You are on page 1of 15

Can. Geotech. J. Downloaded from www.nrcresearchpress.

com by University of Western Ontario on 09/14/16


For personal use only.

. Crawford, R.

annin, and C.

Abstract: A section of Highway 97, west of Vernon, B.C., is located over a soft-to-firm,
compressible, silty clay subsoil. In addition to an extensive site investigation, the performance of two
test embankments was observed for 2 years before construction was begun on the highway grade
between them. When the highway fill reached a maximum thickness of about 10 m a failure
occurred. The design was then changed to include berms on either side, but a second failure occurred
when the grade was rebuilt. An undrained back-analysis of the first failure shows the influence of
various variables on the factor of safety and illustrates the difficulty of choosing appropriate strength
values for design when the site has a strong crust overlying a weaker layer and there is potential for
progressive failure. Observations of settlements, pore pressures, and lateral movements in the subsoil
describe the performance of the embankment during construction and reveal the importance of strain
softening as a factor in the failures. Comparisons with a variety of similar failures in Canada,
Scandinavia, and southeast Asia provide some guidance for future construction over similar
materials.
Key words: case history, embankment failure, field observation, pore pressures, stability, strength,
undrained analysis.

RCsumC : Une section de l'autoroute 97 B l'ouest de Vernon, Cofombie Britannique, est situCe sur
un substrat d'argile silteuse, compressible, de consistance molle B ferme. Aprtts une
reconnaissance de site approfondie on a observC le comportement de deux remblais d'essai sur la
partie de l'autoroute situCe entre eux, pendant les deux ans qui ont prCcCdC la construction.
Lorsque le remblai de l'autoroute a atteint une Cpaisseur maximale d'environ 10 m, une rupture
s'est produite. On a alors change, le profil transversal pour inclure des bermes de chaque cBtC
mais une seconde rupture a eu lieu lorsque la fondation a CtC reconstruite. Une analyse B rebours
de la premikre rupture avec une hypothkse de non-drainage illustre l'influence de diverses
variables sur le coefficient de sCcuritC et montre qu'il est difficile d'obtenir des valeurs
appropriCes de rCsistance pour le calcul lorsque le site prCsente une croQte ferme surmontant une
couche plus faible avec possibilite de rupture progressive. Des observations de tassement,
pressions interstitielles et mouvements latCraux dans le sous-sol dCcrivent le comportement du
remblai pendant la construction et rCvklent I'importance du ramollissement comme facteur de
rupture. Des comparaisons avec diverses ruptures de mEme type au Canada, en Scandinavie et en
Asie du sud-est apportent quelques fils directeurs pour les constructions futures sur de tels
matCriaux.
Mots cle's : Ctude de cas, rupture de remblai, observation sur le terrain, pressions interstitielles,
stabilitC, rksistance, analyse en condition non-drainCes.
[Traduit par la redaction]

Introduction
A new roadway intersecting Highway 9 7 from the west at
the north end of Vernon, B.C., is located over a silty clay
subsoil. T h e s i t e w a s first explored b y t h e Ministry of
Transportation and Highways in 1960, but the decision to
Received May 13, 1994. Accepted November 17, 1994.

C.B. Crawford and R.J. Fannin. Department of Civil


Engineering, The University of British Columbia,
Vancouver, BC V6T 124, Canada.
C.B. Kern. Ministry of Transportation and Highways,
Victoria, BC V8W 3E6.

proceed was delayed until 1985; some new site investigations w e r e carried o u t at that time. O n t h e basis of t h e
available soil information it w a s decided t o extend t h e
investigation by constructing t w o test embankments along
the centre line of the proposed road, o n e with wick drains
to a depth of 24 m (West Abutment fill) and o n e without
drains (Waterline fill). T h e site conditions and the perform a n c e of t h e t w o test e m b a n k m e n t s a r e described i n a
recent paper (Crawford e t al. 1992).
The observations of settlements and pore pressures under
the test embankments indicated that the embankment
between them could b e built without the expense of wick
drains. Construction, with piezometric monitoring, began

Can. Geotech. J. 32: 271-284 (1995). Printed in Canada 1 ImprirnC au Canada

Fan. Geotech. J., Vol. 32, 1995

Fig. 1. Site plan showing location of test fills and failure zones.

Can. Geotech. J. Downloaded from www.nrcresearchpress.com by University of Western Ontario on 09/14/16


For personal use only.

LIMIT OF

WATERLINE
TEST FILL

KEY : @ 1960 BORINGS


@ 1985 BORINGS
1990 BORINGS

SCALE

Fig. 2. Profiles of.water contents, Atterberg limits and shear strengths. See text for explanation of
points and lines.

*-

STIFF

-?,

(;

/'A
O

\-I

1985

0 TESTS

960 TESTS

- . .. .

TO
VERY STIFF
CLAY

20

40

60

WATER CONTENT (%)

80

100

0
20
40
60
80

FIEU) VANE SHEAR STRENGTH (kPa)

in December 1988, but when the fill reached a maximum


height of about 10 m on June 30, 1989, a failure occurred
on the north side. Berms were then placed on either side and
construction was begun again in August 1989 at a very
slow rate; however, a larger failure occurred on both sides

in March 1990. Finally the road was completed successfully


using light-weight fill. This paper describes conditions
leading up to the failures, and the design and construction
procedure used to bring the project to a successful completion.

Crawford et al.
Fig. 3. CPTU (cone penetration test with pore pressure) profiles.
STATION 25

+ 60

STATION 28

Can. Geotech. J. Downloaded from www.nrcresearchpress.com by University of Western Ontario on 09/14/16


For personal use only.

