A failed raft foundation on soft clays—investigation and analysisby S. BUTTLING> Ii L. A.

WOOD'ntroduction

THE INTERACTION OF soils and struc-tures has become a subject of some inter-est in "recent years and much of the workhas been reviewed in a state of the artreport by the Institution of Structural En-gineers (1977).

The applicability of computational tech-niques to this study has given rise toprograms utilising finite and boundaryelement techniques (Wood, 1977, 1978 and1980) aimed at predictions of settlement,deformed shape and bending momentsand taking account of the relative stiffnessof soil and structural elements.

However, case studies which allow thesemathematical models to be proven arerare, as detailed and reliable data on thesoils, the structure, the loading and set-tlement history are not often available.One such case study is described here-after.

Site descriptionA grain silo, built on a raft foundation

over soft compressible soils, has under-gone unacceptable total and differentialsettlements. The cylindrical silo, 11m indiameter and 13m high to eaves to hold1000 tonnes of grain, was constructedin 1977 on a reinforced concrete raft 13.4min diameter.

In the first two years of loading it hadsettled about 500mm and had adverselyaffected adjacent structures. A detailedsite investigation was therefore carried outto establish the relevant soil properties.

From the depot records, details of load-ing and settlement history were availableand the collected data has all been ana-lysed to compare the theoretical modelswith the field behaviour. This has thenbeen extended to predict future behaviourand analyse the effect of some remedialproposals.

peat layers near the surface and shellsnear the base, followed by medium densesand and silt to about 20-22.5m.

One of the three boreholes penetratedthese granular deposits and proved theunderlying Chalky Boulder Clay to 25m.A later probing 190m to the south eastfound the Boulder Clay at about 17m andit is therefore possible that the silts andsands represent a buried river channelof Pleistocene age.

In addition to SPT and CPT tests andstandard U100 sampling of the stiffer clays,piston sampling of the extremely wet andsoft organic silty clays was attemptedin each hole in order to obtain undistur-bed samples foi laboratory testing. Inmost cases the sample tube penetratedunder its own weight, and at some depthsthe sample fell from the 1 000 x 100mmtubes on extraction. Eighteen piston sam-ples were obtained from the three holesbetween the depths of 1 and 11 metres.

In connection with further proposeddevelopment, and having the benefit ofthe samples and tests from these threeholes, the rest of the site was tested us-ing the Fugro electric friction cone at sixpositions, including one alongside bore-hole 1 for calibration. All showed thepresence of at least one peat band by amarked increase in friction ratio, and mostcould not penetrate the silty sands atthe base of the Fen Deposits below 11m.One which was pushed through confirmedthe stiff clay at about 20m, while anothershowed a thinner layer of silty sand andthe clay at 17m.

N

Eight trial pits were also dug by mac-hine to allow closer examination of thesurface profile and also extraction of300mm diameter samples in short sectionsof steel casing pushed in and then ex-cavated by hand. Fig. 2 shows the gen-eralised borehole profile and a typical elec-tric cone read-out.

Soil dataOedometer tests were carried out on

75mm samples, and these took longer thanusual as limiting rates of settlement werenot achieved in 24 hours. Rowe cell testswere executed on 250mm specimens cutfrom the 300mm samples and includedseveral unloading cycles to relate to theloading history of the silos.

The m„values obtained from these testsare shown in Fig. 2 along with the un-drained cohesion values. Some of thesewere determined triaxially but others weretoo soft for preparation and had to beevaluated using a laboratory shear vane.

From these results the undrained elasticmodulus E„was determined and for thedrained case v'as assumed to have thevalue of 0.1 leading directly to the elasticrelationships E' 1.36En and

E'.98/mv.Eu and E're also shown inFig. 2.

Loading historyThe usage of the silos for the storage

of grain means that they are loaded rela-tively quickly during the period of theharvest, and that this load is reducedslowly during the year according to de-

Geology Ilt soil investigationThe site is in the Fenlands and in terms

of solid geology lies near the junction be-tween the Oxl'ord Clay and the CorallianBeds of the Middle Jurassic Series. How-ever this is overlain by an unknown thick-ness of the East Angl:an Chalky BoulderClay, covered in turn by Recent Fen De-posits. These last, which contribute signi-ficantly. to the fertility of the local soil, con-sist predominantly of soft s Ity clays andpeat layers resulting from several marinetransgressions in the area. They are highlycompressible soils, normally consolidatedand, having a very high water table, aremainly saturated.

The site investigation was carried out byGround Engineering Ltd. who put downthree boreholes with a light cable percus-sion rig at the positions shown in Fig. 1.These revealed that the stratigraphy wasreasonably consistent and that the sur-face layer of silty clay loam was about1m thick. Below this was 9 —10m of softto very soft grey organic silty clay with

rconingfon Phillips rg Associates, Brldport, DorsetDDepartment of Civil Engineering, Queen MaryCollege, London.

40 Ground Engineering

Fig. 1. Site plan

DDi g 2 ggi BZ!

