Engineering Structures 49 (2013) 743755

Contents lists available at SciVerse ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Structural assessment of a modern heritage building

Stefano Sorace a,, Gloria Terenzi b
a
b

Department of Civil Engineering and Architecture, University of Udine, Via delle Scienze 206, 33100 Udine, Italy
Department of Civil and Environmental Engineering, University of Florence, Via S. Marta 3, 50139 Florence, Italy

a r t i c l e

i n f o

Article history:
Received 23 January 2012
Revised 16 November 2012
Accepted 3 December 2012
Available online 30 January 2013
Keywords:
Structural assessment
Seismic assessment
Modern architectural heritage
Steel structures
R/C structures
Glazed faades
Linear analysis
Non-linear analysis
Seismic retrot

a b s t r a c t
A structural assessment study on Palazzo del Lavoro in Turin, a masterpiece by Pier Luigi Nervi, was
carried out within a National Research Project dedicated to the analysis of modern heritage architecture
in Italy. Based on the original design documentation collected through records, a complete nite element
model of the building was generated. The study included detailed models of the main structural members, represented by monumental reinforced concrete columns, a mushroom-type steel roof and reinforced concrete ribbed gallery slabs, and the main non-structural systems, constituted by continuous
gallery-to-roof glazed faades. The results of the linear and non-linear analyses developed by these models, aimed at fully understanding the original design concept of the various members, as well as at evaluating their current static and seismic safety conditions, are reported in this paper. The non-linear
computations include a buckling analysis of the slender steel beams constituting the roof, and an integral seismic pushover analysis of the monumental columns. The results of the analyses highlight safe
conditions and good performance objectives in general, but for some important exceptions. Indeed, the
roof beams failed to pass the verications on global and local panel exuraltorsional buckling, and some
cantilever beams of the gallery oors showed poor shear resistance. Retrot hypotheses are also formulated for these elements, so as to help the entire structure to comply with the requirements of the new
Italian Technical Standards.
2012 Elsevier Ltd. All rights reserved.

1. Introduction
Growing attention is currently being devoted to the study of
modern architectural heritage, and particularly to the edices built
from the aftermath of the Second World War until the late 1960s.
Indeed, that was a very prolic period for architecture and structural engineering, which produced signicant theoretical and technical advancements in both elds. As a consequence, a global
enhancement of the construction industry was reached, and a great
number of exemplary masterpiece structures were designed and
erected worldwide. This important stock of buildings is now over
50 years old, and may require important structural maintenance,
repair and/or rehabilitation interventions. In view of this, careful
evaluation and verication analysis strategies are needed to check
the actual safety conditions of these skilled engineering works, and
to plan possible retrot solutions. At the same time, the development of assessment analyses of these outstanding buildings offers
a protable chance to improve the knowledge on the characteristics of their constituting materials, structural details and construction work procedures, as well as on the calculation methods
Corresponding author. Tel.: +39 432 558050; fax: +39 432 558052.
E-mail addresses: stefano.sorace@uniud.it (S. Sorace), terenzi@dicea.uni.it
(G. Terenzi).
0141-0296/$ - see front matter 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.engstruct.2012.12.012

originally adopted for their design. Moreover, these analyses can

provide a better understanding of the original conception of the
structures enquired, and evaluations about the attainment of the
theoretical and technical objectives targeted in their design, which
was carried out without the help of computer software. As a consequence, new contributions to the critical interpretation of the
activity of the master structural designers of the 20th century
can be derived, which are of potential interest also to researchers
working in the eld of history of modern architecture, as well as
to scholars and engineers working in the eld of structural assessment and rehabilitation.
An Italian masterpiece belonging to this stock named Palazzo
del Lavoro in Turin, designed by the world-famous structural
engineer Pier Luigi Nervi, is examined in this paper. The building,
an external and an internal views of which in its current conditions
are shown in Fig. 1, constituted the most important exhibition hall
erected for the celebrations held in Turin for the rst centenary of
the Unity of Italy, back in 1961. The structure was designed in 1959
and completed by spring 1961, after 16 months only. This represented a really challenging enterprise, which can still arouse admiration, especially when the short construction times are compared
to the imposing size of the building160  160 m  m in plan
and considering the strict architectural and functional constraints
imposed on the design, among which the 40 m-long free spans

744

S. Sorace, G. Terenzi / Engineering Structures 49 (2013) 743755

Fig. 1. External and internal views of the building in its current conditions.

required between each vertical structural element. The solution

devised by Nervi consisted in a mesh of 16 reinforced concrete
(R/C) monumental columns with variable sections along the height
(that is, 20 m from ground to the base of the roof), constituting the
most prominent example of Nervis principle of uniform resistance [1] applied to vertical members in his works. Each column
supports a steel mushroom-type roof panel with 16 radial beams
spanning from the center. The panels are mutually separated by
a 2 m-wide joint covered by a glass skylight. This solution, illustrated in the roof plan in Fig. 2, confers a suggestive monumental
look to the building.
The remaining structural elements also remarkably contribute
to the elegant and monumental appearance of the building. The
most important elements are the R/C ribbed slabs constituting
the two perimeter gallery oors. A plan of the upper oor is shown
to the right of Fig. 2. The design solution for the slabs, traced out
following the analytical equal-stress lines of their plate model, is
also a typical feature of Nervis style [2], and was applied to other
famous structures of him. An original drawing of the formworks,
specially designed to the purpose, and a view of the intrados of
the slabs, with the steel roof in the background, are displayed in
Fig. 3.
Among the secondary structural elements of the building, the
continuous glazed faades, also visible in the images in Fig. 1, represent a much advanced technical solution for the time too, as they
are an early application of the curtain wall concept, with remarkable global dimensions (free height equal to around 16 m, from the
rst gallery oor to the top, and total surface greater than
12,000 m2).
This paper offers a synthesis of the structural and seismic
assessment analyses carried out on the building, which make a part
of the studies developed within a National Research Project nanced by the Italian Ministry of Education, University and Research dedicated to innovative structural designs and the
correlations between the leading engineering and architectural
activities in Italy, during the 1950s and 1960s. The computational
models generated for the analyses, as well as the verications car-

Fig. 2. Plans of the roof and the upper gallery.

ried out for all members, were entirely based on the original design
documentation collected through extensive record research, because eld testing activities have never been developed on the
building structure. However, the original documents, also including the certicates of the structural materials tested during the
various stages of the construction works, are exhaustive enough
to x with certitude all input data for the numerical assessment
enquiry.
The following aspects of the study are presented and discussed
in the next sections: a modal analysis of the entire building; basic
resistance and global/local buckling verications of the steel roof
beams, with comparisons of the results derived from the normative
expressions of the critical stress of panels and the global lateral
torsional buckling resistance of beams with the corresponding nite element buckling computations; linear and non-linear seismic
analyses of the R/C columns, the latter being carried out by an
unconventional integral pushover approach, with the numerical
model constituted by a full mesh of solid octahedral smeared
cracking concrete elements with embedded steel reinforcements; the analysis and verication of the ribbed R/C gallery oors,
including an evaluation of the correlation of their equal-stress line
original conception to relevant nite element solutions; and the
seismic analysis and evaluation of the glazed faades, developed
by referring to non-structural performance limitations specially
formulated to the purpose.

