Available via license: CC BY-NC 4.0
Content may be subject to copyright.
Life-Cycle of Structures and Infrastructure Systems – Biondini & Frangopol (Eds)
© 2023 The Author(s), ISBN 978-1-003-32302-0
Open Access: www.taylorfrancis.com, CC BY-NC-ND 4.0 license
Failure analysis of ageing RC bridges: The cases of the Polcevera
viaduct and the Caprigliola bridge
N. Scattarreggia, A. Orgnoni & G.M. Calvi
University School for Advanced Studies IUSS Pavia, Pavia, Italy
R. Pinho
Department of Civil Engineering and Architecture, University of Pavia, Pavia, Italy
D. Malomo
Department of Civil Engineering, McGill University, Montréal, Canada
M. Moratti
Studio Calvi Ltd, Pavia, Italy
ABSTRACT: Many existing bridges now have a life of more than 50 years and the recently
occurred failures underline the need for many of these structures to be assessed and, if required,
retrofitted, in order to avoid catastrophic collapses. The Applied Element Method is herein used
to show as advanced numerical modelling that accounts for as-built structural details, construction
stages, material deterioration effects and structural modifications over time, have the potential to
predict the failure mechanism and study the sensitivity of complex structural schemes to external
actions. To this end, the collapse of the Polcevera viaduct in Genoa (August 2018), and of
the historical multi-span arch bridge of Caprigliola (April 2020) are reproduced. Compari-
sons between numerical results and evidence, indicate that an accurate knowledge of the as-
built structural details and conservation may permit to point out potential criticalities and
analyse the robustness of existing bridges to avoid the premature end of their life.
1 INTRODUCTION
After the Second World War the Italian government announced the need for (re-)constructing the
infrastructural system of the country, introducing guidelines to regulate material and construction
demands that promoted the use of affordable materials such as concrete. Therefore, many of the
existing arch bridges, both in masonry and concrete, that were destroyed during the war were
then rebuilt mainly using concrete. The evolvement from a national emergency situation to a new
economic and commercial development stimulated and pushed for the introduction of innovative
techniques, construction methods and structural systems. Prestressed concrete technology would
have still been regarded as being in its infancy at the time, however, skilled designers such as Levi,
Cestelli-Guidi, Pizzetti, Oberti and Zorzi started nonetheless to apply it to relatively long-span
bridges (Iori & Poretti 2009). Solutions to problems related to creep, temperature variations,
strand relaxation, redistribution effects in statically indeterminate structures, nonlinear and ultimate
response, were only intuitively considered, and sometimes simply overlooked. Between 1956 and
1964 the 760 km “Autostrada del Sole” linking Milan to Rome and Naples was designed and built,
with a total of 853 bridges and viaducts (in addition to 572 overpasses) and 38 tunnels needing to
be constructed. It was only years later, with the development of dedicated software, that it became
possible to model most of the aforementioned complex time-dependent effects.
Today, the road network of Italy counts around 839,629 km with approx. 60,000 bridges and
viaducts, and is managed by four main operators. Almost 40-50% of these bridges have reached
DOI: 10.1201/9781003323020-482
3927
their design life (ANAS 2021), which was typically between 50 and 100 years, and the majority are
reinforced concrete (RC) and precast concrete (PC). In addition, the lack of government-enforced
universal guidelines for the conservation of such complex infrastructural system, and thus the het-
erogeneity of the strategy adopted by each operator, has undoubtedly not contributed to an opti-
mum management of the road network. Indeed, the critical condition of this aging infrastructure
has recently been propelled onto the international stage due to the collapse of the Polcevera via-
duct in Genoa, in August 2018, and is also well epitomized by the fact that a complete registry of
existing bridges is currently still missing. Since then, however, further local and global failures of
other Italian bridges of varying typologies, dimensions, and relevance have occurred, including
the bridge of Caprigliola (Massa e Carrara, Italy), of significant historic value and strategic
importance, which collapsed into the Magra river in April 2020. Considering the limited 2-year
timespan from 2018 to 2020 alone, a relatively large number of local and global bridge failures
have been reported in Italy (see Figure 1), which is quite alarming and highlights the need for an
urgent assessment and rehabilitation of the road infrastructures.
