Journal of Constructional Steel Research 57 (2001) 15–28

www.elsevier.com/locate/jcsr

Strengthening a reticulated spherical dome

against local instabilities
Ronaldo C. Battista *, Michèle S. Pfeil, Eduardo M. Batista
COPPE/Federal University of Rio de Janeiro, Civil Engineering Program, C.P. 68506,
CEP 21945-970 Rio de Janeiro, Brazil

Received 3 November 1999; accepted 6 March 2000

Abstract

The strength capacity of a reticulated spherical dome is generally associated with inelastic
buckling of its slender members and more often of the partially restrained connections between
members. These instability aspects were focused in the theoretical and experimental work
performed to analyze the structural behaviour and to design the strengthening details to upgrade
a large steel double-layer reticulated spherical dome.
The paper outlines the main steps taken on the extensive stability and safety analyses of
the as-built structure. Moreover it reports on the most relevant findings from tests carried out
on a full-scale model of a typical substructure module, to investigate the collapse mechanisms
displayed by the buckling-prone connections. A proposal is made for the use of a simple and
rational expression to estimate the connection strength in the first designing stages. Finally, it
is shown how structural upgrading was done by strengthening a few of the semi-rigid connec-
tions with specially designed fit-in bolted reinforcement details.  2000 Elsevier Science Ltd.
All rights reserved.

Keywords: Reticulated spherical dome; Partially restrained connection; Inelastic buckling; Local insta-
bility; Structural strengthening; Steel structure

1. Introduction and description of the structure

The strength capacity of a steel double-layer reticulated spherical dome, 10 meters

high and 100 meters diameter at its base, as shown in Fig. 1, could be lost through

* Corresponding author. Tel.: +55-21-560-5315; fax: +55-21-280-9545.

E-mail address: [email protected] (R.C. Battista).

0143-974X/01/$ - see front matter  2000 Elsevier Science Ltd. All rights reserved.
PII: S 0 1 4 3 - 9 7 4 X ( 0 0 ) 0 0 0 1 0 - 9
16 R.C. Battista et al. / Journal of Constructional Steel Research 57 (2001) 15–28

Fig. 1. Double-layer reticulated spherical dome. (a) 3D model; (b) vertical cross-section with overall
dimensions.

inelastic buckling of some of its partially restrained (PR) connections if the newly
constructed structure were loaded by several combinations of the full service load-
ings.
The steps performed [1–4] to upgrade this double-layer reticulated spherical dome,
which had about seven thousand members, comprised:

1. the stability and safety analyses of the structure under static and dynamic loads.
2. laboratory tests on a typical substructure module to investigate the collapse mech-
anisms displayed by the connections.
3. performance tests of the reinforcement details for strengthening some PR connec-
tions, carried out on the experimental module.
 R.C. Battista et al. / Journal of Constructional Steel Research 57 (2001) 15–28 17

4. the final design and detailing for fabrication and installation of the reinforcements
on several PR connections of the actual structure.

The analyzed structure, whose 3D finite element model is also shown in Fig. 1,
is composed of tubular members arranged in pyramidal modules on a trapezoidal
base. The reticulated spherical dome lay on pin-bearings on a crown of 36 radial
reinforced concrete frames sitting on pile foundations. These bearings allow for free
rotation of the end joints in the radial planes.
The tubular members are connected by joints composed of welded slender steel
plates where the ends of the concurrent members are bolted. The base plates of these
joints are cold-stamped to give the needed shape at each spatial connection, allowing
for adjustment of straight bars along circumferential or curved radial directions in
the reticulated spherical dome. The primary chord members’ end have a transition
flat folded plate fitted in an axial slit and welded to the tube, as shown in Fig. 2.
The secondary tubular members (diagonals and circumferential chords) have cold-
flattened ends with folded edges cut off for stress relief.

Fig. 2. Typical Connections B and C at lower layer (see also Fig. 1(b)).
18 R.C. Battista et al. / Journal of Constructional Steel Research 57 (2001) 15–28

2. Structural modeling and stability analysis

For a comparative analysis, two models of the reticulated dome structure were
used: a space truss and a space frame structure. In the latter model, the end rotations
of a meridian chord member, around one parallel, were released to simulate the
existing semi-rigid bolted connections.
The geometric deviations from the perfect dome were measured in the actual struc-
ture with topographic optical instruments, and then taken into both numerical models.
The following loadings were considered in the analysis: self-weight of the structure
and dead loading from substructures, roofing and other appendages installed by the
time of its completion; additional permanent loading that was to be installed in the
structure once in service; variable gravity load from several equipments; wind load-
ing according to pressure coefficients obtained from wind-tunnel tests and also tem-
perature variation effects. Loadings were combined in accordance to the Brazilian
Code for steel structures NBR8800/86 [5].
Several structural analyses were performed: (i) linear static for all loading combi-
nations; (ii) geometrically non-linear static analysis of the imperfect structure under
the most adverse loading combinations; (iii) free vibration analysis, (iv) dynamic
response for human loads on the roof maintenance footbridges hanging from the
structure.