>

SAND,
SILT
& CLAY

+ 60

SAND,
SILT
& CLAY

SILTY
CLAY

1i
SILTY
CLAY

TIP RESISTANCE

FRICTION RATIO

PORE PRESSURE TIP RESISTANCE

Site conditions
The final alignment of the road fell between two earlier
proposals, and consequently the site investigations were
located on either side of the as-built centre line. The locations of the two test embankments along the new road
alignment are shown in Fig. 1 in relation to the 1960 bori n g ~to the south (H5, H6, and H7) and the 1985 borings to
the north (TH1, TH2, TH3, and TH28). The groundwater
table (GWT) was observed to be between the surface and
a depth of 2 m.
The natural water contents and Atterberg limits from
the two sets of borings are given in Fig. 2n. Each solid
circle represents an average of 10-15 measurements of
water content and each solid horizontal line represents the
average of three Atterberg limit tests (one from each of
the three test holes) done in 1960. The open circles represent average water contents of the samples from the four
boreholes made in 1985. In September 1990, soil samples
were obtained from four boreholes at stations 27+25 and
27+75, (see Fig. 1). Atterberg limit tests on these samples are shown as broken horizontal lines and average
water contents are shown as solid triangles.
In the 1960 site investigation field vane shear strengths
were measured to depths of more than 30 m, and approximately 9 0 unconfined compression tests were made on
Shelby tube samples obtained to the same depth. Most of
the unconfined tests failed at more than 5 % strain and
even after the specimens that failed at high strains were
eliminated, the remainder still gave much lower strengths
than the field vane values. It was concluded that the laboratory specimens were somewhat disturbed, and for stability

FRICTION RATIO

PORE PRESSURE

analyses more confidence was placed in the field vane


strengths. A few unconfined strength tests were made in
1985 with similar results. There were no tests made to
support an effective stress analysis of stability.
Values for field vane strengths, measured in 1985, are
shown with star symbols in Fig. 2b, and the range of values is given by the horizontal lines. The values to a depth
of 20 m are averages for boreholes TH1, TH2 and TH3,
and below 20 m they are single values from TH28. Mean
values obtained in 1960 are described by the broken curve.
Remoulded vane tests indicated sensitivities of 2-4. Application of the correction proposed by Bjerrum (1972) to
the field vane measurements would reduce their values by
15-20% between depths of 5-15 m. There is no obvious
explanation for the low strengths below elevation 370 m,
but they appear to fall within the range above elevation
370 m, and in any case this layer is below the failure level.
T h e only consolidation tests were made in 1960 on
specimens that had been obtained with 3.5 in. (1 in. =
2.54 cm) diameter Shelby tubes and which appeared to
have suffered considerable disturbance. The interpretation
of these test results indicated an overconsolidation of about
85 kPa below a depth of 9 m and a greater overconsolidation above this depth. Although this is a rather high
degree of overconsolidation at a site where almost normally consolidated conditions may have been expected,
observations on the test embankments reported below indicate an even greater degree of overconsolidation. There
is no geologic evidence that the site has been loaded by
preexisting sediments, so the overconsolidation was attributed to an historic lowering of the GWT to a depth of 9 m
with desiccation above this level.

274

Can. Geotech. J., Vol. 32, 1995

Fig. 4. Longitudinal section through the embankment.

FAILURE

Can. Geotech. J. Downloaded from www.nrcresearchpress.com by University of Western Ontario on 09/14/16


For personal use only.

WEST

FAILURE

- 3 0 JUNE

1 0 MAR. ' 9 0

'89

SURVEY STATION

EAST

Table 1. Imposed loading and soil response.

Station

Original
surface
elevation

Settlement
(m)

Fill
thickness"
(m)

Au,
(kPa)

11

(kPa)

Au
(kPa)

Au,
(kPa)

AL~Au,

35
168
166
182
179
174

84
151
155
193
193
207

0.42
1.11
1.07
0.94
0.93
0.84

Test embankments

Before first failure-June


26+20
27 25
27 + 30
27+75
27 + 80
30+50

390.2
390.8
390.8
390.5
390.5
391.9

0.1
0.4
0.4
0.5
0.5
0.8

4.4
7.9
8.1
9.9
9.9
11.0

89
160
164
201
201
223

Before second failure-March

30, 1989
125
264
262
275
272
- 281

10, 1990

NOTES:
An,, applied stress due to embankment: ri, pore-water pressure measured at elevation 380 m;
Au, assumes ground-water table initially at a depth of 1 m; An,, change in vertical stress at elevation 380 m.
"Includes settlement.
In 1986 before construction of the test embankments,
two cone penetration tests were made along the centre line
at Stations 25+60 and 28+60. The CPTU (cone penetration
test with pore pressure) profiles shown in Fig. 3 indicate a
significant variation between the two locations in the upper
8 m but more uniform conditions below this depth.

Test embankments
The two test embankments provided useful information
on the consolidation properties of the ground and on the performance of wick drains but gave no information on the
ultimate strength of the subsoil. They were built during a

4 week period without any sign of distress: a linear relationship was observed between applied load and settlement under both embankments during fill placement, with
most of the compression occurring in the upper 4 m. The
larger (West Abutment) embankment, with wick drains,
was built to a height of 12.4 m, 1 m higher than the final
roadway elevation. It settled 3 m within the first year, but
pore pressures in the ground dissipated quickly through
the wick drains. The smaller (Waterline) embankment was
built in two sections to a height of 6 and 11.4 m, respectively. The maximum settlement after 1 year was only
0.7 m, but in the absence of wick drains the pore pressure
dissipation was much slower. Based on observed settlements

Crawford et al

Can. Geotech. J. Downloaded from www.nrcresearchpress.com by University of Western Ontario on 09/14/16


For personal use only.

Fig. 5. Photograph of first failure looking east.

Fig. 6. Cross section showing probable failure surface of 30 June 1989 at station 27+80.
I

405

P/

FINISHED PAVEMENT EL. 401 rn


1

I
FAILURE EL.
1

POST FAILURE SURFA

29

ill

SURFACE

DISTANCE FROM CENTRE UNE (m)

of the test embankments, it was concluded that the subsoil below the crust was overconsolidated by about 145 kPa
(Crawford et al. 1992). The implications of these observations will be discussed later in the paper.

Observations during road


construction
There was regular monitoring of surface settlement plates
during construction at locations shown in Fig. 4. Double-tube

pneumatic piezometers placed at elevation 380 m along


the centre line at stations 27+30 and 27+80 gave piezometric levels until the first failure on June 30, 1989, when
they had to be replaced. There were also piezometers in
the failure zone at stations 27+25 and 27+75 that confirmed the readings on the others. The deeper piezometer
tips were all placed at elevation 380 m, and some shallow ones were at elevation 387 m. There was one slope
indicator casing to a depth of 18 m in the failure zone at
station 27+80, 33 m north of the centre line. A longitudinal

,Can Geotech. J., Vol. 32, 1995

276

Fig. 7. Height of fill, settlement, and piezometric surface at centre line during construction at
station 27 +30.