Thus, as underlined in recent guidelines e.g., CNR (2021), research is still strongly needed to
investigate resistance, robustness and potential progressive collapse scenarios of deteriorated
bridges, considering the interaction between continuous damage due to material degradation
and discontinuous damage due to accidental events. As such, the present work aims to provide
a contribution to a better understanding of the sensitivity to various external loadings (both
standard and exceptional, due to natural and anthropic hazards) and input data (e.g., as-design
vs. as-built structural details) of two structural scheme typologies; the arch bridge used in Capri-
gliola and the cable-stayed system adopted for the Polcevera viaduct. Use is made of the rela-
tively innovative Applied Element Method (AEM, Meguro & Tagel-Din (2002)) structural
modelling and analysis approach (as implemented in the computer program ELS (ASI 2020)),
which has been shown to be capable of accurately modelling the response of bridge structures
from their elastic behaviour up to debris distribution (e.g., Malomo et al. (2020)).
2 THE RC CABLE-STAYED BRIDGE OVER THE POLCEVERA RIVER:
51 YEARS OLD
2.1 Overview of its history
The viaduct over the Polcevera river in Genoa, Italy, was built between 1963 and 1967 and is
widely known as Morandi bridge, taking the name of its designer, Riccardo Morandi. The com-
plete viaduct had a total length of about 1100 m, with each of its twelve support points being
Figure 1. Reported bridge failures in Italy in the 2-year timespan between 2018 and 2020. (Red coloured
dots indicate failures reported by national media).
3928
numbered sequentially from the Savona end (Figure 2), with piers 9, 10 and 11 constituting the
so-called “balanced systems” passing over the Polcevera valley and railway tracks below.
The original design of the entire viaduct is described in e.g. Morandi (1967), being then
revisited, in particular for what concerns its construction and loading history, in the work by
Orgnoni et al. (2023a, b). As is common for structures of this type, the Polcevera viaduct was
object of a number of maintenance operations during its lifetime (Rosati et al. 2020), among
which, e.g., the substitution of the original steel guardrails with New Jersey barriers, the enlarge-
ment and strengthening of sidewalks in combination, and also addition of new asphalt layers.
2.2 The collapse in 2018, after 51 years of operativity
It was the balanced-system number 9 (shown in Figure 2) that collapsed on the 14
th
of
August 2018. As is often the case in the aftermath of a structural collapse, the identification of
the triggering event behind the observed stay failure remains an open question, with various
qualitative hypotheses having been thus far formulated and discussed. Initial studies by Calvi
et al. (2019) and Malomo et al. (2020) suggested the failure of one of its stays as a likely initiator
of the observed collapse, an hypothesis that was subsequently confirmed by video footage (la
Repubblica 2019) that clearly showed that the failure mechanism initiated with the occurrence
of a rupture at the top end of the SE stay (see Figure 2). Subsequent forensics studies by Rosati
et al. (2020) identified a number of previously unknown construction defects (e.g. heavily clut-
tered and congested layout of stays‘cables) and consequent (non-visible) material deterioration,
underlining the noticeable mismatch between as-designed and as-built structural details. Pinho
et al. (2023) have shown the influence of such construction defects on the residual capacity of
the ruptured stay, highlighting as their combination with equally unforeseen exceptional load
scenarios (such as e.g. impact of a falling coil and/or strong differential thermal effects) on the
specific day of the collapse, could have indeed triggered the observed collapse.
2.3 Collapse analysis through the AEM-based model
The developed AEM bridge model shown in Figure 3 consisted in the assembly of approx.
100,000 elements for a total of 600,000 degrees-of-freedom. It was built based on pre-collapse
building inspections, thus accounting for the as-built bridge geometry such as e.g. the twist of
the antenna around its vertical axis and the stays in their catenary shape. The model featured
a precise definition of acting loads, geometry, materials, and state of conservation at the time
of the collapse (Scattarreggia et al. 2022a).
