You are on page 1of 12

Raft analysis and design-some practical

1. A. Hooper, MEng, PhD, DIC, CEng,


Ove Arup Partnership

MIStructE, MICE

The determination of raft thickness normally governed by the is need to restrict the maximum shear stress in the concrete to an acceptable level and is generally undertakenhand as calculation. form a The of construction geographical and locationthe of structure also may influence the choice of raft thickness. Where material costs are high, for example, a cellular raft may be more economical than a monolithic raft. Forbasementfoundationssubjecttoanappreciable net reduction in vertical loading and to the hydrostatic pressures resulting from a high Introduction watertable,somegeneralthickeningoftheraftovertheminimum Experience has shown that raft analysis a subject-area where structural is structural dimension may be appropriate togive an additional margin of designers tend to flounder. In particular, it is not uncommon for project safety against uplift or flotation. For thinner rafts, it may be better to engineers to spend a great deal of time on a raft design and still not thickentheraft locally-e.g. beneath widely spacedcolumns so asto produceanefficientoreconomicsolution.This, in turn,hasoften accommodate the localised shear forces while retaining a desirable degree discouraged the use of rafts as a means of foundation support in favour of raft flexibility between columns. In all cases, of course, there should of other systems which, although frequently less elegant andmore always be adequate space for fixing the steel reinforcement. expensive, are much easier to design. Having established the initial raft thickness, then the main object of the Although general design is covered by the provisions of CP20041, for subsequent interaction analysis is to obtain realistic estimates of the soil example, no detailed guidance is given for raft foundations. The main andraftdisplacements,together with theraftbendingmomentfield. problem is, of course, how to represent the soil strata and how to couple Other useful results, such as raft contact pressures and shear forces, as the soil and structural models. It is one of a class of problems in what is well as the forces in any superstructure includedthe foundation model, in usually referred to as soil-structure interaction, and a considerable body arealsocomputed.Theanalysismayberepeatedforraftsofother of work has been carried out on this important and fascinating subject2J. thicknesses, but this is not normally necessary. Most of the work has centered on developing analytical and numerical It is usuallythelong-termstructuralresponsethat is ofprincipal methods for the structural analysis of raft foundations, in some cases concern, and the corresponding analysis relates to drained soil conditions. augmentedbyfieldobservations; relatively littlehas been saidabout Where structures are founded on thick deposits of clay, the short-term applications to structural design. response may also be of interest and the analysis is based on undrained soil properties. In either case, the computed flexural mode of the raft is Theobjectiveofthispaper is to outlinethebasisoffoundation governed by the classical theory of thin elastic plates, although a reduced interaction analysis, and to illustrate its application to design practice by Youngs modulus is used to take account of creep deformation of the summarisingthepertinentfeaturesassociated with severalfoundation concrete. Uncracked section properties are assumed but,if necessary, the design projects. effect of flexural cracking in the reinforced concrete can be included by Basis of interactive approach is means of a simple iterative procedures.6. In all cases, horizontal slip As structural engineers progressively have moved from away hand allowed to occur between the soil and the base of the raft. It has been calculations to computer-based methods of structural analysis, it shown theoretically that the interfacial shear stresses required to prevent or grillage a profoundeffect in reducingcomputedraftbending frequently occurs that raft foundations are modelled as a plate slipcanhave on linear elastic springs. This method can give good answers in certain moments7, but it is doubtful whether such stresses can be sustained in cases,butitsperformance is problem-dependent.Themethod is not practice. It is therefore safer to neglect these shear stresses in design. sufficiently general practical for purposes and lead grossly can to By including stiffnessboth raft any the of the and significant inaccurate results on the unsafe side. The main reason for that there is this superstructure in the analysis, the computed differential settlements can is no interaction or cross-coupling between the springs. There are also be used directly with some confidence to assess the effect raft of significant problems associated with relating the spring stiffness to soil distortion on architectural finishes or the redistribution of superstructure properties. Many other spring models of varying complexity have been forces. Even with a fairly coarse finite element mesh, these results are to developed over the past decades, but the presence of additional elements likely to be sufficiently accurate for the necessary serviceability checks it even moredifficulttorelate be made. In practice, of course, some of the distortion and possible tilt used to achievecross-couplingmakes model parameters to soil behaviour. will be built out during construction. A far more rational approach to represent the soil by a layered elastic is The method of dealing with computed raft bending moments is not continuum,in which fullaccount is takenoftheinteractiveresponse always so straightforward, although no insurmountable problems have been encountered in practice.This is largely becausetheapproach to through the soil. This approach has been found to give consistent and design is made as simple as possible, i.e. the analysis based on working is reliable results for a wide range of foundation interaction problems, and loads and the required quantities of flexural reinforcement are based on was used in theprojectworksummarised herein4. The soil layers are the computed moments used in conjunction with CPl148. Alternatively, assumed to be horizontal, and the calculation of vertical strain at any given point is based on a Boussinesq stress distribution together with the CP1 lo9 may be used, on the basis of appropriate values for the various partialsafetyfactors.However,there is unlikely to be compatibility soil stiffness parameters at that point. The vertical layer displacements between the two methods in cases where the analysis is non-linear, and correspondingtothesestrainsarethensummedto give thesurface furtherwork is requiredto resolve thematter.Witheithermethod, displacements. This simplified method of dealing with layered stratagives considerable and care experience is needed in the interpretation of satisfactory answers in most practical cases, and is closely analogous to the well-known Steinbrenner method of settlement calculation used for computed moments, especially with regard to the density of the finite many years. element mesh and the nature of the applied loading.

Synopsis This paper describes theway in whichthe principles of soil-structure of interaction have been applied in the analysis and design raftsand other surface foundations. Various techniques of modelling soil the and structure are outlined, as is themethod of couplingthe foundation structure to a layered soil continuum. The general approach to design is also discussed, and summaries are given eight structural design qf projects ranging from simple strip footingsto complex raft foundations.

