the Mobr—Coulomb failure criterion, where the critical state frietion angle (ey) is related tothe peak dilation angle (y) and feietion angle (@") as follows sin T= sing’ sin y (@) aay 8) sings = (/) The ratio of the plastic volumetric strain rate to plas shear strain rate is sin yp, Where yin is the mobilised dilation angle and can be obtained from equation (3) by setting @'to the mobilised friction angle (g). Plastic shear strains are derived by assuming that plastic volumetric changes are zero, (g) Dilation ceases when the void ratio increases from its initial value (@jje) 10 @ nominated maximum void ratio (Cou). This study assumes void ratio limits which are typical of sub-angular to sub-rounded uniformly graded sand (nay = 0-78, ein = 0-49), Prediction of CPT 4, Yu & Mitchell (1998) show that experimental g- data can be predicted numerically to a good accuracy using cavity expansion theory. To simulate the penetration process, a spherical cavity expansion was modelled in the Plaxis 2D (version 2012) FE program, following similar procedures to those described in Xu & Lehane (2008). The limiting cavity pressure (Pn) Was measured and the value of qo then derived using the following relationship proposed by Ran- dolph et al. (1994) 1. = Pg + tang" tan 60°) @ A 2D axisymmetric FE mesh comprising 15-node triangu- lar elements, similar to that described in Tolooiyan & Gavin (2011), was set up as shown in Fig. 1. Vertical and horizon- tal fixities were applied to the base and horizontal fixity was applied to the left and right boundaries. The mesh is 10 m wide and 21m deep, and comprised a Im thick linear elastic dummy layer, a 20m thick weightless soil layer and a weightless linear elastic cavity cluster with a radius (ap) of (0-1 m. Spherical cavity expansion was simulated by applying positive volumetric strain to the cavity cluster, and the cavity pressure-radial displacement variation during expansion was deduced by averaging the output from selected displacement nodes and stress points around the cavity cluster. Cavity expansion analyses at different depths were performed by varying the unit weight of the dummy layer to generate the initial stress conditions in the soil layer corresponding to the desired depth. This technique minimises the influence of boundary effects when simulating cavity expansion analyses. at shallow depths, To verify that the cavity expansion mesh has been set up correctly, the closed-form solutions of Yu & Houlsby (1991) were compared with the Plaxis 2D predictions for a Mohr-Coulomb soil model. For this calibration exercise three spherical cavity expansions (MC1, MC2, MC3) were paisymmetic our layer — e Sol ayer — E g Cay custer (a= om) Tom 1. Spherical cavity expansion mesh (Plaxis 2D)8s SURYASENTANA AND LEHANE performed using the parameters listed in Table 1. Fig. 2 shows that the Plaxis 2D predictions are in good agreement ‘with the closed-form solutions. The differences between the closed-form and Plaxis 2D predictions for the limiting cavity pressure (7yiq) are minimal, with the greatest differ- cence being only 1-3%, as shown in Table 2. To verify the validity of using cavity expansion to predict CPT qc, the ge profile predicted from the FE analyses using the HS soil model was compared to CPT qe profiles meas- ured at a sand test-bed site located at Blessington, County Wicklow (Tolooiyan & Gavin, 2011). A calibration proce- dure was undertaken using Plaxis 2D's SoilTest facility 10 ‘Table 1. Material parameters for FE cavity expansion verification po:kPa | &:MPa| v | c’ | 9°: degrees | y: degrees mer | 120 | 30 [oz[o2] 40 ° mez | 120 | 10 [02/02] 42 2 9 | so |o2zlo2| 40 ° 1500 — Praxis 20 an + Yu Hoult (1991) sco ° 10000 pea — Praxis 20 == Yu Houlsby (1991) © ‘000 2800 2000 & 1500 S — Pisis20 1000 == YA Hoult (1901) 00 i 2 3 7 3 3 aa, @ Fig.2. Comparison of Plaxis 2D results