INTERNATIONAL JOURNAL FOR NUMERICAL AND ANALYTICAL METHODS IN GEOMECHANICS, VOL.
18, 25-47 (1994)
FINITE ELEMENT MODELLING OF ROCK-SOCKETED PILES E. C. LEONG School of Civil and Structural Engineering, Nanyang Technological University, Singapore 2263 M. F. RANDOLPH Geomechanics Group, The University of Western Australia, Nedlands. Western Australia 6009, Australia SUMMARY Rock socketed piles have a number of features which differentiate them from other types of piles. The generally stubby geometry leads to more even distribution of capacity between shaft and base. However, the low ratio of pile modulus to rock modulus leads to high compressibility and this, coupled with a tendency for the load transfer response along the shaft to exhibit strain-softening, gives rise to an overall response where the shaft capacity may befully mobilized, and potentially degraded, before significant mobilization of base load. The paper presents results of finite element analyses of the response of rock-socketed piles, with particular attention to the shaft response with and without intimate base contact. The shaft interface uses a model, developed from principles of tribology, that includes dilation (and strain-hardening) prior to peak shaft friction, followed by strain-softening at larger displacements. The results of the study are shown to be consistent with field measurements, and to capture effects of the absolute pile diameter on the peak shaft friction. It is also shown that intimate base contact mitigates significantly the degree of strain-softening of the shaft response. 1. INTRODUCTION For many piles in soil, the base capacity is a small proportion of the total capacity, and under working load conditions little load is transmitted to the base. For typical factors of safety against failure, deflection of the pile will be small, and relative slip (if any) between the pile and the soil will be confined to a small region near the ground surface. With rock-socketed piles, the situation is rather different owing to the generally lower embedment and stiffness ratios compared with coventional piles. The low embedment ratio leads to the base contributing a greater fraction of the total capacity, resulting in much higher mobilization of the shaft capacity under working conditions. The low stiffness ratio gives greater compressibility, even at moderate embedment ratios, which can lead to significant relative movement between the shaft and the surrounding Element tests of the response along the shaft of rock-socketed piles indicate a strain-softening response.. It is therefore important to consider rather carefully the integrated response of the rock-socketed pile when determining suitable working load levels. Depending on the separate rates of strain-softening of the shaft response and mobilization of the base capacity, the overall response may show a plateau, or even a decrease in load-carrying capacity, at the stage where the shaft capacity becomes fully mobilized, even though substantial reserve capacity may be available at large displacements. rock. I CCC 0363-906 1/94/0 10025-23 0 1994 by J ohn Wiley & Sons, Ltd. Received 11 January 1993 Revised 27 July 1993 26 E. C. LEONG AND M. F. RANDOLPH This paper presents results of finite element studies of the response of rock-socketed piles, All analyses were conducted using a modified version of AFENA.3 Details are included of infinite elements, that were introduced in order to minimize the size of the discretized domain, and of the joint elements used for the interface between the pile and the surrounding rock. The interfacial response is based on a model developed by Leong and Rand~l ph,~ and analyses are presented for rock-socketed piles with no base, and for complete socketed piles. These analyses illustrate the important effect on the shaft response of the stress field resulting from load transmitted at the base of the socket. Parametric studies and comparisons with field data are included, showing the effects of geometry and stiffness on the stress mobilization down the rock-socket. 2. MODELLI NG OF UNBOUNDED DOMAINS The rock-socketed pile problem, typical of many geotechnical problems, involves an unbounded domain. It is a common practice in finite element modelling of these problems to truncate the finite element mesh at a distance deemed far enough so as not to influence the near field solutions. These truncations are usually determined by trial and error until an acceptable solution is obtained. Such a method places a heavy demand on computer resources, both memory and time, as solutions for the far field which are of no interest are generated as well. I n the Iast decade or so, considerable effort has been concentrated on modelling unbounded domains. Several approaches have been used: analytical mapping of the far-field s~l uti on;~ mapping the exterior domain onto an interior finite one;6 boundary integral method^;^. infinite elements;-* continuous elements; infinite boundary element;20 equivalent springs.21 (Sharan, 1992). Of these techniques, the use of infinite elements with finite elements appears to be the most popular. There are basically two methods in the formulation of infinite element^.'^The first method is the direct approach, or the displacement descent method, where the natural co- ordinate is extended to infinity in the required direction while keeping the standard mapping function well defined. The unknown variables are expressed in terms of descent shape functions which decay asymptotically to zero at infinity. The second approach is the inverse method, or the co-ordinate ascent method, where the domain of the natural co-ordinate is maintained as usual while ascent mapping functions are employed to cause the physical co-ordinate to exhibit singular behaviour at infinity. CurnierI4 has shown that the two approaches are equivalent provided that they are consistently applied to a linear isoparametric element. However, in the direct approach, a special quadrature formula is needed to accommodate an infinite domain of integration. The inverse method is favoured by many researchers. 