261 ‘Third International Canterence on Numericel Methods In Geomechanics /Aachon /26 April 1979 A method for predicting the effect of piles on slope behaviour RK. ROWE University of Western Ontario, London, Canada H.¢.POULOS University of Sydney, Australia a amreopuerzon In the past, piles have been used to in crease Slope stability in very soft clay (e.g. Brons, 1972) and to stabilise land~ slides by dovel action (e.g. Fukuoka, 1977). However, despite the groving acceptance of piles at a method of stabiliaing « sol) mass, there has been very little research into’ the usefulness and applicability of these procedures. Indeed, present. desien ethods appear to be based upon limited expirical experience and relatively simple snaiytical techniques (e.g. see Poulos and Davis, 1979; Fukuoka, 1977; de Beer, 1977). Jn this paper, a finite elenent tock pique for the malysis of the undrainea behaviour of soil slopes reinforces by Pile groups will be proposed. This tech- ‘ique hes the advantage that it is poss- ible to include the effects of a failure within soll or pile as vel as soil flow past the piles within « single analysis. ‘Thus the éeformation and stability of @ slope ney be assessed for various pile farrangenents and stiffnesses, while the influence of interaction between adjacent piles is isplicisly ineluded in the analysis. Application of the proposed technique MIL be Hlustratea by considering the effects of pile arrangement and stiffness upon the performance of an idealised slope. 2 METHOD OF ANALYSIS, ‘The application of the finite element meth od to the prediction of slope stability for an elasto-plastic material is quite streightforvard (e.g. Snivbahn and chen, 1976). However, extension of the method to allow the analysis of a slope reinforced by piles is complicated ty the three aim ensfonal nature of the pile group. In principle, thie type of problen could be analysed ueing three dimensional finite Glenents, but the cost of performing such sn analysis vould be prohibitive. Alter natively, if the pile group extends parallel to the slope for a reasonable Aistance then the general problea may be analysed, in an spprozimate sense, by using @ conventional tvo dimensions) (plane strain) soil-structure interaction an~ alysis, provided cone allowance is made for ‘the thie dimensional nature of the real situation. In particular, it is necessary to consider: 1. The correlation between the elastic response of a pile group and the idealised Plane strain pile: 2. The determination of the liniting presure viich may act on the piles within ‘group and the provision for soil flov "ehrough” the plane strain pile, corres ponding to #lip oF flow of the soil between and past the piles once this limiting pressure is stained; and 3. Provision for the shielding effect of piles vithin a group upon the liniting pressure the pile vill experience once the Boil beging to flow past the pile. Tn the proposed analysis, the soil and ‘the piles are considered separately, being related by nodsl forces whieh ensure cou~ patibility and equilinriun requirenents are satisfied. Thus, if the Liniting pressure acting on the pile at a point has hot been reached, then the Sisplacenens of Both soil and pile must be identical. Once ‘the Limiting pressure is reached at a point ‘then flow of the soil around the pile will occur, and this may be simulated by sLlov- ing slip of the soil past (i.e. through) the pile at that point, as illustrated in Pig.d. ‘The mmerical formulation for this analysis is similar to thet presented, by, Howe et al (2978) for the general analysis 1073of sofl~structure interaction problems. A brief description of the anelysis is given below, 2.4 Muneriest formation Consider @ structure) ody embedded in an elasto-plastic sofl mass. The composite ‘ody can be separated into two componente, namely the soil and the structure. The ‘tractions acting upon soi] and structure may be considered to take the form of nodal forces. Within any load increnent ‘the nodal forces acting on the soil (or structure) will consist of know nodal forces, due to applied traction acting Girectiy.on the body, snd unknown nods) forces Ey acting Detveen soil and structure 2.2.2 Analysis of the eof Supposing for the monent thet {nerenental nodel forces acting between cil and structure vere also knows, then the init~ ial stress approach (Zienkievier et aly 1969) could be used to analyse the soil, and the vector of nodal dispiacenente at the soii-structure interface p, may be weitten in the fore af) 2 5) 2 gO) ay where by is the vector of nodal displace ments at the soil-structure interface due to the known node) forces acting on the soil (in- cluding residual forces arising from the dnitial stress form wlation); is the vector at inerenentel nodal forces due to the struct ure acting on the soi]; is a matrix of influence co- efticients, which oan be dever- mined initially and vill remain constant for each iteration and osdstep while the geometry of ‘the problem renaine unchanged te. 1 (superscript) denotes the "sth" estinate of that variable. 