415

JAN.

FEE. IMAR.

APR. $MAY IJUNE 'JULY

AUG. 'SEPT. I OCT. 'NOV. ' D E C .

1 JAN.

FEE. 'MAR. I

1989

1990 c,$
xm

410

Ur:

ZE

PIEZOMETRIC SURFA

Can. Geotech. J. Downloaded from www.nrcresearchpress.com by University of Western Ontario on 09/14/16


For personal use only.

-E
-z

(TIP EL. = 380 rn)

...... i
.........
....;

405

400

395

TIME ZERO : 10 DEC. 1988

..... E z3
....,..' 3
& 7
....
2 Om
.............

: " Y

..........................

..............................:
/

("

HEIGHT OF FILL

.................
'II

-E
I-

5
2
W

0-*

, c P I E Z O M E T R I C SURFACE

......................................
I

ORIGINAL GROUND EL. = 390.8 rn

0.5-

c n .
1.o
0

100

200

300

400

500

TIME IN DAYS

section through the centre line of the roadway in Fig. 4


shows the two test fills, the settlement of the ground, the
elevation of the embankment surfaces, and the extent of
the two failures that occurred on June 30, 1989, and on
March 10, 1990. Actual values of the applied stresses, settlements, and pore pressures at several locations, at the
time of the first failure, are listed in Table 1. The maximum
values under the test embankments and for stations 271-30
and 2 7 + 8 0 at the time of the second failure are also
reported.

First failure (June 30, 1989)


Although construction of the road was started in December
1988, only 1 m of fill was placed before March 1989.
Thereafter fill was added at a rate of about 1.5 m per
month until mid-June, when the rate of placement was
increased. At approximately 11 a.m. on June 30, 1989, a
failure occurred between stations 27+20 and 281-30, as
shown in Fig. 5. An eyewitness reported that a crack
appeared along the centre line and within half a minute
the north half of the road settled 4 m as approximately
3500 m3 of material moved laterally. Subsequent surface
surveys and borehole observations indicated the probable
sliding surface shown in Fig. 6. It is not known where the
failure began, but it was observed that the piezometer at station 27+75 stopped functioning on June 27 and those at stations 27+30 and 271-80 functioned until failure occurred
on June 30.

Loads, settlements, and pore pressures


The loading curve during construction, the settlement, and
the piezometric surface at the centre line for station 27+30
are shown in Fig. 7. The piezometric surface is based on

pore pressures measured at the deeper piezometer tip located


at elevation 380 m. Also shown is the lower piezometric
surface measured at elevation 387 m, station 271-25, indicating much freer drainage from the upper soils. At station 2 7 + 3 0 the grade elevation at failure was 398.5 m,
7.7 m above original ground elevation, but the total thickness of fill was 8.1 m includin the 0.4 m of settlement. The
fill unit weight of 20.3 kN/m resulted in an applied stress
of 164 kPa at the ground surface. Similar curves for station
27+80 are shown in Fig. 8 where a total of 9.9 m of fill
was added resulting in an applied stress at the surface of
201 kPa, and at station 30+50 outside the failure region,
where a total of 11 m of fill was applied as shown in
Fig. 9. Since the GWT is thought to vary between the surface level and a depth of 2 m, the full weight of granular
fill was used for the estimation of the applied load. Stress
increases at several locations under the embankments were
calculated using the Newmark chart given in Lambe and
Whitman (1969), and as shown in Fig. 10, the width of
the embankment has an important influence on them.
From day 89 (March 9th) until day 130, a total of 3.1 m
of fill (applied stress due to embankment, Au, = 63 kPa)
was added at station 27+30 accompanied by an increase in
the piezometric surface of 6.6 m (change in pore-water
pressure, Au = 65 kPa). During the next month there was
little increase in load or piezometric surface, but from
day 165 to 192 an increase of 1.4 m (Au, = 2 8 kPa) in
the embankment height increased the piezometric surface
by 2.4 m (Au = 2 4 kPa). During the next 10 days the
embankment level was increased by 1.7 m (AD, = 35 kPa),
the piezometric surface rose by 4.4 m (Au = 43 kPa), and
failure occurred on day 202. As shown in Fig. 7, the piezometric surface at elevation 387 m had also increased rapidly
during the last 10 days. The settlement measurements had

Crawford et al.

Fig. 8. Height of fill, settlement, and piezometric surface at centre line during construction at
station 27+80.

1 JAN. FEB.[MAR.APR.

410

MAY JUNE

1 JULY

TIME ZERO : 09 DEC., 1988

405

-E
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by University of Western Ontario on 09/14/16
For personal use only.

-z

1989

PIEZOMETRIC SURFACE
(TIP EL. = 380 m)

AUG.
1 SEPT.
1 OCT.
1 NOV.
1 DEC.
I JAN.I rrs. I MAR.I
\i.''.'''

"
m

i'

...........:

I r n

HEIGHT OF FlLL

F 400 4

-I
LU

395

--E
2
L

FINAL PAVEMENT

ORIGINAL GROUND EL. = 390.5 m

.
0.5
U

E :

E :
l.O

EL.

1.5
0

-\
!
I

200 TIME

100

IN DAYS 300

400

500

Fig. 9. Height of fill, settlement, and piezometric surface at centre line during construction at

station 30f50.

,..".
........

............ ........
.........

TlME ZERO : 06 DEC. 1988

PHREATIC SURFACE

..............................TJ'! ......EL.

380 m)

.......................

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.................................

HEIGHT OF FlLL

I
100

200

I
300

I
400

500

TlME IN DAYS

responded normally to the loading until about day 185


when the rate of settlement increased rapidly. At station 27+80, the response to loading was similar (see Fig. 8)
but the increase in rate of settlement began a little earlier,
on day 170.
The excess pore-water pressures just before failure at
a common elevation of 380 m at several locations along

the centre line between the test embankments are given


in Table 1. These values are compared with the increases
in vertical stress at elevation 380 m by the ratio AulAu,
(where Au, is the change in vertical stress); a ratio used by
Hoeg et al. (1969) to express the transfer of stresses to the
pore water as loads are applied. Values at stations 25+50
and 2 8 + 3 5 under the test embankments, and at station

,Can. Geotech J , Vol. 3 2 , 1995

Fig. 10. Stress increases beneath original ground surface


caused by embankment loads.