2. Modal analysis of the building

A modal analysis of the entire building was developed as a rst
step of the assessment enquiry, in order to evaluate its general dynamic characteristics [3]. The analysis was carried out by a nite
element model generated by the SAP2000NL calculus program [4],
where all the structural elementsmesh of light alloy proles supporting the glazed faades and pennon beams constituting the
vertical load bearing and bold bracing system of the faades includedwere reproduced. Views of the model without the faade
elements and the perimeter beams of the mushroom roof panels,
and its complete layout, are shown in Fig. 4. Each monumental column is reproduced by 7 frame elements with different cross sections, as discussed in Section 4.1, and each radial beam of the
roof panels by a single frame element with variable section. Ten
frame elements are used to model each of the 20 steel frames constituting the circular drums situated on top of the columns, as described in detail in Section 3. The mesh of perimeter steel edge
beams connecting the free extremities of the radial beams of the
mushroom panels is made of 420 frame elements. In total, the
assembly of the 16 R/C columns and relevant mushroom panels
includes 3364 frame elements. The slabs of the two gallery oors
are reproduced by a mesh of inner square and outer rectangular
shell elements, with 1 m  1 m and 1 m  2 m sides, respectively,
for a total of 12,240 elements for the two oors. The longitudinal,

745

S. Sorace, G. Terenzi / Engineering Structures 49 (2013) 743755

Fig. 3. Original drawing of a formwork and intrados view of a R/C ribbed slab of the gallery oors.

Fig. 4. Internal and complete views of the global nite element model of the
building.

transversal, internal perimeter and cantilever beams of the oors

are modelled by a set of 5720 frame elements. Each one of the
256 R/C bearing columns supporting the oors is constituted
by a single frame element. The glass panes of each faade are
modelled by a mesh of 650 shell elements with average sides of
2.5 m; 720 vertical and 504 horizontal frame elements are
incorporated to reproduce the supporting light alloy proles, and
320 frame elements for the 32 pennons.
The results of the modal analysis show that the rst two modes
are mixed rotational around the vertical axis z/translational along
the two main directions in plan, x and y. Both modes, whose shapes
are plotted in Fig. 5, feature a vibration period of 1.36 s; effective
masses associated to the relevant translational component are
equal to around 20.8% of the total seismic mass of the building;

and effective masses associated to the rotational component are

equal to around 7%. The rst two modal shapes are dominated
by the deformation of the 16 couples of R/C columns and supported
roof panels, which respond independently one from another, as a
consequence of the existing separation joints at roof level. The
rst two building periods of 1.36 s correspond to a mean of the
different rst vibration periods of the internal, side and corner
columnsroof panels couples taken separately, the differences
being determined by the interaction with the bracing system of
the faades. The three periods are equal to: 1.44 scentral couples,
1.35 sside couples (interfaced on one side with the faades), and
1.27 scorner couples (interfaced on two sides).
Concerning the superior modes of the building, crowds of 4
through 6 modes are repeatedly observed in correspondence with
specic vibration periods, with negligible associated masses, as
these secondary modes are essentially related to local response
effects. Several dozen modes are required to gradually nd significant mass contributions and, in total, 83 modes are needed to activate a summed mass greater than 85% (the minimum share
required by the Italian Technical Standards [5] for the development
of a modal superposition seismic analysis) along the two directions
in plan, and around the vertical axis.
Portions of the global model of the building, and more rened
models of the main structural members taken separately, were
used subsequently to carry out the linear seismic assessment enquiry reported in Sections 4.1, 5 and 6.
3. Analysis of steel roof beams
The 20 cantilever steel radial beams forming the corolla of each
one of the 16 mushroom panels of the roof have xed-end bolted
connections to a circular drum, constituted by 20 rectangular steel
frames, 2800 mm high and 1900 mm wide. As shown in the original structural design drawing in Fig. 6, each frame is supported by
a triangular steel platewith a 1500 mm-long vertical side and a
1900 mm-long horizontal sideplaced over a 200 mm-deep
groove on the upper section of the R/C column. The I-section

z
y

Fig. 5. First and second modal shapes of the global nite element model of the building.

746

S. Sorace, G. Terenzi / Engineering Structures 49 (2013) 743755

effects of seismic action, vertical component included (with the

multipliers of gravitational loads normatively xed at 1 for this
type of combination).

Circular steel drum

3.1. Bending and shear resistance and lateraltorsional buckling

verications

Inner hole of R/C column

Triangular steel plates

The verication of resistance to the in-plane bending moment

at the ultimate limit statesreferred to the effective properties of
Class 4 cross sections, to which the considered members belong
according to Eurocode 3 Part 1-1 rules [7]fails to be met. Indeed, the ratio of the design value of bending moment to the corresponding design resistance is signicantly greater than 1 (and it
reaches 1.57 for the xed-end section) along over 3/4 of beam
span. On the other hand, the shear resistance test is met for all sections. The verication at the serviceability limit states concerning
vertical deection, developed according to the current Italian Technical Standards [5] (as Eurocode 3 devolves this specication to the
National Annexes), is widely met too.
The verication of beam resistance to lateraltorsional buckling
was carried out by considering the only effect of the major axis
bending, since the compression axial force induced by the slope
of the center-line of the beams is very low (with a maximum of
22 kN at the xed-end section). The relevant verication formula
is:

Groove on R/C column

Fig. 6. Original design drawing of the vertical section of the connection system of
the steel roof beams to the upper zone of the R/C columns.

welded beams, which are joined on their free end to a continuous

C-shaped steel edge beam outlining the square perimeter of the
mushroom panel (Fig. 7), are 2800 mm to 700 mm high, and their
top and bottom anges are 690 mm to 200 mm wide. The beams
have three different spans, ranging from 15,750 mm (type 1 beams,
orthogonal to the C edge prole) to 20,250 mm (type 3 beams,
close to the diagonal of the square). The constituting steel is equivalent to the current S235JR type, as it is typical of medium-to-high
rise Italian steel structures built in the late 1950s through the late
1960s [6], with nominal yielding and ultimate stress values
fy = 235 MPa and fu = 355 MPa, respectively. The web of beams is
very thin (5 mmtype 1 through 7 mmtype 3), which results in
great slenderness of cross sections, especially in the areas close
to the xed end. The web is subdivided in 13 (type 1 beam)
through 17 (type 3) panels by a set of vertical stiffening plates
welded to the web and to the top and bottom anges. The different
web thickness and stiffener spacing values determine a very similar resistance of the three types of beams to bending and shear
stresses, as well as to local and global buckling, as planned in the
original design of the metallic roof (carried out by engineer Gino
Covre, who worked with Pier Luigi Nervi for this part of the building structure). In view of this, the nite element and verication
analyses are synthesized below for type 1 beams, whose dimensions are reported in Fig. 7, as they are also exhaustively representative of the remaining two beam types.
It is noted that the analyses reported below are referred to the
effects of gravitational loads, constituted by the dead and live
(snow-related) loads of the roof, plus the self-weight of the beams.
Indeed, the stress states deriving from the normative probabilistic
combinations of these loads at the ultimate limit states (with multipliers of dead and live loads equal to 1.3 and 1.5, respectively) are
greater than the ones obtained from the combination with the