2.1.2 Anelysis of the structure In general, if the structure is separate from the sci, 1t will be a mechanien so 262 ‘that the solution for deflection would contain a muster of arbitrary rigid ody notions. This aizrieulty ean be overcome ‘by applying sufficient artificial con straints to renove the indeverminancy. ‘nen this is done, the actual incremented @isplacenents Q, of the structural inter face nodes can Ge obtained from the diss Placenents of the restrained structure ba, plus the arbitrary rigid bogy motions, a2 follovs: be bye tad @ where @ is the vector of incremental rigid body notions, and is of order ps AAs a (wx p) matrix releting the interface deflections to rigia ‘body motions; p fs the mumber of artificial re straints applied to the structure, m is the mumber of degrees of free don st the soil-structure inter Face. The initial stress method can also be gpplied to the structure, and the vector Opr may be written (1) yp (nen) (on). 3 de Rox 2B “a a) where the definitions ,y. Tg, parallel those of Bix, Tigt When equstions (2), (2) ané (3) are combined it is found’ that ay) 5) yp eg) OM? wn gl gpg) = ix!) a If there are m unknown nodal forces and P rigid body motions, then Equation (4) Fepresents = equations in (n+p) unknowns ‘The renaining equations can be derived from the observations that the structure must be in a state of equilibrium. The equations éf equilibrium corresponding to the artificial restraints can be writven in the form fant), A By () 1074whore jy ie the vector of imerenental Joads Afplied to the serscture; and vnere Tor convenience, St has been assumes that the applied loads j, are located at the soll structure intefface. The more general case say be obtained bya single modhfie~ ation of the proguct AE. Equations (li) and (5) aay be written in ‘the simple interaction equation (nea) iz, en ana For a soil-structure interface at whfch 20 slip oecurs the vectors 9, 0") ana 6,78) ay be Taken as quel and’ thus enyatifn (Pay be etvedite aeterine B1GHFna tN “Ouse these guastities abe knove fhe atapnscenenes lain teen he sol ass ang the stricture may be airectly determined. The process ia then repested until convergence is achieved. ‘This approsch derives te goer from the fact that it is only necessary to trisngi- arise the elastic stitmess motrix (for oth soft and structure) once for al Iterations and loadsteps white the probes configuretion rensins constant. Problens in vaich slip occurs at the interface can be solved ueing this method. ‘The most common criterion for alsp 42 that once the nodel force reaches’ « spec- ified value It remains constant (nore complicated eriterta are considered ty Fowe et al, loc.cit).” Suppose that this pode foreé is Fy! and the corresponding holed deflection isp". Onee the eritiea) Yalue is reached Fy'/witt venain constant bo that #," vill bE zero cor all fuvure Inerencat! unless untoading occurs. ‘Thus fone unknown 3 lost fron (6), and replaced By (by = by)» the relative aisplacenent beewedn stfucture and soll, 0 (€) may 411, be solved to obtain by (04) ana teen? 2.2 Correlation between 34D and plane stress pile stiffness Jn approxizate correlation between the 263 elastic response of s real pile group with ‘the idealisea (plane strain) pile group coud be obeaned ty ‘iL. Determining the lateral response of @ row of the real pile group fron an elastic pile group analysis (e.g. Poulos, 1971) 2, Determining the 1oteral response of an idealized plane strain (contsmuous) Bile rev for a range of stifmesses (2.8. from an elastic finite elenent anslysie) ana 3. Bguating the lateral response of the real pile row vith the lateral response of ‘the idealised pile row to give an equiv alent stiffness of the plane strain piles. Unfortunately, the sbove procedure vill not engure that all response character~ istics of the pile row are reproduced, es only one parancter, the stiffness of the plane strain piles, ean be adjucted. Thus, ‘8 plane strain pile row which under horiz- fontal load has an equal head deflection to ‘the pile rov will not necessarily have the sane head rotation under horizontal load, or the sane deflection and rotation under ‘moment loading. For practSeal purposee, St appears that es good a procedure as any is to Gistribute the pile stiffness uni- Somly along the plane strein rov, but to ‘ake the loading per unit run as the single pile loading divided by the single pile Sidth or diameter. The use of this pro- cedure tends to underestimate deflections fang rotations for flexible piles, but gives quite close agreenent with the elastic pile row solutions for piles of mediun to low Tlextbiiity. 