12* l4 l 7 chiefly because (a) the same mapping is used for both geometry transformation and for transforming the unknown function variable from the local to the global co-ordinate system and (b) the usual Gauss-Legendre integration can be used for both the mapped infinite element and finite element. In the present paper, a mapped infinite element based on the inverse method is incorporated into AFENA. For completeness, the mapped infinite e ement is described here. Figure 1 shows local co-ordinates, ( and q are related to the global co-ordinates, x and y , by the singly and doubly infinite elements in two dimensio d s, takkn from Marques and Owen.15 The 2 ( X j - X i ) x - (2Xi - X j ) 2 ( Y j - Y i ) Y - ( 2 Y i - Yj ) l =1- q =1 - which allows [ and q to approach unity as x and y , respectively, approach infinity. The subscripts i and j in equation ( 1) relate to the inner nodes of the infinite element, where node i is at the boundary of the main mesh. The spacing between nodes i and j is determined by a pole node, o, such that x j - x i =xi - x , . The optimum position for the pole node for the FINITE ELEMENT MODELLING OF ROCK-SOCKETED PILES 27 Shape functions NI =$( l - C) ( I- q) ( - l - C- q) N2 =!j(l -C2)(l -9) N3= $(I +C)(l -q)(-l+C -q) N4 =f(1 +C)(l - q2) Ns =f ( l - C) ( 1- q2) (a) Singly infiniteelement 1 2 Mapping hnctions Shape functions (b) Doubly infiniteelement Figure 1, Two-dimensional singly and doubly infinite elements derived from eight-noded Serendipity isoparametric element (Reference 15) rock-socketed pile problem was found to be along the outer edge of the pile, the axis of symmetry and the upper ground surface.22 The performance of the infinite elements was evaluated by Leong for a variety of axisymmetric problems under elastic and elastic-perfectly plastic conditions. Figure 2 shows different mesh arrangements that were used for the rock-socketed pile problem. The resulting elastic flexibilities are shown in Figure 3, from which it may be seen that it is sufficient to limit the discretized zone to 3 4 times the pile radius around the pile, and a similar distance below the pile, bounded by infinite elements (see Figure 2). Although it may be argued that the mapped infinite element may be less accurate when compared to the infinite boundary element or the equivalent springs method,2 its ease of 28 E. C. LEONG AND M. F. RANDOLPH 28R Figure 2. Finite element for study on mesh truncation formulation and implementation into a general finite element code makes it an attractive compromise. To date, publications of finite element solutions of pile problems have mostly been obtained using a truncated mesh. 2.1. Elastic response The elastic response of a pile socketed into rock has been investigated using the finite element method by previous researcher~. ~~- ~~ Such problems have been previously studied by extending the finite element mesh to some distance. For example, Donald et ~ 1 . ~ ~ used a half-width of 20R and a depth of 50R, where R is the radius of the pile. The use of infinite elements can give superior accuracy, with fewer degrees of freedom. Figure 4 shows calculated flexibilities for a range of embedment and stiffness ratios of the pile, allowing for a change in soil modulus at the level of the socket base. I t may be seen that, for the FINITE ELEMENT MODELLING OF ROCK-SOCKETED PILES 0. 8 - 0.7- 0.6 - 0. 5- 0. 4- 0.3- 0.2- 29 0.1- L O U m 1 LI - -B - Mesh 2 . -x- - Mesh 3 - -- - Mesh 4 - Finite +Infinite elements I I I 1 I I I , I 1.0 I I I I 0.9 l Q I -* - Mesh 1 F E J E m Figure 3. Comparison of finite element mesh truncation with use of infinite elements for a rock-socketed pile stiffness ratio of E, / E, =10, where E, is Youngs modulus of the pile and Em is Youngs modulus of the surrounding material, the flexibility coefficient becomes nearly constant for embedment ratios of LID greater than about 10. Also shown for comparison on Figure 4 are flexibility coefficients from Donald et a1.,23 using a rather coarser mesh. As expected, these solutions generally give lower flexibilities, the difference being most obvious for EJ E, =10. 3. J OI NT ELEMENTS TO MODEL PILE-ROCK INTERFACE The load transfer behaviour of a rock-socketed pile through side-shear is dependent on the behaviour of the pile-rock interface. The pile-rock interface usually behaves differently from either the pile or rock material. In finite element modelling, special elements have been used to model the pile-rock interface. Osterberg and Gill24 have used spring-loaded linkage elements. Rowe and Pells26 have used a series of dual nodes based on the compatibility condition at the pile-rock interface. Donald et aLZ3 have modelled the problem by assigning a different set of material properties to the layer of finite elements adjacent to the pile. Another element commonly used to model interface behaviour is the Goodman-type joint element. Donald et aLz3 have also used Goodman-type joint elements to model the pile-rock interface but found the joint elements near the pile base behaved erratically on reaching incipient slip. The extent of the use of special elements to model the pile-rock interface varies. Osterberg and Gill24 have placed their spring-loaded linkage elements along the edge of the pile (side and base), Rowe and Pells26 placed the dual nodes only along the side of the pile while Donald et were not specific, but implied that the joint elements were just down the side of the pile. I t is not clear how the finite element models handled element connectivity for these elements, particularly at the base of the pile. 30 E. C. LEONG AND M. F. RANDOLPH 1 .o 0.9 0.8 ,Q 0.7 $ 0.6 w 0.5 s 0,4 G 0.3 0.2 0. I L. 0 L 3 I U c d - t 5 I 2 3 5 10 Donald et a1 (1980) - Finite +Infinite elements 1 T l Z ' I ' I ' I 2 4 6 8 10 12 14 16 18 20 Embedment ratio L/D (a) E,/E, = 10 1.04 ' I . ' " " ' I " ' I " " ' 0.9- 0.8 - ," 0.7- $ 0.6- 05- 3 0. 4- Lc Donald e t al. (1980) 0 - 03- 2 3 0.2- 5 0.1 - 10 5 L. 0 cd 1 U 8 - Finite +Infinite elements I ! . I ' ~ ' l ' l ' I 3 I ' ' ' 1 0 2 4 6 8 10 12 14 16 18 20 Embedment ratio LI D (b) E,/E, = 1000 Figure 4. Settlement influence factors, I,, for complete piles in two layer systems In the present study, six-noded Goodman-type joint elements were used to model the pile-rock interface behaviour, while eight-noded isoparametric quadrilateral elements were used to model the rock mass. Infinite elements were used to model the unbounded domain. To maintain element connectivity, the joint elements were extended down to the bottom edge of the finite element mesh, including the use of a one-dimensional infinite element to extend the joint into the FINITE ELEMENT MODELLING OF ROCK-SOCKETED PILES 31 unbounded domain. Leong has shown that the joint elements within the continuum have negligible effect on the continuum response. 