for (a) MCL, (b) MC2 and (6) MC3 with closed form solutions (Yu & Houlsby, 1991) ‘Table 2. Summary of FE cavity expansion verification results, pv: kPa] py (closed-form) | ps (Plaxis 2D) | tp: — pry % mei | 120 3006 040 me2| 120 S908 10 cs | 90 2602 13 find the optimised HS soil model parameters that yield the best fit between the model predictions and the laboratory oedometer and triaxial compression test results. The opti- mised parameters obtained from this calibration procedure are listed in Table 3, which differ slightly from those described in Tolooiyan & Gavin (2011). A comparison of the FE model results using the optimised parameters and those described in Tolooiyan & Gavin (2011) can be seen in Fig. 3. Using these optimised parameters, spherical cavity expansions were simulated to obtain the predicted CPT q. profile at Blessington, This predicted profile is compared ‘with measured q. profiles in Fig. 4, where it can be seen to provide a reasonable estimate (but slight underestimate) of the measured CPT ge traces. The greater underestimation of the ge predictions at shallow depths is potentially due to an underestimate of the peak fiction angle, which Doherty et al. (2012) estimate could be as high as $4° at 1m depth. Prediction of lateral pile response ‘The prediction of lateral pile response was obtained using the Plaxis 3D Foundation (version 22) FE program. A 3D finite-element mesh comprising 15-node wedge clements was set up, as shown in Fig, 5. Vertical and horizontal fixties ‘were applied to the side boundaries. The mesh is 100m x 100m wide and 79 m deep, and comprises a linear clastic, solid pile fully embedded at the centre of the mesh under free head condition. No installation effects were considered (ie. the pile was “wished into place’) and hence the results can be considered more applicable to bored piles than to driven piles (although litle distinction with respect to pile type is made in practice). The lateral soil resistance (p) is calculated as the first derivative of the pile shear force (Q) with respect to depth (2). The py variation during the load application was deduced by repeating the calculation of the soil resistance (p) at each load inerement To verity the ability of the Plaxis 3D model to predict the ‘Table 3. Material parameters for FE qe predictions verification Property Vale Material type Hardening soil Drainage type Drained KPa 97000 kPa 25000 Eo kPa 200000 Your: KN 20 degrees a4 e 0 ve degrees 66 , 02 m os R 09 Tensile strong o 0373 oan 0733 100NUMERICAL ow oo g 50 5 00 a0 eoared : = Care a == Tolooiyan & Gavin (2011) obo bao one baad as sean a Pa § 0 a0 wo] esi ated alaoiyan & Gavin (2011) ‘0008 01 Ail strain © 0018 ‘0020 Fig. 3, C and rarison of FE calibration results of (a) oedometer test ial compression test with laboratory measurements ae MPa 20 © Praxis 20 cen cpt2 opts opts Depth: m Fig. 4. Comparison of Plaxis 2D predictions with feld measure lateral response of a pile using the HS constitutive model, calculations were performed using the mesh shown on Fig. $ near dimensions reduced by a factor of 2) for the land field tests reported by Reese er al. (1974). soil model parameters used for this verification are listed in Table 4. The E%} (kPa) profile is DERIVATION OF CPE-BASED P- ¥ CURVES FOR PILES IN SAND 189 109, Volume pte in NA ARK] Y Ky YU x NV - Uf Uf! a i Yf NS Uf Y ff Uf ]] ff YL Mf S A SN ) Mf Fig. 8. Lateral pile test mesh (Plaxis 3D foundation) ‘Table 4. Material parameters for FE lateral pile response predietions ve Property Vahue Soil parameters Material ype Hardening soil Drainage spe Drained gf: ka 25 00 (2 = depth in meres) eS Ea ey 3B Yui Nim? aos oF degrees 9 ‘ ° ‘degrees ° : 033 Phreatic level ° rameters ope Massive circular Mater type Lincar elastic Drainage type Nomporous Epson KPa 24 y 2s Dim oot im 21 assumed to be 25000: (where = is the