A number of numerical modelling studies, referenced throughout the present manuscript,
have been carried out on this bridge by means of linear and nonlinear models, all reciprocally
cross-validated, to investigate the impact of construction errors on the bridge conservation and
stress-strain state at the time of the collapse, as well as the potential triggering factor of the
stay’s failure. The results in Figure 4, taken from Scattarreggia et al. (2022a), show in particular
Figure 2. (a) Longitudinal schematic of the viaduct, (b) pre-collapse photos of the bridge and of pier 9.
3929
that the AEM model with as-built details (also including material deterioration) led to a failure
mechanism noticeably different with respect to that obtained in initial studies by e.g. Calvi et al.
(2019). Indeed, the model with the as-built details duplicated the collapse of the bridge portion
in the immediate moments after the stay failure (see the comparison between numerical and as-
observed counterpart in Figure 5), a result that could not be obtained in the earlier preliminary
work by e.g. Calvi et al. (2019) and Malomo et al. (2020), where the actual as-built construction
details, loading history and state of conservation were not yet fully known.
3 THE MULTI-SPAN RC ARCH BRIDGE OF CAPRIGLIOLA: 71 YEARS OLD
3.1 Overview of its history
The multi-span arch bridge over the Magra river, was first built in 1908, as a link between the
two municipalities of Caprigliola and Albiano Magra, in Tuscany, Italy (see e.g. Ceradini
Figure 3. Mesh discretization (Δ indicates the average smallest and larger mesh dimension) adopted for
various bridge parts, reinforcement details and restraints adopted in the AEM model.
Figure 4. Frames of the collapse simulation induced by a stay failure, considering both the initially
assumed (top) and as-built (bottom) structural details assumptions.
3930
(1951)). It was designed by the Italian engineer Attilio Muggia, one of the Italian pioneers in the
field of reinforced concrete. At that time, with its 260-m length, constituted by five 51-m spans,
the bridge had the record of the longest bridge ever built worldwide. Unfortunately, it was bom-
barded during the Second World War, and it thus had to be rebuilt between 1945 and 1949.
The new design was inspired by the engineer Robert Maillart, consisting of five variable-thickness
slender RC arches that were supported on the existing original piers (Figure 6). It involved the
change from a fixed-arch static scheme to a three-hinges arch bridge, as well as an increase of
permanent loads due to carriage enlargement to meet the increase of traffic with respect to the
time of its first construction. The static scheme change aimed at achieving a theoretically per-
manent state of compression in the arches, which would be a statically favourable characteris-
tic, and deemed appropriate by the designers to manage potential settlements of the original
foundations of the bombarded bridge. Subsequently, during the 90’s, to adapt the bridge to
the road traffic of the twentieth century, the carriageway was enlarged by about 2 m (from
8 m to 10 m) and its thickness increased by about 0.35 m (as estimated from the post-collapse
remains).
Figure 5. Comparison between the observed (top) and numerical collapse (below).
Figure 6. (a) Bridge over the Magra river built in the early ‘900, (b) bridge rebuilt in 1940’s, and
(c) schematic of the whole rebuilt bridge.
3931
3.2 The collapse in 2020, after 71 years of operativity
On April 8
th
, 2020, around 10:20 am, after 71 years of operativity, the bridge suddenly col-
lapsed (see Figure 7). Although the bridge was generally crossed by more than 2,000 vehicles
per day, there was little traffic on that occasion due to travel restrictions in place due to the
COVID-19 pandemic, and only minor injuries to two truck drivers were recorded. According
to an eyewitness, who was driving on the bridge a few seconds before the event, localized
damage, not well identified, would have spread rapidly starting from the Caprigliola side,
causing a progressive domino-type of collapse.
