LOAD CARRYING BEHAVIOUR OF THRUST RINGS

R. Ofner, R. Greiner

Abstract: The design of thrust rings for the axial anchorage of

penstocks may be carried out on the assumption of different
pressure distributions between the ring and the surrounding
concrete. This results in a different design. In this paper the
load carrying behaviour of thrust rings is presented on the basis
of FE-results. Finally a concept for the design of thrust rings
with a more realistic pressure distribution between the ring and
the concrete is proposed.

1 Introduction
Penstocks for hydro power plants have to be anchored at certain places to transmit
axial forces into the surrounding concrete. This anchorage is carried out by thrust
rings. They are often made of flat steel plates. For the design of these rings two
different pressure distributions between the ring and the concrete may be assumed
(Fig.1). On the one hand the pressure distribution results in a clamped restraint of the
ring into the pipe (1) and on the other hand the pressure distribution results in a
clamped restraint of the ring into the surrounding concrete (2). The load carrying
capacity of the thrust ring are NRd,1=0.67·σc,1·h respectively NRd,2=0.25·σc,2·h.

Fig.1 Thrust ring with different assumptions for the contact pressure distribution

These different assumptions lead to a different design of the thrust rings. There is no
rule which pressure distribution should be used, e.g. in relation to the height over
thickness ratio h/t of the ring or in relation to the ratio of the thickness of the pipe to
the thickness of the ring tp/t. The choice of the pressure distribution, therefore, is
rather randomly selected. To find out which pressure distribution should be used for
the design some different configurations have been investigated.
2 Results of the investigation
The load carrying behaviour of thrust rings made of flat steel plates are investigated
on the basis of the Finite Element Method with ABAQUS [5]. The axisymmetric solid
FE-models include - among other parameters - the contact behaviour between the
pipe and the concrete. For the contact no friction is assumed. The material behaviour
of the concrete is defined by a parabolic stress-strain curve according to [1] but
without a crack model.

The first investigation concerns the distribution of the axial force if several thrust rings
are arranged in a row (Fig.2). In this example the ratio of the mean value to the
highest ring force βLf is 0.62. This means that the design of thrust rings can be on the
unsafe side if the axial force is distributed evenly in a row of thrust rings.

Fig.2 Uneven distribution of the axial force for thrust rings in a row

This effect is also known from long bolted connections. There a reduction factor βLf
for the shear resistance of the bolt is introduced [2]. The reduction factor is in
between 1.0 and 0.75. This rule can be the basis for an adaptation to a row of thrust
rings. There are additional parameters which influence the uneven loading, e.g. the
more complex geometry, the concrete and the contact behaviour. For a general
conclusion further investigations are necessary to determine important parameters
and their effects.

The next investigation concerns the load carrying behaviour of a single thrust ring.
The aim is to find out a realistic pressure distribution between the ring and the
concrete. Four configurations are presented in the following. The first axisymmetric
solid FE-model is shown in Fig.3. The ring has an h/t-ratio of 6 and the ratio of the
thickness of the pipe to the thickness of the ring tp/t is 1.2. The design values of the
internal pressure (p=γF·11.2=16.8N/mm2) and the axial force (N-NRd) are in
accordance with an actual example. N results from the pressure end load. It is
obvious that the pressure distribution in Fig.3 is significantly different to the
assumptions presented in Fig.1. The contact pressure has approximately the shape
of a triangle over the half height of the ring and a concentrated contact force K at the
outer edge of the ring.
A variation of this example is carried out for an axial force N=NRd=2100N/mm and
therefore (N-NRd)=0. This result as well in a triangular contact pressure over the half
height of the ring and a comparable concentrated contact force K but the MISES
stress at the inner surface of the pipe is reduced from 370N/mm2 to 300N/mm2 and
the value at the fillet increased from 330N/mm2 to 400N/mm2.
Fig.3 Contact pressure CPRESS [N/mm2] of a thrust ring h/t=370/62, pipe tp=74mm

In order to estimate the influence of the h/t-ratio a further geometry is analysed

(Fig.4). Now the ring has an h/t-ratio of 3 and the tp/t-ratio is 1.2 again. The internal
pressure and the axial force is the same as in Fig.3. The shape of the contact
pressure is triangular as before and also the length of the contact is comparable. This
means that the length of the contact is not so much in relation to the height of the ring
but more in relation to the thickness of the ring. Unlike the geometry before the
concentrated contact force K at the outer edge of the ring vanishes. This effect is
supported by the rather high tp/t-ratio.
Fig.4 Contact pressure CPRESS [N/mm2] of a thrust ring h/t=185/62, pipe tp=74mm

The next variation of the geometry is show in Fig.5. The h/t-ratio of the ring is 6 but
now the tp/t-ratio is 0.6. The internal pressure p and the axial force N are halved but
NRd is defined as before. The contact pressure remains triangular over a length of
about 150mm and as expected the concentrated contact force K increases due to the
lower clamping effect of the ring into the pipe.
An analysis with a linear elastic material behaviour of the concrete results in a peak
contact pressure at the fillet which is four times higher than the peak contact pressure
with the elasto-plastic material behaviour. The length of the contact is halved.