depth in metres) which is equal to the equivalent linear soil stiffness profile proposed by Dodds & Martin (2007) for the Mustang Island test location. The variation of soil stiffness with depth was modelled in the FE analysis by dividing the soil pro! into 22 uniform sand layers, each 1 ick. Owing to limited reported data on the sand. the friction angle was simply assumed to be the peak angie with a dilation angle of zero The predicted lateral pile load-displacement curve ob- tained in the FE analysis is shown in Fig. 6, where itis seen to be a close match to the measured response, Fig. 6 also shows that the pile bending moments at a selected lateral Toad level are very similar to those measured by Reese er al (1974). The agreement evident in Fig. 6 confirms that the ‘numerical approach employed is suitable for the analysis of laterally loaded piles.190 00 250 % 200 2 iso z 3 100 — Pax 20 2 Reese otal. (1974) ° — —_ ow sa Deflection at groundline: mm @ fending moment: kN m =100 0 100200 400500 Depth below groundine: m | — Plas 30 © Roose ota. (1974) SURYASENTANA AND LEHANE where D is the pile diameter p is the calculated lateral resistance (per metre run) at depth, =, and displacement, qe_is the computed cone resistance and oj is the vertical elfective stress at depth, = {A total of 110 FE analyses were carried out, comprising 100 lateral pile test simulations and ten cavity expansion simulations, The test cases investigated are shown in Table 5, which comprises ten different sets of soil parameters and ten different pile diameters. All analyses assumed a dry sand condition (with soil unit weight, 7), a linear elastic (solid) pile and fully rough conditions at the pile-sand interface. ‘The rather high upper bound, ¢., employed of 40° could represent that exhibited by angular carbonate sands (Lehane et aly 2012). The parameter range in Table 5 is generally consistent with the proposals of Al-Defae er al. (2013) for HS model parameters applicable to sands. "The parameters listed in Table 6 were held constant to restrict the amount of computing required to a reasonable level. It is noted that while E, and ES! wore varied throughout the study, the analyses assumed ‘fixed (but typi cal) ratios between these parameters and the selected ES! value. The adopted E,/Es) ratio of unity is supported by validation exercises performed by Brinkgreve er al. (2010), although Al-Defae et al. (2013) propose a ratio of 1-25. Limitations of the HS model itself, as well as the need to hold certain parameters constant (including )), are recog- nised when assessing the precision of the relationships developed in the following. It is also noted that, as sand conditions were assumed dry, the term 72 is used in the following to represent 0; in equation (3) Post-processing of the lateral pile loading results involved double differentiation of the computed pile bending moments to derive the net pile lateral forces (p). The collated results ‘Table 6. Constant parameters for all test eases » Bee EEE Property Value Soil parameters Ke 04s * ® : i Fig. 6. Comparison of FE predictions of (a) pile displacements © ee and (by bending moments with the field measurements of Reese et 7 Ka at (1974) fou A FINITE-ELEMENT ANALYSES Fe a Dimensionless analysis indicated that the following rela. SESS tionship should be investigated for the parametric study of Sore Gare eane Pile parameters PL (He ¥ = 5) ne 24 kN? ao 1(sb°a) o # ° ‘Table 5, Test cases investigated in the numerical study ei Dd. Eg: MPa @'idegrees | yrdegrees | pi:degrees | Pile diameters: m os 097 100 Bo 15 2 5 05 097 100 506 15 40 8 os 097 0 39 15 2 5 05 097 0 506 5 40 5 oo | 06 60 a7 a | 3 5 06 062 0 486 Bb | 40 5 07 028 0 361 2 5 o7 028 0 37 5 0 5 07 028 20 361 5 2 3 07 028 20 87 5 | 4 3NUMERICAL DERIVATION OF CPT-BASED P-Y CURVES FOR PILES IN SAND 191 comprised p/(yD) values for 2/D values