Beruflich Dokumente
Kultur Dokumente
849-855, 1997
Q 1997 Civil-Comp Ltd and Elsevier Science Ltd. All rights reserved
Pergamon Printed in Great Britain
PII: SO0457949(96)00420-g GM%-7949/97 $17.00 + 0.00
Abstract-The present paper describes numerical techniques for the computational simulation of the
membrane method for shape determination of arch dams and subsequent stress analysis of the dam shell.
The method is based on the nonlinear computation of membrane deformation using the finite-element
method. Techniques are presented for the automatic mesh generation for the membrane and for the shell
problems. The presented software package allows the computational analysis of all steps from the shape
determination to the stress analysis. Examples are included. 0 1997 Civil-Comp Ltd and Elsevier
Science Ltd.
1. THE IMPORTANCE OF THE SHAPE in the slopes of the valley, which is loaded by water
pressure (simulating the hydrostatic load) and by
Arch dams are the only dams which resist water forces acting vertically upwards, which correspond to
loading by the compressive strength of the concrete the self-weight of the dam. The deformed membrane
from which they are built. In all other types of dams has the form of a funicular of the loads, since it has
the self-weight and the shear strength play an no bending stiffness.
important role. If all forces are reversed, the equilibrium is
If adequate bulk is given to the arch dam, it is preserved and the tensile stresses become compres-
theoretically possible that the whole concrete of the sive. On a first view this would mean that in a dam
dam works under the same stress (for a given shell with that form, it is possible to have only
loading), which describes the most rational utilization compressive stresses which do not vary in thickness.
of the material. This means that there are no bending However, this is not completely true, since, as the
moments installed in the shell, since otherwise the shell has bending stiffness, its deformation induces
stress would vary in its thickness. However, this can bending moments, especially in the abutments.
only happen if the arch dam has the shape of a However, the experience collected using this
funicular of the loading (hydrostatic load and dead method and testing the obtained forms shows that
weight) and the thickness varies according to the forms may be obtained which are generally free from
normal forces in the shell. Now, it must be pointed tensile stresses [ 1,2].
out that contrary to a load distributed on a line, for The experimental membrane method has the
a load distributed. on a surface-like the water disadvantage of being very time and cost expensive,
pressure in the upstream fact of a dam-there is an so that, from the results obtained by this method,
infinity of funicular forms. empirical rules were developed, in order to avoid the
The search of a shape which satisfies this condition experimental shape determination.
in the most demanding loading case-full reservoir-
and also behaves well in other loading conditions, 2.2. Computational simulation of the membrane
namely an empty reservoir, is of a crucial importance method
for the design of an economical and safe dam. After the appearance of the computer and the
development of numerical tools for structural
analysis suitable to be used on it-like the
2. GENERAL CONSIDERATIONS ABOUT THE
MEMBRANE METHOD finite-element method-it became possible not only
to calculate stresses and displacements, but also to get
2.1. Physical description the membrane shape by numerical means [3]. This
The membrane method has become the most shape depends on the form of the membrane fixation,
successful way to model the form of an arch dam. the thickness and the rheological behaviour of the
However, before the computer was available, this membrane material.
technique had to be used experimentally. The test The numerical determination of the membrane
consists of a rubber membrane initially flat, fixed in deformation is much faster and cheaper than the
a frame with the form of the abutments of the dam experimental one and, as opposed to the experimental
FE-analysis
Non-linear
for the
FE-analysis
calculation of
Automatic mesh for the iterative
stresses and
generation for calculation of
displacements
the membrane the membrane
in the shell
equilibrium
(commercial
configurations
program)
Fig. 1. Software modules for the numerical shape determination and stress analysis of arch dams.
12311 ypy )
2 z
4 b)
Fig. 2. (a) Definition of the valley form and macro elements for FE mesh generation; (b) angle G(and /3
for definition of the initial membrane configuration.
tests, in the computational simulation it is very easy and 14, respectively. These nineteen points define
to vary the problem parameters, especially the the eight macro elements shown in Fig. 2a. Each
constitutive law of the membrane material. It is even macro element is then divided in four or sixteen
possible to use a fictitious material behaviour, which quadrilateral membrane elements with eight or nine
does not exist in real materials. This gives the nodes.
numerical simulation many more possibilities than Once defined, the nodal point coordinates in the
the experimental way offers. xy-plane the coordinates z are calculated. In the first
stage it is assumed that the membrane lies in an
inclined plane, defined by a rotation CIof the vertical
3. THE PROGRAM plane xy around the z-axis (Fig. 2b). On performing
3.1. Outline a projection in the z-direction, the cylindrical surface
defined by angle p is then obtained (Fig. 2b). The
The computer program developed in the Depart-
objective of this procedure is to achieve a
ment of Civil Engineering of the University of
double-curvature shell, even when the valley slopes
Coimbra for the automatic shape determination and
have a constant inclination. A perspective of the
structural analysis of arch dams consists of the
membrane start configuration in a trapezoidal valley
modules represented in Fig. 1.
is shown in Fig. 3.