2.1. Global stability analysis

Non-linear analyses were performed for increasing values of the load factor combi-
nation parameter g. The responses showed that the structure displays a quasi-linear
behavior under both combined service load (g=1) and design load level (g=1.3). These
non-linear analyses of the imperfect structure indicated buckling loads equal to 24
times and 4.5 times the service load for the truss and the frame model respectively.
In both analysis, for a certain loading combination the theoretical onset of buckling
is thought to be associated with the first elastic instability occurring for one of the
degrees of freedom in one of the almost seven thousand members of the imperfect
structure. Numerically, this corresponds to a negative (or close to zero) eigenvalue
of the stiffness matrix.
The interpretation of these results must always be taken with care because of these
numerical and some other inherent modeling uncertainties:

앫 plate joints were not discretized with finite elements and therefore all modeled
members are 5 to 10% longer their actual counterparts;
앫 because all connections are perfect pinball joints in the truss model no local bend-
ing effects are considered and therefore this model is not reliable for a stability
analysis.
앫 small local bending effects yielded by geometric deviations in members or joints
are inadequately considered in the space frame model as the plate joints and the
transition plates were not modeled.

The joints and the members ends composed by slender plates are the elements
 R.C. Battista et al. / Journal of Constructional Steel Research 57 (2001) 15–28 19

most prone to collapse due to the additional bending effects produced by local eccen-
tricities and misalignments of the compression chords. Hence, the buckling load of
the imperfect space frame structure estimated according to a criterion associated with
the first elasto–plastic instability of a joint or connection could be lower than that
obtained under the usual elastic compression member buckling criterion.

2.2. Local instability: collapse of joints and bars

The collapse load of compression members is in general associated, in regard to

their slenderness, to the elastic or inelastic instability as a strut. Instability estimates
depend on the effective buckling length of each member, which in its turn depends
on the flexibility and/or on the collapse mechanism of the joint — member end
connections. But these connections do not have an experimentally validated math-
ematical model for estimation of its collapse loads. As such, they cannot be included
in a numerical model in order to perform a geometric and material non-linear analysis
of the structure. With the purpose of fulfiling these lacking aspects and to investigate
and describe the behavior and local instability of these joints and connections, an
experimental program was carried out for testing a typical substructure module.

3. Experimental program

Experimental tests [2] on a 1:1 scale model of a typical substructure module shown
in Fig. 3 were carried out to assess the performance of the structural joints in respect

Fig. 3. Substructure module tested at COPPE’s Laboratory of Structures; at left, the bearing joint API;
at the center, the J2A joint where the observed collapse took place (see Fig. 4 for joint notation).
20 R.C. Battista et al. / Journal of Constructional Steel Research 57 (2001) 15–28

to deformation, strength and stability. A typical pyramidal substructure module with

trapezoidal base adjacent to a bearing (see Fig. 1(a)) in the most stressed region of
the space structure was selected as the structural model for experimental tests. In
this way tests were performed on typical joints under similar bracing conditions of
the prototype joints.
The tests were carried at COPPE Laboratory of Structures; the test setup is shown
in Fig. 4. A careful assemblage of all components of the test model was done in
order to reproduce the geometry of the as-built structure. A calibrated wrench was
used to turn all nuts and bolts to the right torque, following the ASTM A325 Specifi-
cation: around 0.27 kN·m for the 5/8 inch bolts at ends J2A and J3A of strut type
F1; and around 0.37 kN·m for the 3/4 inch bolts at ends J2A and API of strut type
G1 (see Fig. 4).
Several tests were performed on the same module with replacement of the dam-
aged joints and members. The servo-controlled hydraulic actuator (depicted in Fig.
4) applies compression load in member F1 and, with its position in relation to the
model, simulates the presence of the lower radial chord member adjacent to F1
beyond the joint J3A, yielding a force distribution similar to the one in the as-
built structure.
The monitoring of the model behavior was done by on-line measurements of the
applied force and corresponding actuator’s stroke and of the vertical displacement
of the central joint J2A (in Fig. 4) whose collapse mechanism was focused.