between 0-4 and 4, and y/D- values between O01 and OL in soils with 4,/(0/2) tatios between 38 and 400. It is noted that excluding clfort involved in post-procs ch lateral pile loading analysis required 8 h of computation time. Power law relationship The collated FE results were used initially to deduce the functional relationship between p/(y=D) and q./('2). ¥/D. ’D employing the power law format, similar 1 that used in {he empirical CPT-based py relationships of Novello (1999) and Dyson & Randolph (2001). The functional form of the power law relationship is as follows OO) ® where ba unknown pranters leesqats msi dermis the anno panes tau ds ied he flowing beset atop yo" 0 ‘The p/(y=D) values calculated using equation (7), for siven ./(/2) and v/D ratios, are compared in Fig. 7 with p/GyzD) salues computed at the same ratios in the FE analyses. This comparison shows a reasonable fit to many of the FE computations, although it is clear that certain esti ates of p/(y/=D) are overpredicted by nearly a factor of 2 ‘These overpredictions correspond primarily with FE results at large y/D values. Given that equation (7) is based on the same power law a PratctedPiyz0 10} oo 70 a0 Measured Pty20 Fig. 7. Comparison between FE results and predictions obtained using equation (7) ‘Table 7. CPTbased p-y relationships framework as the existing CPT-based p-y selationships, a comparison between these relationships was carried out to identify any significant differences. To allow for a direct comparison, each relationship was rearranged into its com- ponent form, as shown in Table 7. The comparison. shows that equation (7) is in good agreement with the Novello (1999) relationship, which was derived largely on an intui- tive basis from the back-analyses of centrfuge-scale lateral load tests in calcareous sand. However, the empirical rela- tionship proposed by Dyson & Randolph (2001), which was also assessed from centrifuge model pile tests in calcareous sand, prediets that p is independent of depth (2) and has a larger diameter dependence than the FE predictions: these differences can have a very significant impact on evaluated poy curves Exponential relationship ‘One major shortcoming of the power law relationship is its inability to model the limiting nature of the p-y curve. This limitation explains the tendency seen in Fig. 7 for overpredictions at large displacements, but can be overcome ‘easily with an exponential framework. The functional form Of the exponential relationship is as follows eraunE ® D) is the normalised ultimate soil resistance and 2 isa Soil stiffness decay coefficient, py/(y=D) should be independent of displacement and therefore a function only of g./(72) and =/D. A review of the collated FE results indicated that the following power law format is a suitable functional form for both the normalised ultimate soil resis- tance and soil stiffness decay funetions 0) (ao) where Ry, Ro. ay. bi. a3, ba, €2 are unknown parameters, To determine the unknown parameters (Ri. Ro, di, Dis & tn, c2) oowlioear rogation malyaea were cared sat tng, the least-squares method. These yielded the following best- At relaionsips ap «ay Relay Eon rig form) Eon emponet fom) Equation) 2(%)""(8) pa aausteroros™ore Novello (1999) ‘ws *eDy* Dyson & Randolph (2001)* oy anDy" p= 135(q)” * R factor for preinstalled piles,192 SURYASENTANA AND LEHANE From equation (8), the complet then be determined as 7 “af Figure 8 compares the FE results with those calculated using equation (13). Evidently, a good fit is obtained for all cases. A comparison of Fig. 7 with Fig. 8 provides an indication of the superior format of equation (13) over equation (7), sxponential relationship can 3) FIELD TEST VALIDATION To assess the general applicability of Equation (13), the predicted lateral pile response obtained using p-» curves calculated using this