The StructuralEngineerlVolume 62A/No. 8lAugust1984

233

Reinforcement design is usually based on either the principal bending across the site, although numerical methods have devised for dealing been with inclined layers and other forms of horizontal heterogeneityls. orthogonal moments augmented the by twisting moments, or the momentslo, depending on structural the layout the and pattern of Suitable values of Poissons ratio may be assigned to various types of in computed results. For the particular plate bending finite elements used soil, and the elastic moduli may be estimated from the results of shear It the present work, these moments have been either the element centroid strength or penetration tests carried out during the site investigation. will normally may also be possible to augment these results by means of more elaborate values or the average nodal values, noting that the latter or pressuremeterapparatus. pick up higher peak values in areas of concentrated applied loading. In in situtestsusinglargediameterplates Where the reliability of thesoil moduli is questionable, then the effects of such cases, however, the actual raft bending moments will be lower than It is the theoretical values obtained from a classical elastic analysis because of using upper and lower bounding values may warrant investigation. normally found that, although the maximum computed settlements are the finite thickness of the structural members and because of local yield of directly proportional bounding to these values, sensitivity the of the reinforced concrete in the vicinity of the re-entrant corners. differential settlements - and hence bending moments - to the different Whichever values of computed moment are chosen for design, their soil moduli is very much lower. accuracy and reliability will greatly depend on the mesh density, as the is For caseforanystressanalysisbasedonthefiniteelementmethod. Inputdataforthestructuralmodeltaketheusualformofnodal example,wheretheraft is supportingagrid of columns,there is no geometry, element topology, member properties, applied and loads, possibilityofcomputingareasonablydetailedbendingmoment field where the 2-dimensional elements usedto model the raft, walls, and floor slabs,arebasedonthe classical theory thin of elastic plates. This unless there are a sufficient number of raft finite elements between the modelling can encompass raft foundations of virtually any plan shape columns. Where it is not possible to have such a fine mesh, then the results of the coarse mesh analysis must be followed by supplementary and any distribution of vertical applied load, supported on a layered soil continuum.Padandstripfoundations,as well as groundslabs and calculations to determine the local bending effects in these areas. This procedure is usually necessary for asymmetric foundations, where there pavement structures, can also be handled as they are simply particular are generally too few nodes available to model the structure in detail formsofraftfoundation.Thepresenceofanysuperstructure(walls, because of computer storage restrictions. Work currently in progress to is columns, beams, slabs) also can be taken account; into this is an developandtestsolutionmethodsthatenablelargerproblemsto be important consideration because, in many cases, superstructure the solved at reasonable cost, thereby allowing the use of much finer meshes. strongly influences behaviour. raft Extensive plotting facilities are To some extent, the actual stresses in any raft are bound to depend on essential to the analysis, both in checking input data and in presenting the method and sequence of construction; these effects are difficult to results in an intelligible and digestible manner. assess but probably modify only the detailed pattern of stresses rather Continuation of theanalysisbeyondthelinearelasticrangeallows thanthebroadstructuralresponseoftheraft.Insomecases,other modelling of soil yield at the base of the raft and also separation of the For example, the requirements factors affect the reinforcement detailing. contact surfaces arising from uplift forces. Piled rafts may analysed by be for watertight concrete in a basement or fluid containment structure may modelling the combined raft and pile group by a plain raft located at or near pile base level. Indeed, for large pile groups, interactive analyses influence the size and spacing of steel reinforcementl1.12; similarly, for large concrete pours, where special provisions may be required to containbased on discrete pile elements areimpracticalatpresentbecauseof the shrinkage and temperature f0rces13~14. Where the structural geometry excessive computer requirements. rafts storage Steppedcan be is particularly complex, as with multilevel rafts interconnected by vertical satisfactorily modelled only the step size is small in comparison with the if walls or inclined members of appreciable span, it may be advisable to principal raft dimensions. Where this not the case, the computed results is specifyequal topandbottomreinforcementbasedonthemaximum will require careful post-processing to correct or make allowance for the of whetherthis is shown by to stepped construction. It may be necessary supplement such calculations computedbendingmoment,irrespective analysis to be sagging or hogging. by a plane strain analysisof a section through the step or, in extremis, by The development of more sophisticated design procedures is clearly a fully 3-dimensional analysis. possible, but whether these can be entirely justified in the light of the Because the settlement profile for a raft of zero stiffness be readily may many approximations inherent in the analysis is open to question. It is obtained during the initial part of the analysis, it is a simple matter to essential to recogniseparticular the difficulties uncertainties and carry out a traditional settlement analysis at very low cost. Generally, associatedwithfoundationinteractionanalysis,andtheseshould be differential settlements will be considerably overestimated,a but reflected in a reasonably simple and conservative approach to design. reasonably accurate value of maximum settlement will be obtained for a given soil stiffness profile. This part of the program has been adapted to been found operate on some of the larger desktop computers, and has Outline of computational method particularly useful by those who, at least in the early stages of design, The basic method of interactive raft analysisis simple, and the principal need to carry out only a simple settlement analysis. steps are: -foranygivenraftplanlayout,computeflexibilitymatrix of soil Project summaries continuum What follows is a brief description of eight project raft analyses takenin -invert to give soil stiffness matrix chronological order, compiled originally to provide general guidance on -compute stiffness matrix of structure the scope and capabilities of foundation interaction analysisls. In each -add soil and structure stiffness matrices and solve for displacements case, the results presented are confined to those directly relating to the under given applied loading principal objectives of the analysis. The foundations vary greately in size -from the displacements, compute moments, forces and rotations in and complexity,fromsimplestripfootingstolargeraftfoundations, structure, together with interfacial contact pressures althoughthe use ofplain in situreinforcedconcreteconstruction is -where necessary, continue analysis into non-linear range by means assumed throughout. Some analyses were completed in a matter of days, of iterative procedures, compute and corresponding forces and while others were carried out over a period of many weeks. displacements In every case, the plate bending finite elements used to model the raft Thus by means of the hybrid technique of representing the structure by or three-nodedtriangles,andthe areeitherfour-nodedquadrilaterals or boundary elements, the finite elements the and soil by surface assumed elastic parameters for the concrete are E, = 15 GPa and v = 3-dimensional problem is essentially reduced to a 2-dimensional one to 0.15. No iterative analyses to investigate flexural cracking were carried enable practical problems tobe solved. Currently, a mainframe computer out in these examples, although non-linear analyses were performed in is required for the analysis, although the necessary space and storage two cases to assess the effect of local yield of the soil on foundation requirements are likely to be available on the largest desktop computers behaviour. The vector plots of principal bending moment M l , M 2 ) in ( relate the near future. to the element centroids arrowheads and denote hogging (negative) of little than more Input for elastic model data the soil consist moments. specifying the variation with depth of Youngs modulus and Poissons Theflexuralresponseofaraft is largelygoverned by its relative ratio.It is assumedthatthesoil is horizontallylayeredand is either stiffness, K, defined here as E, (1 - v2)t3/E(l - v,2)B3 for a rectangular isotropic, or transversely isotropic with a vertical axis of symmetry, the (or circular) plan shape, where t and B denote the thickness and breadth latterbeingparticularlyusefulinmodellingheavilyoverconsolidated (or radius), respectively, and E and W denote isotropic the elastic clays where the horizontal stiffness typically is about double the vertical is parameters of the soil. Where the soil taken to be transversely isotropic, stiffness. It is further assumed that the layers are of uniform thickness soil the vertical, horizontal and shear moduli are denotedby E,, Eh, andGvh,