250

1390

500

938

1046

1149

where MEd, Mb,Rd are the design value of the moment and the design
buckling resistance moment, respectively, with Mb,Rd expressed as:

Mb;Rd vLT W y

fy

cM1

being Wy = Weff,y for Class 4 sections (Weff,y is computed by determining the effective section as a function of the reduction factor
q for the compressed portion of the web and the compressed
ange), fy = 235 MPa, as noted above, and cM1 = 1.05; vLT is given
by the following relation:

vLT

1
q
ULT U2LT  k2LT



where ULT 0:5 1 aLT 
kLT  0:2 
k2LT , aLT is an imperfection factor, equal to 0.76 for welded I-sections with height-to-base ratio
q
W y fy
, and Mcr is the elastic critical moment
greater than 2, 
kLT
Mcr
for lateraltorsional buckling evaluated according to the following
expression in Annex F of Eurocode 3 Part 1-1 [7]:

10

11

1556

1629

12

13

1641

1015

2821

2820

2300

255

2200

M Ed
<1
Mb;Rd

275

1887
518

887

1158

1289

1325

1492

1610

15750

+20,0

2500

Fig. 7. View of a type 1 steel roof beam (dimensions in millimeters).

S. Sorace, G. Terenzi / Engineering Structures 49 (2013) 743755

2s
3
 2
pEIz 4
k
Iw kL2 GIT
2
M cr C 1
2
C 2 zg  C 3 zj  C 2 zg  C 3 zj 5
kw I z
p EIz
kL2
4
where C1, C2, C3 = coefcients that depend on loading conditions
and end constraints; L = length of the beam between lateral constraints; Iw = warping constant; kw, k = effective length coefcients;
zg = zazs, with za = coordinate of the point of application of load,
and zs = coordinate of the center of torsion; zj = [0.8(2bt  1)hs]/2,
with: bt = Itc/(Itc + Ift), Itc = moment of inertia of the ange in compression with respect to the minor axis of the section, Ift = moment
of inertia of the ange in tension with respect to the minor axis of
the section, and hs = distance between the centers of torsion of the
anges.
By applying the relations above, the ratio of MEd (equal to
2414 kN m) to Mb,Rd (1415 kN m) results to be equal to 1.706,
and thus the verication inequality (1) is denitely not met. The
unsafety factor is obtained by inverting the ratio between the
two moments (Mb,Rd/MEd = 1/1.706), i.e. 0.586.
3.2. Web panel buckling verications
The web panels are much more sensitive to buckling than the
ange plates are, as a consequence of the high slenderness of the
web determined by the geometrical characteristics of the beams.
The verication analysis is carried out in this case by referring to
the criterion proposed in a previous edition of the Italian Standards
for steel structures [8], where the effects of normal and shear stresses are jointly considered, assuming an ideal critical stress rcr,id to
be compared to the design ideal stress computed according to Von
Mises rule. The expression of rcr,id is derived from Massonnet normal critical stressshear critical stress domain [9] as follows:

rcr;id

q
r21 3s2
r


2  2
1w r1
3w r1
 rcr
 rcr sscr
4
4

where r1 = 130.9 MPa and s = 20.1 MPa are the design normal and
shear stress values; rcr = krrcr,0, scr = ksrcr,0, being kr, ks the normal and shear stress buckling factors, and rcr,0 the elastic critical
plate buckling stress of the equivalent orthotropic plate, expressed
t
2
p2 E
as rcr;0 121
m h , with t = plate thickness and h = plate width (or
mean width in case of variable section); and w is a coefcient dening the linear variation of normal stress over the section, which can
be set as equal to 1 in this case, by neglecting the very little contribution of the axial force to r1, quantied by a normal stress of
1.6 MPa. Panel 4 (Fig. 7) results to be the most critical one of the
13 web panels of type 1 beams. Considering its geometrical characteristics (base = b = 1050 mm, h = 2445 mm, t = 5 mm), kr = 27.6,
ks = 33 and rcr,0 = 0.78 MPa values come out, from which
rcr = 21.6 MPa and scr = 25.9 MPa are derived. By applying formula
(5), rcr,id results to be equal to 22.2 MPa.
The values of the normal and shear stress buckling factors are
computed in [8] as a function of the aspect ratio a = b/h (whose
average value is equal to 0.427 for panel 4) according to the
expressions

kr 15:87
ks 4

5:34

a2

1:87

a2

8:6a2

a<1

a6

2
3

6
7

which provide good analytical approximations of TimoshenkoGere

[10] original instability curves for linearly varying (with w 6 1) normal stress, and uniform shear stress distributions, respectively. The
difference between the rcr,id and rcr values above (22.2 MPa against

747

21.6 MPa) is so little because of the great prevalence of r1 over s,

which generates poor inuence of shear stress in the critical stress
interaction domain.
A second observation concerns rcr, which is greater than the value of 19.1 MPa derived by the Eurocode 3 Part 1-5 [11] formula

kp

s
fy

rcr


b=t
p
28:4e kr

q
 is the web width, e 235, and k = 23.9 for w = 1. The
where b
r
fy
difference between the two rcr estimates obtained from [8,11] is
caused by the two kr values adopted (27.6 [8]formula 6against
23.9 [11]). Indeed, unlike Standards [8], Eurocode 3 Part 1-5 [11]
prudentially assumes the minimal theoretical value of 23.9corresponding to a = 2/3 in formula (6)for any aspect ratio of panels,
when w = 1.
3.3. Finite element buckling analysis
The nite element model of type 1 beams generated for the
buckling analysis is constituted by a mesh of quadrilateral isoparametric shell elements with an average side of 150 mm. This dimension determines a number of constituting elements of each beam
panel varying from around 80 to around 120, which is generally
deemed appropriate for an accurate simulation of local buckling effects in laterally loaded stiffened or non-stiffened plates [12,13].
Fixed end restraints are imposed to the internal end section of
beams, connected to the steel drum, whereas only the lateral displacements are blocked on the tip end section, so as to accurately
reproduce the restraint offered by the perimeter C-shaped edge
beam of each mushroom roof panel. These boundary conditions
are obtained by introducing displacement restraints for all three
axes of the local coordinate system in the nodes of the shell elements situated on the internal end section of beams, and displacement restraints acting only along the horizontal axis in the nodes
of the elements placed on the opposite end. The buckling analysis
is developed in SAP2000NL [4] by a classical eigenvalue
formulation:

K E  kK G fv g f0g

where [KE] and [KG] are the elastic and geometric stiffness matrixes
of the structural element or system, k is the generic eigenvalue, and
{v} is the corresponding eigenvector. The solution of Eq. (9) provides
the instability factors ki and the instability modal vectors {vi}. The
least of the ki multipliers computed by the program represents
the rst (or critical) eigenvalue k1. If k1 is greater than 1, no buckling
occurs under the imposed loads.
The rst mode buckling conguration of type 1 beams resulting
from the analysis, displayed in Fig. 8, highlights that the maximum
lateral deformation is achieved in panel 4, consistently with the
analytical assessment predictions commented in Section 3.2. The
relevant k1 factor is equal to 0.259. By multiplying this value by
the maximum von Mises ideal stress obtained in the central zone
of the panel for the rst buckling mode deformed conguration,
equal to 90 MPa, the following nite element critical ideal stress
estimate rcr,id,FE is deducted: rcr,id,FE = 23.3 MPa. This value is close
to the rcr,id normative estimate of 22.2 MPa given by formula (5),
with a percent difference not exceeding 5%. Similar correlations
are obtained for the subsequent local buckling modes too (the second mode achieves the maximum lateral displacements in panel 5,
the third mode in panel 3, etc.), as the differences between rcr,id
and rcr,id,FE never exceed 5%. The seventh and eighth buckling
modes are the rst two involving a global (lateraltorsional) instability deformed shape. The maximum lateral displacements and
stresses are reached in the eighth mode, visualized in Fig. 9 with
an amplication factor of 5000. The horizontal projection is also

748

S. Sorace, G. Terenzi / Engineering Structures 49 (2013) 743755

can represent a credible balance point between the need to reach

accurate results and to restrict the computational effort.
3.4. Proposal of a retrot intervention
Based on the results of the assessment analyses, a simple retrot solution was proposed, which consists in strengthening the
beams by a line of horizontal steel plates placed at mid-height of
the cross sections, plus a diagonal plate positioned in the lower half
of the panels that proved to be the most sensitive to buckling (the
rst 7 out of 13, in the model represented in Fig. 8), all welded to
both sides of the web, as shown by the modied nite element
model in Fig. 10. The rst buckling factor in strengthened conguration grows from 0.259 to 1.33, guaranteeing a satisfactory safety
margin with a low-impact intervention. The same retrot solution
is extended to type 2 and type 3 beams too, involving the rst 8 out
of 15, and the rst 9 out of 17 panels, respectively, and producing
similar increases in the corresponding rst buckling factors.
Fig. 8. Deformed shape of type 1 beams obtained for the rst buckling mode.

4. Analysis of R/C columns

DEFORMED BEAM AXIS

UNDEFORMED BEAM AXIS

Fig. 9. 5000-Times magnied deformed shape of type 1 beams obtained for the
eighth buckling mode.

plotted in this drawing, showing that the deformed shape corresponds, as for the seventh mode, to the rst theoretical global
buckling mode of the beams. The k8 eigenvalue is equal to 0.524,
which must be compared to the unsafety factor Mb,Rd/MEd = 0.586
resulting from the lateral/torsional buckling verication discussed
in Section 3.1. The difference between the two values is around
12%, and the numerical result in this case is more conservative
than the normative factor estimate.
As is known, the data obtained from a computational analysis
are always a function of the geometrical dimensions of the mesh.
In view of this, mesh-sensitivity was investigated by varying the
sides of the shell elements by factors 2, 1.5, 0.75 and 0.5 with respect to the reference average dimension of 150 mm. As a general
result of this enquiry, no appreciable inuence on eigenvalues and
eigenvectors was observed when passing to the most rened
meshes. A trend towards a progressive rise in eigenvalues emerges
when increasing the sides (e.g., rcr,id,FE in panel 4 becomes equal to
24.1 MPa and 25.3 MPa for mesh factors 1.5 and 2, respectively),
even if the shapes and the hierarchy of buckling modes are kept
unchanged. Based on these observations, the average sides of the
shell elements assumed for this analysis appear to be the greatest
values compatible with the accuracy of the solution, and thus they

The shape of the cantilever monumental columns constantly

varies from the base (cross-type section with 6 m-long and 1 mwide sides) to the top (circular-type Section, 2.5 m wide), as illustrated by the photographic image of the building interiors in Fig. 1,
and by the sequence of geometrical cross sections along the height
reproduced in Fig. 11. This variable shape was designed in order to
obtain nearly uniform resistance members with respect to the
combined effects of bending moments and axial force, as observed
in the Introduction. The top section, reduced to a diameter of 2 m,
is prolonged for further 1.6 m to form the groove where the triangular steel plates supporting the circular drum of the mushroom
roof are located (Fig. 6). The original drawings of the structural sections at the xed-end base, at an intermediate height and on top,
displayed in Fig. 12, show an eccentric inner hole (also visible in
the vertical section in Fig. 6), where a spiral steel staircase to access
the roof, and a conductor pipe, are housed.
The foundations of the columns consist in cross-type plinths
with 10.4 m-long and 2.4 m-wide sides, supported by groups of
16 pedestal (Franki)-type reinforced concrete piles with a diameter
of 500 mm (Fig. 13). The verications carried out both on the piles
and the plinths, based on the complete geotechnical data of the
soil, which are not reported here for brevitys sake, are largely
met for all static and seismic load design combinations.

Fig. 10. Deformed shape of type 1 beams obtained for the rst buckling mode in
retrotted conditions.

S. Sorace, G. Terenzi / Engineering Structures 49 (2013) 743755

Fig. 11. Sequence of the geometrical cross sections of the R/C columns.

4.1. Linear analysis

The geometry and the cross sections of the linear nite element
model of columns, generated by SAP2000NL [4] too, are illustrated
in Fig. 14. As mentioned in Section 2, the model is subdivided in seven portions, which correspond to the seven cast zones constituting the columns. Each portion consists of a frame element with
constant section, equal to the average section of the cast zone. This
type of discretization was necessary because the section-builder
option offered by the program, which was used to reproduce the
rather complex geometry of the columns (internal hole included),

749

does not allow generating continuously varying sections along

the axis of a frame member. The reinforcing bars were not included
in the model.
The pseudo-acceleration design response spectrum adopted for
the seismic analyses carried out at the basic design earthquake level (BDE, with a 10% probability of being exceeded over the reference time period VR = 200 years xed for the building, obtained by
multiplying the assumed nominal structural life of 100 years by a
coefcient of use equal to 2, as imposed by Standards [5] for strategic buildings) is plotted in Fig. 15. The spectrum is referred to the
city of Turin and C-type soil conditions (deep deposits of dense or
medium-dense sand, gravel or stiff clay from several ten to several
hundred metres thick, according to the identical soil classications
of Standards [5] and Eurocode 8 [14]), as resulting from the geotechnical testing campaigns enclosed to the original design documentation. A behavior factor q equal to 1.5 was selected for the
BDE to scale the ordinates of the corresponding elastic response
spectrum, in consideration of the low-ductility inverted pendulum
structural conguration of columns. The vertical loads considered
in the analysis are the ones transferred by the steel roof, plus the
self-weight of the columns.
The concrete used for the columns, as well as for the gallery
slabs, was 680-type, featuring a high characteristic value of compressive strength fck for in situ cast R/C structures at the time, equal
to 42 MPa. Indeed, in the 1960s throughout the 1970s, similar fck
values were more typical of concretes used in prefab R/C buildings
[15,16]. The reinforcing steel was R50/60-type, with nominal
yield stress fy of 370 MPa and limit stress fu equal to 545 MPa.
The results of the linear assessment analysis are demonstratively displayed, for the three cross sections shown in Fig. 12, in
the diagrams plotted in Fig. 16, reproducing the axial forcebending moment interaction domains of the sections and the represen-

Fig. 12. Original drawings of the structural sections at the base, an intermediate height and the top of the R/C columns.