100t 250t

Qgj BH

~k>,++ -ITg 2

BH2I

Dcg

Dutch cone

r ggg

ggg + )TP7

y / JTP8~DC 3 DC 4/~

Legend Borehole Trial pit

TopsoilF irm becomingsoft clay

Peat

~10 6 0 FRkn/m'c

100 200 250~ 0 20 40 0k

(182)i'04(I 011

~ ~

4 60

~ ~

m /MN

1 2 3

~ k ~Undersideof silos

XXX Xg.'i'ery

softbecoming softto firm clay

Loose tomediumdense sand

Dense to verydense sand

-8E

Cl

a —10

c0 -12CI>

00)o —14

ao—16

-18

cu

Legend

~ Borehole 1

E +Assumed ~ E„

E„=1.36E'ln

V

Assumedincompressiblebelow 11 Om

Very stiffbecoming hardclay

-20

-22

+ Borehole 2

~ 250mm dia.trial pit specimens

Fig. 2. Soil properties

mand. The dead weight of the concreteraft and the steel silo are reasonablyknown, and the depot records of grain in

and out over their own weighbridge givea good record of the variation of loadwith time.

A settlement datum was not reliablyestablished as no problems were antici-pated, but the 1000tonne silo was firstloaded with 920 tonnes of barley over atwo week period in August 1977. In Sep-tember the first draw-off was attemptedand it was noted that the central dischargeauger had buckled. Subsequently the load-ing was reduced and levels were takenaround the base.

The variation of load with time, andthe level records, are shown in Fig. 3along with the idealised curve used inanalysis. It can be seen immediately fromthese curves that the response betweenloading and settlement was extremelyrapid, and that the recovery on unloadingwas minimal.

Analysis and predictionsIt was accepted that the lack of an

original datum level for settlement wasa disadvantage which would make an ab-solute correlation between computed andmeasured settlements impossible. How-ever, it was anticipated that the behav-iour of the silo under cyclic loading couldbe studied reliably, and that the influenceof the loaded raft on adjacent structures,which was also of extreme importance,could be examined. A summary of thevarious analyses and the results obtainedfrom these is given below.

Firstly the performance of the 1 000tonnesilo during its initial loading was consider-ed under undrained and then drained con-ditions. For the former E„was derived asabove and v„was given the value 0.5. Forthe latter E'as 0.98/m„and v'as 0.1as stated above. It was found that theratio E„/c„was 55 and that the oedometer

results led to an overestimate of E'y afactor of 2.

As shown in Fig. 2 the average c„ inthe upper soil layers was about 15kPabut the average bearing pressure on theunderside of the silo base was about60kPa. Thus it was clear that local shearfailure would occur in the soil, and thiswas allowed for in the model by placingan upper limit of 90kPa on the contactpressure. The computed radial settlementsfrom these analyses are shown in Fig. 4,where it can be seen that all four profilesare very similar.

The effect of allowing for local shearfailure has been to increase the total set-

i

1978 (1878

10rr l5

l0

0I

8 12Time lmonths)

I

16I

20

200

Undrained

E 400E

cDrained

~ 600-—V)

Fig. 3. Measuredand computed loadsand settlementagainst time

800

Computed

--——Measured

July, 1982 43

tlement for the undrained case by 40mmand for the drained case by 75mm. Bend-ing moments corresponding to these de-flected shapes were also virtually iden-tical with maximum sagging of 150Nm/mat the centre and maximum hogging at170Nm/m at a distance of 1.5m from theedge. The maximum computed edge set-tlement of 535mm agrees well with themeasured value of 500mm since the lat-ter does not include that which had oc-curred before monitoring commenced. Forthis reason it was accepted that thechosen elastic parameters gave reason-able estimates of the stiffness of the soil.

Time-dependent behaviour was then

200E

C

400

Extent of live load and slabRing beam= =

0 1 2 3 4

-----Elastic soil responseI

Contact pressure limited to 90kPa

Undrained

6(m) ~

-100

E 0

1 000t silo~Axis of silos

16 24

Undrained

Drained

32 (m)

Drained

Fig. 4. Computed settlements of 1 000 tonne silo Fig. 5. Computed effect of 1 000t silo on 100t and 250t silos

considered, but problems were experi-enced arriving at suitable c values. The75mm specimens had indicated about 1m'/year but the 250mm specimens gave50m'/year for the soft clay and as highas 133ms/year for the peat. The analysiswas therefore carried out assuming thatdrainage was vertical with free drainageboundaries at the top and bottom of the11.0m thick layer; the initial rise in porewater pressure was equal to the increasein vertical stress and c„was given thevalue 35m'/year throughout the depthduring loading and 4m'/year during un-loading. The soil stiffness was also in-creased during unloading and reloadingcycles.

The computed settlements of the siloedge over the two cycles of loading aregiven in Fig. 3 along with the results ofthe static analysis. The effects of localshear failure are included and the mea-sured settlements plotted for comparisonhave assumed as a datum the computedsettlement for 1$ months.