234

The StructuralEngineerlVolume 62A/No. 8IAugust 1984

Paper: Hooper

5 Bays at 3-2m

/ c

7 Bays at m 3.2

yc

(a) Plan

Fig 1. Silo raft .foundation,Douala


l

Joint

3200

3200

14 10

3960
Inonlyonecasehas it beenpossible to followupthefoundation analysisanddesign with settlementmeasurements.Butsomeofthe projects are located in parts of the world where the prospects of obtaining accurate reliable and settlement records a overperiod of years are extremely low. In other cases, the general economic climate has been such thatclientsunderstandablyhavenot beenprepared to payforsuch measurements, and most professional institutions and associations have been short of funds.However,sufficientlygoodcorrelationhasbeen obtained between observed computed the and behaviour of other foundations to justify a reasonable measure of confidence in the present approach, without necessarily succumbing to the utopian expectation that calculations can precisely model practical construction. TABLE I-Effective flange widths .for beams used to model wa1l:floor system L-beam

Dimensions in mm

5 Blinding 0

(b) Section A-A respectively, while v,h and vh denote the Poissons ratios in the principal planes. In the present examples, the soil stiffness parameters relate solely to drained conditions. A straightforward approach is generally adopted to take account the of stiffening effect of the superstructure. For flexible structures with widely spaced columns and few loadbearing walls and partitions, the B superstructurestiffnessmaybeeitherignored in theraftanalysisor included simply adding cumulative by the bending stiffness of the t ) and modelling then this floorslabstothat of the (thickness raft composite structure as a monolithic equivalent raft of thickness T.If T are columnshortening is takenintoaccount,thenlowervaluesof obtained. Any such thickening of the raft naturally reduces differential settlements increases and bending moments, although required the moments in the actual raft are proportionately lower, and are obtained by factoring the computed values for the equivalent raft by ( t / 7 ) 3 . At the opposite extreme, the system of crosswalls and applied loading may be such that a realistic estimate of foundation behaviour can be obtained only by directly modelling the lower part of the superstructure, as well astheraft itself.This approach was adoptedinoneofthe followingexamples,wheretheraft,basementwalls,andgroundfloor slab, were modelled elements as finite to representcellular the substructure. Fortunately, more this elaborate form of structural modelling is needed only in a minority of practical cases. On most occasions, the only significant additional stiffness is provided by loadbearing walls (e.g.corewalls,crosswalls,basementretaining walls) and it is normally sufficient to model these walls and their floor connections by beam elements joined to the raft in the plan positions of the walls. Itis usually necessary to consider only one or two storeys above raft level, and the simple rules devised to establish beam flange widths are given in Table 1. These effective flange widths depend on whether the floorslab is continuous on either side of the or only on one wall side: they also depend on the wall spacing and span (or the distance between points of contraflexure), so that the lowest value given in the table is used in the analysis. When calculating the second moment of area of the beam, the width of the raft section is taken to be the same as the flange width, although strictly the moment of the raft about own neutral axis should its be deducted from the total as it is already included in the plate bending finite element.

T-beam b

12d

B/2

b is the wall thickness d is the thickness of floorslab L is the effective span of wall B is the wall spacing

Silo foundations, Douala, Cameroun Structure. As part of a major extension to an existing brewery, the silo structure comprisescellular foundation a raft (Fig 1) supporting a 25m-high bin steel superstructure malt for storage. bin a Each has capacity of 100 t of malt, and is constructed from a plain steel hopper base and corrugated steel side-panels bolted on to steel stanchions. The raft is founded near the ground surface on sandy clay and strata, at a silt level well above the water table. The structure to be built in two phases, is separated by a period of several years, and the two component rafts are to be butt-jointed.The firstphasewasconstructed duringthe15-month period commencing October 1977. Design ana/ysis. Preliminary hand-calculations were carried out in order toestimatethemosteconomicformofraftconstruction,havingdue regard to the isolated location of the site and the high cost of materials. Cellular construction of the form shown in Fig l(b) turned out to be most suitable, with fill material contained in the raft cells used to support the formwork for the upper concrete slab. The dimensionsof the deeper raft section supporting the elevator tower were fixed because of machinery prerequisites,buttheremainingraftsectioncouldbevariedtosuit structural design requirements. By using known the results a for uniformly loaded rectangular raft on an elastic layer, it was possible to estimate the sensitivity of raft distortions and moments to variations in relative stiffness, hence raft and determine optimum the structural configuration. Hand checks were also made to ensure that shear stresses in the concrete were acceptable.
235

The StructuralEngineerlVolume 62A/No. 8lAugust 1984

Paper: Hooper

1 2 3 4 5 6 7 %

7 r- l 7
---

220kNm 220kNm
A

f
I

220kNm kNm 220

190kN 190kN

B-

- .J

1 (Raft only)