equation was compared against the measurements obtained in an independent field study car- ried out in Hampton, Virginia (Pando et al., 2006). This field study included instrumented lateral tests in-medium dense sand (with a water table at 81m depth) on three 17 m Jong piles comprising a prestressed concrete (PC) pile with side width of 061m, a plastic pile (PP) with D=0:59 m and a fibre-reinforeed polymer (FRP) pile with 062 m. Fig. 9a) shows the q. profiles at the location of the FRP and PC piles; no CPT data are presented at the location of the PP pile but, based on their respective locations at the test site and inferences from assessed soil stiffness parameters given in Pando et al. (2006), it is assumed that ground conditions at the locations of the PC and PP piles were very similar. The reported. moment dependence of the flexural rigidity (EI) of each pile is presented on Fig. 9(b) The p-y curves were derived using equation (13) at depth intervals of 03m along the test piles and then input into the Oasys ALP program (Oasys, 2010) to predict the lateral pile response. The square PC pile (of ‘width 0-61 m) was represented approximately as a pile with the same diameter. Similar to many other commer- cially available laterally loaded pile programs, the ALP program combines elastic beam elements and” non-linear p-¥ springs to model lateral pile-soil interaction. The predictions needed to be performed in an iterative manner to ensure that the computed bending moments were consistent with the flexural rigidity characteristics. shown in Fig. 9(b). Additional predictions for comparative pur- poses were also performed using p-» formulations recom- mended by API (2011) for the friction angles assessed by oo so 40 0 60 70 60 Measured 2D Fig. 8. Comparison between FE results and predictions obtained ~~ PC ple testiocaion — FRP ple test locaton Depth below original groundine: @) FRP ple — Pe ple Plastic ple Flvaral sites, El Mm ‘ooo 800 Moment: kN m ) ‘20 400 “000 ¥200 Fig. 9. (a) CPT q, a test pile locations; (b) flexural characteristics of test piles (Pando et al, 2006) Pando et al. (2006), namely and 9’ = 35° below 10m, Figure 10 shows the comparison of the measured and predicted pile load-displacement curves at the original ground level. These predictions have been adjusted for the initial ‘seating load” of 11 KN applied in the field tests. It is clear from Fig. 11 that predictions based on equation (13) provide a good match to the observed load displace- ment response. Fig. 10 also shows that API (2011) pre- dicts a stiffer Toad—displacement response than that measured: this is especially significant for the PC and PP piles. Comparisons between the measured lateral pile deflection profiles and those predicted using the p-y curves derived using equation (13) are presented in Fig. 11 for three separate load levels, Eneouragingly, it is evident that the calculated and measured profiles are very similar for all eases (apart from the under-prediction of displacements for the PC pile). This agreement provides additional corrobora- tion in support of the general form of equation (13) and the use of a direct CPT g. approach for the analysis of laterally loaded piles in sand. Equation (13) has also been shown by the present authors (in a submission forthe 3rd international symposium on cone penetration testing, which is being held Las Vegas in May 2014, titled ‘Verification of numerically 33° in the upper 10mNUMERICAL DERIVATION OF CPT-BASED P-¥ CURVES FOR PILES IN SAND 193 ae a = Equation (19) Measured 2 Detection at origina grounding: mm @ — Equation (19) + APL Measured racer eee eer eer eee) Deflection at origina grounding: mm o 6 — Equation (13) - APL Meosured 2 40 SCO Deflection at original grounding: mm o Fig. 10. Predicted and measured ipton, Virginia, US) hhead load-displacement (a) FRP pile; (b) PC piles Lateral