Fig. 13. Original drawings of the foundations of the R/C columns.

750

S. Sorace, G. Terenzi / Engineering Structures 49 (2013) 743755

Fig. 14. Geometry of the linear nite element model of the R/C columns.

conrm that the columns have an approximately uniform resistance along the height, also according to the most recent normative
verication criteria.

0.5

Pseudo-Acceleration [g]

0.45

Design
Spectrum
BDE Level
q = 1.5

0.4
0.35

4.2. Non-linear analysis

0.3
0.25
0.2
0.15
0.1
0.05
0

0.5

1.5

2.5

3.5

4.5

Period [s]
Fig. 15. BDE-scaled pseudo-acceleration design response spectrum.

tative points of the maximum combined effects derived from the

nite element computations. Similar results are obtained for the
remaining sections, omitted herein for brevitys sake. The graphs
in Fig. 16 highlight that the verication points largely remain within relevant safety domains, and their distance from the edges of the
domains is comparable in the three cases. As this holds true for all
the remaining sections, the results of the linear analysis seem to

Based on the outcome of the rst-level linear assessment analysis, a second-level step was started, which consisted in a nonclassical pushover analysis, carried out by an integrally non-linear
model generated by the ANSYS calculus program [17]. The model is
constituted by a full mesh of 8400 solid octahedral smeared cracking concrete elements with embedded steel reinforcements that
can be freely oriented with respect to the global coordinate system.
No reductions to simplied models were considered in this enquiry, as the uniform resistance columns should ideally reach the
rst signicant cracked congurations, and then the plasticization
of vertical reinforcements, simultaneously in several sections along
the height. This full-cracking application offers a more direct and
realistic simulation of the evolution of the non-linear response of
columns as compared to models including lumped plastic hinges
or ber-composed plastic zones, but it requires a greater computational effort and proper checks on the stability and accuracy of the
solution.
The WillamWarnke triaxial failure domain [18] is adopted to
model the ultimate compressive, tensile and mixed compressive
tensile triaxial ultimate response of the concrete material. The
classical DruckerPrager yield criterion [19] is assumed by the

Fig. 16. Axial force-bending moment domains of sections in Fig. 12 and verication points derived from the analysis.

751

S. Sorace, G. Terenzi / Engineering Structures 49 (2013) 743755

3000

2500

Base Shear [kN]

program for plastic deformations. A bilinear strain-hardening elasto-plastic behavior is assigned to reinforcing steel. The main
mechanical parameters of the concrete model are as follows:
sto = shear transfer coefcient for an open crack, stc = shear transfer
coefcient for a closed crack, ftd = uniaxial cracking design stress,
fcd = uniaxial crushing design stress, fcb = biaxial crushing design
stress, Ec = Young modulus, and mc = Poisson ratio. The parameters
of reinforcing steel are: fy = yielding stress, sh = kinematic strain
hardening ratio, Es = Young modulus, and ms = Poisson ratio. The
parameters dening the surface of the DruckerPrager domain
are: c = cohesion, / = friction angle, and w = dilatancy angle. The
following basic values were adopted in the analysis: sto = 0.3,
stc = 0.85, ft = 1.7 MPa, fcd = 23.8 MPa, fcb = 1.2fc, Ec = 35,600 MPa,
mc = 0.2, fy = 321.6 MPa, sh = 0.015, Es = 206,000 MPa, and ms = 0.3,
according to the characteristics of the materials; and
c = 2.12 MPa, / = 30 and w = 0 (associated ow rule), from literature suggestions concerning the plasticity domain for concretetype elements [18,20,21].
The gravitational loads are the same as introduced in the linear
analysis but, unlike the linear model, the geometry reproduces
the real one with continuity. The horizontal load for the development of the pushover process was applied to the top of the column. Geometrical non-linear effects were taken into account, in
view of the expected high maximum displacements. As for all
types of incremental analysis, the critical parameter for the convergence and the accuracy of the numerical solution was represented by the number of sub-steps to be developed in the
ramped loading process within any single load step, with the latter xed at 10 mm. A displacement-based criterion for convergence control was adopted, with a tolerance of 5%. The
following numbers of sub-steps were nally selected, after several
tentative choices: 50 (corresponding to 0.2 mm) for steps 1
through 13, characterized by moderate cracking effects in the
concrete elements; 200 (0.05 mm) for steps 1427extensive
cracking in the tension zones; 300 (0.033 mm) for steps 2870
softening response phase. These data conrm general suggestions
[20] about the preferable values (ranging from 0.1 mm to
0.01 mm) of the displacement increments in full-cracking/crushing problems when the non-linear behavior of a signicant portion of the model is activated. Further increases in the number
of sub-steps in the more accentuated non-linear response phases
did not show any practical impact on the accuracy of the solution.
Indeed, by amplifying the number of sub-steps by a factor up to
10, that is, by assuming up to 2000 sub-steps for steps 1427,
and up to 3000 sub-steps for steps 2870, differences no greater
than 0.1% on base shear were found.
For the assumed set of mechanical parameters, the pushover
analysis was concluded at the end of step 70, corresponding to
a top displacement dtop of 700 mm and a drift ratio (ratio of top
displacement to column height) dr equal to 3.5%. This was xed
as the numerically determined structural collapse condition. The
only two parameters not related to the specic characteristics of
the constituting materialssto and stcwere varied in their
technical ranges of interest (sto from 0.2 to 0.4, stc from 0.65 to
0.9) to check their inuence on response, which resulted to be
negligible.
The capacity curve obtained from the analysis by plotting the
reaction force in the xed-end base (base shear) as a function
of top displacement is displayed in Fig. 17. A median vertical
section reproducing the cracked conguration of the model at
the end of the last step of the pushover analysis, and a
view orthogonal to the loading direction showing the distribution of the axial stress in reinforcing bars, are shown in
Fig. 18. The following observations can be drawn from Figs. 17
and 18.

2000

1500

1000

500

100

200

300

400

500

600

700

800

Top Displacement [mm]

Fig. 17. Response curve obtained from the pushover analysis.