2.3 Determination of the 2intting presture between pile end soil ‘The value of the Limiting soll pressure whieh can act at amy particular point long an isolated pile depends both upon the shape of the pile and the depth below the surface to the point in question. A Sunnary of the various solutions for the ultinate soil presture acting on a pile at considerable depth belov the surface of Purely cohesive soil hae been given by Broms (1964). For a circular pile, s Lliniting pressure of a/e = 9 spears to be generally accepted, ithough the value may vary between 6 and 12 depending upon pile shape and roughness. ‘The variation in ultimate bearing pressure with depen ranges from g/e = 2 at the cur~ face to the "deep" pile value at a depth of approximately 3 pile diameters (e-6- Poulos and Davis, Joc.cit). In the pro- posed analysis, the pressure of the soi) acting on the pile is manifest as nodal 1075sey ~ Rough Rigic: 000 00m Fig.l Finite elenent idealisation of soil slope oe ta aad taoaal ‘tm soured group: 4) get tow ot aco, Pig.2 Boundary conditions for pile groupe 1076 Pig.3 Brficiency of a 3xinfinite pile 264forces required to ensure compatibility of pile snd soit deflections, Accordingly , Limiting nodal foree, eorresponaing to the Limiting pressure over @ region of the pile, can de specified for each node and once this value is atvained then compat Sbility requirements are no longer enforced ‘at that point, thereby alloving the soil to flow past the pile. ‘The value of the Limiting nodal force may very from point to point, alloving freedon to select the Gistribution of ultinste pressure consider fed appropriate, ‘The shielding effect of piles within a large group upon the ultinate pressure may be deternined by performing a separate analysis. For exenple, a group (idealised fas plates) consisting of three anfinite Tous as indicated in Fig. 2, vas analysed using boundary conditions similar to those suggested by Yegian and Weight (1973). The efficiency of @ pile Ny, is defined as the ratio of the mean pregoure g, acting fon the pile at ecllepse to the mean pressure acting on an daclsted pile, Le. am wt te) between piles and B ie the diameter of the pile. co) The efficiencies obtained Zrom an elasto- plastic analysis of the 3 x infinite pile group (see Rove (1978) tor full details of this analysis) are plotted against spacing s/B in Pig. 3. The Deheviour of the front (or rear) pile within the group was 2erge~ ly unaffected by adjacent piles. However, ‘the central pile was significantly shiclé- ed ty the leading and end piles, resulting in a much lover efficiency. ‘A comparison between the shielding effect predicted for eiroular piles by Yegian and Wright (oc.elt) and that pre Aicted ty the authors for these slender rectangular (plate) piles indicated agree ment to better than 10%. Thue the ultimate pressure at a point on an isolatea pile ay be approxinstely modified for the effect of group interaction by using the efficiency factor ny, Table 1. Soil parameters for the lope a a} ‘Type _ (See Fig-4] (Fa) 08a) Hoa EB 9000 OB 50 geneous Eeyered ow 9000 0.18 30? 28000 0:48 60 Table 2, Pile cases considered ‘ase File Tenath(a) Pie boundary (from top of conditions slope! op Batten x ° To pile Boo free sree c 20 free pinned D 20 To lsteral inne Aisplacenent E20 free fixes 43 PROBLEM DESCRIPTYON AND FINITE ELUTE IDEALIGATION A hypothetica? 20m high 1:2 slope in soft etay (layer (a) resting on 2 10x thick clay stratum (layer (b)) and underlain by fa rough rigia bese, was considered (see Fig.l). Ivo soi) profiles vere examined, nanely, a homogeneous soil mass and a two layered #08) mass. The soil paraneters for these two profiies are given in Table? and the finite elenent {desiieation of the S081 mase is shovn in Fig.1. To {lustrate the effect of the piles upon the performance of the slopes, = particular pile group consisting of three Tous of piles with a cemsre to centre spacing Of om vas examined. For this spacing, two pile éianeters, a, vere con sidered; namely, 2=0.5e and €21.0n, corresponding to'a dimensionless spacing 8/a of four and two respectively. The effect of shielding upon the wleinate bear ing pressure (taking q/e=9 for an isolated Pile) wae determined from Fig.3- ‘The piles within the group were subject toe variety of restraint conditions as shown in Fig. and sumsrised in Table 2 For purposes of comparison, three stif?— nesses (SI) were considered for the ideal- ised piles:- EL 2.1 x 10% EE = 1.2 x 208 EI = 8:4 x 10° Tie corresponds stirmess. o facilitate the comparison of the relative performance of aifferent pile aystens, an increnental density construct ion sagience vas adopted. ‘This approsch hae the advantage of simplicity and does not create serious difficulties ven com paring the relative performances of aiff~ ferent systens. Hovever, careful consider~ ‘tion mist be given to the physical smpli- ations of any construction sequence when fis desired to predict actual deforsst— fons for comparison vith experimental observation. Tn addition, consideration must be given to the method ty which the piles are in ain? fm dae? fn nas /a to 8 N00 fold range