Fig.5 Contact pressure CPRESS [N/mm2] of a thrust ring h/t=370/62, pipe tp=37mm
An extreme variation of the geometry is presented in Fig.6. The h/t-ratio of the ring is
12 and the tp/t-ratio is 0.6 again. As before the contact pressure is triangular but the
length of the contact is reduced to about 120mm. The concentrated contact force
decreases in comparison with the example in Fig. 5 however the length of the lever
arm is twice as before. The result of the linear elastic analysis shows a significantly
reduced length of the contact as already mentioned for the example in Fig.5. Further
the contact at the upper surface indicates a low contact pressure over a length of
about 130mm. The concentrated contact force K at the outer edge still remains there
due to the reduced deformation of the concrete at the edge.

Fig.6 Contact pressure CPRESS [N/mm2] of a thrust ring h/t=740/62, pipe tp=37mm
The parameter variations have shown that the concentrated contact force K can drop
down to zero but the shape of the contact pressure remains approximately a triangle
over a length which does not correspond with the height of the ring. A conclusion for
the design of a thrust ring with a flat plate can be found by predefining an h/t-ratio of
about 4 to 6. A lower ratio will hardly activate a clamping effect into the concrete and
this can be unfavourable for the pipe. A higher ratio will hardly increase the load
carrying capacity of the ring because the contact length depends mainly on the
thickness of the ring and not on the height of the ring. In chapter 3 a formula will be
proposed for the h/t-ratio of the thrust ring.
In addition to the presented results some other parameters have been varied, e.g.
the radius of the fillet. This part of the member is of importance due to the high
concrete pressure. The reinforcement near the fillet must fulfil the requirements for
the local compressive strength of the concrete. Another attention has to be turned on
the stress in the steel at the fillet. This stress depends mainly on the fillet radius. The
question is how this stress which results from a solid FE-model should be interpreted
in respect to the fatigue check (nominal stress or hot spot stress) and to the choice of
the detail category. Here is still a lack of experiences and of regulations.

3 Proposal for the design of a thrust ring

Finally a proposal is worked out for the design or at least the pre-design of a thrust
ring without an FE-analysis. The basis for the mechanical model is the contact
pressure distribution presented in chapter 2. This contact pressure distribution and in
addition the axial force and the internal pressure are applied on a shell model without
contact to a surrounding concrete. Of course, a detailed stress field near the fillet
cannot be calculated with a shell model but the advantage is that the member can be
calculated by formulae. The arrangement of the mechanical model with its
parameters is given in Fig.7.

Fig.7 Parameters of the shell model and the loading situation

For the formulae of the proposal some assumptions are necessary. One assumption
concerns the strength of the concrete. The local compressive strength fcd,L [N/mm2] is
specified in standards [1] for central loading on a local area AL which is surrounded
by an area A but not for the arrangement of a thrust ring. Nevertheless this
specification will be adopted for the thrust ring (Eq.1).

fck A f
fcd,L = ⋅ = ck ⋅ 2 = 2 ⋅ fcd (1)
γc A L 1.5

The next definition refers to the length c [mm] of the contact pressure. The formula
for the length is derived from the following assumption. The stress in the ring near the
fillet adds up from the contraction due to the internal pressure and the Poisson effect
of the axial force as well as from the bending stress due to the contact pressure
(Fig.7). If we assume that the bending stress in the ring due to the contact pressure is
limited by fyd/3 to give enough space for the other stress components then the length
of the contact is determined by Eq.2. The same consideration is applied for the pipe
and therefore the minimum of (1 ; √2·tp/t) is added. Thus the length c relates to the
thickness of the pipe for a tp/t-ratio less than 1/√2.
Another predefinition is the height of the thrust ring which is proposed in advance by
the fixed value h=2·c (Eq.2). This leads to an h/t-ratio of about 4 to 6. In this proposal
only the thickness t of the ring has to be given and the other data, the height h, the
capacity NRd and the MISES stresses, are determined by Eq.1 to Eq.9.