A remarkably smooth shape of the capacity curve emerges, as a

consequence of the high number of sub-steps adopted in the
analysis.
The curve is rather linear up to around 1500 kN (with dtop = 20
mm and dr = 0.1%), that is up to around 60% the maximum base
shear, equal to 2390 kN; then, cracking begins to develop significantly in the elements situated on the tension side, and the
curve visibly gets non-linear.
This second response phase goes on up to a force of 2200 kN,
with corresponding top displacement of 110 mm (dr = 0.55%),
when the rst plasticization of reinforcing bars occurs.
The plasticization then increases, determining nearly a plateau
zone extended from around 250 mm to around 450 mm; the
maximum shear force is reached for dtop = 300 mm (dr = 1.5%).
A softening branch follows, featuring a strength degradation of
around 0.2 kN/mm up to the last two steps, where degradation
reaches accentuated values of 0.5 kN/mm (step 69) and
2 kN/mm (step 70), while it does not mean a sudden drop in
strength in proximity to the numerical solution divergence point.
Cracking extends rather uniformly over the tension side,
whereas crushing is never attained. The maximum compression
values in the external ber of base section are no greater than
0.25 fc. This indicates that concrete is far from ultimate strength
conditions on the compression side of the columns at the last
step of the analysis.
Plasticization of reinforcing bars in tension is disseminated over
about 1/3 of the height. The extent of the macroscopic plastic
hinge zone of columns determined via numerical analysis
shows that the hypothesis of uniform resistance is not veried
in the non-linear eld, although cracking is spread rather uniformly along the height. The greatest rotations are obtained
during the entire loading process in the cross section situated
at a distance of 3.3 m from the base, which practically coincides
with the mid-height section of the plastic hinge zone.
These observations underline that numerical collapse is not
determined by a failure of the constituting materials, but by an
excessive deformation of the octahedral elements in various portions of the mesh. Deformation is not sensitive to the number of
sub-steps, which was increased further to a value of 10,000 in
the 70th step, without any practical consequences. Moreover, the
response curve highlights satisfactory behavioral capacities of columns, with an elastic response for rather high base shear values,
and good ductility resources. A formal interpretation of the results
of the pushover enquiry is presented in the next section.

752

S. Sorace, G. Terenzi / Engineering Structures 49 (2013) 743755

Fig. 18. Cracked conguration of the model and stress distribution in reinforcing bars at the end of the last step of the pushover analysis.

4.3. Formal seismic assessment based on the pushover analysis

response
The pushover response was evaluated for the four reference
seismic levels established by Standards [5], that is, in addition to
the basic design earthquakeBDE, the frequent design earthquake
(FDE, with a 81% probability of being exceeded over the reference
time period VR), the serviceability design earthquake (SDE, with a
50%/VR probability), and the maximum considered earthquake
(MCE, with a 5%/VR probability). Relevant top displacement demands dtop,FDE, dtop,SDE, dtop,BDE, and dtop,MCE were calculated by
referring to the site-dependent displacement response spectra prescribed for the city of Turin [5], scaled at the amplitude of the four
earthquake levels. The maximum response displacements came
out for the internal columns, as they are characterized by the highest vibration period as compared to the remaining columns, as
noted in Section 2. However, due to the little differences existing
with the periods of the side and corner columns, little differences
with the relevant displacement demands resulted too. The values
calculated for the four seismic levels proved to be: dtop,FDE = 19.8
mm (drFDE = 0.1%), dtop,SDE = 25.9 mm (drSDE = 0.13%), dtop,BDE =
52.9 mm (drBDE = 0.26%), and dtop,MCE = 59.7 mm (drMCE = 0.3%).
The response was assessed by referring to the four classical
structural performance levels considered by Standards [5], i.e.,
similarly to most performance-based international Seismic Standards, Operational (OP), Immediate Occupancy (IO), Life Safety
(LS) and Collapse Prevention (CP). The top displacement limitations
for the four levels were xed as follows. At the OP structural limit
state, the response must be elastic, although not strictly linear.
Looking at the response curve in Fig. 17, the OP-related limit can
be assumed as a rounded average of the values corresponding to
the end of the linear and non-linear elastic stages, dtop = 20 mm
(dr = 0.1%) and dtop = 110 mm (dr = 0.55%), that is, dtop,lim,OP = 60
mm (drlim,OP = 0.3%). The 0.3% drift value also coincides with the
non-structural performance limit normally adopted by international Seismic Standards and Recommendations for the assessment
and rehabilitation of existing buildingsamong which ASCE/SEI
41-06 [22]for the OP level in the presence of damageable nonstructural elements directly interacting with the main structural
members (in this case represented by the glazed faades interacting with the side and corner columns, as discussed in Section 6). At
the IO structural limit state, little and easily reparable damage is
accepted, with very limited reductions in terms of resistance and
horizontal stiffness. This condition practically corresponds to the
early plastic response stage following the rst plasticization of

the bars in the critical zone of columns (which begins at a displacement of 110 mm, as noted above, but develops appreciably beyond
120 mm). Therefore, the top displacement limit for the IO level
dtop,lim,IO can be xed at 120 mm (drlim,IO = 0.6%).
The LS and CP structural limit states are assessed in terms of
plastic response levels. Detailed criteria for the evaluation of R/C
columns are offered in [22], where a series of acceptable limits
for the plastic rotation angles are formulated as a function of the
geometrical and reinforcement characteristics of the members, as
well as of the axial force computed from the analysis. Although
these limits are suggested for columns belonging to frame structures, in the absence of specic indications for the special case of
free-standing cantilever columns, they can be reasonably extended
also to this type of elements. Based on the characteristics and the
axial force values calculated for the columns of Palazzo del Lavoro,
the suggested limits of plastic rotations result to be equal to 0.012
radiansLS, and 0.016 radiansCP, respectively. By assigning the
two values to the median section of the plastic hinge zone situated
at the height of 3.3 m, where the maximum response rotations are
recorded in the pushover analysis (Section 4.2), the following top
displacements are derived: 472 mm and 606 mm. These values
are then assumed as the displacement limitations for the LS and
CP performance levels, dtop,lim,LS and dtop,lim,CP, with corresponding
drift limits drlim,LS = 2.36% and drlim,CP = 3.03%, respectively. It is
noted that these drift limits fall within the reference ranges ([2
3%]LS, [34%]CP) typically proposed for R/C structures, when
no direct correlation with member rotations or other local response parameters is formulated.
By comparing the top displacement demands relevant to the
four reference seismic levels with the limitations established for
the four basic response limit states, a remarkably high seismic performance of the columns emerges. This is a consequence both of
their overstrength factors, already highlighted by the linear dynamic analysis, and of the low seismicity of the site of the building.
Concerning the latter, the numerical enquiry was completed by
examining what the performance of columns would be should
the building be located in a high seismicity zone in Italy, rather
than in Turin, i.e. a low-to-moderate seismicity area. In this hypothetical situation, the spectral ordinates would be up to three times
greater than in the Turin spectra, averagely for the four seismic levels. By assuming this mutual rounded amplication factor in the
computation of displacement demands, they would increase to
dtop,FDE,inc = 59.4 mm (drFDE,inc = 0.3%), dtop,SDE,inc = 77.7 mm (drSDE,inc =
0.39%), dtop,BDE,inc = 158.7 mm (drBDE,inc = 0.79%), and dtop,MCE,inc =
179.1 mm (drMCE,inc = 0.9%). For these magnied values, the

S. Sorace, G. Terenzi / Engineering Structures 49 (2013) 743755

assessment analysis gives the following results: dtop,FDE,inc = dtop,lim,OP,

dtop,SDE,inc < dtop,lim,IO,
drBDE,inc < dtop,lim,LS,
and
dtop,MCE,inc < dtop,lim,LS, where the conditions expressed by the second, third and fourth inequalities are met with wide margins. This
corresponds again to a very satisfactory performance, giving rise to
three diagonal (FDEOP, SDEIO, BDELS), and an over-diagonal (MCELS) correlations between performance levels and earthquake levels. This underlines again that the columns were
conceived with considerable structural redundancyalso observed
in similar columns included in other outstanding R/C buildings designed by Nervi in the same periodwith the aim of emphasizing
their monumental look and warranting prudentially wide safety
margins with respect to the higher performance demands that
could have been imposed in future generations of Technical Standards, rather than optimizing their original dimensions.