4. PROPERTIES OF PILE, ROCK MASS AND PILE-ROCK INTERFACE 4.1. Pile The pile is assumed to be elastic with Youngs modulus, E, =35 GPa and Poissons ratio v, =0.2. The diameter of the pile is assumed to be 1 m unless stated otherwise. Effects of embedment ratio, LID, are investigated for values of LID =2, 5 and 10. 4.2. Rock mass There are several strength criteria for intact rocks, e.g. the Mohr-Coulomb criterion, Griffith- type criteria and empirical criteria. However, rock masses generally have naturally occurring discontinuities such as joints, seams, faults and bedding planes. Therefore, the properties of a rock mass may be very different from that of the intact rock. One way of accounting for these discontinuities is by assigning equivalent elastic properties to the rock mass as suggested by Goodman and Duncan.28 For example, for a rock mass with three orthogonal discontinuity sets, the equivalent elastic properties are 1 1 -+-+- G, Siksi Sj ksj (3) for i =x, y, z withj =y, z, x and k =z , x, y; where E, , G, and 0, are the elastic properties, Youngs modulus, shear modulus and Poissons ratio, respectively, of the intact rock; k, and k, are the values of shear and normal stiffness of the discontinuity, respectively, and S is the joint spacing. It is important to estimate the stiffness of the rock mass accurately so that the stresses obtained in any analysis involving dilatancy at the pile-rock interface may be predicted correctly. When high stresses are developed through dilatancy, the intact rock may crack and may show reduction in ~tiffness.~ However, such behaviour is not considered in the present study. The rock mass was assumed to obey a Mohr-Coulomb criterion with a shear strength, c, =200 kPa, a friction angle, 4, =30 and zero dilation. These properties correspond to those reported by Williams3 for sample SM7 of a highly weathered mudstone, which had an unconfined compressive strength, qu, of 750 kPa. Poissons ratio, v,, for the rock mass was assumed to be 0.3. To investigate the effects of rock mass stiffness on the load transfer behaviour of rock-socketed piles, two different values of Youngs modulus, Em =500 MPa and Em =50 MPa, were used. These effectively give relative stiffness ratios, E, / E, =70 and 700, respectively, for a pile with E, =35 GPa. The effective unit weight of the rock mass was assumed to be 23 kN m-3. The initial stress state of the rock mass was assumed to be at KO condition with KO as unity unless stated otherwise. 4.3. Pile-rock interface The pile-rock interface in many respects resembles a rock joint and, as such, arguments presented for rock joints are assumed to be equally valid for the pile-rock interface. The simplest 32 E. C. LEONG AND M. F. RANDOLPH constitutive relation that may be used to model slip at the pile-rock interface is the Mohr-Coulomb model. However, this model does not account for post-peak softening behav- iour, a phenomenon commonly observed in the side-shear response of rock-socketed piles. To model this response, Rowe and Pells26 assumed that slip will occur only when the mobilized shear stress, T, reaches the peak shear strength given by ( 5 ) z =cp +(r, tan 4p where cp is the peak interface adhesion and bp is the peak friction angle. Once slip occurs, some or all of the bond at the pile-rock interface will be lost and the shear stress will reach a residual value given by where the subscript r denotes residual values. If all the bond at the pile-rock interface is lost, c, becomes zero. The degradation from the peak parameters (cp, 4p) to the residual parameters (c,, br) is assumed to be linear with relative displacement. Rowe and co-workers26,31 have presented a detailed parametric study on rock-socketed pile behaviour using this model. Ooi et al.,32 described an approach similar to the one used by Rowe and his coworkers, modified from that of Pease and K ~l hawy. ~~ They assumed that the degradation of c and 4 is related to the plastic shear displacement, up, in the following manner: For up <ur. T =c, +(r, tan 4, (6) c =cp(l - U P / d ) tan 4 =tan 4p - (tan c$p - tan 4,)uP/ur (7) (8) and for up >ur, c =o (9) 4 =4 r (10) where ur is the plastic shear displacement at which the interface reaches the residual condition. Leong and Randolph4 have presented a version of the Mohr-Coulomb model, which allows for gradual mobilization of peak friction and dilation, and also has strain-softening behaviour. The model, shown schematically in Figure 5, has been derived from tribological considerations. Minimal elastic movements (a) occur until the underlying frictional resistance is mobilized, after which the resistance increases to a peak due to dilation and ploughing (b). At the same time, wear takes place along the interface, and particles are gradually removed, leading eventually to strain-softening (c) and finally steady-state sliding (d). The model may be summarized by a failure criterion of (1 1) 1 - tan 4 tan i,b T =0, tan(4 +$) +- I where 4 is taken as 4r, the residual friction angle, the dilation angle, i ,b, is given by tan $ =tan t,hoexp( - nx/S)[I - exp( - i x/S)] ( 1 2) and the ploughing parameter, p. is given by p =poexp( - mx/S)[l - exp( - j x/ S) ] In equations (12) and (13), the parameter S is a length which reflects the wavelength of the interface roughness, while x is the relative displacement along the interface. The parameter j controls the development of ploughing resistance, while the parameters n and m control the degradation of resistance due to wear. Equation (12) also defines the flow rule for the interface. FINITE ELEMENT MODELLING OF ROCK-SOCKETED PILES 33 7 a micro shear displacement b. increase in sliding resistance due to dilation and plough resistance c degradation of surface roughness by wear d steady sliding Sliding Distance Figure 5. Shear stress : sliding distance response for irregular rock surfaces (Reference -4) Figure 6 shows a comparison of this model with constant normal stiffness (CNS) tests from Lam,34 for the case of a confining spring stiffness, K =350 kPamm-'. Parameters for the model were: 4r =23.5". p,, =0.41; m/S =n/ S =05 mm-'; i / S =1000 mm- '. h may be seen that the model allows reasonably close fitting of the responses measured over a wide range of initial normal stress levels. Further examples of comparisons of the model with experimental data have been given by Leong and Rand~l ph.