ple detection: mm 12.0 10.20 20 40 50 60 70 60 Lateral pile dotocton: mm 10.0 1020 30 40 80 60 70 &D derived CPT-based p-y curves for piles in sand”) t0 lead 10) ‘good predictions of other lateral pile load test data including, tests in fully submerged conditions CONCLUSIONS ‘The difficulties in selecting appropriate stiffness and strength parameters for the prediction of lateral pile response prompted a numerical investigation of a CPT-based p-y formulation for the analysis of laterally loaded piles in sand. ‘The FE computations performed to predict the ge profiles and the p-y curves adjacent to laterally loaded piles (with diameters ranging from 0:5 m to 5m) used the non-linear, clasto-plastie HS constitutive soil model. The computations: involved q- and py predictions in soils with a wide range of stiffnesses and strengths (as listed in Table 5), but the parameters listed in Table 6 were held constant so. that ‘overall computation time required for the study was realistic, Regression analyses of the extensive series of computa tions, assuming a power law format, uncovered an expres- sion (equation (7)) which is in good agreement with the ‘empirical expression derived by Novello (1999) using experi- ‘mental data for laterally loaded piles in calcareous sand. An improved exponential format for p-y curves is proposed (equation (13), which is shown to provide a better match 10 ‘the computations as well as incorporating the tendeney for lateral pressures to reach a limiting or ultimate value. The applicability of equation (13) to laterally load tests in the field is verified by the good predictive match obtained with the measured response of three piles in Hampton sand. Equation (13) provides important insights into the relative effects of various parameters controlling py curves in sand. Withstanding the simplifications involved in its derivation, the paper provides strong support for a changeover to a CPT. ge-based p-y formulation for sands. It is hoped that addi- tional calibration against more case history data will assist With the refinement of equation (1 NOTATION 42 current radius of spherical cavity ‘4 initial radius of spherical cavity cohesion D_ pile diameter D, relative density E_ drained secant Young’s modulus Lateral pile dolection: mm 40.0 10 20 30 40 50 60 70 80 a 2 | ‘ fe, Measured Measures Measures : (eet = 516589 (eeeita=si2m — § 8 tera = 481889 if ~ Meares Nosoures 7 Mauss 3 (latraticad = 1448KN) 3 8} (ateraltoad = 1442KN) {lateral load = 140-1 kN) Big] | - Monee Biol f- Momues Bo = Nesauee : (essed - 230-19 i (tire - 2050) (las — 20 8) fwtnee stom 2] (emibus” 5120 (reo asamp Eu Euston 13) Suj | —Emotoncy Hi — Euan (1 foots” weeny SM) feat” a2 & {ond ose 40-140) ws} | — Eamon i 6 — Exanton 13) 6 — Exuomen (2) (ertesa” 20-109 aeeree? z00 Gere” 20e0y 8 a wei 2 i 2» 2 ° Fig. 1. Predicted and measured displacement profiles at Hampton, Virginia: (a) FRP pile; (b) PC pile: (c) plastic pile198 SURYASENTANA AND LEHANE EGE value detrmined when g = 50% of 9¢ at ps EE, oedometer loading modulus at Per EG unloading modulus a pr nial void ratio ea marin vod eatio qo minimum void rato TK carh pressure cootcien pile length ‘m Exponent controlling stress level dependence of siffness 1p lateral soil resistance per unit ile length Pi caxity expansion limit pressute reference confining pressire deviator tess rmasimum deviator stress Steady-state penetration resistance faire ratio local pile displacement depth below ground surface Sil unit weight te ile nit weight "Poisson ratio ‘oly insite horizontal stress ‘8, erica elective stess (at depth, (peak friction angle Constant volume fiction angle dilation angle REFERENCES, AbDefae, A. H., Caucis, K. & Knappett, J. A. (2013). Aftershocks and the wholesife seismic performance of granular slopes. Géorechnique 63, No. 14, 1230-1244, hp: /ds.doi.ong/10.1680) weot 12 