4. Experimental results

Fig. 5 shows typical structural responses in terms of the apparent shortening of

the lower chord (i.e. the radial displacement of joint J3A in the direction of a sub-

Fig. 4. Setup for the experimental tests on a substructure module carried out at COPPE Laboratory
of Structures.
 R.C. Battista et al. / Journal of Constructional Steel Research 57 (2001) 15–28 21

Fig. 5. Variation of radial displacement of joint J3A with increasing compression load. (a) test where
collapse occurred at the connection between strut F1 and joint J2A. (b) test where collapse occurred at
the connection between strut G1 and joint J2A (see also Fig. 4).

sequent strut F1 in the actual 3D structure, or equivalently the actuator’s stroke

during tests) for increasing compression load as measured in two of several perfor-
med tests.
In Fig. 5(a) it can be observed, on early stages of loading, cycles of charge and
discharge that were applied to accommodate the bolted connections. The occurrence
of two elastic–plastic mechanisms of collapse can also be easily observed: the first
associated to the development of a yield line at the transition plate in F1 member’s
end at joint J2A; and the second related to the development of a yield line at the
base plate of the J2A joint. These collapse mechanisms are illustrated in Fig. 7.
Fig. 5(b) shows another structural response related to the occurrence of collapse
in the connection between the strut G1 and J2A joint. In this particular test both
joint J2A and strut F1, that collapsed in the previous test, were substituted for new
22 R.C. Battista et al. / Journal of Constructional Steel Research 57 (2001) 15–28

ones, while the apparently not damaged strut G1 was kept in place. Residual stresses
and geometric deviations imposed by loading in previous tests led to the collapse
of the more robust but slightly damaged transition plate in G1 member’s end connec-
ted to joint J2A. The behaviour was characterized (see Fig. 5(b)) by a cascade of
collapses leading to the final sudden structural collapse. It may be observed that
during the formation of the successive collapse mechanisms under increasing load-
ing, the overall stiffness was kept nearly the same, contrary to the behaviour observed
in the test response shown in Fig. 5(a).
Fig. 6 shows in details one collapsed joint where it can be seen, with the aid of
the illustrations of Fig. 7, the formation of two yield lines: the first initiated in the
transition plate close to the tubular members’ end; the second quasi-circular, whose
formation is induced by the first one, is located at the base plate of the joint.

5. Safety analysis

The resistance of the structural members and their bolted and welded connections
were calculated according to the Brazilian Code NBR8800/86 except the resistance
related to the observed mechanisms of collapse, for which the experimental results
were used.

Fig. 6. Collapse mechanisms at joint J2A, characterized by the formation of two yield lines.
 R.C. Battista et al. / Journal of Constructional Steel Research 57 (2001) 15–28 23

Fig. 7. Elasto–plastic mechanisms of collapse of a typical joint and transition plate at member’s end.
(a) elevation. (b)model for strength calculations. (c) plan view.

5.1. Strength of compression members

The buckling length of the compressed members was initially obtained through a
refined finite element model of a strut composed of the tubular member, the transition
end plates and half-length of plate joints. The ends of the strut (at the center of the
plate joint) were considered either free or fixed, as shown in Fig. 8(a) and (b).
Another model shown in Fig. 8(c) was envisaged in which the tubular member is
hinged in both ends at yield line 1 developed during the observed collapse mech-
anism. Table 1 shows the results for slenderness ratio l of two typical members
obtained with the described models.
The refined FEM model with hinges at the center of plate joints is certainly a
conservative model since the diagonals reaching the joint do offer some rotational
restraint to the chords ends (see Fig. 7(b)). The use of this model to calculate the
compressive strength capacity of the members would lead to an extensive reinforce-
ment of the structure components.
As shown in Table 1 the slenderness for the fixed-ends model is smaller than with
the hinged-ends at the yield lines. The latter model was then adopted to calculate
24 R.C. Battista et al. / Journal of Constructional Steel Research 57 (2001) 15–28

Fig. 8. Structural models for buckling length calculations. (a), (b) struts discretized by finite elements.
(c) analytical model.

Table 1
Slenderness ratios l(=L/r)a

Model with
Tube L (mm) L/r Refined FEM model hinges at yield
lines (Fig. 8(c))
hinges at joint fixed at joint
center (Fig. center (Fig.
8(a)) 8(b))

B1 5583 184.6 187.6 133.9 169.4

D9 5960 136.2 137.8 101.5 124.3

a
L=distance between joint centers; r=radius of gyration of the member.

the strength capacity of the members since it is consistent with the experimentally
observed primary collapse mechanism, and as such is the most rational model.