fyd ⎛ tp ⎞
c = t⋅ ⋅ min ⎜ 1 ; 2 ⎟ and h=2 ⋅ c (2)
3 ⋅ fcd,L ⎝ t ⎠

With some further assumptions for the K/NRd-ratio the resistance NRd [N/mm] of a
single thrust ring is determined by Eq.3. The concentrated contact force K is at the
outer edge of the ring and Kp corresponds to the radius of the pipe. A comparison of
Eq.2 and Eq.3 with FE-results is given in Tab.1.

fcd,L ⋅ h 1 Kp 0.2 K K 1
NRd = ⋅ with = and = p ⋅ (3)
4 Kp NRd ⎛t ⎞
2
NRd NRd 1 + h
1+ 1+ 2 ⋅ ⎜ p ⎟
NRd r
⎝ t ⎠

Fig.3 Eq.2 and Eq.3 Fig.5 Eq.2 and Eq.3

c 170 185 150 156
K/NRd 0.048 0.044 0.095 0.098
NRd 2100 2070 2100 1650
Tab.1 Comparison of Eq.2 and Eq.3 with Fig.3 and Fig.5

The next set of formulae applies to the calculation of the axial stress σx and the hoop
stress σφ at the points 1 to 4 (Fig.7) resulting from the contraction due to the internal
pressure p (Eq.5) and due to the axial force N (Eq.6). The sign ± in Eq.5 and Eq.6
refers to the inner/outer surface. These equations are derived for a cylindrical shell
with a radius r and a wall thickness tp. The mechanical model for the ring is not an
annular plate but has only a ring area A=h·t. For the stresses due to the axial force
the mean value of N and (N-NRd) is used (Eq.4).
 1
κ= with LE = 1.56 ⋅ r ⋅ t p + t
LE ⋅ t p
1+
h⋅t (4)
p ⋅r N − 0.5 ⋅ NRd
σpφ0 = respectively σNφ0 = ν ⋅ with ν=0.3
tp tp

σpx,1,2 = ±1.82 ⋅ κ ⋅ σpφ0

σpφ,1,2 = + (1 − κ ) ⋅ σpφ0 ± ν ⋅ 1.82 ⋅ κ ⋅ σpφ0 (5)
σpφ,3,4 = + (1 − κ ) ⋅ σpφ0 and σpx,3,4 = 0

N − NRd N
σNx,1 = + ∓ 1.82 ⋅ κ ⋅ σNφ0 and σNx,2 = + ∓ 1.82 ⋅ κ ⋅ σNφ0
tp tp
σNφ,1,2 = +κ ⋅ σNφ0 ∓ ν ⋅ 1.82 ⋅ κ ⋅ σNφ0 (6)
σNφ,3,4 = +( κ − 1) ⋅ σNφ0 and σNx,3,4 = 0

Eq.8 is used for the calculation of the stresses resulting from the bending moment of
the contact pressure and the concentrated contact force. The sign ± in Eq.8 refers to
the inner/outer respectively upper/lower surface. The moments are calculated as if
the ring is a cantilever beam (Eq.7). This is obviously a simplification of the annular
plate. The difference of the moments resulting from a shell analysis and Eq.7
depends on the h/r-ratio. For an h/r-ratio less than 0.2 the difference of the moments
is less than 10%. A shell analysis can be carried out e.g. by [3], [4] or by a software
for axisymmetric shells. In an axisymmetric shell analysis K is used whereas Kp is
used in Eq.7. The stress σφ in Eq.8 comes from the effect of a shell due to σx.

1 ⎡ h2 ⎛ 3 ⋅ t p ⎞ ⎛ tp ⎞ ⎤
M1 = + ⋅ ⎢ fcd,L ⋅ ⋅ ⎜1+ ⎟ − Kp ⋅ h ⋅ ⎜1+ ⎟ ⎥ = −M2
2 ⎣ 24 ⎝ h ⎠ ⎝ 2 ⋅ h ⎠⎦
h2
M3 = + fcd,L ⋅ − Kp ⋅ h (7)
24
m3 ⎛h ⎞ Kp ⋅ h
M4 = + fcd,L ⋅ − Kp ⋅ ⎜ + m ⎟ with m =
3 ⋅h ⎝2 ⎠ fcd,L
6 ⋅M
σMx,1,2 = ± 2 1,2 and σMφ,1,2 = ν ⋅ σMx,1,2
tr
(8)
M 6 ⋅ M3,4 M M
σx,3,4 =∓ and σ φ,3,4 = ν⋅σ x,3,4
t2

The related stresses of Eq.5, Eq.6 and Eq.8 are added up and the verification is done
by the MISES stress (Eq.9) for each point i=1 to 4 at both surfaces (Fig.7, Tab.2).