~ In a recent publication by Desai and Ma,35 a disturbed-state concept was described for the modelling ofjoints and interfaces. In this concept, a yield function, a critical state and a disturbed state are defined. Besides the usual elastic stiflness, k, and k, , five parameters are needed to define the yield function, four parameters for the critical state and four parameters for the disturbed state. The Leong and Randolph4 interfacial model uses only six parameters. However, it may be of interest to point out that the disturbance function, D, proposed by Desai and Ma3' is of a similar form to equations (12) and (13), and is given by D =D, [l - exp( - K ~ E ) ] (14) where D, is the ultimate or critical value of D, K and R are material parameters and t D is the trajectory of plastic shear strains. In addition to the failure criterion, it is necessary to assign values to the elastic stiffness parameters, k , and k,, for the pile-rock interface. Goodman et al.,27 have suggested that these parameters may be measured, k, from a direct shear test and k , from a compression test, with each being the initial slope of the respective stress-displacement curve. The incremental stress-strain relationship is given by where C,, and C,, are the cross-stiffnesses of the joint. Under elastic conditions, C,, =k, and C,, =k, , and C,, and C,, are usually assumed to be zero.27 However, when slip occurs, C, , and 34 E. C. LEONG AND M. F. RANDOLPH 596 200 0 10 20 30 Shear displacement (rnm) Figure 6. Comparison of predicted values using Leong and Randolph model with experimental data for K =350kPamm- C,, become non-zero if the joint exhibits dilatancy or strain softening.32, 36 These values are then dependent on the values of k, and k, . Sun et aL3 suggested a way of measuring the stiffness components of a rock joint through a shear compliance test and a normal compliance test. The stiffness terms of equation (15) are then given by an inversion of the compliance matrix obtained from the tests. However, Sun et al. cautioned that the two pairs of compliance so obtained are stress dependent and must be used with care when predicting rock joint behaviour in a general stress path. Interestingly, Leichnitz3* arrived at a constitutive relationship for rock joints by explicitly determining the stiffness terms of equation (15) from shear tests using curve-fitting procedures. Thus, the stiffness matrix deter- mined by Sun et al. is not the elastic stiffness matrix. The normal stiffness of rock joints and discontinuities is a function of the relative stiffness, geometry of asperities, normal stress and joint fill material. The effective modulus, E, for elastic contact of two joint surfaces with no fill material may be given by the well-known Hertzs solution where 1 ( 1 - v:) (1 - v;) E El E2 +- - -- - where Ei and vi are the elastic properties of the materials making up the joint. Usually, the normal stiffness of a rock joint is measured in terms of joint closure. Goodman39 suggested a hyperbolic function linking normal stiffness to joint closure under increasing normal stress. A hyperbolic function between normal stiffness and joint parameters of aperture strength and roughness has also been suggested by Bandis et aL4 Swan41 showed that the normal stiffness can be related to the normal stress by simple relations through assumed distribution functions for the asperity heights. A review on shear stiffness for rock joints can be found in References 40 and 42. Shear stiffness was also found to vary with the relative stiffness, geometry of asperities, normal stress and joint fill material. In summary, it seems likely that the values of elastic joint stiffness, k, and k,, may be functions of stress levels, in a similar manner to their counterparts, the shear modulus and Youngs modulus of a continuum. However, unlike a continuum, quantification of joint stiffness is difficult as it is FINITE ELEMENT MODELLING OF ROCK-SOCKETED PILES 35 dependent on the properties of the joint fill material and also on the physical characteristics of the joint surfaces. In the finite element analyses that follow, k, and k, are assumed to be G/ t and E,/t respectively, where t is assumed to be 1 per cent of the pile diameter, 0, and G and En are the shear and constrained moduli of the pile-rock interface, respectively. Young's modulus of the pile-rock interface is assumed to be 1 per cent of the mass modulus, Em. These values of k, and k, are used in all subsequent analyses reported in the paper. Strictly speaking, k, and k, should be measured elastic parameters as suggested by Goodman et al." In their finite element modelling of the rock socketed pile problem, Donald et aLZ3 used a value of 50 MPam-' for k, in their joint element, reportedly measured from laboratory tests. Based on t =0.01 m(0.010), this would imply a value of 5 MPa of the interface shear modulus, which is only 0.1 per cent of a typical Young's modulus for the rock mass of Em =500 MPa. By contrast, Bandis et aL4' and K ~l hawy ~' , ~~ reported values of joint stiffness; k, and k,, which for t =0.01 m would give Young's modulus for the j6int in the region of 10 per cent of the intact rock modulus, E, . 5. COMPUTED RESPONSE OF ROCK-SOCKETED PILES Construction techniques for rock-sockets can play a significant role in determining the perform- ance, both from the point of view of the roughness of the sides of the shaft (which will affect the pile-rock interface response), and also the tendency for the boring operation to leave a soft layer of debris at the base of the rock-socket, necessitating large displacements to develop any effective end-bearing. The latter aspect may be addressed by specific measures to clean the base of the socket (often involving manual removal of debris). Alternatively, it is quite common for the load-bearing capabilities of the base to be ignored, and the design of the rock-socket is based entirely on the shaft performance. The first series of analyses concentrates on such 'shaft only' rock-socketed piles. In the finite element model, a 'soft' layer of elements is placed below the pile tip. This layer, 0.150-0.250 thick, is given a Young's modulus, Esoft, of 0.1 per cent Young's modulus of the rock mass, Em. Osterberg and Gillz4 used Esoft =Em/3000, while Rowe and PellsZ6 used Esoft =EJ15. The performance of the rock-socketed pile will be compared for two different models of the shaft interface: (a) a simple Mohr-Coulomb model with constant angle of dilation, and (b) the model of Leong and Randolph4 outlined earlier. The parameters for Mohr-Coulomb interface are c' =100 kPa, 4 =0 or 30", $ =0 or 5" , while those for the latter model are (based on parameters deduced from tests on sample SM7 of Williams44) 4, =26", $o =185", p,, =062, rn/S =0.08 mm-l, n/S =0.32 mm-' and [/S =5 mm-'. 