753

Fig. 20. Original design drawing of a slab eld, and tensile stress lines with
background draw of the ribs.

5. Analysis of R/C gallery oors

A set of symmetrical technical joints separates the R/C gallery
oors in four identical angular zones and eight identical central
zones. As mentioned in Section 4.1, concrete and steel are of the
same types as used for the R/C monumental columns. The most
stressed members belong to the four angular portions, the original
design drawing and nite element model of which are displayed in
Fig. 19. The model, generated again with SAP2000NL [4], is more detailed as compared to the one incorporated in the global model of
the building by which the modal analysis was carried out, illustrated in Section 2. Indeed, in this case the geometry of the shell
elements simulating the at slabs exactly follows the shape of
the formworks. Furthermore, the ribs are expressly modelled by
a set of additional frame elements. Based on these renements,
the model shown in Fig. 19 totals 11,264 shell and 3266 frame elements (the model of the same corner zone included in the global
model of the building discussed in Section 2 is composed of 846
shell and 384 frame elements).
The diagrams of the bending moment on the beams resulting
from the analysis are also drawn in Fig. 19, showing high peak negative and positive values on the robust perimeter beams of the
internal square slab elds. The equal tension stress lines on a slab
eld are plotted in Fig. 20 in superimposition to the plan of the ribs,
traced out in the background, showing a remarkable correlation
between the computational solution and the original design of R/
C members. The verications carried out on the various elements
always gave positive results for bending moments, while a lack
of shear resistance was found in some terminal sections of the longitudinal and internal perimeter beams, as well as in the cantilever
beams. A carbon ber reinforced plastics (CFRP) U-jacket was proposed as strengthening solution for these members, as illustrated
in Fig. 21 for the cantilever beams. One 0.165 mm-thick sheet reinforcement was sufcient for all members, except for the longitudi-

Fig. 21. CFRP-based U-jacket strengthening of the cantilever beams.

nal beams, where a double sheet was required. Here too, the
interventions are characterized by a low architectural impact,
and they are respectful of the monumental value of the building.

6. Analysis of glazed faades

The rst gallery oor subdivides the faades in two independent
zones in vertical direction, with interstory heights of 4 m (ground
zone) and 16 m (top zone), respectively. The top zone can be considered as a continuous faade supported by a number of vertical
steel pennon beams with rounded triangular section, which act
both as vertical load bearing and bold bracing elements. Each pennon beam, hinged at the bottom to a triangular cantilever slab
belonging to the rst gallery oor and on top to the perimeter
beam that connects the external mushroom panels, is bonded to
the horizontal light alloy proles of the faades (which constitute

Fig. 19. Original design drawing and nite element model of an angular portion of the gallery oors, and bending moment diagrams on the beams resulting from the analysis.

754

S. Sorace, G. Terenzi / Engineering Structures 49 (2013) 743755

Fig. 22. Global and detailed views of the deformed conguration of the faades obtained in the analysis at the FDE level.

a mesh with sides of about 2.5 m, as observed in Section 2) by short

and rigid steel beams.
These curtain wall-type faades are among the most sensitive
non-structural elements to in-plane and out-of-plane interstory
drifts. In order to avoid local damage to the glass panes and the
supporting proles, the maximum values of both types of drifts
should be constrained within 5 of the interstory heights
[23,24]. Therefore, this can be assumed as the limit threshold for
the non-structural OP performance level, drNS,lim,OP. Beyond this level of deformation, a rst series of hairline cracks appears on the
glass panes, reparable without interrupting the use of the building
until the drift approximates 1%. This value can be adopted as the
limit for the non-structural IO performance level, drNS,lim,IO. Irreparable cracks are observed on the surface of the panes for drifts
equal to about 1.5%, and cracks extended to the entire surface at
about 2%, with the latter value marking the acceptable boundary
for the LS non-structural performance level, drNS,lim,LS. The glass
panes become fallout-prone for drifts around 3%, which can be
xed as the upper limit for the CP non-structural performance level, dNS,lim,CP.
The evaluation of the performance of the glazed faades was
carried out by modal superposition analyses, scaling the input response spectra to the four normative earthquake levels discussed
in Section 4.3. The analyses were developed by the complete nite
element model of the building described in Section 2, combining
the contributions of the 83 modes needed to activate a summed
mass greater than 85% along the x and y axes in plan, as well as
around the vertical axis z, as discussed in Section 2.
The maximum drifts were obtained in the upper area of the faades, caused by horizontal displacements of the rst and second
gallery oors always notably greater than those of the roof panels.
Furthermore, the out-of-plane components were always around
two times greater than the in-plane ones. As way of example of
the results obtained from the analysis at the FDE earthquake level,
the maximum deformed conguration of the model along x direction in plan, and a zoomed view of a terminal zone, are displayed in
Fig. 22. These images highlight the differential displacements between oors and roof, which induce the most accentuated outof-plane deformations. The deformed congurations derived from
the remaining levels of the seismic action are totally similar from
a qualitative viewpoint.
The maximum out-of-plane drift values computed for the four
earthquake
levels
are:
drfa,FDE = 1.28%,
drfa,SDE = 1.55%,
drfa,BDE = 2.19%, and drfa,MCE = 2.39%. By comparing these demands
with the limitations established for the four basic non-structural
limit states, the following inequalities are found: drfa,FDE <
drNS,lim,LS,
drfa,SDE < drNS,lim,LS,
drfa,BDE < dNS,lim,CP,
drfa,MCE <
dNS,lim,CP. The resulting performance is relatively poor, with local
(FDE) and extended (SDE) irreparable cracks on the glass panes
only satisfying the requirements for the Life Safety non-structural
performance level, as concerns the two lower earthquake levels,
and very severe damage, with drifts just below the limit relevant

to the Collapse Prevention state, for the upper earthquake levels.

Therefore, the faades represent the most vulnerable elements of
the building, as a consequence of the marked differences existing
between the displacements of the gallery oors and the roof.

