Geotechnical Aspects of Underground Construction in Soft Ground ~ Viggiani fed) ‘© 2012 Taylor & Francis Group, London, ISBN 978-0-415-68367-8 3D finite element analysis of deep excavations with cross-walls S. Rampello & S. Salvatori “Sapienza”, University of Rome, Kraly ‘Ground movements induced by deep open excavations retained by diaphragm walls may ally reduced using sacrificial cross-walls installed as props between the retaining walls. In this paper, a 3D finite clement study is presented in which the retaining walls and the cross-walls are modelled using brick clements and accounting for the interfaces between the panels of diaphragm walls and of cross-walls, Influence of cross-walls spacing and length in reducing. the horizontal deflection of diaphragm walls and the ground settlements behind the excavation is evaluated. Companion plane strain analyses, in which an homogeneous equivalent medium is considered in between the retaining walls, are also presented to assess the capability of 2D analyses to- predict the performance of deep excavations in which cross-walls are used asa mitigation measure to reduce wall dellections and ground movements, | INTRODUCTION In recent years, cross-walls have been often used in deep excavations in urban Taiwan area to reduce ‘Constructions of deep box excavation in soft diaphragm wall deflections and associated ground ground generally result in ground movements movements. Well documented case histories per- ‘that can induce significant damage to adjacent mitted to assess the effectiveness of cross-wall ‘buildings and services. To reduce their impact, — reducing excavation-induced displacements (Hsiung ‘box exeavations are usually retained by stiff dias etal, 2001; Qu etal, 2006), provided that the joints phragm walls and props installed as the exeavation between the perimeter diaphaga walls and the progresses. However, vertical stress assaei- _ eross-walls are well constructed (Hsieh et al, 2008). ated to excavation induces surface settlementseven ‘The beneficial elfects were evaluated by comparing when the retaining walls are prevented from mov- horizontal wall deflections monitored at locations ing horizontally and deep-seated inward displace- affected by the presence of eross-walls with those ments of the walls that cannot becontrelled by the measured at sites with no special support (Hsiung props that are installed within the excavation itself et al,, 2001) or computed under plane strain eondi- (Burland et 1979). tions-assuming ne cross-walls (Ou et al., 2011). To minimise the impact of deep excavations in Effectiveness of cross-walls in reducing wall ‘urban areas, inward displacements of diaphragm deflections and ground movements has been walls and surface settlements around exewations recently studied by Merrit et al. (2010) via 3D may be reduced installing an internal support sys» finite element analyses in which the retaining walls tem below the formation level, prior to excavation. and the props were represented by shell clements Such a system of deep in situ props may be consti-. and the cross-walls by membrane elements; spe- tuted by tunnelled struts (Stevenset al., 1977; Bailey sifically, influence of horizontal axial stiffness of stal., 1999), ground treatment methods such as jet eross-walls relative to that of perimeter retaining ical 2009) and jet walls was assessed, -grouting, used to replace the soil below the finalexca~ "This paper deals with « 3D Finite element study n kevel oF sacrificial eross-walls (Hsiung etal., of deep excavations in which the diaphragm walls 2006; Hsich ct al., 2008). Cross-walls and the cross-walls are modelled using brick cle- can be formed by jet grouted columns or unrein- — ments and the contacts between the corresponding forced panels installed with diaphragm walling panels are represented by interface clements, equipment. They are installed between the perim- -Exeavations were modelled without cro: eter diaphragm walls before the start of exeavation and with crossewalls of different length, and are excavated out with the soil down to the at different spacing. To assess the eapability of 2D depth of the final excavation level. At theend of the analyses to predict the performance of deep exea- jon sequence, the only portion of the eross- vations with sacrificial cross-walls, plane tte the horizontal defleet analyses were also carried out in which an equiva- of the diaphragm walls is that eventually extending lent material is used to represent the sei below the bottom af exew ‘erasg-walls within the