5
I I

. . I
4 1

Fig 2. Load cases .for Douala raft analysis

Fig 4. Results .for machine strip .foundation

0 1 E
CI

2
1

3
I

Grid line 4
I

5
I

6
I

7 8
1

E T V = = = - -+

20

--

60 I

rAverage
S
Envelope :all load cases

Having sized the raft in this way, a full interactive analysis was carried 1 raft,for outforthepurposesofdetaileddesign.Takingthestage example, the mesh (56 nodes, 42 rectangular elements) matched the basic gridlayoutofthe cells, althoughthe silo structurewassubsequently 1). The cellular raft was reduced tothe extended one (Fig by bay equivalent monolithic section, with thicker elements used to model the stifferintakestructure.Theoverallstiffeningeffectofthesteelwork superstructure was estimated to be low, although thesilo bins themselves were considered by the manufacturer to be rather sensitive to differential 10m-thick upper layer ( E = settlement. The soil strata were modelledaby 8 MPa, U = 0-2) followed by a stiffer layer ( E = 24 MPa, U = 0-2) of equal thickness underlain by a rigid base. Seven load cases were considered, so as to cover the worst anticipated combinations of full and empty silo bins (Fig 2). Plotting the results of these elastic analyses along selected orthogonalsectionsenabledthe design envelopes of raft settlement, bending and twisting moment, and shear force, to be determined. These plots were prepared by hand at the posttime of the analysis, but would now be arranged as part of the processing of results by the computer. Results along the axis of symmetry of the raft are shown in Fig 3. The local bending of the 3.2m square groundslabs spanning between the cross-beams was checked by hand, basedon the raft contact pressures obtained in the main analysis.As the raft is moderately stiff relative to the soil ( K = 0 . 7 ) , computed pressures beneath the outer bays are typically double those on the inner bays. In view of possible consolidation of the fill materialwithinthecellularraft,thecorrespondingtopslabwas designed to take its own self-weight in addition to the imposed loading. No discernible effects of differential settlement on the silo structure have been observed. Indeed, the design settlement criteria were probably too severe because, at the end of construction, the bolted connections betweenthe steel binswereleft slack while the silos were filled and emptied. Only after this proof loading, and the consequent bedding-in of the joints and preloading of the soil, were the bolts fully tightened.

Machine .foundations, Popondetta, Papua New Guinea Structure. As part of the civil engineering works for a palm oil factory, which the mill buildingcontains a series ofparallelstripfoundations supportrotatingmachineryonplinths.Eachstrip is 14-6mlong,lm

Fig 3. Results .for beam strip along gridline B-B: Douala raft analysis

236

The Structural EngineerlVolume 62A/No. 8lAugust 1984

raper: nooper

(a) Basement plan

-L
F g 5. Central Bank, Lagos i

wide, and 0 - 5 m thick, increasing to a thickness of 1 3m at the plinths. At thecentreofeachplinth,theverticaldesignload is 90kN,anda horizontal applied load of kN acts at a distance of 1 -45m above the top 80 of the plinth. Trial pits at the site confirmed the presence a substantial of thicknessmedium sand. of dense Construction place took during 1978-1979. Design analysis. The principal design requirement was to obtain realistic estimates of differential settlement along the length of the foundation in order to comply with shaft alignment tolerances. A secondary aim was to provide an independent check on the design bending moments obtained previously by hand calculation. The foundation was modelled by28 rectangular elements (58 nodes) spanning the full width, as no half-symmetry facilities were available at the time the analysis was carried out. Element properties varied to model the stepped cross-section, and each horizontal force was converted to an applied moment, based on alever arm of 2.75m. The magnitude of these horizontal applied forces is small compared to the frictional forces that are capable of being mobilised on the base of the foundation. The soil model was taken as a 50m-thick homogeneous elastic stratum with E = 25 MPa and v = 0.3. Computed values of settlement and longitudinal bending moment from anelasticanalysisareshown in Fig4.Theaverageverticalcontact pressure is only 25 kPa, and the computed differential settlements well are within the limiting value specified by the machinery manufacturer. The computed bending moments are also compatible with those obtained by hand calculation.

E-MPa

200

400

Dense sand

Stiff clay

dense sand
Hard clay

Dense to very

Very dense sand

Central Bank,Lagos, Nigeria Structure. A proposed extension to an existing bank comprises a 32-storey officetowerandthreesmallerbuildings,eachhavingseveralcarpark floors. The 120m-high tower measures60m x 30m in plan, and its design of weight is 120 OOO t. There is a single-storey basement, where the layout the internal walls is designed provide to security vaults and access corridors (Fig 5(a)). Thereis a ground level banking hall, and carparking on the first three floors is linked directly to the corresponding floors on the adjacent buildings. Detailed structural designs were prepared from December 1978 until May 1979, when the project was shelved. The tower i s to be founded on a 2.5m-thick raft connected to a group of400mmdiameter prebored driven steel-cased piles at 1 -2m centres. There are 1200 such piles, extending to 37m below ground level. The soil strata consist of interbedded sand and clay layers, and the ground water level is close to the ground surface. A significant feature of the superstructure which affects the foundation design is the truncation of the longitudinal spine walls at mezzanine level, leading to very high loads being transferred to the columns and gable walls at the edge of the raft, as well as to the central core (Fig 5(d)). Clearly, the substructure forms a vital link this transfer of load asit will in
(b) Layered soil model

(c) Substructure model

The Structural EngineerlVolume 62A/No. 8lAugust,1984

237

Paper: Hooper

ROOF LEVEL 32

78

PLANTROOM 31

30
29 28

27
26

2 5
2L

2 3 2 2
21 20 19 18 17 PLANTROOM 16 15 11 13 12 11 10

Spot values in mm

(a) Raft settlements

0 9
08

05MNmlm

---Internal walls

07
06 05 PLANTROOM OL 03 02 01 MEZZANINE GROUNO FLOOR BASEMENT
0
1

(b) Principal raft bending moments

F-*
I
I

I t
mMPa

+
I

t-+--Q
I
I

t -

_ _ - internal walls

(d) Section A-A

(c) Basement wall membrane stresses Fig 6. Results .from Central Bank .foundation analysis

act as a contiguous multicell box comprising the raft, the 900mm-thick outer the walls,750mm-thick inner walls, and900mm-thick the groundfloor slab. Furthermore, the wall structure above mezzanine level will provide additional restraint to substructure distortion. Design analysis. The principalaimofthefoundationanalysiswasto identify in some detail the mode of load transfer through the substructure were to the soil, and to check that the resulting stresses and displacements commensurate with the reinforced concrete design. Preliminary elastic analyses were based on the 2.5m-thick raft, first unstiffened and then a stiffened by beam elements, and gaveuseful indication of thelikely raft settlements and bending moments. Quarter-symmetry was assumed, and the piled raft was replaced by a plain raft located at pile base level, a simple, yet effective, approach which has proved to be very useful in the analysis of large piled raft foundations. The soil continuum beneath the
238