Similarly to what happened after the collapse of the Morandi bridge, many qualitative
hypotheses were formulated on the possible underlying collapse causes, among which (as
described in Scattarreggia et al. (2022b); i) materials degradation, ii) foundations subsidence,
iii) excessive loading, iv) a slow slope slippage on the Caprigliola side. However, to date, the
actual cause, or, perhaps better and as usually happens, the contributing causes to the collapse
are still unknown. Notwithstanding, from a preliminary analysis of Figure 7 it is readily possible
to note that the residual rotations, with respect to the transversal axis of the bridge, of pier P2
and P3 towards abutment A2 are a clear indication that the progressive collapse of the bridge
could not have started from the side of abutment A1, given that the piers could only have
rotated towards the direction from where arch support became absent, and would suggest that
something must have instead occurred on the opposite side, i.e. on Caprigliola side.
3.3 Collapse analysis through the AEM-based model
The AEM-based numerical model consisted of an assembly of approx. 100,000 elements for
a single bridge span, shown in Figure 8 below. The model has been cross-validated using trad-
itional 3D and 2D FE models, i.e. a 3D model in MidasFEA (MIDAS IT 2021), which eigen-
value analysis results were compared with onsite dynamic characterisation tests, and a 2D
model in VecTor2 (VecTor Analysis Group 2020), as shown in Scattarreggia et al. (2022b).
The analyses carried out indicated that the vertical load capacity of the bridge was well above
the stresses induced by today’s traffic loads, and hence notwithstanding the fact that the con-
struction of the additional slab in the 90’s had led to a considerable increase in stresses close
to the pier-arch connections (+50%), the bridge was still capable of supporting a vertical load
significantly higher than that acting at the time of the collapse. The potential deterioration of
Figure 7. Post-collapse debris distribution: (a) top view, (b) global perspective view from the South.
3932
the steel and concrete materials, hypothesized both independently as well as in combination,
was not a critical factor for the collapse of the bridge under the action of vertical loads
(assumed to be uniformly distributed).
On the contrary, small settlements/rotations at one end of an arch potentially induced on the
piers in the riverbed by the erosive action of the river water, or near an abutment due to the
thrust of slow slope slippage present on the Caprigliola side, would have had the potential to
generate localized damage at critical sections near the pier-arch connection capable of inducing
the failure. Such failure potential would have been further increased by the unintended variations
in the height of the throat of RC hinges, which would have considerably reduced their rotational
capacity compared to what it should have been for a 3-hinged arch bridge (theoretically iso-
static), leading to a concentration of damage in correspondence of the critical section at pier-arch
connection. Figure 9(a) shows the scenario following a failure in the span on the Caprigliola side
at the most stressed section as inferred from the analyses considering small displacements/rota-
tions of the pier/abutment on the Caprigliola side (as potentially induced by the recorded slow
slope slippage). The final configuration of the debris predicted by the model (at 7.5 seconds in
Figure 9(a)) is compatible with that observed on site (Figure 9(b)).
4 CONCLUSIONS
The AEM seems to be a computational efficient numerical analysis method to study the
response of quite complex and large RC bridge structures taking into consideration as-built
Figure 8. (a) AEM model of a single bridge span, with the nomenclature of main constituting compo-
nents, and (b) comparison between original drawing reinforcement details and numerical counterpart.
Figure 9. (a) Predicted progressive collapse mechanism of the entire bridge as triggered by a failure close to
the pier-arch connection potentially caused by slope slippage at the abutment A2 and (b) observed debris.
3933
structural details as well as changes in loading and materials during the life of the structures.
The application to the cable-stayed Polcevera viaduct and to the arch bridge of Caprigliola
highlighted the criticalities of such bridge typologies. For what concerns the Morandi bridge,
the innovative idea by Morandi of introducing only four main post-compressed stays, seemingly
ingenious and effective, when combined with construction errors (that became known only after
the collapse) and the consequent (non-visible) material deterioration throughout its life rendered
the structure more vulnerable to potential exceptional load scenarios. In the case of the Capri-
gliola bridge, the idea of changing the initial fixed arch scheme into a three-hinges arch configur-
ation, theoretically appropriate to render the bridge less sensitive to imposed foundation
movements, when combined with initially unforeseen changes in the arch hinges throats and the
years long slow slope slippage behind one of the abutments did inevitably render the bridge
more susceptible to the type of complete collapse that was observed.
REFERENCES
Applied Science International LLC (ASI). (2020). “Extreme Loading for Structures (2020).” Durham
(NC), USA.