The thickness must be entered as data to the FE
3.2. Starting parameters and mesh generation membrane analysis, since the corresponding program
The form of the valley is defined by the coordinates computes the dead weight from a given thickness
of eleven points distributed as shown in Fig. 2. distribution. In the standard version of the program
The middle point, 6, is the origin of the reference the thickness is assumed to have a linear variation
system. Points 12 to 15 are uniformly distributed between top and bottom. However, it is simple to
on the line between points 1 and 11. The position program other laws of variation [4]. The membrane
of points 16 to 19 may be on any point on the mesh generation program interpolates this data and
lines between 3 and 13, 5 and 13, 7 and 14 and 9 puts it in a form suitable to be used in the FE
analysis.
Fig. 3. Membrane start configuration in a trapezoidal valley. Fig. 4. Exponential material law.
3(-l, 1)
W, 1)
W,l)
Computation of membranes of arch dams 851
u I v
1(-l, -1) L-$-q-7(1, -1)
7 Fig. 7. Mesh of 9-node membrane elements for a
symmetrical valley.
Fig. 5. Isoparametric membrane element of 9 nodes.
ay ay
--
membrane mesh by calculating the coordinates of the all, aq2 ay aZ ay aZ
supplementary nodes. In order to obtain this, the nx = aZ aZ =as,all,-Jy&
--
direction cosine of the normal to the middle surface aq, aq,
of the membrane must be calculated in each node.
-aZ -az
This is done by using the form functions of the
membrane elements. aql h2 az ax -- aZ ax
In the case of the 9-node element the form function 4 =
-
ax
ax =all,all,- aq2alt,
for the coordinate x takes the form (Fig. 5):
allI &i
--ax ax
ax ay ax ay
atl, aq2 =--_--
nz= --ay ay ati aq2 afh aq,’
The constants C,,, . . , Cs, are calculated from afh aq2
the conditions x(-- 1, - 1) = XI, . . . , x( 1,l) = ~9,
yielding The normalization of this vector gives the direction
cosine of the normal to the membrane. The
Cl, = x5 coordinates of the shell are then calculated, using the
shell thickness in the point under consideration
Ch = f(x* - x1) (Fig. 6).
compressive stresses
5. CLOSlNG REMARKS
surface, large elemlents can be used in these directions. April 1987, (Ed. Laginha Serafim and R. W. Clough) A.
A. Balkema, 1990.
The bad aspect ratio is then caused by the small
2. M. Costa Pereira and J. Laginha Serafim, “Design of
thickness, when compared to the other two arch dams using membrane shapes”, in Arch Dams,
dimensions of the elements. Besides, as four elements International Workshop on Arch Dams, Coimbra 5-9
were used in the thickness, to capture the bending and April 1987 (Ed. Laginha Serafim and R. W. Clough) A.
shear deformation, the aspect ratio is even worse. A. Balkema, 1990.
3. P. G. Smith, “Membrane shapes for shell structures”,
Despite this, no numerical problems appeared in the Ph.D. dissertation. Univ. of California. Berklev. 1969.
FE analysis of the dam shell, which can be explained 4. C. M. A. F. C. F: Sa Marques, V. D&s da Sl’lva and
by the fact that the elements are not distorted. M. H. F. Melao Barros, “Finite element model for the
However, acceptalble results were also obtained with analysis of arch dams and optima1 design by the
membrane method”, Inr. Symp. Arch Dams, Nanjing,
two or even one element in the thickness.
China, October 17-20, 1992.
5. H. Wiistenberg, “FEPS 3.3-finite element programming
system, element library”, ICA Report No. 22. Stuttgart,
REFERENCES 1986.
6. Jean Martin, “The search of shapes for arch dams”
I. A. Portela Santa’s, M. C. Matos and .I. Laginha Serafim, in Arch Dams, In!. Workshop on Arch Dams, Coimbra
“Membrane shapes for arch dams”, in Arch Dams, 5-9 April 1987 (Ed. Laginha Serafim and R. W. Clough)
International Workshop on Arch Dams, Coimbra 5-9 A. A. Balkema, 1990.