7. Conclusions
The assessment study carried out on Palazzo del Lavoro allowed improving the knowledge on its constituting materials and
construction details, as well as getting a better understanding of
the design concept of its structural members, and checking the degree of correlation of the computed stress states to the original calculations and the geometrical shapes conceived by Pier Luigi Nervi
and co-workers. At the same time, the actual safety conditions of
the building and the seismic performance of the main members
were evaluated according to the criteria of current Technical Standards, and retrot hypotheses were proposed for the steel roof
beams and some beams of the R/C gallery oors, which failed to
pass some of the verications. Specic remarks deriving from the
results of the study are reported below.
Neither type of buckling verications carried out on the beams
of the mushroom steel roofglobal lateraltorsional and local
web panelwas met. The values of the critical stress and the
unsafety factors obtained from the normative formulas were
close to the values deducted from the nite element analysis.
This is an interesting result of this section of the study since
the output of buckling calculations developed by commercial
calculus programs is generally limited to the list of buckling
eigenvalues and eigenvectors, with no direct indications about
the critical stress values to be considered in structural verications, which leaves a considerable margin of uncertainty for
professional users. For web panel verications, the best correlation to the values derived from the nite element analysis is
particularly obtained by the formula included in a previous edition of the Italian Standards on steel structures, where the
effects of normal and shear stresses are jointly considered in
the calculation of the ideal critical stress, unlike in the latest
Standards edition.
The unsafe conditions of the steel beams coming out from the
buckling analysis could be easily overcome by the low-impact
and respectful retrot intervention proposed, which allows
reaching over 500% increase of the rst buckling factor, as compared to the initially computed value, for all beams.
The pushover analysis carried out on the monumental R/C columns by an integrally cracking/crushing computational model
showed remarkable response capacities of these members, with
top displacement demands constrained below the limitations
established for the Operational performance level up to the
highest normative seismic input amplitude. The rst reason of
this very good performance lies in the wide safety margins

S. Sorace, G. Terenzi / Engineering Structures 49 (2013) 743755

assumed in the original calculation of columns, also conrmed

by the results of the linear modal superimposition dynamic
analysis, which highlights that the original design attempted
to emphasize the monumental look of columns also by increasing their cross-sections (although by not impairing elegance,
thanks to a rened and daring development of geometry along
the height), rather than by optimizing their dimensions. The
second reason of the high performance of columns is represented by the low seismicity of the site of the building. To test
the inuence of this parameteralso considering that the
design had necessarily included, at the time, only a gross estimation of seismic forcesthe assessment analysis was repeated
by hypothesizing to ideally locate the building in a high seismicity zone in Italy. This originated a satisfactory output too,
consisting in three diagonal (FDEOP, SDEIO, BDLS) and
an over-diagonal (MCELS) correlations between performance levels and earthquake levels.
At the nal stage of the pushover process, the extent of yielding
in the reinforcing bars on the tension side of columns is limited
to about 1/3 of the height. This underlines that Nervis design
hypothesis of uniform resistance, approximately conrmed by
the linear analysis, is not veried in the non-linear eld.
The equal tension stress lines-governed design concept of the
at slabs of the R/C gallery oors was precisely conrmed,
instead, by the computational solution. The CFRP U-jacket
strengthening solution proposed for the nominally unsafe rib
members is non-invasive and respectful of the monumental
value of the building, as in the case of the interventions outlined
for the steel roof beams.
The seismic assessment analysis of the glazed faades, developed by referring to non-structural performance limitations
specially formulated in the study, revealed that they represent
the most vulnerable building technology included in the edice.
Since effective retrot interventions on relevant load bearing
and bracing systems are difcult to be carried out at reasonable
costs and limited cosmetic impact, and considering that the curtain wall faades have also rather limited thermal and acoustic
insulation capacities as compared to the current standards, a
complete technologicalstructural redesign of the faades
appears to be necessary in any possible future architectural
refurbishment projects of the building.
In addition to the detailed information gathered on Palazzo del
Lavoro, the study offered new contributions to a critical interpretation of Nervis design activity, as well as to an analysis of the
modern heritage R/C structures belonging to the same period.
Acknowledgements
The study reported in this paper was nanced by the Italian
Ministry of Education, University and Research within the PRIN

755

2008 Project (Research Programme Conceiving structures: engineering and architecture in Italy in the 1950s and 1960s. A multi-disciplinary research). The authors gratefully acknowledge
this nancial support.
References
[1] Nervi PL. Scienza o arte del costruire? Rome (Italy): Edizioni della Bussola;
1945 [in Italian].
[2] Nervi PL. Aesthetics and technology in building. Cambridge (USA): Harvard
University Press; 1965.
[3] Sorace S, Terenzi G. Structural and historical assessment of a modern heritage
masterpiece. The Palazzo del Lavoro in Turin. In: Proceedings of the
STREMAH; 2011. p. 22132.
[4] SAP2000NL. Theoretical and users manual. Berkeley (USA): Computers &
Structures Inc.; 2012.
[5] Technical Standards on constructions. Italian Council of Public Works. Rome
(Italy); 2008 [in Italian].
[6] Sorace S, Terenzi G. Fluid viscous damped-based seismic retrot strategies of
steel structures: general concepts and design applications. Adv Steel Constr
2009;5(3):32239.
[7] Eurocode 3. Design of steel structures. Part 1-1: General rules and rules for
buildings. EN 1993-1-1 2005. Bruxelles (Belgium).
[8] Steel structures instructions for design, construction, testing and
maintenance. CNR 10011. UNI 1997. Milan (Italy) [in Italian].
[9] Massonnet C. Rapport sur le thme II C: Poutres de grandes dimensions me
mince. In: Proceedings of the 8th AIPC Congress; 1968. p. 157208.
[10] Timoshenko SP, Gere JM. Theory of elastic stability. 2nd ed. New York
(USA): Mc-Graw Hill; 1961.
[11] Eurocode 3. Design of steel structures. Part 1-5: Design of plated structures. EN
1993-1-5 2006. Bruxelles (Belgium).
[12] Mukhopadhay M, Mukherjee A. Finite element buckling analysis of stiffened
plates. Comput Struct 1990;34(9):795803.
[13] Chin C-K, Al-Bermani FG, Kitipornchai S. Finite element for buckling analysis of
plate structures. ASCE J Struct Eng 1993;119(9):104868.
[14] Eurocode 8. Design of structures for earthquake resistance. Part 1: General
rules, seismic actions and rules for buildings. EN 1998-1 2004. Bruxelles
(Belgium).
[15] Palermo A, Carabellese A, Toniolo G. Numerical validation of pseudo-dynamic
and quasi-static cyclic tests on full-scale precast industrial building
prototypes. In: Proceedings of the 8th Pacic Conference on Earthquake
Engineering. Paper No. 183.
[16] Sorace S, Terenzi G. The damped cable system for seismic protection of frame
structuresPart I: General concepts, testing and modelling. Earthq Eng Struct
Dyn 2012;41(5):91528.
[17] ANSYS. Engineering analysis system theory/users manual. Canonsburg
(USA): Swanson Analysis System Inc.; 2011.
[18] Willam KJ, Warnke EP. Constitutive model for the triaxial behaviour of
concrete. In: Proceedings of the IABSE seminar on concrete structures
subjected to triaxial stresses; 1974. p. 130.
[19] Drucker DC, Prager W. Soil mechanics and plastic analysis for limit design. Q
Appl Math 1952;10(1):15765.
[20] Chen WF. Plasticity in reinforced concrete. New York (USA): Mc-Graw Hill;
1982.
[21] Borri A, Sorace S. FE analysis strategies for structural materials with small
tensile strength. ASME J Pressure Vessel Technol 1993;115(3):15663.
[22] ASCE/SEI 41-06. Seismic rehabilitation of existing buildings. American Society
of Civil Engineers, Structural Engineering Institute. Reston (USA); 2006.
[23] BS 6262-4. Glazing for buildings. Code of practice for safety related to human
impact. BSI committee B/520/4 2005. London (UK).
[24] ASTM E1300-04. Standard practice for determining load resistance of glass in
buildings. West Conshohocken (USA): ASTM International; 2004.