exca 3052 PROBLEM DESCRIPTION 2 Ground conditions adopted in the analyses are representative of these encountered in the histori- cal centre of Rome, close to the Ti wer. The soil profile shown in Figure 1 consists of medium dense gravelly Made Ground, with an average thickness of 9 m, overlying recent alluvial deposits of Pleistocene age, They consist of slightly over- consolidated clayey silt, about 6 m_in thickness, medium dense silty sand, 15 m thick, normally consolidated silty clay, 27 m in thickness and sandy gravel, of 6 to 8 m thick. The gravel is underlain by a thiek layer of stiff and overeonsolidated silty clay of Phocene age. Figure | show typical profiles of tip resistance g, measured by CPTU tests and of small-strain shear modulus G, as obtained by Cross Hole tests. The profile of pore water pressure was abserved to be hydrostatic with hydraulic head at about the bot- Soil prafile and constitutive modet | behaviour of 1g an elastic-plastic rate independent model with isotropic hardening (Hardening Soil, HS; Schanz et al. 1999), available in the model library of the code Plaxis The model is capable of reproducing soil non-linearity due to the occur- rence of plastic strains from the beginning of the loading process. u (KPa) 0 200 40060 18 pe + o 0 10 ten] = Se ' tes ow ws = os i Figure | ao ‘The elastic behaviour is defined by isotropic elas- ticity through a stress-dependent Young's modulus: pn [ Scott + ay" y “got gf +p) © where cis the minimum principal effective stress, eis the cohesion, g’ is the angle of shearing resistance and p’ 100 kPa is a reference pres- sure; £™ and 1 are model parameters, The model has two yiekl surfaces, f, and {. with independent isotropic hardening depending, on deviatoric plastic strain 77 and on volumetric plastic strain e?,, respectively. The deviatorie hard- ening rule is related to parameter £%,, while the volumetric one is controlled by parameter E%,,. Both of them are given by expressions similar to Eq. (I) but, in contrast to E% they are not used within a concept of elasticity. The flow rule is associated for states lying on the surface f., while a non-associated flow rule is used for states on the surface f Under monotonic loading, HS model _can account for non-linear stress-strain behaviour, with tangent initial modulus equal to £% and for the occurrence of irreversible strains from the beginning of the loading process. Upon unloading, the model assumes elastic behaviour with Young's modulus £% thus reproduci nificant change n stiffness, Go (MPay e 0 200 400600 ‘Vertical section of the exexvation with soil profile. 506Since £” represents the tangent initial Young's modulus of the stress-strain curve, it has been related to the shear modulus at small strains G, obtained from Cross Hole tests. In particular, vale ues of £™ and m were obtained by best fitting the cross-hole test results using Eq. (1) and assuming v= 0.2. The continuous line in Figure | represents the prediction of G, obtained with the values of 4p, B% and m reported in Table 1, Specitically, the values of o’, needed for computing £” were obtained using the values of the coefficient of carth pressure at rest K, listed in the table. It is worth mentioning that OCR has to be regarded as a Yield Stress Ratio (¥SR) defined nthe framework of strain hardening plasticity, so that values of OCR > 1 can be specified also for geologically normally consolidated soil layers exhibiting a yield stress larger than thein situstress; this was the case for the granular sails such as the made ground and the layers of sind and gravel, normally consolidated silty clay, 20 for the stiff ever-consolidated clay deposit and 10 for the remaining soil layers: B*,JE"., = Land an angle of dilataney at failure y = 0 were adopted for all soils these values provided a satisfactory fitting with the stress-strain curves observed in the trlaxial tests. Analyses were carried out in terms of effective stresses modelling the clay layers as undrained and the other soils as drained. 