Fig. 11. Measured lateral deflection at'station 2 7 f 8 0 ,


33 m north of centre line.
DEFLECION (mm)
RIGHT (SOUTH)

Can. Geotech. J. Downloaded from www.nrcresearchpress.com by University of Western Ontario on 09/14/16


For personal use only.

LEFT (NORTH)

30+50, are also provided. The ratios of AulAu suggest


that failure began near station 27+25, although as mentioned earlier the piezometer at station 27+75 ceased to
function 3 days earlier than the others.
The West Abutment test fill had shown that the high
pore pressures could have been relieved through wick drains,
but observations on the smaller test fill had indicated that
the required embankment could be built without this added
expense. There was considerable concern about the high
pore pressures being observed at stations 27 + 30 and 27 80,
but at the same time, some of the concern was being alleviated by observations at station 30+50 (Fig. 9) where a
total of 11 m of fill (Aa = 223 kPa) had been added without any indication of failure. An important difference, however, is that at station 30+50 the last 30% of the load was
added over a period of 80 days whereas at station 27+80 the
final 30% was applied during a 10 day period.

Lateral movements
The lateral movement relative to the bottom of a slope
indicator casing at station 27+80, 33 m north of centre
line is shown in Fig. 11. There was relatively little lateral
movement at the toe of the slope until the major construction work began at the middle of March, but a trend had
developed by mid-April. The last observation, on June 19,
showed an accumulated movement of 6 mm into the slope
at a depth of 4 m and movement of 10 mm away from the
slope at a depth of 12 m. Although the lateral movement is
very small it coincides with the beginning of an increase in
the rate of settlement shown in Fig. 8. It is thought that
lateral movements would have increased significantly as
the rate of settlement increased but unfortunately there are
no measurements after June 19 to confirm it.
Probable reasons for failure
This postfailure assessment suggests that two main factors contributed to the discrepancy between design and
performance. Firstly, the observations on the smaller test fill

8 MAR. ' 8 9

1 8 APR. ' 8 9

were somewhat misleading because, although it was built


to a height of 11.4 m at station 25+50, the stresses in the
ground decreased rapidly with depth as shown in Fig. 10.
Secondly, the test results obtained in 1960 and 1985, as
plotted in Fig. 2, were taken as representative of the soil
conditions between the two lines of boreholes in assessing the stability of the embankment to be built along the
new alignment. But subsequent observations throw some
doubt on this assumption. For example, the CPTU tests
in Fig. 3 show a greater variation across the site, especially in the depth and strength of the crust, than assumed
from the earlier investigations and also the sampling in
1990 revealed a higher plasticity index and therefore a
lower corrected vane shear strength in the vicinity of the
failure. The stronger subsoils toward the east are also confirmed by the larger load at failure at station 27+80 and the
even larger load without failure at station 30+50.

Second failure (March 10, 1990)


An analysis of the failure on June 30, 1989, indicated that
the embankment could be rebuilt with the addition of berms
along both sides. After an interval of 10 weeks, construction was begun again with berms 3 m high and 10 m wide,
except in the failed zone where the berm was 5 m high
and it extended 5 0 m north from the centre line as shown
in Fig. 6. Since the first failure, the piezometric surface
at station 2 7 + 3 0 had lowered about 4.5 m as illustrated
in Fig. 7. Unfortunately, the piezometer at station 27+80
malfunctioned until January 1990. The subsoil conditions

Crawford et al.

Can. Geotech. J. Downloaded from www.nrcresearchpress.com by University of Western Ontario on 09/14/16


For personal use only.

Fig. 12. Photograph showing final reconstruction.

in March 1990 were much more complex than they were


before the first failure and this prevents a meaningful
analysis of the second failure. Probing after the first failure showed that the granular fill had dropped as low as
elevation 386 m, 5 m below the original ground surface.
On October 2 3 the grade at station 2 7 + 3 0 had been
restored to the height at which failure had occurred on
June 30, without raising the piezometric surface. But the
addition of fill in January 1990 caused an immediate porepressure response and an increase in the settlement rate (see
Fig. 7). During the following month the fill was held constant,
and on February 22 the grade was raised a further 0.4 m
when, because of a sudden increase in the settlement rate
and rapidly rising pore pressures, all construction was stopped.
During the night of March 10, following 16 days of rest,
the embankment failed a second time with movement in
both directions from the centreline between stations 26+50
and 28+20. The measured piezometric surfaces before failure at both stations 27+30 and 27+80 (Figs. 7 and 8) were
at the same level as they were just before the first failure,
but the piezometer at station 27+30 appears to have been
constricted after February 26 because no increase in pore
pressure was observed during the next 12 days.

Final embankment design and


construct ion
After the second failure serious consideration was given
to the use of wick drains covering an area 30 m wide and

180 m long, but probing with an auger rig in June 1990


indicated that 6 or 7 m of augering through granular fill
would be required for the insertion of each wick drain and
the cost of the operation was considered to be prohibitive.
There was also a need to adopt a solution that would permit an early opening of the roadway.
Finally it was decided to maintain the berms on both
sides and to use light-weight fill to limit subsurface stress
increases under the road. The first step was to excavate
the existing fill to a level about 1 m above original ground
elevation and to replace it with a 2 m thick layer of polystyrene foam blocks covered with polyethylene sheathing.
Hog fuel (wood waste) was placed on top of the foam to an
elevation 1.5 m below final grade. Granular material was
used to complete the embankment, including the side slopes.
The polystyrene blocks (unit weight 0.25 k ~ / m were
~ )
placed on a compacted sand levelling course as illustrated
in Fig. 12. The hog fuel (unit weight 10 k ~ / m was
~ ) carefully wrapped all around with 10 mil (1 mil = 25.4 pm)
black polyethylene sheathing to prevent aerobic decay and
stop leachate from entering the groundwater. When the
reconstruction began in August 1990, the piezometric surface at station 27+30 was at elevation 401 m and it reached
a maximum of 402.6 m as the fill was raised to elevation
398.3m in October 1990 and then construction was suspended for the winter. From mid-June until mid-July 1991,
the embankment was raised to its final elevation at 399.5 m
but the light-weight fill reduced the applied stress to about

,Can. Geotech. J . , Vol. 32, 1995

105 kPa. The piezometric surface peaked at 402 m, about


5 m below its level at the time of the earlier failures.