5. 1. Load transfer through side-shear only The average shear stress responses of the different models using displacement loading on the pile head are shown in Figure 7. The model of Leong and Randolph4 exhibits the characteristic load transfer curves often observed in socketed piles where the load is taken in side-shear only. The displacement to peak shear stress is about 0.0080, which is a typical rate of mobilization of peak shear stress in piles. This shows that the estimated values of joint stiffness are realistic. The stress profiles at the pile-rock interface for both types of models are shown in Figures 8 (c-4, zero dilation), 9 (c-4, $ =5" ), and 10. As can be observed, the shear stress is relatively uniform over the central part of the pile regardless of the model used. The dilatant Mohr-Coulomb model and the Leong and Randolph model show the normal stress increasing once the relative slip exceeds 2 mm. However, the dilatant Mohr-Coulomb model shows the normal and shear stresses increasing at a constant rate with slip, which is unrealistic in the 36 E. C. LEONG AND M. F. RANDOLPH 4 0 0 / " " " " -. I h m e 2 Y I- l l L/D = 2 ,eong and Randolph model 0 10 20 30 40 Pile head displacement (mm) Figure ?. .4verage shear stress response of side-shear only rock-socketed piles 0 2 0 4 0 6 0 8 - N E 1 0 12 1 4 1 6 2 0 ( a) Shear stress (b) Normal stress Figure 8. Stress profiles at pile-rock interface of a side-shear only rock-socketed pile using c-c#-$ Mohr-Couloinb model physical situation. Interestingly, the Leong and Randolph model shows an increase in normal stress past the peak shear displacement of 0.0080. After the peak shear displacement, the normal stress reduces slightly at the pile head while it increases over the lower part of the piie and finally reaches an equilibrium condition (see Figure 10(b)). However, the shear stress, decreases after the peak shear displacement and reaches a residual shear stress. The limit loads for the non-dilatant Mohr-Coulomb models may be estimated directly from the initial stress conditions. The effect of taking the rock mass Young's modulus, Em, as homogeneous (Em =50 or 500 MPa), or proportional to depth (Em =5002 MPa) is shown in Figures 11 and 12. While the average shear stress response is only slightly affected, there is a much more significant effect on the stresses induced along the interface. This shows clearly that the increase in normal stress, and FINITE ELEMENT MODELLING OF ROCK-SOCKETED PILES 37 1 (kPa) 'Jn (kpa) 600 500 400 300 200 LOO 0 100 200 300 400 500 600 0 0 2 0 4 0 6 + Applied displacement 0 8 - n + m o x X 6mm N o + m n x c 8mm z 1 0 o + m o x I 2 1 4 1 6 I 8 2 0 (a) Shear stress (b) Normal stress Figure 9. Stress profiles at pile-rock interface of a side-shear only rock-socketed pile using c-+$ Mohr-Coulomb model 1 (kPa) 5n ( k P 4 600 500 400 300 200 100 0 100 200 300 400 500 600 0 0 2 0 4 0 6 Applied displacement 0.8 6 N . E 1 0 1 2 1 4 1 6 1 8 2 0 (a) Shear stress (b) Normal stress Figure 10. Stress profiles at pile-rock interface of a side-shear only rock-socketed pile using b o n g and Randolph model hence the peak shaft friction, is strongly affected by the stiffness of the surrounding rock. This point is emphasised further by the response for Em =50 MPa (E,/Em =700) shown in Figure 11, which gives peak shaft friction a factor of 5 lower than for Em =500 MPa. The effect of embedment ratio was investigated using LID =10, compared with LID =2, as shown in Figure 13 for the Leong and Randolph model. The average shear stress response for LID =10 exhibits a similar brittle response to that for LID =2, even though there is a more gradual mobilisation of shear stress. Note that the initial average normal stress at the pile-rock interface is higher by a factor of 5 for the case with LID =10. However, as may be seen from the stress profiles in Figure 14, there is much less dilation before failure, and the peak average shear stress is only about 25 per cent greater than for the case of LID =2. 38 E. C. LEONG AND M. F. RANDOLPH I L/D =2 0 10 20 30 40 50 Pile head displacement (mm) Figure 11. Effect of rock mass Young's modulus on average shear stress response of side-shear only rock-socketed piles 1 (kPa) 1.1 1.6 I B 20 (a) Shear stress Applied displacement X Bmm n 10mm I - X 30mm 0 + a 0 40mm 0 + B Initial stress + m +% a (b) Normal stress Figure 12. Stress profiles at pile-rock interface of a side-shear only rock-socketed pile using Leong and Randolph model, for Em =5002 MPa Laboratory and small-scale field tests show that the pile diameter may have an effect on the load transfer behaviour of pile^.^'.^^Randolph4' showed that if the shear zone thickness at the interface is constant, the additional shear stress, AT (over and above that due to the in situ normal stresses) is given by Aw Az=4Gsi n$tan#- D where G is the shear modulus, $ the dilation angle, #the friction angle, Aw the relative displacement between pile and surrounding material and D the pile diameter. FINITE ELEMENT MODELLlNG OF ROCK-SOCKETED PILES 400 300 - - m n. Y EOO- 100- 39 ' " " " " ' ' ' I ' L/D = 10 L/D =2 , ' I ' I ~ 1 ~ / . Figure 13. Effect of LID ratio c 0 2 4 1 E - N 6 I (kPa) I 500 400 300 200 LOO (a) Shear stress 100 ZOO 300 400 500 600 700 800 Applied dspl acement rl 5mm (b) Normal stress and Figure 14. Stress profiles at pile-rock interface of a side-shear only rock-socketed pile using Leong and Randolph model for L/ D =10 The effect of pile diameter was studied by keeping the ratio L/D at two and increasing the scale of the finite element mesh. A parametric study using pile diameters, D, of 0.5, 1,2, 5 and 10 m was performed. To ensure comparable magnitude of stress, the rock mass was assumed to have an initially uniform isotropic stress of 20 kPa. The effect of pile diameter on the average peak shear stresses, Z, for the various pile diameters is shown in Figure 15. I t may be seen that there is a significant effect of the pile diameter, in keeping with field results such as those discussed by Fahey et aL4* However, in contrast to the simplistic analysis of Randolph4' that indicates an increase in shear stress that is inversely proportional to D (equation (17)), the computed results indicate a variation that is closer to D-0.4. 