σMISES,i = σ2x,i + σ2φ,i − σ x,i ⋅ σφ,i ≤ fyd (9)

Example
The data of the example are summarized in Tab.2 (proposal) and in Fig.3 (FEM). The
length of the triangular contact pressure according to Eq.2 (185mm) is in a good
correlation with the result in Fig.3 (170mm). The difference of the maximum MISES
stress between Eq.9 (463N/mm2) and the solid FE-model (370N/mm2) is rather high
(25%) but this must be expected due to the significant difference of the two
mechanical models and due to the high stress gradients near the fillet. The position
of the maximum stress is at the inner surface of the pipe in both models. For a further
comparison the result of an accurate shell analysis [5] is given in Tab.2 in the column
MISES in italic type (438N/mm2). In the shell analysis the mechanical model of the
ring is an eccentrically attached annular plate to a cylindrical shell. The contact
pressure distribution is according Fig.7 and Eq.3.

steel stresses due to

yield stress fyk [N/mm2] 690 N/mm2 p N M Σ MISES
partial factor γs 1.1 1 inside σx 257 150 94 501 463
poisson ratio ν 0.3 σφ 376 9 28 412 438
concrete 1 outside σx -257 220 -94 -131 310
compression fck [N/mm2] 25 σφ 221 30 -28 223 297
partial factor γc 1.5 2 inside σx 257 178 -94 342 349
load σφ 376 9 -28 356 346
internal pressure p [N/mm2] 16.8 2 outside σx -257 248 94 84 248
axial force N [N/mm] 15800 σφ 221 30 28 279 271
partial factor γF 1.5 3 upper σx 0 0 -149 -149 315
σφ 298 -40 -44 214 347
contact ratio Kp/NRd 0.052 3 lower σx 0 0 149 149 262
σφ 298 -40 44 302 266
geometry geometry 4 upper σx 0 0 36 36 253
pipe radius r [mm] 1937 σφ 298 -40 11 269 235
pipe thickness tp [mm] 74 4 lower σx 0 0 -36 -36 267
ring thickness t [mm] 62 σφ 298 -40 -11 247 260

N-NRd c c
upper surface
result 1 3 4 K result MISES
inside outside surface
contact length c [mm] 185 2
pipe max 463
ring height h=2c [mm] 370 lower surface ring max 314
fcd,L
ring capacity NRd [N/mm] 2069 N steel fyd 627

Tab.2 Input data and results for the example according to the proposal

An outlook for a further concept which provides a higher level of exploitation of the
thrust ring capacity is presented in Fig.8. For the example in Fig.3 the reference load
(internal pressure and axial force) is increased until the limit state. A limit load factor
of 2.3 can be achieved under the condition that the concrete does not crack. The
material behaviour of the concrete is defined by a parabolic stress-strain curve
according to [1] but without a crack model (a crack model without reinforcement
results in a load factor well below 1.0). Fig. 8 presents the state at the load factor of
2.0 (p=2.0·16.8 N/mm2). In the pipe and in the ring distinct plastic zones exist. The
length of the contact is 185mm which is comparable with the length at the load factor
of 1.0 (Fig.3) but the shape of the contact pressure tends from triangular to parabolic.
In further investigations attention should be turned on the behaviour of the reinforced
concrete close to the thrust ring and if a parabolic contact pressure is acceptable.

Fig.8 Contact pressure CPRESS [N/mm2] of a thrust ring h/t=370/62, pipe tp=74mm

Conclusion
The FE-results in chapter 2 provide an inside into the load carrying behaviour of
thrust rings. In chapter 3 a concept for the design is proposed. In this proposal only
the thickness of the ring (Eq.2) has to be given and the other data, the height, the
load carrying capacity and the MISES stresses in the ring as well as in the pipe are
determined straightforward by the formulae presented in chapter 3.

References
[1] EN 1992-1-1, Design of concrete structures
[2] EN 1993-1-8, Design of steel structures, Design of joints
[3] Guggenberger W., Linder C., Elastic stress analysis of axisymmetric shells,
ECCS International Conference, Prague, October 2003
[4] Markus G., Theorie und Berechnung rotationssymmetrischer Bauwerke, 1976
[5] ABAQUS/Standard, Abaqus Inc., RI USA

Authors
Dr.techn. Robert OFNER, Univ.-Prof. Dr.techn. Richard GREINER
Graz University of Technology, Institute for Steel Structures and Shell Structures
Lessingstrasse 25, A-8010 Graz, AUSTRIA
Phone: +43 316 873 6201, FAX +43 316 873 6707,
E-mail: robert.ofner@tugraz.at, r.greiner@tugraz.at