Professional Documents
Culture Documents
AND D A M 26 COFFERDAM
By G, Wayne Clough 1 and Thangavelu Kuppusamy, 2
Downloaded from ascelibrary.org by Technische Universitat Munchen on 07/13/15. Copyright ASCE. For personal use only; all rights reserved.
Members, ASCE
ABSTRACT: Replacement of the old Lock and Dam 26 on the Mississippi River
involves construction of one of the largest systems of cellular cofferdams ever
built. Preliminary analyses using conventional theories led to sometimes con-
tradictory recommendations about the cofferdam design. To help resolve out-
standing issues, instrumentation was placed on the first-stage cofferdam, and
finite element procedures were developed for analysis of the cofferdam. Three
different two-dimensional finite element procedures were generated, including
axisymmetric, vertical slice, and generalized plane-strain models. The alterna-
tive approaches allowed prediction of cofferdam response for the most impor-
tant design conditions and at the critical locations. In all models, allowances
were made for nonlinear soil response, slippage on the sheet-pile-soil inter-
faces, staged construction simulation and seepage effects where appropriate.
Further, a means was provided to accommodate the effect of sheet-pile inter-
lock yielding, which led to an orthotropic response of the cellular cofferdam.
Predicted behavior from the finite element models was found to be consistent
with observed trends. The models appear to have considerable potential for
use in future designs.
INTRODUCTION
In 1981, the St. Louis District of the U.S. Army Corps of Engineers
began construction of a $900,000,000 replacement for the old Lock and
Dam 26, located near Alton, Illinois, on the Mississippi River. The plans
called for the use of a three-stage cellular cofferdam to allow construc-
tion of the new lock and dam in the dry. The first stage of the cofferdam
was designed following conventional approaches and, by January, 1984,
was in place and dewatered with dam construction under way. The cof-
ferdam consists of 45 circular cells of 63 ft (19 m) diam and 60 ft (18.2
m) height above the dredgeline, joined by 43 connector cells (see Fig.
1). Design of the cells called for 35 ft (10.6 m) of sheet-pile embedment
into the sandy river bottom, and use of high-strength steel sheet piles
along the common wall between the connector and circular cells. Be-
cause it was concerned by contradictions between existing design tech-
niques for cofferdams and wanted to reduce embedments and eliminate
the need for costly high-strength sheet piles along the common wall, the
Corps of Engineers instituted a large-scale instrumentation program for
the first-stage cofferdam (1,3).
The instrumentation program was designed to monitor key aspects of
the soil-structure interaction process; namely, strains in the exterior and
'Prof, a n d Head, Dept. of Civ. Engrg., Virginia Polytechnic Inst, a n d State
Univ., Blacksburg, Va.
2
Assoc. Prof., Dept. of Civ. Engrg., Virginia Polytechnic Inst, a n d State Univ.,
Blacksburg, Va.
Note.—Discussion open until September 1, 1985. To extend the closing date
one m o n t h , a written request m u s t be filed with t h e ASCE M a n a g e r of Journals.
The manuscript for this p a p e r w a s submitted for review a n d possible publication
on February 6, 1984. This p a p e r is p a r t of t h e Journal of Geotechnical Engineer-
ing, Vol. I l l , N o . 4, April, 1985. ©ASCE, ISSN 0733-9410/85/0004-0521/$01.00.
Paper N o . 19648.
521
FIG. 1.—Aerial View of Stage 1 Lock and Dam No. 26 Replacement Cofferdam and
Mississippi River
interior cell walls, earth pressures acting on the cells, and cell-wall de-
flections. Many of the parameters that were to be measured could not
be predicted by conventional cofferdam analysis procedures. This led to
a program to develop finite element procedures that could be applied to
the Lock and Dam 26 replacement (R) stage-one cofferdam before the
construction was complete. This paper describes the results of that ef-
fort, as well as subsequent studies done to refine the initial methodol-
ogy.
Although the cellular cofferdam is a three-dimensional problem, lim-
itations in time and funding dictated that some form of two-dimensional
finite element approach be taken. Following discussions with the Corps
of Engineers and its consultants, it was decided to use three different
two-dimensional models, each of which could best treat a particular as-
pect of cofferdam behavior. Utilization of the results from the various
analyses would allow an insight to be developed into the three-dimen-
sional response of the system. A consistent philosophy was used to de-
velop all the finite element codes, in that a capability was provided to
handle the sequential nature of the cofferdam loading, nonlinear soil
behavior, and imperfections in the sheet-pile system response as a result
of interlock yielding and possible initial out-of-round sheet-pile aline-
ment.