raft is represented by a series of horizontal elastic layers whose moduli (Fig 5(b))are mainly based on the shear strength and standard penetration test results given in the site investigation report. A Poissons ratio of 0.2 is assumed for all strata. Analyses were also carried out to examine the effect of using different soil moduli, of and dispensing with the assumption of quarter-symmetry. These initial results indicated that the raft moments and shear forces couldbeaccommodated,albeit with some difficulty. Thecomputed differential settlements of the unstiffened raft were unacceptably high, but this wasto be expected in view of the relatively low raft stiffness ( K = 0 . l). The additionbeam of elements substantially reduced these settlements, and also the corresponding raft bending moments, but this formofsubstructuremodelling was felt to beoversimplified in this particular case. Accordingly, the full cellular action of the combined raft,

The StructuralEngineerlVolume 62AINo. 8lAugust 1984

7 Bays at 4.8m 'c /

(a) Basement plan

Fig 7. Office tower, Lagos

300mm outer walls

basement walls and groundfloor slab modelled by replacing the beam was elements by vertical wall elements and horizontal floor elements capable of sustaining both bending and in-plane stresses. The finite element model (104 nodes, 78 plate elements, 38 wall elements) is shown in Fig 5(c), with thegroundfloorslabomittedforclarity,andthe vertical loads were applied at groundfloor level. Results from the subsequent elastic analysis are partly summarised in sagging and hogging raft Fig 6 , which illustrates the complex mode of curvature. Because of the high average level of applied loading (573 kPa excluding raft self-weight), the analysis was continued into the non-linear region to examine the effect of local yield of the soil on the structural response of the foundation. This was achieved by specifying a limiting 0 raft contact pressure of 9 0 kPa, a value based partly on intuition and partly on the anticipated behaviour of the pile group. The result was to decrease the transverse sagging curvature and to increase the longitudinal hogging curvature, especially towards end the each of foundation. However, it was considered that this hogging curvature would be much reduced in practice because of the additional restraining effect of the stiff superstructure walls.

l 2 MNm/m

(b) Raft moments without beam elements

Office to wer, Lagos, Nigeria Structure. One arm of this L-shaped building is a 23-storey tower block, 85m high, with corner core walls and a single-storey basement (Fig 7(a)). Mosttower fourth of above floor the level consists of office accommodation: remaining the structure is a two-storey split-level carpark having plan dimensions of 20m x 30m. The tower block foundation is a 1 5m-thick raft on a regular grid of 450mm diameter bored piles formed within layered sands and clays to a depth of 35m. There are 340 piles, and the spacing is generally 1.7m, decreasing to 1.2m in the lift core area. The founding level of these piles is the same as that for the bank building described previously, whose site is adjacent to that for the present structure. Construction commenced at the end of 1978. Design analysis.The objective of theanalysis was to provide an independent check on the raft settlements and bending moments obtained in the local design office using other methods. The asymmetric raft foundation was (70 modelled by 54 rectangular plate elements nodes), with beam elements used to represent the stiffness of the substructure walls and groundfloor slab. The superstructure consists mainly of waffle floorslabs supported by widely spaced columns, and was not consideredin the analysis to provide any additional flexural restraint to the raft. For calculation purposes, the raft was located at the level of the pile bases, and the layered soil model was the same as that used in the previous analysis of the bank building foundation. Computed values of principal raft bending moment obtained from two elastic analyses, with andwithoutthestiffeningbeamelements,are shown in Fig 7. These results indicate that the effect of the substructure is to change the distribution of raft bending moments without appreciably

changing magnitude the the of maximum moment. maximum The computed moments-including both the element centroid values shown and the also average nodal values-were rather smaller those than obtained in the main design. Ideally, asymmetric foundations of this type need to be modelled by a finer mesh in order to provide a sufficiently detailed moment field for reinforcement design. Largely as a result of the low relative stiffness of the raft (K = 0.04) and the low average applied loading (347 kPa, including raft self-weight), compared with the previous bank tower, the specified limiting contact 0 pressure of9 0kPa was exceeded at very few nodes, enabling the effect of local yield on raft behaviour to be neglected. The computed maximum

The Structural Engineer/Volume 62AINo. 8/August 1984

239

Paper: Hooper

(a) Rail track section

Fig 8. Railway depot, Hona Kong


Dimensions in mm

and minimum settlements were 42mm and 22mm for the unstiffened raft, and 41mm and 24mm for the stiffened raft. As a result of this analysis, the connections between the tower block and the carpark were redesigned of to take account the anticipated differential settlement between the two structures.

Railway depot, Hong Kong Structure. The depot is located at the northern end of the mass transit railway system and provides stabling and maintenance facilities for 30 eight-car trains. The six maintenance sidings, 200m long and spaced 7m apart, consist of continuous reinforced concrete rail track of stepped cross-section (Fig 8(a)). The steel rails are supported on concrete plinths at 1120mm centres. track The is foundedon 3m of compacted fill overlying weathered rock, and was built during 1980. Design analysis. The objective of the soil-structure analysis to obtain was a realistic settlement profile of the track structure, together with design moments and shear forces for reinforcement detailing. This followed a separate analysis by others based on linear elastic spring supports which gave unacceptably high values of track settlement and bending moment. The local flexural mode of this continuous track was examined by the quarter-symmetry modelling of a rectangular portion of track subjected totheverticalaxleloads (240 kNoneachaxle)fromonefour-wheel bogie. The loads were assumed to be static, although some allowance was madeforinertialandbrakingforces.Inthefiniteelementmodel (94 nodes, 95 elements), the element thicknesses across the section correspond to those shown in Fig 8(a), but the presence of the concrete plinths is ignored. The free boundaries of the model are sufficiently remote from the axle positionsas to have little effect on the local bending inheavily the ( E = 2 MPa) loadedareas.Thesoil is modelledasaIm-thicklayer overlying a 7m-thick layer whose elastic modulus varies linearly from8.8 MPa at the top to 25.6 MPa at.therigid base, with v = 0 . 2 throughout. in Figs 8(b) and (c). The Results of theelasticanalysisareshown computed settlements were considered satisfactory, the and bending moment field formed the basis of the reinforced concrete design. Housing development, Grangemouth, Scotland