in 1077 265serted and the condition of the enbanixent oF slope at the tine of insertion, For cuanple, significant negative friction forces may develop between pile and soil Af end bearing piles are inserted before Anerementing the soit density. Fer the purposes of this analysis, it was sssuned ‘het no negative friction developed be- ‘tveen pile and soil, since this vas con sidered to be the most realistic situation. However, it should be recognised thst the Gevelopnert of negative skin friction may significantiy alter the performance of the pile-soil aysten. It is anticipated thet negative friction vould have a beneficial ‘effect upon the performance of the siope provides it does not induce failure within ‘the piles. Finally, in the following analyses, st ‘bas been assured that no failure would take place vithin the pile. This simpiity- Ang assumption was made to reduce the number of parameters to be studied, but it is recognised that in practice the noent resistance of the pile my be a major factor influencing the performance of the pile reinforced slope. The provision for the formation of plastic hinges vithin the piles is quite stralgntforvard, 3+1 Effect of Piles on the deformation and stability of the slope The results obtained from the analyses of slope Debaviour nay be subaividea into four series as indicated in Table 3. In Particular, Table 3 summarises the reduct- Hon in the Lateral displacement (8,) at Locations "X' and ‘Y" (shown in Fig.1) a yell as the reduction in vertical heave (Ry) at location 'Y" obtained for the various cases where: Gieplacenent obtained for particular case with piles ‘Geplecenest of the same point when there are no pfles To allow 4 reasonable comparison to be ma‘ between the effect of various pile arrange ments upon the performance of the slope. the displacement ratios (Ry, Ry) and max imum bending moment Mya, were Compared for 2 typioal value of cotfunit weight Y= 16 kx/m’. Sintlariy, the maxim wom Ente within the plles were compared for © loading corresponding to a lateral dis~ placement st the point 'X" of 250mm, vhich ‘generally corresponded to the situation Just prior to collapse of the unreinforced Slope, Pinslly, the inerease in the factor of safety (relative to that for an unrein- forced slope) 1s also shown for the var~ Aous pile arrangenents. In fact, the unit Ry or By = 266 weight for failure of the homogeneous Slope was found to be y = 16.5 kl/a? whieh ney be directly compared vith values of Y= 16.8 lil/e” obtained tron Taylor's charts, indicating agreenent to within better thon 2. The plastic region and velocity field at collapse for the 1:2 un- Pelntorced homogeneous slope are shown in Fig. 5. Fig.6 shove the deformed profiie of the ponogeneous slope with Y= 16 ki/x for ‘the four cases considered in series 1. Fron thie, s¢ can be sean that pile stiff nese significantly alters the lateral dis- Placenent of the elope, and the heave et ‘She toe of the slope; “hovever, the rs of improvenent with increasing pile stiff- nese 4¢ relatively sloe. It vill be app recieted from Pig.5 that the failure of the slope is associates with a deep seated novenem of the soil. Accordingly, the introduction of piles has the grestest effect on the deformation at the toe of the slope and the least effect upon the lateral movenent of the upper portion of the elope. This situation arises because the piles (even for large stiffneasas) tend to nove ‘with the eo1, and although they offer bone restraint to the deep seated movenent, ‘the rotation of the pile sbout the base stil allovs quite significant deformation of the slope. In fact, for pile arrange- ments such 8¢ shown in'Fige. 25 snd 20, ‘the piles moved with the soil mass to such fan extent that failure of the slope could bbe deened to have occurred before any evidence of soil flow past the pile. ‘The effect of pile stiffness upon the lateral deformation of the seit adjacent to the leading row of piles may be assessed from Fig.7. It can be seen that incressing pile stiftness significantly decreases the lateral deforsation near the centre of the layer, but has very ittle effect on the Geformation at the top of the slope. On tthe other hand, the bending monexts within ‘the pile (see Fig.6) increase dramatically With increasing eviftnese. Finally, it 8 of some interest to note that « twentyfold Increase in pile stiffness omy increased the slope stability by about 95. ‘The effectiveness of piles in reducing ‘the deformation and increasing the stab- Alsty of slopes depends both on the pile stiftnese and the fixity of the pile. To GLustrate thie point, Figs. 9 and 10 show ‘the slope profile obtained for various pile arrangements vere the idealised (plane strain) pie hee qn equivaient stitr— ness EI= 2.1 x 10" kiln®/m (series (2)) fang 4.2 x 10° kit-m*/a (series (3)) respect ively. Tae corresponding lateral deform ation (for y= 16 bil/m®) at the location Of the leading pile is show in Fig.7. 1078