The bases for the finite element models are described, and the results
522
them. While the methods presented in the paper were generated spe-
cifically for the Lock and Dam 26 (R) first-stage cofferdam, they are gen-
eral, and can be used to advantage in the design or analysis of any cof-
ferdam problem. Presently analyses are being made for the second stage
of the Lock and Dam 26 problem, scheduled for construction in 1985.
BACKGROUND
The foundation for the Lock and Dam 26 (R) cofferdam consists of a
series of dense-to-medium dense sand layers approximately 70 ft (21.2
m) thick, overlying rock. Although minor distinctions can be made be-
tween the sand layers, for practical purposes they have very similar
523
In this study, three different finite element models were used, two of
which were developed specifically for this effort. The models are re-
ferred to as axisymmetric, vertical slice and generalized plane strain, and
they are described briefly in Table 1, where the advantages and disad-
vantages of each are given. Details are provided in Ref. 5. The finite
element meshes for the axisymmetric, vertical slice and generalized plane-
strain models are given in Figs. 2, 3 and 4, respectively. These are useful
in visualizing the respective differences between the models.
The axisymmetric model provides a complete description of the pro-
cess of filling an isolated cell. In contrast to the previous work of Kit-
tisatra (8), allowances were made in this effort for sequential staging of
the fill load, nonlinear soil behavior, slip on the interface between the
soil and sheet piles, and yielding in the interlocks of the sheet piles.
Although only clamshell filling techniques were used at Lock and Dam
26, analyses were performed for both chamshell and dredge-filling cases.
Dredge filling differs from clamshell filling in that the soil is loaded into
the cell full of water, and only after a period of time is the water drained
out of the fill. In the dredge-fill simulation, it was assumed that the fill
is placed, entirely saturated, to the top elevation, followed by a draining
524
-Interface Elements
- $ - E I 370
of main and arc cells under a uniform filling condition. As shown in the
mesh for this case (Fig. 3), only a quarter of the main cell and arc cell
need be represented, since there is symmetry around both center lines.
The generalized plane-strain model provides data on interlock tensions
and cell deformations for the main and arc cells, the common wall, and
the critical wye section where the two cells are joined. The model is
unable to account for the lateral support of the foundation soils, and
thus is applicable only for the upper two-thirds or so of the cells where
the dredgeline effects have little influence. However, this is not a major
drawback in terms of predicting maximum interlock tensions, since these
are usually observed at about the level of the lower one-third point (9).
Also, if anything, the generalized plane-strain model provides a con-
servative answer, and the degree of conservatism can be closely checked
by the results of the axisymmetric analysis where dredgeline effects are
accommodated.
The program used for the analyses is SOIL-STRUCT, a code developed
by the senior writer and his co-workers over the past decade. For the
vertical slice model, no modifications were required. Special adaptations
were needed for the axisymmetric and generalized plane-strain models.
Verification studies were performed and are presented in Ref. (5).
SOIL RESPONSE MODEL
The soil model used in all cases was a nonlinear elastic version along
the lines of that described by Duncan, et al. (4). In this approach, the
shear and bulk modulus values of the soil are varied to accommodate
the changes in shear and confining stresses, which occur during the step-
by-step simulation of the construction loading process. The modulus of
each element in the mesh is modified after each step of loading, and the
entire stiffness matric reformulated. Parameters required for the soil model
were obtained based on the substantial database provided by the Corps
of Engineers (5), and interpreted using the procedures described by
Duncan, et al. (4). Table 2 presents the values used for each of the foun-
dation layers and the cell fill.
PS X-32./
PS-32
r
FIG. 5.—Interlock Pull Test Results From Ref. 13 (1 in. = 2.54 cm; 1 kip/in. =
17.5 kN/m)
the fill in a series of 10 layers, with allowance for differences in fill unit
weight above and below water level.
Construction loading in the generalized plane-strain model involves
applying the vertical load above the assumed elevation of the plane in
a series of steps, the number of which vary depending upon the amount
of load to be applied. The actual number of steps is chosen to insure
convergence for the nonlinear soil model.
The vertical slice model presents the largest problems in construction
sequence simulation because it allows a reasonable representation of the
entire process of construction. This involves cell filling, berm placement,
staged removal of the interior water, staged institution of the dewatering
system and mobilization of seepage patterns, opening of the excavation
for the dam foundation and, finally, application of flood stage loadings.
Details are given in Ref. (5).