Can. Geotech. J. Downloaded from www.nrcresearchpress.com by University of Western Ontario on 09/14/16


For personal use only.

Stability analysis of embankments


The American Society of Civil Engineers (ASCE) Specialty Conference on Performance of Earth and EarthSupported Structures held at Purdue University in 1972
created a great deal of interest in the design and performance of earth embankments. Research activity was further
stimulated and supported in Canada by important questions
concerning the proposed construction for the huge James
Bay hydroelectric project. This resulted in several important papers in the Canadian Geotechnical Journal in the
late 1970s and early 1980s. One of the later papers (Tavenas
and Leroueil 1980) together with the discussions by Ladd
and Foott (1980) and Mesri and Choi (1981) provided an
overview of the state of the art, and in the closing discussion Tavenas and Leroueil (1981) stated that "a rational
stability analysis should account for the full stress-strain
behavior of the foundation up to failure" but "such an
analysis requires an ability to account for such phenomena
as principal stress axis rotation, strain softening, and progressive failure, none of which can be properly described
within our present technology".
Although all of this research advanced the understanding of embankment performance, there was considerable
support for using simple procedures for ordinary engineering works on soft clay foundations. These include the
undrained analysis for "end of construction" stability based
on field vane tests with Bjermm's (1972) correction applied.
It is recognized that this method is empirical and open to
criticism but more sophisticated analyses will be justified
only for critical projects that merit and can support extraordinary investigations.
In a study of two test embankments in Norway, Hoeg
et al. (1969) found that for low loads (up to 113 to 112 of failure load) there was a corresponding linear increase in settlements and in pore pressures. After some critical stress
level was reached there was a nonlinear increase in the rate
of settlement and in the rate of pore-pressure development.
When the final load was added to the embankment, the pore
pressures continued to increase for several days. It was postulated that at the critical stress level there were areas under
the embankment where local shear stresses exceeded the
undrained shear strength leading to progressive failure with
plastic deformation along narrow slip bands. The Vernon
case is very similar to this Norwegian experience.
In a case study by Eide and Holmberg (1972), three
embankment failures in the soft Bangkok clay were reported
in which, based on field vane measurements of shear
strength, the calculated factor of safety was 1.5. The failures
generally occurred several days after loading was finished
but while pore pressures were still increasing. This led to the
construction of a test embankment with a variety of cross
sections and construction methods: one section with a berm
and one without and a third section with bamboo reinforcement. In the test embankment, failures also occurred
some days after last filling and following the appearance
of longitudinal cracks near the centre line. The bamboo
reinforcement increased the stability by preventing lateral

spread at the base of the embankmknt. Undrained stress


analyses of failures based on field vane tests gave safety
factors of about 1.5, in agreement with the earlier failures.
Further vane tests during construction of the highway
showed insignificant gains in strength after 1 year. It was
concluded that the value of the shear strength that could
be mobilized by the Bangkok clay was dependent on the
testing rate of strain and on the orientation of the failure
plane. This observation and others led to the development
of the correction to vane tests proposed by Bjerrum (1972).
Dascal et al. (1972) calculated a factor of safety of 1.6
at failure of a test embankment on a soft, normally consolidated lacustrine clay where the failure occurred slowly
over a half hour period. They concluded that the overestimation of the factor of safety was partially due to the
"rate effect" on shear strength measured by the field vane
and to progressive failure along the failure surface. In a
subsequent investigation of the failure of a test embankment
over a soft sensitive marine clay, Dascal and Tournier
(1975) observed that at a certain stress level there was a
marked increase in the ratio AulAu, and this was attributed to a local yielding phenomenon where shear stresses
exceeded the undrained strength. The authors introduced a
new correction factor for sensitive clays that added the
influence of "progressive failure7' to the "rate effect" and
"anisotropy" factors of Bjerrum.
A comparison of predicted with observed performance
of a test embankment built on soft Malaysian marine clay
was described in a recent paper by Indraratna et al. (1992).
The boundaries of a yield zone approaching the critical
state were determined by a finite element analysis and
found to agree well with the actual failure surface beneath
the embankment. It was concluded that tensile strains at
the base of the embankment contributed to observed longitudinal cracking before failure and that the use of internal reinforcement such as geogrids would probably allow
construction of higher embankments.
In an extensive study of the analysis of embankment
stability during stage construction, Ladd (1991) reviewed
three approaches: (i) total stress analysis (TSA),(ii) effective stress analysis (ESA), and (iii) undrained strength
analysis (USA) and concluded that the USA method is
superior to the others. Peters et al. (1992) noted that the
USA method may be difficult to apply "because it is necessary to constantly update the strength profile to account
for consolidation," and they proposed a fourth method,
the undrained effective stress analysis (UESA), that avoided
this requirement. Another factor that may often need to
be considered in selecting the most appropriate method of
analysis occurs in cases such as described by Eide and
Holmberg (1972) where the expected strength increases
do not occur during early stages of consolidation.
Stermac et al. (1967) described an experience in which
the proposed stage construction of an embankment had to
be abandoned because after 2 years, and about 8% compression of an underlying layer of varved clay, the excess
pore pressure had dissipated by only approximately 25% and
there was a slight decrease in the undrained strength. In
discussing the paper Crawford and Eden (1967) noticed
that when, after 2.5 years, the fill was reduced to one half
its height the pore pressures immediately decreased to less

Crawford et al.
i

Fig. 13. Angular ranges for anisotropic strength.

Can. Geotech. J. Downloaded from www.nrcresearchpress.com by University of Western Ontario on 09/14/16


For personal use only.