40 E. C. LEONG AND M. F. RANDOLPH m e s - I 300 - 200 - 100- 0 1 2 3 4 5 6 7 8 9 1 0 D (m) Figure 15. Effect of pile diameter on average peak shear stress of side-shear only rock-socketed piles using Leong and Randolph model The results in Figure 15 were obtained keeping all the rock and interface parameters constant. In reality, it is likely that some of these parameters will have a scale dependency (particularly the parameters +o and po, Reference 22). If this scale dependency is allowed for, the reduction of peak shear stress with increasing pile size will not be as marked as that shown. However, the effect may still be significant, and should be borne in mind in the design of rock-socketed piles, or the interpretation of small-scale pull out tests on grouted bars. 5.3. Load transfer through side-shear and end-bearing Piles which carry load in both side-shear and end-bearing will be termed complete piles in the following discussion. It is a common practice in pile design to superimpose solutions from side-shear and end-bearing to obtain the response of a complete pile. It has been pointed out that such solutions are not valid, as the stress distributions in side-shear only piles and complete piles are different.23. 31 Such indiscriminate superposition in cases with post-peak strain-softening response in side-shear may result in undue conservatism. Finite element analyses were performed for the response of a complete pile with LID =2, using an incremental stress loading. The load responses for the c-4 (non-dilatant) Mohr-Coulomb and the Leong and Randolph models are shown in Figures 16 and 17, respectively. The base loads were obtained by integrating the vertical stresses at the mid-level of the bottom layer of pile elements (0.05D from the pile tip). The error between the side-shear load, obtained from the integration of the side-shear stress, and the difference of the applied load and base load is less than two per cent. The complete pile response is strain-hardening although the side-shear exhibits strain-softening behaviour for the latter model. The average shear stress mobilized along the shaft may be obtained from the bearing stress, q, by 5 =q/(4L/D). Thus, the peak shaft capacity corresponds to an average shear stress of about 235 kPa for the complete pile, compared with about 255 kPa for the shaft-only pile. The stress profiles for the complete pile with the Leong and Randolph model are shown in Figure 18. Comparison with the corresponding profiles for the side-shear only piles, Figure 10, shows that the stress distributions are different as expected. There is an interaction of the FINITE ELEMENT MODELLING OF ROCK-SOCKETED PILES 41 Pi l e head displacement ( mm) Figure 16. Load response of a complete rock-socketed pile using c-&Mohr-Coulomb model Pile head displacement (mm) Figure 17. Load response of a complete rock-socketed pile using Leong and Randolph model end-bearing and side-shear load transfer mode in a complete pile. The effect of the load transferred to the pile base is to limit the increase in normal stress, and hence shear stress, near the pile base. In fact, tensile failure at the pile-rock interface near the pile base is observed. The percentage of applied load that is transferred to the pile base, up to the peak shear stress is about 50 per cent for the Mohr-Coulomb model, and 60 per cent for the Leong and Randolph model. It is encouraging to note that the shear stress profiles are similar to those measured by Williams et a1.,44 reproduced in Figure 19. The above analysis was repeated for 1 m diameter socketed piles with LID of 5 and 10 using the Leong and Randolph interface model. The response for the latter case is shown in Figure 20. Again, the base load was obtained by integrating the vertical stresses at the mid-level of the bottom layer of pile elements, 0. 250 from the pile tip. Hence, the base load is slightly over- estimated, with the difference of the applied load and base load underestimating the side-shear 42 E. C. LEONG AND M. F. RANDOLPH 1 (kPa) O n @Pa) 0 0 2 0 4 0 6 0 I MPa 0 Applied stress, q ,--. N 5 1 0 1 2 1 4 1 6 I 8 2 0 (a) Shear stress (b) Normal stress Figure 18. Stress brofiles at pile-rock interface of a complete rock-socketed pile using Leong and Randolph method I (MPa) Figure 19. Non-uniform development of side resistance from test pile (after Williams et load. The difference, however, is less than six per cent. The load that is transferred to the pile base reduces from 60 per cent for LID of 2, to 38 per cent for LID of 5 and to 20 per cent for LID of 10. Interestingly, the side-shear response in the complete pile has become more plastic (with less strain-softening) compared with the side-shear only socketed pile (see Figure 13). This is due to a difference in the mobilization of the side-shear stress. I n the complete pile, the shear stress is not fully mobilized near the pile base. It is tempting to assume that, in the case where debris leads to a softer base response, some strain-softening may be observed in the overall response. However, analyses conducted with a layer, 0.15D thick, with reduced modulus Esof, =0.1E, do not show any such strain-softening. FINITE ELEMENT MODELLING OF ROCK-SOCKETED PILES 43 * M L * 4 Figure 20. Load response of 20 I8 Total Pile head displacement (rnm) a complete rock-socketed pile with LID =10 using Leong and Randolph model 0 Integrated from side shear 0 5 10 15 20 25 Pi l e head displacement (rnrn) Figure 21. Load response of a complete rock-socketed pile with a soft base layer of E,,, =0.1E, using Leong and Randolph model Figure 21 shows the results of an analysis with LID =2, from which it may be seen that the deduced shaft response is very similar to the case without the soft layer (see Figure 16). 