Calvi, G. M., Moratti, M., O’Reilly, G. J., Scattarreggia, N., Monteiro, R., Malomo, D., Calvi, P. M. &
Pinho, R. 2019. “Once upon a Time in Italy: The Tale of the Morandi Bridge.” Structural Engineering
International, 29 (2): 198–217. Taylor and Francis Ltd.
Ceradini, G. 1951. “Esperienze su recenti ponti italiani di cemento armato: Esperienze sul ponte sul
ume Magra, fra Caprigliola ed Albiano.” Giornale del Genio Civile, 1951.
Consiglio Nazionale delle Ricerche (CNR) 2021. CNR DT-214/2018. Guide to Design of Structures for
Robustness. Rome, Italy.
Iori, T. & Poretti, S. 2009. “The golden age of ‘Italian style’ engineering.” Proceedings of the 3rd
International Congress on Construction History. Cottbus, Germany.
Malomo, D., Scattarreggia, N., Orgnoni, A., Pinho, R., Moratti, M. & Calvi, G. M. 2020. “Numerical
Study on the Collapse of the Morandi Bridge.” Journal of Performance of Constructed Facilities, 34
(4): 04020044. American Society of Civil Engineers (ASCE).
Meguro, K. & Tagel-Din, H. 2002. “Applied Element Method Used for Large Displacement Structural
Analysis.” Journal of Natural Disaster Science, 24(1), 25–34.
MIDAS IT. 2021. “Midas FEA NX - Advanced Nonlinear and Detail Analysis System.”
Morandi, R. 1967. “Il viadotto del Polcevera dell’Autostrada Genova – Savona.” L’Industria Italiana del
Cemento, XXXVII, 849–872.
Orgnoni, A., Pinho, R., Moratti M., Scattarreggia, N. & Calvi, G. M. 2023a. “Critical review and model-
ling of the construction sequence and loading history of the Morandi bridge.” International Journal of
Bridge Engineering, 10 (3).
Orgnoni, A., Pinho, R., Calvi, G. M., Moratti, M., & Scattarreggia, N. 2023b. “Numerical dynamic
characterization of concrete bridge stays.” Earthq Eng Struct Dyn, submitted for publication.
Pinho, R., Scattarreggia, N., Orgnoni, A., Lenzo, S. G., Grecchi, G., Moratti, M. & Calvi, G. M. 2023.
“Forensic estimation of the residual capacity and imposed demand on a ruptured concrete bridge stay
at the time of collapse.” Structures, submitted for publication.
la Repubblica. 2019. “Ponte Morandi, le immagini inedite della tragedia: il crollo ripreso dalla videosor-
veglianza - YouTube.” Accessed September 2, 2021. https://www.youtube.com/watch?
v=a-LfXohbn0U&ab_channel=LaRepubblica.
Rosati, P., Losa, M., Valentini, R. & Tubaro S. 2020. Secondo incidente probatorio - Procedimento Penale
N. 7998/18 R.G.G.I.P. (N.10468/18 R.G.N.R.), Tribunale Ordinario di Genova, Genoa, Italy.
Scattarreggia, N., Galik, W., Calvi, P. M., Moratti, M., Orgnoni, A. & Pinho, R. 2022a. “Analytical and
numerical analysis of the torsional response of the multi-cell deck of a collapsed cable-stayed bridge.”
Eng Struct, 265: 114412. Elsevier.
Scattarreggia, N., Salomone, R., Moratti, M., Malomo, D., Pinho, R., & Calvi, G. M. 2022b. “Collapse
analysis of the multi-span reinforced concrete arch bridge of Caprigliola, Italy.” Eng Struct, 251:
113375. Elsevier.
Scattarreggia, N., Orgnoni, A., Pinho, R., Moratti, M. & Calvi, G. M. 2023. “Numerical modelling of
the impact of a falling object on a bridge deck”. Eng Fail Anal, submitted for publication.
VecTor Analysis Group. 2020. “VecTor2 – NLFEA program for the analysis of two-dimensional
reinforced concrete membrane structures.”
3934