k'
(b) Computed settlements

[120kN wheel loads1


I

.
Free boundary

0 1 0 kNm/m 0
U

4 Axle loads
(c) Computed principal bending moments

Structure. An extensive development of two-storey brickwork housing occupying a 0 . 7 ha sitein Grangemouth, some 30 km west of Edinburgh, (a) Half-plan was built during 1982-1983, the of start construction having been postponed for 2 years for economic reasons. The site located within the is Joints Dwarf omitted walls alluvial plain of the River Forth, and borehole records show at least 30m of soft clay and silt overlying stiff clay and dense granular material, with the water table about l m below ground level. Substantial variations in wall androofheightarereflected in theunevenfoundationloading, 30-60 kN/m. Several alternative which is mostly within range the foundation schemeswere examined during the initial stages of design, and a cellular raft construction turned out to be the most economic, provided that member sizes could be kept small, as indicated in Fig 9. Design analysis. In view of the thick deposit of compressible soil beneath the raft, the main foundation design problem was to limit differential settlements and hence avoid cracking of the unreinforced brickwork. The Fig 9. Cellular raft, Grungemouth housing groundslab was modelled in the usual way (91 nodes, 67 plate elements),
n

Half-symmetry

240

The Structural EngineeriVolume 62A/No. 8IAugust 1984

raper: nooper

L5p-L --t

225

.:.... ,....

.;--..< __._.

..,* l . ; : , (

....j,....,

~........__ ,::. ...-,..


?.!..I

L . :

...... '. :

:ii:

C'.. .,.>.;:.: i v.- ..:.....

1
II

'Blinding concrete

Dimensions in mm
(b) Section A-A

-Half-symmetry ... --

l i l l i i
Beam zlement Plate
e p n tII

l u l,,,n 1

I U 1
(c) Finite element model

I U I O I

Unloaded raft

--- Loaded raft

assuming half-symmetry of the structure, and the interconnecting system of dwarf walls was represented by a network of 113 beam elements having No the appropriate stiffness properties (Fig 9(c)). allowance was made for the stiffness of the brickwork superstructure, and the imposed loads were applied directly to the groundslab. The soilwas modelled as a 40m-thick heterogeneous elastic layer underlain by a rigid stratum, and the variation of drained modulus with depth was estimated from the results of cone penetration tests carried out during the site investigation. The modulus was takenas2MPatoadepth of6mbelow groundlevel,thence increasing linearly to 8-8 MPa at40m depth. A Poisson's ratio of 0.25 was assumed throughout the layer. Theinitialintention was tosupportthehousingmoduleonfour separate rafts, butt-jointedto each other andwith shear key connections. However, the first two foundation analyses showed that these joints could A be eliminated, leadingto a single continuous cellular raft. third analysis showed that it was possible to replaceprestressed the concrete groundfloor slabs with timber floors without increasing member sizes, thereby reducing costs further simultaneously still and satisfying an architectural preference for this type of flooring. Some of the results of this third elastic analysis are shown in Fig 10. Maximumvalues of angulardistortionareapproximately 1/500 and occur in the end regions of the raft in the longitudinal direction. These values were deemed acceptable, especially as the in-plane stiffness of the brickwork will further reduce raft distortion, The principal moments in the groundslab resolved their were into orthogonal twisting and components design for purposes, additional calculations with hand carried out to check the two-way spanning of the groundslab between walls when subjected to the computed contact pressures. The bending momentscarriedbythedwarfwalls,togetherwiththecorresponding plots of torque and shear force, provided the necessary information for detailing the wall reinforcement. The foregoing interactive analysis demonstrates how raft foundations of supportinglow-rise canprovideapracticalandeconomicmeans brickwork structures poor on ground, suggests they and that are particularly effective in reducing differential settlements. A programme of settlement measurements has been implemented in order to check the validity of the design, and provide useful field data on actual structural to behaviour.

Officetower, Cairo, Egypt


0 1QOvmrn

(a) Raft settlements

Structure. The planned layout of this 23-storey structure, located in the commercial centre of Cairo, includes four levels of shopping, two plantroom floors, and 16 office floors. The 85m-high tower weighs 32 800 t (dead plus reduced live load, excluding raft self-weight), and there is a single-storey basement for carparking. The superstructure consists mostly