+go0

<

0.44
10

20

30

SILTY CLAY SUBLAYER ( P a )

- 45 O
- 90 O

Fig. 14. Undrained stability analyses. (I), isotropic


strength; (A), anisotropic strength.

c,=

SHEAR STRENGTH OF
SILTY CLAY LAYER

than half their previous value and despite the resulting


drop in hydraulic gradient the observed rate of dissipation
increased. This was attributed to a termination of the strain
mechanism that had been generating the excess pore pressures, preventing any gain in the undrained strength. They
also reported an experience with an earth fill founded on a
sensitive marine clay at Kars, Ontario, where, 3.5 years
after loading and about 4% consolidation, there was no
increase in the vane strength of the subsoil.
A similar example comes from a test site near Stockholm where Holtz and Lindskog (1972) reported that
10 years after a test embankment was built, the glacial
clay subsoil had compressed almost 4% and there was no
gain in strength measured with the field vane. There appear
to be few published records where undrained strength is
monitored during and after consolidation in the field, and
yet it is normally assumed that a gain in strength will
occur under such circumstances. Such an assumption ought
to be suspect because natural clay soils are known to have
a structure, formed on a geologic time scale, which breaks
down under strain, causing a loss in strength.

Undrained analysis
The postulated geometry of the first failure of the embankment (see Fig. 6) shows most of the slip circle to pass
more or less horizontally between elevations 382 and
376 m. A circular surface matches well with the physical
evidence from pre- and postfailure observations, and preliminary back-analyses indicated a circular surface yielded
lower factors of safety than a noncircular surface. The
profile of undrained shear strength with depth (Fig. 2)
indicates a relatively strong crust layer between ground

40

Sb

surface and elevation 382 m, overlying a weaker sublayer


in which the undrained strength increases linearly with
depth.
Undrained slope stability analyses were performed to
determine the factor of safety of the slip circle for a range
~ ~ ~Designs
~
1992).
of soil properties, u s i n g - x s (1.S.
Although no strength data exist to support an effective
stress analysis, Simons (1976) has observed that a total
stress analysis can yield a good indication of stability conditions. The angle of shearing resistance of the embankment fill was taken to be 33" or 35", and it was assumed the
water table was at the ground surface. The thickness of
the crust was taken to be 8 or 6 m; the strength of the
silty clay crust was varied between 5 0 and 7 0 kPa, and
the strength of the silty clay sublayer varied between 10 and
5 0 kPa. Most analyses were made for isotropic strength
properties denoted by (I). Since anisotropy may influence
the available shear strength of both normally and overconsolidated soils depending on the intersection angle
between the potential failure surface and the plane of deposition, a series of analyses was made for the case of
anisotropic strength denoted by (A), using the strength
parameter ranges shown in Fig. 13. The anisotropic strengths
were selected based on relationships between field vane, triaxial compression and extension, and direct simple shear
data for similar clays reported by Bjerrum (1972), and
between triaxial and plane strain conditions (Ladd 1991).
The results (Fig. 14) describe the sensitivity of the factor of safety to each of these variables. The strength of
the embankment fill material exerts no significant influence, and the selected range of strength anisotropy some
influence, on the factor of safety. The value of crust thickness becomes more important with increasing difference
between crust and sublayer strength. The factor of safety is
more sensitive to a unit change in strength of the silty
clay sublayer than that of the crust.
According to Fig. 2, the average vane strength of the
crust is about 70 kPa and the Bjerrum correction factor is
one. The soft-to-firm silty clay sublayer has an average
vane strength of approximately 38 kPa and a minimum of

$Can.Geotech J., Vol. 3 2 , 1995

282

Fig. 15. Excess pore pressures at elevation 380 m caused by surface loads at station 27+30.
I

I SECOND FAILURE
MAR. 1 0 ,

Can. Geotech. J. Downloaded from www.nrcresearchpress.com by University of Western Ontario on 09/14/16


For personal use only.

FIRST FAILURE

DAY 283

371

50

100

150

200

250

APPLIED SURFACE PRESSURE (kPa)

26 kPa; the corrected values (Bjerrum 1972) are reduced to


31 and 21 kPa, respectively. In a reevaluation of Bjerrum's
work, Aas et al. (1986) included the influence of overconsolidation ratio and aging to arrive at a revised correction that, in this case, reduces the corrected vane strength
from 38 to 25 kPa. The undrained analysis (Fig. 14) shows
that for a crust strength of 7 0 kPa and a sublayer strength
of 31 kPa the factor of safety is about 1.24 and for 21 kPa
it is about 1.08. Mesri (1975) reworked the Bjerrum correction factors and proposed that, for both young and aged
clays, s, = 0.22 ah irrespective of plasticity. Taking the
overconsolidation stress to be in the range 85-145 kPa,
from laboratory tests and interpretation of field settlement
records, respectively, and u:, = 84 kPa at a depth of 12 m,
y i e l d s u; values from 1 6 9 to 2 2 9 k P a a n d therefore
undrained shear strengths between 37 and 5 0 kPa and factors of safety of 1.35 and 1.56.
Progressive failure occurs in strain softening soils when
the segment of the failure surface below the embankment
that is sheared in compression is strained beyond its peak
strength and loses resistance before the available strength
is fully mobilized along the segment of the failure surface
that is nearly horizontal and that sheared in extension (neither of which exhibit a peak strength). L a d d ( 1 9 9 1 )
describes how anisotropy influences the maximum average resistance mobilized along a potential failure surface;
when strain compatibility is considered, progressive failure
was seen to reduce the available resistance by an order of
10% f o r a normally consolidated clay. In an effort to
account for progressive failure, Dascal and Tournier (1975)
proposed revisions to Bjerrum's correction factor that also
led, in this case, to a corrected average vane shear strength
in the sublayer from 38 to 25 kPa.

General discussion
In an invited lecture to the ASCE Conference on Stability
a n d P e r f o r m a n c e of S l o p e s a n d Embankments-2,
Morgenstern (1992) noted that "while the progress over
the past 25 years is encouraging, much more remains to
be accomplished in order to reduce the landslide devil's
laugh to a smile." Full-scale field cases of failure a r e
important reference points for engineering practice a n d
there are several features of this case that can contribute to
future understanding.
It is seldom justified to carry out sophisticated analyses
of highway embankments, but it is always wise to provide sufficient instrumentation to monitor performance at
difficult sites. Such instrumentation adds little to the overall cost but contributes enormously to the understanding
of the state of stability during and after construction. In
particular, the development of pore pressures during loading and at failure is of considerable interest. It has been
established that when shear stresses approach the ultimate
resistance of a strain-softening clay, the structure of the
clay begins to break down and the load is progressively
transferred to the pore water. This results either in continuing consolidation at constant effective stress (shown
for example by Crawford and Bozozuk 1990) or in a progressive shear failure, as illustrated by several case records
described earlier. The transfer of stresses from the soil
structure to the pore water is indicated by the ratio of
changes in pore pressure to changes in the applied stress,
as shown for station 27+30 in Fig. 15. It should be noted
that the change in pore pressure is measured at elevation
380 m, as in Table 1, while values of applied pressure of the
embankment (Au,) are calculated at the original surface

Can. Geotech. J. Downloaded from www.nrcresearchpress.com by University of Western Ontario on 09/14/16


For personal use only.