20 The excavation consists of an elongated box with a ength (66.8 m) to width (33,6 m) ratio of about 2 and a depth of about 40 m. Dimensions of the box excavation are typical of planned metro stations in the historical centre of Rome. The exeavation is retained by diaphragm walls and by horizontal propping levels installed during the top down exca- vation sequence. In the analyses the excavation was carried out in twelve stages, with seven levels of props con- structed as exeavation progresses. The wall behaves as an embedded cantilever up to the excavation level at +7.9 m o.d. (8.9m depth), at which the first propping level is installed. The sequence of exeava~ tion and propping then continues to the final exea- vation level at -23 m od. (40 m depth). Figure 2 depicts schematically the final stages of excavation: 610 64 mof soil were excavated after the construc- n of each propping level constituted by the oor slabs, with the exception ef temporary steel props installed at stage 9 at an elevation of -16.6 m od.. ‘The perimeter retaining walls reach the deep layer cof stiff clay, to prevent from uplift the excavation Excavation model and structural properties Stiffness decay with shear strain was described using ratios of EVE", = 13 for the Table 1. Soil properties * e ¢ cl Soll (Nien) (kPa) F) OCR Ky (MPa) on MG ISS 1S 29 20 052 20 06 sit 190-426 24 ORS 22007 Sand 19.0 us 0.49 2700 OS neslay 18.6 4 Boil 0.60 1 09 Genel 19.0 35 5 043 2100 OS ocelay 2.0 30-25 ORD 4D OB 1 t a eee tog Sintra] 6 Figure 2. Excavation stages: cross-wall length L, 42.5 my. 307bottom and to minimise water flow within the ‘excavation; they are assumed to be formed from reinforced concrete for a length of $5 m and from uunreinforced conerete for the final portion, 10 m long. Diaphragm wall panels, as well as cross-wall panels, were assumed to be 1.5 m thick and 2.8 m wide. Three lengths were considered for the sac- rificial cross-walls: £., = 38 m_that is equal to- the maximum depth of excavation (23 m od.) L,, = 42.8 m and 47.0 m, for which the cross- wall toc is installed 4,$ and 9.0m below the final level (=27.5 m o.d. and -32.0 m od.), respectivel Three values of eross-wall spacing were also. assumed in the analyses (Fig, 3): s= 84, 11.2 and 224 m, with two, three and seven panels of dia- phragm walls included in between the cross-walls, The ratio between the area occupied by the cross- wallsand that of the soil within the excavation was. Addy = 21.7, 134.and in the three cases. ‘Using symmetry, a portion of the box was con sidered in the analyses extending between the cen= tre line of adjacent cross-walls as shown in the plan view of Figure 4 for a cross-wall spacing s= 8.4m, so that a current central portion of the box excava- tion was modelled. The 3D finite element mesh extends t 100 m behind the diaphragm walls, that is 2.5 times: the excavation depth, and up to the exeavation eens tre line in front of it. Its height is about 2 final exeavation depth (76.8 m). A 3D view of the domain analysed for eross-wall spacing s-= 11.2 m 168. Figure 3. Analysed schemes of cross-walls systems. Figure 4, Plan view of the analysed domain fer cross ‘wall spacing s= 8.4 m, PT HT aT Figure §. 3D view of the analysed domain for stoss- ‘wall spacing s= 112m. Both soil and structural elements were disere- ised with 1Senoded linear strain wedges; 10240, 141.44 and 37056 clemenis were used for the three valucs of cross-wall spacing. All the structural elements were modelled as incar clastic materials with a Poisson coefficient 0.15. The unit weight was y= 24 and 25 KN/m” for unreinforced and reinforced concrete. ‘Cross-wall panels were assumed to be formed from unreinfoteed concrete of strength (and stiff rness) lower than that of perimeter diaphragm, walls; specifically, the cross-walls Young's modulus was 60% of that of diaphragm walls, £ =0.6 E,.. 7 508Stiffness of the floors of the box construction, acting as propping levels, was reduced to account for ‘the presence of the lift and stair openings also used vo extract the soil during excavation: openings ‘was assumed to be 40% of the floor area in tot Floor slabs were connected to the diaphragm walls through linear elastic interface elements, 0.2m thick, with a stiffness reduced by about an onder of magnitude to minimise the transmis- sion of bending moment (o the retaining wall (E= 13 6Pa Table 2 summarises the propertics of structural elements. ‘The temporary stccl props installed at the elevation of -16.6 m od. were also modelled using, ‘brick clements, assuming a Young's modulus of 3.GPa to represent tubular steel props. Soil-structure interfaces were modelled using finite elements 0.2 m thick (Fig. 4}. The same eon- stitutive model adopted for the soils was used, but with reduced strength and stilTness parameters (Table 3). Specifically, a reduction factor used for c”and tan @ while the reducti equal to R,? for the Young's modulus, Stiffhess profile of soil- structure interfaces, and ‘he stiffness decay with shear strain were deseribed using the same values of 7 JEM and y adopted for the