5.4. Comparison with field data Donald and co- wor ker ~~~- ~~ have presented results from both finite element analysis and field loading tests of rock socketed piles. However, no direct comparison of the two sets of results was attempted. At the time, the finite element analysis did not model the strain-softening behaviour in the side shear response that was observed in the field tests. The present model has been used to back-analyse one of the pile load tests (M8) presented by Williams et ~ 1 . ~ ~ As mentioned in the original publications of Donald et al., some of the relevant 44 E. C. LEONG AND M. F. RANDOLPH information for the rock mass were missing, and best estimates were made by Donald et al. in the light of their experience. The same estimates have been adopted here, taking Young's modulus of the rock mass as 610 MPa, Poisson's ratio as 0.3, a unit weight of 23 kNm-3, and KO =1. The rock mass was assumed to obey the Mohr-Coulomb yield criterion with c =1.5 MPa, and 4 =t,b =20". The only difference in the present analysis is that the dilation angle has been taken as I) =0". The other relevant parameters for the proposed model are (br =26", I),, =20, n/S =0.32 mm-' i / S =5 mm-' and pLp =0.1. The parameter m is assumed to equal 44, and the shear and normal stiffness are taken as G/t and E, / t respectively with t =0.01 m. These parameters are very similar to those used to back-analyse the results of direct shear tests on mudstone samples reported by Wi l l i am~.~' Details of these may be found in Reference 4. The concrete pile was assumed to have a Young's modulus of 35 GPa and Poisson's ratio of 0.2. The finite element results have been obtained using the Leong and Randolph model for the rock socket interface. The finite element mesh comprised 286 eight-noded quadrilateral elements with infinite elements at a distance of three times the radius at the side, and a distance of four times the radius below the pile base. Two analyses are presented here, one with the base of the rock socket in direct contact with the rock mass, and one with a soft base layer of thickness 0.140, with Esof, =0.01 Em, immediately below the pile tip. Introduction of the soft layer was found to be necessary in order to match the field response. Figures 22 and 23 show comparisons between computed and measured results. In Figure 22, with no softened base, the magnitude of the base response is overestimated although the side-shear response still agrees well with the field data. By contrast, Figure 23 shows excellent agreement between computed and measured responses of both shaft and base. 6. CONCLUSIONS Modelling the load response of a rock-socketed pile using the finite element method is a challeng- ing problem. The problem may be divided into three modelling aspects: (a) the rock mass, (b) the pile-rock interface and (c) the unbounded domain. In the present paper, the rock mass was assumed to obey a Mohr-Coulomb strength criterion, while infinite elements were used to model the unbounded domain. The use of infinite elements may not be the best method currently available, but the ease of formulation and implementation in a general finite element code renders the approach attractive. The rock-socket performance is dominated by the non-linear response at the pile-rock interface. This remains an active area of research, particularly as high quality field data are scarce. However, the interfacial model of Leong and Randolph4 adopted in the paper appears to have captured the observed load response of rock-socketed piles very well, as indicated by the comparison with the field load test reported by Williams et Further verification with other field data is desired. Several important observations are derived in the present paper. In socketed piles with side- shear only, the average shear stresses are dependent on the embedment ratio, LID, the diameter of the pile, D, and the relative stiffness ratio, EJE,, where E, is the pile modulus. As L/D and D increase, there is a decrease in the maximum average shear stress mobilised along the shaft of the rock-socket. However, no scale effects have been taken into account for the model parameters adopted in these analyses. Allowance for any such effects will result in a more gradual reduction in the average shear stress with increasing pile diameter. As the rock modulus, Em, decreases, the pile capacity is much less as the increase in normal stress due to dilation is reduced. In spite of the strain-softening nature of the side-shear only response, the response of a 'com- plete' pile (including full base contact) gave an overall load response that was strain-hardening, FlNlTE ELEMENT MODELLING OF ROCK-SOCKETED PILES 45 14 I Pile settlement (mm) Figure 22. Comparison of computed load response with field data of Williams et a1.,44 with no soft layer at the pile 0 Field t est M8 - FE results base 10 20 30 40 50 60 Pile settlement (mm) Figure 23. Comparison of computed load response with field data of Williams et al.,44 including a soft layer at the pile base even where a softened layer with modulus Esoft =OlE, was introduced below the pile base. However, the side-shear component still has a strain-softening response, but was more plastic compared with the side-shear only socketed piles. This is due to the interaction of the two modes of load transfer in side-shear and end-bearing, particularly in terms of modification of the stress-field around the lower part of the pile shaft. Overall, it has been demonstrated that strain-softening models of the type proposed by Leong and Randolph4 are capable of replicating a number of features commonly observed in the load transfer behaviour of rock-socketed piles. For complete piles, the distribution of shear stress down the sides of the pile is similar to that observed in the field (eg. Reference 44). The common practice of using load transfer models to estimate the complete pile response ignores interaction 46 E. C. LEONG AND M. F. RANDOLPH between base and shaft load transfer, and will lead to a conservative estimate of the degree of strain-softening of the shaft component. REFERENCES 1. I. W. J ohnston and T. S. K. Lam, Shear behaviour of regular triangular concrete/rock joints - analysis, J. geotech. 2. I. W. J ohnston and T. S. K. Lam, Shear behaviour of regular triangular concrete/rock joints -evaluation, J. geotech. 3. J . P. Carter and N. P. Balaam, AFENA Users Manual. University of Sydney, 1988. 