Crawford et al

of the ground, and the increment of stress at the piezometer level is believed to be about 5% less than the imposed
surface stress.
The ratio of AulAu, was unity during application of the
first 40% of the load. During the application of the next
40% of load the ratio fell as low as 0.7 in periods of little
or no loading, but from day 192 to day 194 the ratio rose
rapidly to 2.5. This rapid rise occurred as the pressure at the
surface increased from 130 to 136 kPa. This is close to
the value for the overconsolidation stress (145 kPa) inferred
from the observations on the Waterline test fill and may
indicate that stresses in the subsoil had reached the normally consolidated range. From day 194 until failure on
day 202 the average value for AuIAu, was about 0.93.
Although the first failure followed quickly after some substantial increases in load, the second failure was preceded
by long periods at rest that were intended to reduce the
risk of failure.
When reconstruction began in July 1989, the fill was
raised easily to its elevation at failure on June 30 with only
slight increases in pore pressure. On day 409 (January 24,
1990) the fill was raised 0.3 m (Ao, = 6 kPa) and left at rest
until day 437 (4 weeks) when the excess pore pressures
had decreased by 14 kPa. On day 438 a further 0.4 m of fill
was added, and during the following 16 days the excess
pore pressures increased steadily from 139 to 156 kPa, and
the second failure occurred on day 454 (March 10, 1990).
T h e sensitivity of the failure region after January 2 2
(day 407) until failure 47 days later is remarkable. The
first small increment of load resulted in a small increase in
pore pressure followed by a substantial decrease and a
negative value for AulAo, = - 1416 = -2.3, but the second
small increment resulted in Aulhu, = 2218 = 2.7 before
failure. These observations indicate that on day 437 the
factor of safety was clearly greater than unity and increasing under an applied surface stress of 203 kPa, but on
day 438 when the applied stress was increased by 4% to
211 kPa the failure process was initiated. This observation is consistent with the development of a yield zone
approaching the critical state along the potential failure
surface similar to the case described by Indraratna et al.
(1992).
The increasing rate of settlement and the rapid rise in
pore pressures under a constant load indicate that the pore
pressures are being generated by shear strains along the
potential failure surface. The observations of lateral deflection (Fig. 11) support the view that most of the vertical
movements are caused by lateral strain at essentially constant volume. The pore pressure on the failure surface at the
moment of failure will increase to that value which is
required to reduce effective stresses to the failure condition
and it is unlikely that a piezometer would capture the value
even if it was located in that zone.

Conclusions
There was a reasonable expectation based on average soil
properties and on the performance of two test embankments, that the approximately 12 m high road embankment
at Vernon could have been built successfully as planned.
However, a rotational failure occurred in construction when

the fill height was nearly 10 m and a second failure during


initial reconstruction. Undrained stability analyses were
made using the postulated shape of the first failure to
quantify the relative influence of the angle of shearing
resistance of the fill, crust thickness, crust strength, sublayer
strength, and strength anisotropy on the factor of safety.
Results show that the use of Bjerrum's correction for rate
effects and anisotropy, together with a lower bound value
of vane strength in the silty clay sublayer, yield a factor of
safety near unity. A progressive failure mechanism will
act to reduce slightly the available resistance and may
explain the onset of failure.
An interpretation of the record of load, settlement, and
pore pressures indicates that in both cases the failure was
initiated by rapidly increasing pore-water pressures. In the
case of the second failure, the transition from state of
increasing stability to progressive instability when the final
small increment of load was applied is instructive. Clearly,
in situations where there is a serious risk of failure, the
pore pressures must be monitored very carefully and used
to control the maximum load and the rate of loading.
Three indications of impending failure existed: (i) increasing rate of settlement, (ii) increasing pore pressures under
constant load, and (iii) some lateral spreading. The measured lateral spreading was an indicator, but the maximum
lateral movement was only 10 mm at the last reading on
June 19. The increasing rate of settlement was a better
indicator, and the increasing pore pressure was probably
the best warning of all. Unfortunately, once a yield zone
begins to form it may not be possible to prevent failure.

Acknowledgements
The authors are grateful to Mr. E.A. Lund, Chief Highway
Engineer of the Ministry of Transportation and Highways
of the Province of British Columbia, for permission to
publish this paper. The research was assisted by an operating
grant from the Natural Sciences and Engineering Research
Council of Canada.

References
Aas, G., Lacasse, S., Lunne, T., and Hereg, K. 1986. Use
of in situ tests for foundation design on clay. ASCE
Geotechnical Special Publication No. 6, pp. 1-30.
Bjerrum, L. 1972. Embankments on soft ground. ASCE
Proceedings of the Specialty Conference on Performance of Earth and Earth Supported Structures,
Purdue University, Lafayette, Ind., June 11-14,
Vol. 2, pp. 1-54.
Crawford, C.B., and Bozozuk, M. 1990. Thirty years of
secondary consolidation in sensitive marine clay.
Canadian Geotechnical Journal, 27: 3 15-3 19.
Crawford, C.B., and Eden, W.J. 1967. The performance
of an embankment on a deep deposit of varved clay.
Discussion. Canadian Geotechnical Journal, 4: 63-64.
Crawford, C.B., Fannin, R.J., deBoer, L.J., and
Kern, C.B. 1992. Experiences with prefabricated vertical (wick) drains at Vernon, B.C. Canadian Geotechnical Journal, 29: 67-79.
Dascal, O., and Tournier, J-P. 1975. Embankments on

Can. Geotech. J. Downloaded from www.nrcresearchpress.com by University of Western Ontario on 09/14/16


For personal use only.