soils. The contacts between diaphragm wall panels and cross-wall pancls, as well as the contacts ‘between cross-walls and diaphragm walls were also modelled through finite elements, 0.2 m thick ‘The mechanical behaviour of these clements was Table 2. Stiffness of structural elements x {MPa} Reduction (GPa) Reinforced perimeter wall 32 Unreinforced perimeter wall 12 Cross wall 12 Floor propping level Rn 06 Bottom slab- x ‘Table Soil structure imterfave properties. te eo me Soil Rhy CF MP) Made ground 0.705 BT Clayey silt 07 6S 188 Silty sand 03 nom nesityelay = 0.7 168K ” Gravel 10 co 210 coc silty slay os ou noise 509 described using a linear clastic—perfeetly plas- tic Mohr—Coulomb model, The shear strength between the panels was characterised by negligible coesion, c= 2 kPa, a friction angle @ = 38° and ancy angle at failure y = 10°, he stiff ness was equal to that of the corresponding panels: F = 36 GPa for contacts between panels of dia- phragm walls and £'= 22 GPa for contacts between panels of cross-walls and between diaphragm walls and ¢ross-walls. 3 ANALYSIS RESULTS 3.1 Threesdimensional finite clement emalyses Refinement of 3D meshes was evaluated through nary analyses carried under conditions (without cross-walls) using the 2D and the 3D suites of the eode Plaxis Soil and structural elements were discretised with 15-noded 3rd order strain triangles in the 2D model and with 13-noded, linear strain wedges in the 3D model, Refinement, of 3D meshes was assumed satisfying when dif- ferences of computed wall deflections were lower ‘than 5% for all stages of excavation. The results of 3D finite element analyses were examined to evaluate the effects of cross-walls spacing and Iength on the horizontal deflections ‘of the perimeter diaphragm walls and the ground settlements behind the excavation. Both of them are in the following referred to the middle section between the cross-walls (dashed linc in Fig. 4). Figure 6 compares profiles of horizontal wall deflections computed for excavation to 40 m depth (stage 9), without eross-walls and with eross-walls of varying length, installed at different spacing. Table 4 summarises the computed values of maximum wall deflections. Figure 6 Eflests of crose-walls spacing and length on, ‘wall deflections,Table 4. Maximum wall deflections. NoCW s=84m s=12m s=24m he “ " 4 = mm mm mun mim wo BT $43. 36.0 2s 10.7 2 ao 203 Inall eases, the horizontal walldisplacements are strongly reduced throughout the excavation depth, while a different behaviour is observed below the bottom of excavation depending on cross-walls Iength, For values of L.. = 38 mm, equal to the maxi- mum excavated depth, a slight influence of cross walls spacing is observed on the displacement profiles and the maximum wall deflection is about 20% lower than that computed without cross-walls, ‘When cross-walls are prolonged below the fins! excavation Ievel, horizontal deflections strongly reduce also below the formation level and the influ ence of crass-walls spacing becomes more evident. For a cross-walls length £,, = 42.5 m (4.5 m depth below the excavation bottom) the maximum wall deflection rediices of about 85% using a cross-wall ing x= B.4-11.2 m and of about 67% using 4m. A further increase in cross-walls length, 47 m, (9.0 m depth below the exeavation bot- tom) providesa further reduction of maximum wall displacement (= 90% for s= 8.41.2 mand 70% for r= 22.4 m), although the benefit of increasing L., reduces for higher cross-walls spacings. Limited n of cross-walls length below the final n level then produces a significant redue tion of wall deflections along the embedded por- ion of the retaining walls Figure 7 shows variation of horizontal deflec tions along the diaphragm wall at ~19 m od,, 3m below the temporary propping level and at the excavation level (-23 m od.) for the case of LL, =42.5 m. Wall displacements smoothly increase with distance from eross-walls with negligible rela- tive displacements between the panels of the dia- phragm walls. Sufficient connection between wall pancls was then provided by the strength properties assumed for the interface elements and by the contribution of both the temporary props and the soil below the excavation level. Similar trends were computed ynal analyses im which the friction and tancy angles were reduced to 30° and 5°, respectively Figure 8 shows the displacement field computed without cross-walls and with cross-walls of dif installed at spacing s = 11. Figure 7. Maximum horizontal deflections of perim- ter diaphragm walls with varying