It is considered that the agreement isacceptable, the main discrepancy being inthe computed rebound which could bebecause, due perhaps to creep and scaleeffects, the laboratory tests do not reflectthe soil response accurately enough inthis respect. It should be noted that ifthe soil stiffness on unloading is increasedby a factor of 100 then a very good cor-

relation can be achieved, but at presentno theoretical justification for such anass mption is known.

he effect of the settlement of the 13.4mdiameter raft on other adjacent structureswas also of interest. There were four 100tonne silos on reinforced concrete ra ts,fand a grain drying building with concretefloor and stanchions on pad footings lo-cated as shown on Fig. 1. A proposal hadalso been made to reduce the rated capacity of the 1 000tonne silo to about 500tonnes and construct an additional 500tonne silo adjacent to the 250tonne siloshown in Fig. 1.

The effects on the 100tonne silos hadnot been monitored in detail, but since theconstruction of the 1 000tonne silo theoverhead conveyor had had to be realignedon two occasions, indicating that signifi-cant movements had occurred. The set-tlements of the smaller silos on their ownwas first computed, and then the addi-tional movement as a result of the 1 000tonne load. This latter is shown in FJg. 5for the settlements along the centre lineof the silos. Radial movements perpendi-cular to this line were also computed.These correlated with the adjustmentswhich had to be made to the equipment.It was also shown that the proposed 500tonne silo on a 14m diameter raft wouldcause the 250tonne silo to tilt towardsit by about 175mm across its diameter.

ConclusionsIt is considered that this case study

demonstrates the potential of a truly three-dimensional analytical model in assess-ing the interaction between a group ofdifferently loaded silos, all on shallowraft foundations. Although the model isapproximate the soil parameters usedwere obtained from standard laboratorytests and the results show reasonableagreement with the measured settlements.

Generally the laboratory m„values havetended to underestimate those achievedin the field. It is always tempting, withthe availability of powerful computationaltechniques, to indulge in over-sophistica-tion of a soil model in relation to thequality and relevance of available soil data.It is felt that the predictive power of themodel employed here is in keeping withthe level of soil information which shouldbe available from a competently designedsite investigation.

ReferencesInstitution of Structural Engineers (1977): Struc-ture-Soil Interaction: A State of the Art Report.Wood, L. A. (1977): "The economic analys'.s ofraft foundations". Int. J. Numerical and AnalyticalMethods in Geomechanics 1, 4, 397-405.Wood, L. A. {1978): "Rafts: a program for theanalysis of soil-structure interaction". Advancesin Engineering Software, 1.Wood, L. A. (1980): Time dependent settlementof structures in New Developments in BoundaryElement Methods, ed. C A, Brebbia CML Publica-tions, Southampton.

8 Trade Literature

Elements of Foundation Design, by G. N.Smith & E. L. Pole. Published by GranadaPublishing Ltd., PO Box 9, Frogmore, St.Albans, Herts AL2 2NF. 222pp, illus, priceE6.95 (paperback), 00.50 (hard'back).

Following Aberfan, Heriot-Watt Univer-sity ran courses for mining engineers toteach them the rudiments of soil mechan-ics. G. N. Smith had to collect a greatdeal of information for these courseswhich he used for his well known "Ele-ments of soil mechanics for civil and min-ing engineers", now in its 4th edition (not

44 Ground Engineering

to be confused with M. J. Smith's "SoilMechanics", also in its 4th edition).

G. N. Smith, having found this success-ful route, continued with other books. Thepresent publication, with co-author E. L.Pole, is one of his latest. It claims to con-centrate on recent developments, mainlyin the field of foundation engineering,avoiding subjects such as earth pressuretheory, sheet piling, etc. already adequate-ly covered in other text books.

The chapter on bearing capacity andsettlement of foundations begins with thefamiliar Terzaghi equations, but quicklymoves to foundations of offshore struc-tures with their problems of size, eccen-tric cyclic loading and scour. On the wayit passes usefully through in-situ testssuch as the SPT, Dutch cone, pressure-meters of Menard as well as the self-boring versions of Baguelin and Wroth,and also plate loading tests.

Piles are dealt with from the fundamen-tals of shaft adhesion and end-bearingusing the work of Meyerhof, Tomlinson,

Vijayvergiya & Focht, Burland, Skempton,Whitaker and Cooke. Consideration isgiven to negative skin friction, tensionpiles, groups of piles and the eccentricand horizontal loading from offshore struc-tures.

A chapter on finite difference numericaltechniques for foundations leads to theproblem of a vertical pile subjected tolateral loading. It is suggested that amodulus of horizontal subgrade reactioncan be obtained from plate loading testsand corrected for depth by a factor. Forthe numerical analysis, the pile is dividedinto a number of equal lengths and equa-tions set up for the load/deffection at eachnode. A knowledge of matrix algebra isrequired (there is a brief explanation inan appendix) and a T.l. programmable 59calculator is recommended for the calcu-lations. The problem of calculating thereaction of the soil on a simple groundbeam supporting two columns is used tointroduce the method.

A chapter on reinforced earth discusses