4. E. C. Leong and M. F. Randolph, A model for rock interfacial behaviour, Rock Mech. Rock Eng., 25(3), 187-206 5. W. L. Wood, On the finite element solution of exterior boundary value problem, Int. j . numer. methods eng., 10, 6. B. Nath, The w-plane finite element method for the solution of scalar field problems in two dimensions, Int. j . numer. 7. C. A. Brebbia, The Boundary Element Methodfor Engineers, Pentech Press, Plymouth. 1980. 8. 0. C. Zienkiewicz, D. W. Kelly and P. Bettess, The coupling of the finite element and boundary solution procedures, 9. P. Bettess, Infinite elements, Int. j . numer. methods eng., 11, 53-64 (1977), eng., diu. ASCE, 115, 711-727 (1989a). eng., di u. ASCE, 115, 728-740 (1989b). (1992). 885-891 (1976). methods eng., 15, 361-379 (1980). Int. j . numer. methods eng., 11, 355-375 (1977). 10. P. Bettess, More on infinite elements, Int. j . numer. methods eng., 15, 1613-1626 (1980). 11. G. Beer and J . L. Meek, Infinite domain elements, I nt. j . numer. methods eng., 17, 43-52 (1981). 12. 0. C. Zienkiewicz, C. Emson and P. Bettess, A novel boundary infinite element, Int. j . numer. methods eng., 19, 13. F. Medina and R. L. Taylor, Finite element techniques for problems of unbounded domain, Int. j . numer. methods 14. A. Curnier, A static infinite element, Int. j . numer. methods eng., 19, 1479-1488 (1983). 15. J . M. M. C. Marques and D. R. J . Owen, Infinite elements in quasi-static materially nonlinear problems, Comput. 16. P. Kumar, Static infinite element formulation, J. Struct. Eng., 111, 2355-2372 (1985). 17. G. R. Karpurapu and R. J . Bathurst, Comparative analysis of some geomechanics problems using finite and infinite 18. G. Donida, R. Bruschi and R. Bernitti, Infinite elements in problems of geomechanics, Comput. struct. 29,(1), 63-67 19. E. A. Rukos, Continuous elements in the finite element method, Int. j . numer. methods eng., 12, 11-33 (1978). 20. G. Beer and 1. 0. Watson, Infinite boundary elements, Int. j . numer. methods eng., 28(6), 1233-1247 (1989). 21. S. K. Sharan, Elastostatic analysis of infinite solids using finite elements, I nt. j . numer. methods eng., 35, 109-120 22. E. C. Leong, A study of internal and external shaft friction ofpiles, Ph. 0. Thesis, The University of Western Australia, 23. I. B. Donald, S. W. Sloan and H. K. Chiu, Theoretical analyses of rock socketed piles, Proc. Int. Con$ on Struct., 24. J . 0. Osterberg and S. A. Gill, Load transfer mechanism for piers socketed in hard soils or rock, Proc. 9th Can. Symp. 25. P. J . N. Pells and R. M. Turner, Elastic solutions for the design and analysis of rock socketed piles, Can. Geotech. 26. R. K. Rowe and P. J . N. Pells, A theoretical study of pile-rock socket behaviour, Proc. Int. Con$ on Struct. Founds. on 27. R. E. Goodman, R. L. Taylor and T. L. Brekke, A model for the mechanics of jointed rock, J . soil mech.found. diu. 28. R. E. Goodman and J . M. Duncan, The role of structure and solid mechanics in the design of surface and underground excavations in rock, Proc. Con$ Struct. Solid Mech. and Eng. Design, Part 2, Paper 105, 1971, 1379-1404. Wiley, London. 29. C. M. Haberfied, The performance of the pressuremeter and socketed piles in weak rock, Ph.D. Thesis, Monash University, Melbourne, Australia, 1987. 30. A. F. Williams, The design and performance of piles socketed into weak rock, Ph.D. Thesis, Monash University, Melbourne, Australia, 1980. 31. R. K. Rowe and H. H. Armitage, Theoretical solutions for axial deformation of drilled shafts in rock, Can. Geotech. 32. L. H. Ooi, G. F. Boey and J . P. Carter, Finite element analysis of dilatant concrete rock interfaces, Proc. 5th Con$ 393-404 (1983). eng., 19, 1209-1226 (1983). struct., 18, 739-751 (1984). element methods, Comput. Geotech., 5, 269-284 (1988). (1988). (1992). 1991. Founds. on Rock. Vol. I Sydney, 1980, pp. 303-316. on Rock Mech., Montreal, Que., Canada, 1973, pp. 235-262. J. , 16. 481-486 (1979). Rock, Vol.1, Sydney, 1980, pp. 253-264. ASCE, 94, 637-659 (1968). J. , 24, 114-125 (1987). Australia on Finite Element Methods, Melbourne, 1987, pp. 157-162. FINITE ELEMENT MODELLING OF ROCK-SOCKETED PILES 47 33. K. A. Pease and F. H. Kulhawy, Load transfer mechanisms in rock-sockets and anchors, Report No. EL-3777, EPRI, 34. T. S. K. Lam, Shear behaviour of concrete-rock joints, Ph.D. Thesis, Monash University, Melbourne, Australia, 1983. 35. C. S. Desai and Y. Ma, Modelling of joints and interfaces using the disturbed-state concept, Int. j. numer. anal. 36. J. Ghaboussi, E. L. Wilson and J . Isenberg, Finite element for rock joints and interfaces, J. soil mech. found. div. 37. Z. Sun, C. Gerrard and 0. Stephansson, Rock joint compliance for compression and shear loads, Int. J. Rock Mech. 38. W. Leichnitz, Mechanical properties of rock-joints, Int. J. Rock Mech. Min. Sci. Geomech. Abst., 22, 313-321 (1985). 39. R. E. Goodman, The mechanical properties ofjoints, Proc. 3rd Congr. ISRM. Vol. IA, Denver, 1974, pp. 127-140. 40. S. C, Bandis, A. C, Lumsden and N. R. Barton (Fundamentals of rock joint deformation, Int. J. Rock Mech. Min. Sci. 41. G. Swan, Determination of stiffness and other joint properties from roughness measurements, Rock Mech. Rock Eng. 42. F. H. Kulhawy, Stress deformation properties of rock and rock discontinuities, Eng. Geol., 9, 327-350 (1975). 43. F. H. Kulhawy, Geomechanical model for rock foundation settlement, J . geotech. eng., div. ASCE, 104(2), 211-227 44. A. F. Williams, I. B. Donald and H. K. Chiu, Stress distributions in rock socketed piles, Proc. I nt. Conf: on Struct. 45. R. J. Jewell and M. F. Randolph, Cyclic rod shear tests in calcareous sediments, Proc. I nt. Conf: on Calcareous 46. A. M. Hyden, J . M. Hulett and A. F. Abbs, Design practice for grouted piles in Bass Strait calcareous soils, Proc. Int. Con$ on Calcareous Sediments, Vol. I , Perth, Western Australia, 1988, pp. 297-304. 47. M. F. Randolph, The axial capacity of deep foundations in calcareous soil, Proc. Int. Con$ Calcareous Sediments, 48. M. Fahey, R. J . Jewell, M. S. Khorshid and M. F. Randolph, Parameter selection for pile design in calcareous Palo Alto, Ca., 1984, pp. 102. methods geomech., 16, 623-653 (1992). ASCE, 99, 833-848 (1973). Min. Sci. Geomech. Abst., 22(4), 197-213 (1985). and Geomech. Abst., 20(60), 249-268 (1983). 16, 19-38 (1983). (1978). Founds. on Rock, Vol. I , Sydney, 1980, pp. 317-325. Sediments, Vol. I , Perth, Western Australia, 1988, pp. 215-222. Vol. 2, Perth, Western Australia, 1988, pp. 837-857. sediments Predictive Soil Mechanics Proc Wroth Mem. Symp., Oxford, pp. 261-278.