Cell //n
420
Fill -s //i
410 •At-Rest M l / It
Pressure// / *
z
I "00 - ill
Active Earth
Pressure/"""-— ^
$ 390
c
J 380
>
' ^7 -
UJ 370
360
350
340
530
• 50 ."is .*:
«
Downloaded from ascelibrary.org by Technische Universitat Munchen on 07/13/15. Copyright ASCE. For personal use only; all rights reserved.
.1,6 •." 5
1*3 v ? :
« .1)2 .iiO . 3 !
• e
.1.2 .40
* — 4- El 402
•^ -^8 i " J!
Note-
8 6 1
Lateral earth pressure coefficient
;39 •J *> bS " •;:
is predicted lateral stress divided
by effective overburden pressure
.38 •F -.35 -.32 -.«
-^•El 370
.29 .29 .29 . 2 9 . 2 ! 7.^0 it.to 2.58 1.56 - J I
a « «
• « « •
.30 JO ..30 .,32 J i 2..70 2,37 I ..63
II
^30 JO ..31 ..32 j : 1..72 ...57 1J6
1 v
420 "* /
\ h
410 L \ / ^PileJJ
Downloaded from ascelibrary.org by Technische Universitat Munchen on 07/13/15. Copyright ASCE. For personal use only; all rights reserved.
^
i U \ V \ V
400
1 l Pile KK--^ ^ v
390 \
ki ^^?E-Ratio=I.O
. < \1x
\ i ^^^"^•*0 Js
380
~E-Ratio=0.03
A,
370
r' ..
^-E-Ratio=O.IO
intfa/Ji,
¥
360
INBOARD K K
350
1
&
CSV) - J-J
340
OUTBOARD
I '
Legend
o Strain Gage Data
from November I3.I98I
• Predicted Value
COD
Strain Gage Locations
532
generated in the upper passive zone nearest the sheet piles. The coef-
ficients in the foundation are all higher outside the sheeting than inside
the sheeting. All these trends are consistent with the relative movements
induced between the sheeting and the soil in the fill and foundation.
Predicted versus Observed Behavior.—From the large differences in
the predicted results shown in Fig. 6 as a result of the E-ratio, it is ob-
viously important to know which best represents the actual behavior.
Simply on the basis of past experience, it is possible to rule out the case
of an E-ratio of 1. Observations of a number of cofferdams have shown
that much larger sheeting deformations occur than those predicted with
an E-ratio of 1 (9,15).
Observed sheeting deflections and measured interlock forces for the
Stage 1 Lock and Dam 26 (R) cofferdam at the end of filling confirm this
(Figs. 8 and 9). [Note that in Fig. 8, data for piles / - / and K-K were
obtained November 17, 1981, with cell 33 filled, a neutral water head,
and the connecting are not in. Also note that the predicted deflection is
based on results of axisymmetric analysis.] While there is scatter in the
field data, it is clear that the E-ratio of 1 leads to prediction of displace-
ments that are too small and interlock forces that are too large. The data
observed seem to further suggest that the case of an E-ratio of 0.03 yields
results more consistent with the average behavior.
Clamshell Filling versus Dredge Filling.—The Lock and Dam 26 (R)
Stage 1 cofferdam was filled by clamshell procedures. However, many
cofferdams are filled by dredging. Axisymmetric analyses were per-
formed to compare the effects of those two options. Dredge filling differs
from clamshell filling in that the soil is loaded into the cell full of water,
and only after a period of time is the water drained out of the fill. In
the dredge-fill simulation, it was assumed that the fill is entirely satu-
rated and placed to the top elevation, that then the water level is drained
back to the river elevation. The results of the analysis showed dredge
filling to produce a more severe loading on the cell than clamshell filling,
with the most critical condition existing after topping out of the cell fill.
With dredged fill, the water pressure before draining exists in full mea-
sure from top to bottom of the cell, along with effective earth pressures.
0 © I 1 «-'° ^0
9
©il
LEGEND- 0W '•8(i3)l
Z.9 - E-Ratio = 0.03
0 -E-Ratio = 1.0
NOTES- s@4'
1. Interlock force expressed
in kips per inch
2. Computed at El 390 @©"-2
§3@4-
420
4I0
400 t-*-r+
390 o* —9—&
380 +t H-
E-Ratio=I.O
-E-Ratio=O.IO
370
-E-Ratio=0.03
oo o
360 XEtO
350
340
330
Legend
o Strain Gage Data
from November 23,198I
H Range of Predicted
QD
Strain Gage Locations
Interlock Force
The predicted behavior from the vertical slice model for the case of
cell filling agreed remarkably well with that found for the axisymmetric
model when comparable E-ratios were considered. Of course, since only
the vertical slice model could simulate stages beyond the filling of the
cells, no other comparisons between models could be made.