5.2. Strength of a typical joint connection

Based on the experimental results, a proposal is made for a simple and rational
expression to estimate the connection strength in the first designing stages. Referring
to Fig. 7 the compressive resistances related to the development of the yield lines
are as follows:
Transition plates — Yield line 1
S1t21
Nres1⫽ fy (1)
2

Joint plate — Yield line 2

 R.C. Battista et al. / Journal of Constructional Steel Research 57 (2001) 15–28 25

alt22
Nres2⫽ f ; a⬇p/2 (2)
4d y
in general resulting in
Nres2⬎Nres1 (3)

The latter expressions were obtained from the equilibrium equation of the plastic
configuration shown in Fig. 7(b):

冕
t/2
St2
Nd⫽Mp where Mp⫽2 fySydy⫽ f (4)
4 y
0

and fy is the yield stress of the steel and the geometrical parameters Si, ti (i=1,2), l,
a e d are described in Fig. 7. To take into account the geometric imperfections
imposed on the connections by fabrication and assembling procedures, the minimum
proposed value of the misalignment is d=t/3, where t is the thickness of the thin-
ner plate.
The strength capacity related to the elastic–plastic buckling of the transition plate
under eccentric compressive force may be taken as
2t1S1 fy
Nresb⫽ f where K⫽ 2
2 y
(5)
1+k(l/4r) pE
Nresb⬎Nres1 (6)
where E is the longitudinal modulus of deformation and r the radius of gyration of
one of the two transition plates. For the buckling limit state the buckling curve a
from the Brazilian code [5] was used for the effective length corrected according to
the experimental results. This buckling curve is the same design curve included in
the Eurocode 3, Part 1.1 [6], for compressed members.

6. Reinforcement of the connections

It became clear from the experimental results and the several analyses performed
that some connections along the lower chord members needed reinforcement to
assure, with the adequate margin of safety, the local and consequently the global stab-
ility.
The typical reinforcement of joints along rings B and C (see Fig. 1(b)) is illustrated
in Fig. 9 and consists of a bolted split sleeve that serves as an extension of the
cylindrical tubular members. This special detail composed of several pieces
assembled by bolts, was conceived to avoid welding which would cause plastic defor-
mation by heating of the thin plates, and risk of collapse of the structure which is
highly stressed under gravity loads.
Figs. 10 and 11 show two views of a typical reinforced joint in the actual structure.
26 R.C. Battista et al. / Journal of Constructional Steel Research 57 (2001) 15–28

Fig. 9. Typical reinforcement of chord member’s end connections. (a) elevation. (b) plan view.

Fig. 10. Side view of a typical reinforced connection between a member end and a joint in the actual
structure.
 R.C. Battista et al. / Journal of Constructional Steel Research 57 (2001) 15–28 27

Fig. 11. A typical reinforced connection between a member end and a joint in the actual structure, seen
from below.

7. Concluding remarks

To upgrade a large spatial steel structure of a double-layer reticulated spherical

dome, extensive stability and safety analysis were performed which involved numeri-
cal modeling and experimental tests on a typical substructure module. These tests
revealed the elastic–plastic collapse mechanisms of the joint to member connections
characterized by the formation of two yield lines.
Based on test results, simple and rational expressions to estimate the connection
strength were proposed and a consistent model to calculate the strength of com-
pression members was adopted. The experimental results were of crucial importance
on the decision process for strengthening the structure and to conceive the type of
detail which should be used to reinforce the actual structure.
The special reinforcement of the joints that was conceived to avoid welding which
would cause plastic deformation by heating of the thin plates, and risk of collapse
of the structure that was highly stressed under gravity loads, is shown and described
in detail.

References

[1] Battista RC. Structural stability and safety analysis of the steel double-layer reticulated dome of a
sport arena. Contract Report COPPETEC ET–150691, Rio de Janeiro, December 1996 (in Portuguese).
28 R.C. Battista et al. / Journal of Constructional Steel Research 57 (2001) 15–28

[2] Battista RC, Batista EM. Experimental tests of a full scale substructure module of the steel double-
layer reticulated dome of a sport arena. COPPETEC Report ET150699, Rio de Janeiro, 1996 (in
Portuguese).
[3] Battista RC, Pfeil MS, Carvalho EML, Batista EM. Upgrading and strengthening of a large spherical
dome. XXVIII Jornadas Sul-Americanas de Engenharia Estrutural, São Carlos, SP, Brasil, September
1997 (in Portuguese).
[4] Battista RC, Pfeil MS, Batista EM, Carvalho EML. Structural upgrading of a double-layer reticulated
spherical dome. SSRC Annual Technical Session, Atlanta, September 1998, in press.
[5] ABNT, NBR8800 — Projeto e execução de estruturas de aço de edifı́cio. Associação Brasileira de
Normas Técnicas, Rio de Janeiro, 1986.
[6] Eurocode 3: Design of steel structures, Part 1.1: General rules and rules for buildings, ENV 1993–
1–1, 1992.