erosswall spacing: a} -19 mod: b) -28 mad, ang Figure 8, Effect of cross-wall length on displacement field at constant crose-wall spacing s= 11.2 my. results refer to the final excavation level at 23 m ‘od. (40 mt depth). Both wall deflections and ground surface settlement profile behind the sxcavation are for the mid-section in between the sross-walls, Table § reports the maximum surface‘Table $. Maximum surface settlements, No cw (mm settlements, obtained at a distance of about 30m from the wall, Consistently cross-walls. also Iy the may tude of surface settlements. Specifically, benefit increases with eross-walls length, settlement reduction varying from about 32% for £,,. = 38 m to about 90% for L. = 42.5-47 m, Surface settlements also reduce with decreasing cross: walls spacing although limited benefit is obtained fors<11.2m. 3.2, Plane strain finite element analyses To evaluate the capability of predicting the per- formance of deep excavations with sacrificial crossewalls assuming plane strain conditions, com- parison 2D finite clement analyses were fin: ried out Inthe analyses, the soil and the ceoss-walls within the excavation were modelled as an equivalent homogenous material with linear elastic behaviour, le the interface between the equivalent mate- rial and the diaphragm walls was described using a lincarelastic—perfectly plastic Mohr Coulomb model. of the equivalent homogencous mate- ed averaging the soil and the cross- walls properties, weighted on the corresponding dimensions occupied in pln. The equivalent ‘Young's modulus was for example evaluated using the relationshi where sand s are the cross-walls thickness and spacing and E., and E, are the crosswalls and soil Young's moduli Preliminarily to this, average soil properties were evaluated for the soil layers included throughout the eross-walls length, considering the layers thick- ness as a weil tor. Due to the assumed lin- car elastic behaviour, soil stiffness was computed using the values of £%, at the mid point of each layer, that can be thought to be representative of stiffness at average strain levels, rather than using the values of £’that were linked to the small-strain shear modulus G,. Similar values of average soil properties were obtained for the different cross- Sul 4.0m walls lengths, so that the following quantities were 8.8 KNim’, £7= 22 MPa, c’= 14 kPa ‘weight and the stiffness of equivalent material were slightly affected by cross-walls length as well, so that a single value of 7,, and E.,’ were assumed for each cross-walls spacing (Table 6). Sirength parameters of interfaces elements between the equivalent material and thediaphragm walls were obtained using relationships similar te Eg. (2) for cand tang, and a reduetion fae tor R,, = 0.7. Xgain, values were obtained irrespective of cross-walls length and spacing thus assuming ¢,, = 9 kPa and @,, = 21°, Stillness of interface elements was instead obtained multiply ing the values of £,, in Table 6 for the reduc factor R,2=0.49. ure 9 compares profiles of wall deflections computed for stage 9, when the final excavation level is reached, under 2D and 3D conditions 1s refer to erosswalls length L.. = 42.5 m mod). Far values of s=8.4 and 11.2m, the plane strain analyses, in which the cross-walls and the soi within the excavation are described by an equiva- Jent material, provide a fairestimate of wall deflee- tions along the excavated depth, while horizontal wall displacements are overestimated below the final excavation levels, when assuming plane strain conditions ically, maximum wall displacements eom- puted in the 2D analyses are of about 2 times greater than (hose computed in the 3D analyses and occur at higher depihs. Significant differences the deflection profiles are instead observed for 5=224m, for which the three-dimensional pattern of behaviour is more relevant. Similarly to the results obtained for the wall deflections, plane strain analyses also provide an overestimate of ground surface settlement pro- behind the wall, as shown in Figure 10 for 42.5 m; again, for s= 8.4 and 1 Ln mum settlements computed under 2D conditions are about ? times greater than the ones computed under 3D conditions Then, plane strain analyses in which an alent material was used to represent the soi the eross-walls within the exca an overestimate of wall deflections and ground movements However, significant differences. in hiv andTable 6. rial, ‘Young's moduli of the equivalent m 84m 12m 24m (MPa) 3980. 