Figs. 12 and 13 present predicted sheeting movements and interlock
forces from the vertical slice analysis for various key stages of loading.
It is apparent that the major sheeting deflections are caused by cell filling
and flood-stage loadings. The flood condition causes the entire cell unit
to tilt inwards towards the dewatered area. The interlock forces shown
in Fig. 13 represent average values between those for the extreme in-
board and outboard sheets. The radial forces in this type of analysis can
only be calculated from the results predicted for the springs that connect
the inboard and outboard sheets, and the springs average the behavior
535
Inboard
FIG. 12.—Deflection of Sheet Piles for Filling, Berm Placement, Dewatering and
Flood Stage, Vertical Slice Analysis (1 in. = 2.54 cm)
Interlock Force,kips/in.
I 2 3 4
430
420
4I0 -
400
390
.£ 380
370
360
350 -
340
FIG. 13.—-Interlock Forces for Filling, Berm Placement, Dewatering, and Flood Stage,
Vertical Slice Analysis (1 ft = 0.305 m; 1 kip/in. = 17.5 kN/m)
536
of the two extreme positions: The stage which produces the maximum
predicted interlock forces in cell filling, while other construction stages
prior to flooding have relatively minor effects. Flooding causes a strong
reduction in predicted interlock forces above the dredgeline, but a small
increase below dredge line. In the end, however, the maximum pre-
dicted interlock forces are still those generated during filling.
In Table 3, the induced changes in cell deflections and interlock forces
in differing key load stages after cell filling, as predicted and observed,
are presented. The trends of the results are identical, i.e., where the
vertical slice analysis predicts a decrease or increase, a decrease or in-
crease is observed in the field data. The magnitudes of interlock forces
and the increments in deflection are less well-predicted. However, the
agreement between predicted and observed values is reasonably good,
and could be refined by adjusting the value of the E-ratio. Also, the
flood-stage modeling procedures can be improved in future work.
Legend:
proaches for the end of cell filling for Lock and Dam 26 (R) conditions.
The methods of Schroeder and Maitland (12) and the Corps of Engineers
yield distributions consistent with the trends of the finite element re-
sults, although the former tends to give magnitudes similar to the case
of E-ratio = 0.1, while the latter yields magnitudes similar to that of E-
ratio = 0.03. Other design methods provide very conservative estimates
of interlock forces, particularly below elevation 380.
The second question, concerning ease of use, may be answered af-
firmatively for the axisymmetric and horizontal slice models. The meshes
for these models can be developed relatively easily by automated com-
puter routines, and the loading sequence data can also be generated au-
tomatically. The user should be primarily concerned with selecting the
proper soil and structure parameters. The vertical slice analysis is much
more of a problem for the user, since the mesh is complex and the data
preparation is challenging. However, recent developments in computer
graphics technology may ultimately reduce user problems associated with
this model to an acceptable level.
538
The results of the finite element model development and analyses lead
to the following conclusions:
Downloaded from ascelibrary.org by Technische Universitat Munchen on 07/13/15. Copyright ASCE. For personal use only; all rights reserved.
ACKNOWLEDGMENTS
The studies described in this paper were sponsored by the St. Louis
District of the U.S. Army Corps of Engineers, and work was done under
the auspices of Shannon and Wilson, Inc., Geotechnical Consultants.
Mr. Richard Freuh of Shannon and Wilson was a significant contributor
to discussions of the methods and interpretation of the results. Edward
Demsky and Thomas Mudd of the Corps of Engineers provided valuable
comments and criticisms. Dr. Mark P. Rossow, a structural consultant
to the Corps of Engineers, proposed the generalized plane-strain model
and aided in its technical implementation. Dr. Lee Schoreder, an instru-
mentation consultant on the project, also made contributions through
discussions with the writers. Dr. Richard J. Finno and Dr. Paul R. John-
ston worked with the senior writer in the first phases of the studies at
Stanford University. Virginia Polytechnic Institute and State University
students who aided in the latter phases include T. Shirasuna, S. An-
navarapu, R. Mosher and S. Campbell. To all these individuals, the writ-
ers express appreciation and thanks.
APPENDIX I.—REFERENCES
1. "Analysis of Instrumentation Data," Interim Report Task 3.1, Lock and Dam
539
APPENDIX II.—NOTATION
E, = tangent modulus;
K0 = lateral earth pressure coefficient;
Km = modulus constant;
n = modulus exponent;
=
pa atmospheric pressure;
540
= friction angle.
541