\
(b) Principal bending moments in groundslab
200mm&300mm

internal walls

-I. --

05 1

MNm

(c) Bending moments in dwarf walls

1
l

\
1-6m square

400mm outer

l
241

Fig IO. Results of Grangemouth raftanalysrs

Fig 11. Basement plan .for qffice tower, Cairo

The StructuralEngineerlVolume 62A/No. 8/August 1984

Paper: Hooper

of coffered floorslabs spanning between octagonal the corner core, peripheral columns, four internal and large columns, precast with concrete cladding to the facades. Detailed structural design took place during the first half of 1980, but the project was shelved later the same year. The piled raft foundation is approximately 30m square, and the raft itself is 1 - 6 m thick (Fig 11). The 10m-long driven piles penetrate soft clay and silt, so that their enlarged bases are formed in the upper region of a thick layer of dense sand. There are 325 of these 520mm diameter endbearing piles, spaced between 1 - 5m and 1 * 8m apart. The water table is within l m of ground level. Design analysis. The objective the of foundation analysis was to determine the pattern of raft settlements and bending moments resulting from the markedly asymmetric loading and to check the raft shear forces 1 in the vicinity of the heavily loaded internal columns. These -6m-square 30 MN, and preliminary hand columns each carry a working load of calculations indicated that a substantial amount of shear reinforcement 1 would be necessary. But the raft thickness was not increased beyond 6m in order to minimisetheextentofthedewateringoperationsduring construction of the basement. The whole raft was modelled (97 nodes, 87 plate elements) because of the absence of structural symmetry, and beam elements were used to represent the substructure walls: any additional stiffness contributed by the structure above groundfloor level was neglected. The piled raft was modelled as a plain raft located at pile base level on the surface of a 34m-thick sand layer underlain by a rigid base. The soil modulus at the upper surface was taken as 100 MPa, increasing approximately linearly to 530 MPa at 34m depth, and v = 0 - 2 was assumed throughout. Two elastic analyses were carried out, with and without the stiffening beam elements. Some results for the latter case are shown inFig 12, with the beam positions indicated on the bending moment plot. The deformed shape of the raft was similar in both analyses, and although the relative stiffness of the unstiffened raft is low (K = 0*02),the substructure walls appear to add little to the global raft stiffness. The principal difference was in the distribution of raft bending moments, although maximum raft moments were similar in both cases. The analysis of the stiffened raft was continued into the non-linear rangebyspecifyingthatraftcontactpressuresshouldnot exceed the average contact pressure (356 kPa) by more than 50%. The result was to significantly increase raft settlements and bending moments in the corner area opposite to that containing the octagonal core, as shown in Fig 12. Changes of this magnitude were not anticipated prior to theanalysis, despite the rather severe contact pressure limit and the absence of any load shedding along the pile shafts. They probably stem from the lack of a crosswall substructure, interconnected with the outer retaining walls, as to the this undoubtedly would provided have substantial stiffening relatively flexible raft. Had the project not shelved in the final stages been of design, this aspect would have been examined in more detail. The design of the raft reinforcement was not entirely straightforward because the computed bending moment fields did not conform to clearly defined patterns. Finer mesh divisions in the finite element model would have helped in this respect, but the necessary computer storage facilities were unavailable. shortcoming This in the analysis applies most to asymmetricraftssubjectedtoheavyconcentratedloads.Underthese circumstances, cautious a approach is called for in design. Thus reinforcement quantities were based maximum on principal bending moments rather than -their orthogonal and twisting components, with shearlinksalsospecified with agenerousmarginofsafety.Inthis connection, improvements in computer output could usefully be made as

(b) Settlements (non-linear analysis)

r - l - l l

5MNm/m

--- Beam positions

(c) Principal bending moments (elastic analysis)

""#

0
Fig 12. Results.from Cairo raft analysis
242

5MNm/m

--- Beam positions

(d) Principal bending moments (non-linear analysis)

The StructuralEngineedVolume 62AINo. 8IAugust 1984

Paper: Hooper

(a) Finite element mesh

66m
Nodes

5354
O

Elements

48

47
I

4 6
I

45
I

44
I

43
I

31

(b) Typical results

Fig 13. Hammersmith raft analysis


detailed hand-checking for shear can be laborious and time consuming, especially when dealing with irregular mesh configurations.

Office development, London, England Structure. Part of the planned development .of a 2 . 4 ha island site at Hammersmith in central London includes anew bus-rail interchange, an extensiveofficedevelopment,carparkingareas,andapubliclibrary. 10 stories high with There are four office blocks ranging from four to single-storey basements. The irregular plan shape of one of the blocks is

shown in Fig 13(a). Apart from the service core, the superstructure is flexible, being mostly of floor-column construction. Preliminary schemes were prepared for both piled and raft foundations, but the developercontractor strongly favoured the latter, chiefly for economy reasons. The basement is 3mdeepandthe15m-thickraft is founded on gravel overlying London clay. The water table occurs the gravel, although the in porewaterpressureswithintheclaystratumarenotexpectedto be hydrostatic because of underdrainage. Design analysis. Theobjective of theanalysis,carriedoutduringthe preliminary planning stage, wasto examine the feasibility of using a large asymmetric raft foundation to support the variable-height office blocks. Settlementsofthefoundationandthesurroundingground were of particular interest, as were the approximate magnitudes of the raft bending moments. The asymmetric raft mesh (99 nodes, 87 plate elements) shown in Fig 13(a) closely matches the plan shape of one of the four office buildings. Only eight additional beam elements were required to model the stiffness of the 200mm-thick basement walls. Design loads carriedby the columns and core walls were applied as nodal forces and correspond to an average pressureof156 kPa overtheentireraftarea.Thesoilstrata were modelled as transversely isotropic elastic layers having a vertical axis of symmetry, with stiffness properties derived from previous retrospective analyses of buildings in London whose settlements have been measured. (E, = 100 MPa), 60m of Theassumedsuccessionwas 6mofgravel London clay (E, = 25 + 22 MPa, where 2 is the depth (in m) below the top surface of the clay), and 20m of Woolwich and Reading Beds (E, = 200 MPa). The remaining parameters E h = 2. 3E,, G,, = 0 . 66E,, = v,, 0.1, and v,, = - 0 - 15, were used throughout. Computed results along a transverse section of the raft passing through thecoreareshownin Fig 13(b) toillustratethe likely magnitudeof settlements and bending moments. The settlement profiles show that the raft is essentiallyflexiblebecauseofitslargeplanareaandtheopen layout of the superstructure. By extending the mesh to well outside the raft plan area and replacing the applied structural loads by the vector of raft reaction forces, the ground surface settlements in the surrounding area were computed. These settlements were acceptable in the design, as were thebendingmomentsandthecorrespondingquantitiesof steel reinforcement.Fordetailedstructuraldesign,afinermeshwouldbe needed to give amoredetailedpictureofthebendingmomentfield. However, the results of this preliminary analysis clearly demonstrated the feasibility of using a plain raft foundation.