P149. API (American Petroleum Institute) 2011), Geotechnical and foun- dation design considerations, ANSUAPL RP 2GEO, Ist eda, ‘Washington DC, USA: API Publishing Services. Broms, B. B. (1964). Lateral resistance of piles in cohesionless Soils. Soil Mech. Found. Div: 90, No. 3, 123-158. Brinkgreve, R. B. J, Engin, E. & Engin, H. K_ (2010). Validation ‘of empirical formulas to derive model parameters for sand. In ‘Numerical methods in geotechnical engineering (eds T. Benz land S, Nordal, pp. 137=142. Rotterdam, the Netherlands: CRC Press Balkema, Doherty, P, Kirwan, L., Gavin, K., Igoe, D,, Tyrrell, S., Ward, D. & ‘O'Kelly, B. (2012). Soil properties at’the UCD geotechnical research site at Blessington. Proceedings of the bridge and conerete research in Ireland conference, Dublin. Dodds, A. & Martin, G. (2007), Modeling pile behavior in large pile groups under lateral loading. Buffalo, NY, USA: MCEER "University at Buffalo, Dyson, G. 1 & Randolph, M. F. (2001). Monotonic lateral loading ‘of ples in calcareous Sand. J Geotech. Geoemiron Engng 127, No. 4, 346-382 Houlsby, G. T. & Hitchman, R. (1988). Calibration chamber tests of| ‘a cone penetrometer in sand. Géotechnique 38, No, 1, 39-44, hp: /dx do. org/10,1680/geot.1988, 38.139. Lee, J, Kim, M. & Kyung, D. (2010), Estimation of lateral load ‘capacity of rigid “shor piles in sands using CPT results, J Geotech. Geoemiron Engng. ASCE 136, No. |, 48-56. Lohane, B. M., Schneider, JA., Lim, 1K, & Morara, G. (2012). Shaft ftietion fiom instrumented displacement piles in an un- cemented calcareous sand. J. Geotech. Geoemiron Engng 138, No. Il, 1387—1368, Novello, E. A. (1999), From static to eyclic p-y data in calcareous sediments. In Engineering for calcareous sediments. (ed K. AlShafei), vol. I, pp. 17-24. Rotterdam, the Netherlands Balkema, O'Neill, M. W. & Murchison, J. M. (1983). 4 evaluation of p-» relationships in sands, report to American Petroleum Institute, Houston, TX, USA: University of Houston. Oasys (2010). Alp software manual. Oasys, London, UK. Pando, M. A., Ealy, C. D, Filz, G. M. E, Lesko, J. J. & Hoppe, E. J, (2006). 4 laboratory and field study: of composite piles for bridge substructures, report to Federal Highway Administration, ‘Charlottesville, VA, USA: Virginia Transportation Research Council Randolph, M. F, Dolwin, 1. & Beck, R. (1994). Design of driven piles in sand. Géotechnique 44, No. 3, 427-448, hap ‘dx.do.org/10,1680/goot 1994.44.3.427, Reese, LC, Cox, W. R. & Koop, FD. (1974), Analysis of laterally ioaded’ piles in sand. Proceedings of the offshore technology conference, Houston, TX, paper OTC 2080, Salgado, R-& Randolph, M. F. (2001). Analysis of cavity expan- sion in Sand. Int. J. Geomeck, 1, No, 2, 175-192. Schanz, T, Vermeer, P. A. & Bonnier, PG. (1999), The hardening Soil model: formulation and verification. In Beyond 2000 in ‘computational geotechnics: en years of Plaxis Inernational (ed, R_B. J. Brinkgreve), pp. 281-296. Rotterdam, the Netherlands Balkema, Tolooiyan, A. & Gavin, K. 2011). Modelling the cone penetration test in sind using cavity expansion and arbitrary Lagrangian Eulerian finite element methods. Comput. Geotech. 38, No. 1, 482-490, ‘Vesic, A. S. (1961). Beams on elastic subgrade and the Winkler hypothesis, In Proceedings of the Sth intemational conference ‘on soil mechanics and foundation engineering, Vol. 1, pp. $45. 4850. Paris, France: Duo, Xu, X, & Lehane, B, M. (2008), Pile and penetrometer end bearing Tesistance in two layered soil. Géotechnique $8, No. 3, 187— 197, htp:/dx do. org/10.1680/geot.2008.58.3.187, Yu, Ho S..& Houlsby, G. T. (1991), Finite cavity expansion in “ilatant soils: loading analysis, Géotechnique 41, No, 2, 173— 183, htp:/dx.do.ong/10,1680/geot.199141.2.173, Yu, H.'S. & Mitchell, 1. K. (1998). Analysis of cone resistance: review of methods. J’ Geotech. Geoensiron Engng 124, No. 2, 140-149,