, C a n . Geotech. J , Vol 32, 1995


soft and sensitive clay foundation. ASCE Journal of
the Geotechnical Engineering Division, lOl(GT3):
297-3 14.
Dascal, O., Tournier, J-P., Tavenas, F., and LaRochelle,
P. 1972. Failure of a test embankment on sensitive
clay. ASCE Proceedings of the Specialty Conference
on Performance of Earth and Earth Supported Structures, Purdue University, Lafayette, Ind., June 11-14,
Vol. 1, pp. 124-159.
Eide, O., and Holmberg, S. 1972. Test fills to failure on
the soft Bangkok clay. ASCE Proceedings of the Specialty Conference on Performance of Earth and Earth
Supported Structures, Purdue University, Lafayette,
Ind., June 11-14, Vol. 1, pp. 151-180.
Hoeg, K., Andersland, O.B., and Rolfsen, E.N. 1969.
Undrained behavior of quick clay under load tests at
Asrum. GCotechnique, 19: 101-1 15.
Holtz, R.D., and Lindskog, G. 1972. Soil movements
below a test embankment. ASCE Proceedings of the
Specialty Conference on Performance of Earth and
Earth Supported Structures, Purdue University,
Layfayette, Ind., June 11-14, Vol. 1. pp. 273-284.
Indraratna, B., Balasubramaniam, A.S., and
Balachandran, S. 1992. Performance of test embankment constructed to failure on soft marine clay.
ASCE Journal of Geotechnical Engineering, 118(1):
12-33.
I.S. Designs (1992). XSTABL, an integrated slope stability
analysis program for personal computers. Interactive
Software Designs, Inc., Moscow, Idaho, U.S.A.
Ladd, C.C. 1991. Stability evaluation during staged construction. ASCE Journal of Geotechnical Engineering,
117(4): 540-615.
Ladd, C.C., and Foott, R. 1980. The behavior of
embankments on clay foundations. Discussion.
Canadian Geotechnical Journal, 17: 454-460.

Lambe, T.W., and Whitman, R.V. 1969. Soil mechanics.


John Wiley & Sons, New York.
Mesri, G. 1975. New design procedures for stability of
soft clays. Discussion. ASCE Journal of the Geotechnical Engineering Division, 101: 409-412.
Mesri, G., and Choi, Y.K. 1981. The behaviour of
embankments on clay foundations: Discussion.
Canadian Geotechnical Journal, 18: 460-462.
Morgenstern, N.R. 1992. The evaluation of slope stability-A 25 year perspective. Proceedings, Specialty
Conference on Stability and Performance of Slopes
and Embankments. American Society of Civil Engineers, Geotechnical Special Publication No. 3 1,
Vol. 1, pp. 1-26.
Peters, J.F., Saucier, C.L., and Oswald, R.H. 1992.
Undrained analysis of slopes based on effective stress
methods. Proceedings, Specialty Conference on Stability and Performance of Slopes and Embankments.
ASCE Geotechnical Special Publication No. 3 1,
Vol. 1, pp. 493-505.
Simons, N.E. 1976. Field studies of the stability of
embankments on clay foundations. In Laurits Bjerrum
Memorial Volume. Edited by N. Janbu, F. Jorstad, and
B. Kjaernsli. Norwegian Geotechnical Institute, Oslo,
Norway, pp. 183-209.
Stermac, A.G., Lo, K.Y., and Barsvary, A.K. 1967. The
performance of an embankment on a deep deposit of
varved clay. Canadian Geotechnical Journal, 4:
45-6 1.
Tavenas, F., and Leroueil, S. 1980. The behavior of
embankments on clay foundations. Canadian Geotechnical Journal, 17: 236-260.
Tavenas, F., and Leroueil, S. 198 1. The behaviour of
embankments on clay foundations: Reply. Canadian
Geotechnical Journal. 18: 462-466.

Can. Geotech. J. Downloaded from www.nrcresearchpress.com by University of Western Ontario on 09/14/16


For personal use only.

This article has been cited by:


1. K. Suganya, P. V. Sivapullaiah. 2016. Role of composition and fabric of Kuttanad clay: a geotechnical perspective. Bulletin of
Engineering Geology and the Environment . [CrossRef]
2. C.T. Gnanendran, Arun Valsangkar, R. Kerry RoweCanadian Case Histories of Embankments on Soft Soils and Stabilization
with Geosynthetics 507-535. [CrossRef]
3. Gianfrancesco Rocchi, Giovanni Vaciago, Maurizio Fontana, Monica Da Prat. 2013. Understanding sampling disturbance and
behaviour of structured clays through constitutive modelling. Soils and Foundations 53, 315-334. [CrossRef]
4. GuangfengQuG. Qu, Sean D.HinchbergerS.D. Hinchberger, K. Y.LoK.Y. Lo. 2009. Case studies of three-dimensional effects on
the behaviour of test embankments. Canadian Geotechnical Journal 46:11, 1356-1370. [Abstract] [Full Text] [PDF] [PDF Plus]
5. Sean D.HinchbergerS.D. Hinchberger, GuangfengQuG. Qu. 2009. Viscoplastic constitutive approach for rate-sensitive structured
clays. Canadian Geotechnical Journal 46:6, 609-626. [Abstract] [Full Text] [PDF] [PDF Plus]
6. Sang-Ok Chung, Ki-Jung Park, Seung-Ho Son. 2008. Calibration and Sensitivity Analysis of the RICEWQ Model. Journal of
The Korean Society of Agricultural Engineers 50, 3-10. [CrossRef]
7. Hag-Gyou Min, Dal-Won Lee. 2008. Prediction of Excess Pore Water Pressure of Reservoir Embankment on Soft Ground.
Journal of The Korean Society of Agricultural Engineers 50, 37-44. [CrossRef]
8. C.T. Gnanendran, A.J. Valsangkar, R. Kerry RoweChapter 27 Case histories of embankments on soft soils and stabilisation with
geosynthetics: Canadian experience 787-812. [CrossRef]
9. R Kerry Rowe, C T Gnanendran, A J Valsangkar, A O Landva. 2001. Performance of a test embankment constructed on an
organic clayey silt deposit. Canadian Geotechnical Journal 38:6, 1283-1296. [Abstract] [PDF] [PDF Plus]

You might also like