The StructuralEnnineerlVolume 62AlNo. 8lAugust 1984

243

Paper: Hooper Book reviews

Acknowledgements Concluding remarks The development work associated with the project analyses was funded Analytical and numerical studies carried out over the past decade or so, together with field measurements in certain cases, have greatly improved by theOveArupPartnership.Variouscontributions were madeby of several years and by the project our understanding of the interaction between structure and soil. One of computer specialists over a period design staff. theprincipalbenefitstoaccruefromthisworkhas been to putthe structural analysis of rafts and similar foundations on a more rational basis. Used properly and sensibly, foundation interaction analysis can be cost effectivein design practice because of savings in time manpower: and laborious hand-calculations are avoided, as is the use of oversimplified soil models. References Specific observations that have become evident during the course of the C P 2004 Foundations, London, British Standards Institution, 1. project work are: 1972 (a) The results of an interaction analysis often indicate that a plain raft Structure-soil interaction, London, Instn. Struct. Engrs, 1977 2. foundation is to be preferred to some more costly alternative means of Hooper, J. A.: Foundation interaction analysis, Developments 3. support, especially for building structures of up to moderate height. (ed. R. C. Scott), London, Appl. Sci. in Soil Mechanics--l (b) In most practical cases, the stiffening effect of the superstructure on Publn., 1978, ch. 5 , pp149-211 or raft flexure can be adequately modelled increasing the raft thickness by Hooper, J. A., and West, D. J.: Structural analysis of a circular 4. by using beam elements attachedto the raft surface. Only a minority of in raft on yielding soil, Proc. Instn. Civ. Engrs, 75, Pt. 2, June 1983, wall, cases is it necessary to usethemoreelaboratecombinationof ~~205-242 column and slab elements for the superstructure modelling. 5. Hooper, J. A.: The effect of flexural cracking on differential raft (c) For simple floor-column construction, superstructure effects are likely settlements, Proc. Instn. Civ. Engrs, 61, Pt. 2, September 1976, to be small. In contrast, the presence of loadbearing internal crosswalls ~~567-574 and external retaining walls may significantly influence the magnitude Hooper, J. A.: Analysis and design of a large raft foundation in 6. and distribution raft of bending moments. the At same time, the Baghdad, Proc. Instn. Civ. Engrs, 74, Pt. 1,November 1983, additional loads taken by these walls may be considerable. In the case of ~~837-869 basement retaining walls, for example, some thickening may be required 7. Hooper, J. A.: Non-linear analysis of a circular raft on clay, to reduce shear stress to acceptable levels and additional reinforcement Gkotechnique, 33, No. 1, 1983, ppl-20 may be needed in the longitudinal direction specifically to accommodate C P l 14 The structuraluse of reinforced concrete in buildings: Part 8. in-plane forces. 2: Metric units, London, British Standards Institution, 1969 (d) Local yield of the soil can have a significant effect on the flexural 9. C P 110 The structural use q f concrete: Part I : Design, materials response of raft foundations, although the best method of choosing a limiting contact pressure remains unresolved. In contrast, partial and workmanship, London, British Standards Institution, 1972 Wood, R. H.: The reinforcement of slabs in accordance with a separation between raft and soil is rarely encountered in the design of 10. Concrete, 2 , No. 2, 1968, predetermined of field moments, building structures. pp69-76: discussion, 2, No. 8, pp319-321 (e) It is generally beneficial to base as much preliminary design work as BS 5337 Thestructural use q f concrete for retainingaqueous possible on simple settlement analyses (zero raft stiffness) before entering 11. liquids, London, British Standards Institution, 1976 into a full structural raft analysis. Computing costs are low using this Guide to the design qf waterproof basements, London, 12. simplified approach, even for fine meshes, and settlements outside the Construction Industry Research Information and Association, plan area of the foundation may be readily computed. (f) In some cases, particularly for asymmetric foundations, the maximum 1978, guide 5 Large pours for RC structures, Proc. ConcreteSociety Symp, available mesh density has not been fine enough to give a sufficiently 13. Birmingham, 1973 detailed ofbending pattern raft moments for purposes, design Birt, J. C . : Large concrete pours-a survey of current practice, necessitating an appreciable number of supplementary calculations. This 14. London, Construction Industry Research and Information problem can be overcome by significantly increasing the available number of nodes used to model the structure, and various solution methods are Association, 1974, report 49 currently under investigation. 15. Interactive foundations analysis of on Hooper, J . A.: horizontallyvariablestrata, Proc. Insm. Civ. Engrs, 75, Pt. 2 (g) Sensible results forpiled raft foundations incorporating a large group September 1983, pp491-524 of piles have been obtained by themethodofanalysingaplainraft 16. Summary raft of design projects based on Hooper, J . A.: founded at or near pile base level. (h) In view of inherent the difficulties in modelling interactive the interactivefoundationanalysis, 1977-1980, London,OveArup structural response of foundations, irrespective of the apparent degree of Partnership, April 1981 (unpublished report) 17. sophistication of the analysis, the conversion of computed moments and Fraser,A., Wardle,J.: R. and L.Numerical analysis of forces to reinforcement quantities should be approached with caution, rectangular rafts on layered foundations, Gkotechnique, 26, No., and the methods used should be kept as simple as possible. 4, 1976, ~ ~ 6 1 3 - 6 3 0

Book reviews
Introductory structural analysis
C. K. Wang and C . G . Salmon (Englewood Cliffs: Prentice-Hall, 1984) 591 .pp, 28.45. ISBN 0 13 501569 3. At an age of rapid developments in computer technology, when structural engineering students frequently misunderstand fundamental structural principles, it is refreshing to see a new undergraduate text which reaffirms the importance of a proper understanding of statics, geometry, and the basic concepts. All in all, this well-presented and well-illustrated book can be warmly recommended to all engineering libraries. The book concentrates on framed structures;
244

imperial units are used and loadings are based on north American practice. Chapters l to 12 cover the classical methods, including virtual work, moment area, Maxwell and Bettis theorems, Muller-Breslau principle, Castiglianos theorems, slope deflection, moment distribution, and the Williot-Mohr diagram which (despite its educational value) is often not taught these days. Chapters 13 to 20 cover the matrix methods, and here the reviewer would have welcomed further topics on computer application. Also, the stiffness method has been treated in a way that is specific to the type of structure under analysis; it would have been better to present a general technique and then show how this can be applied to different types-of structure. The authors have rightly devoted comparatively little space to the flexibility method; this reduced emphasis will be welcomed by many

teachers and students who are at times understandably worried by thoughts of redundancies and cuts! In the early chapters, it is not difficult to detect a Timoshenko flavour, which may be broadly taken as a sign of good quality. However, even Professor Timoshenko made the occasional mistakes, one of which is repeated in this 1984 book! Thus, on page 115, the authors explicitly qualify the virtual-work statement with the phrase For a given elastic structure, which is of course a mistake (see, for example, The Structural Engineer, vol. 61A, June 1983, ~ 1 7 5 ) However, this and the . earlier criticisms are minor compared with the value of another new book that reminds us of the importance of basic principles. F. K. KONG

Continued on Page 248

The StructuralEngineerlVolume 62A/No.8IAugust 1984

You might also like