2065 04 aakNim’) 19:7 Ins md 9) Figuis 9, Horizontal displacement of the diaphragm, walls computed by 3D and 2D finite element analyses Figuse 10. Surface settlement profiles computed by 3D. and 2D analyses: L,. = 42.8 m,5= 11.2 m. profiles of wall deflections were obtained from 2D and 3D analyses when cross-walls spacing was increased to 5 = 22.4, due to the increasing influ: ence of the 3D displacement field. 4 CONCLUSIONS 3D finite-element analyses of deep excavations with sacrificial cross-walls were carried out to assess the effectiveness of eross-walls as a mitigation meas- 512 ure for reducing the horizontal deflections of dia- phragm walls and the ground settlements around the excavation, In the analyses, a portion of the was modelled extending between the & ¢ of two adjacent cross-walls, Both soil and structural elements, including the eross- walls, were described by 15-noded linear strain wedges, using interface elements to model the soil-siructure contacts and the contacts between the structural slements, stich as those between the panels of diaphragm walls.and of cross-walls Specifically, the analyscs were pcrformed to investigate the effects of cross-walls length and spacing on the ground movements induced by the excavation, considering a soil profile typical of central Rome. ‘The results show that sacrificial cross-walls permit to reduce substantially the diaphragm wall deflections and the ground surface settlements behind the perim- eter diophragm walls. Ground movement reduces further prolonging the cross-walls below the fi ng cross-Wall spa observed for values of cross-walls embedded length greater than 4.5 mi and fier eross-walls spacing lower than 1.2m. Comparison 2D analyses were also carried out to evaluate the capability of predicting the per- formance of a deep excavation in which cross- walls are used as @ mitigation measure 10 reduce ground movements. In the analyses, an equivalent material was modelled to represent the eross-walls, and the soil within the excavation. Fer cross-wall spacing s = 8.4 and 11.2 m., a fair agreement was obtained for wall displacements computed throughout the excavated depth, while an overes- imate was obtained for wall deflections computed below the excavation bottom and for the ground surface settlement profiles behind the excavation. On the other side, profiles of wall deflections computed by 2D and 3D analyses differed sub- stantially for s = 22.4 m, for which the 3D pat- terns of behaviour strongly affect the computed displacement field. ACKNOWLEDGEMENTS: ‘The Authors wish to thank Mrs F. Buono and Dr G. Landi for useful information on soil data and Prof, L. Callisto for fruitful discussions and suggestions. REFERENCES Bailey RP, Haris D.L and Jenkings MM. (1999), Design and construction of Westminster station the Jubilee Line Extension. Prac. insta Ci, Eagrs Chi Engng Jubilee Line Extension, 132: 36-46,Burland LB, Simpson B. and St Joha HD. (1979) “Movements around excavations in London Clay. Bree 7th ECSMFE, Brighton, 1: 13-29. Hisich HLS. Lu ¥.C, and Lin TM, (2008) Eifeets of deiails on the behaviour of cross walla Journal of ‘Gee Fanginecring, 3 (2): 58-60. stung BCH, Nash D.E-T., Cheng KHL, Huang CC. and Hwang, R.NH. (2001). The effectiveness of jet-grout slabs and cross-walls in restricting wall movements in deep excavations. Prac. /4ll Southeast asia Geoucch, Engng Conference, Hong Kong, China Merritt A.S., Menkiti C.0., Harris DLL, Zdravkovie L., Potts DM. and Mair RJ. (2010), 30 finite element sanalysisof a diaphragm wall excavation with sacrificial srosswalls. In Geetccdmical Challenges in Megacities ‘GeoMos2010, Fut, Geotech. Canf-, Moscow, Russa Osborne NH. Ng CC. and Cheah CK. (2009). The benefits of hybrid ground treatment in significantly reducing wall movements: A Singapore case history. In Geotechnical aspoets of sndergrennl consiruetion in ‘soft ground. Ng, Huang & Liu (eds), Taylor & Francis Group, London, 447-453. 513 ‘Ou CY, Hsich PG. and Lin YL 2011) Performance ‘of excavation with crass walls. Jounal of Geotechn cal axe! Geoenviroanienial Engineering, ASCE, 137 (1), 94-104, Ou CY, Lin VIL and Hick PG. (2006). Case records of an excavation with eross walls and buttress walls, Journal of Geo Engineering, U2}: 1X6, ‘Sehanz, T., Vermece, P. Bonnice, P. 1999. Formulation, and Verification of the Hardening-Soil Model. Proc. Plaxis Symposium Beyord 2000 in Computational Ge seclmies, Amsterdam, the Netherlands, 281-296, Stevens A., Corbett B.O. and Stecle AJ, (1977). Barbican Arts Centre: the design and construction of the sub- structure. The Structural Engineer, 55 (11): 473-485.