You are on page 1of 26

Application of reliability

analysis to heritage buildings


Difficulties
• Missing geometric data
• Information about the inner core of the structural elements is also missing
• Characterization of the mechanical properties of the materials used is difficult and expensive
• Large variability of mechanical properties, due to workmanship and use of natural materials
• Significant changes in the core and constitution of structural elements, due to long construction
periods
• Construction sequence is unknown
• No distinction between structural and decorative parts
• Existing damage in the structure is unknown
• Regulations and codes are non-applicable
• The behaviour of the connections between masonry elements (walls, arches and vaults) and
masonry elements and timber elements (roofs and floors) is usually unknown
Inferences
Uncertainties
Hence the More complex and Simplified indices
should be
quantitative results accurate methods used in
accounted for and
of structural do not ensure understanding the
hence probabilistic
analysis much be reliable and better crude behaviour of
analysis are better
inspected carefully results the structure
suited
Papers reviewed
• Lourenço, Paulo B., and J. A. Roque. "Simplified indexes for the
seismic vulnerability of ancient masonry buildings." Construction and
Building Materials 20.4 (2006): 200-208.

• Lorenzoni, Filippo, et al. "Uncertainty quantification in structural


health monitoring: Applications on cultural heritage
buildings." Mechanical Systems and Signal Processing 66 (2016): 268-
281.

• Beconcini, Maria L., et al. "Probabilistic reliability assessment of a


heritage structure under horizontal loads." Probabilistic Engineering
Mechanics(2016).
1. Simplified indexes for the seismic vulnerability of ancient masonry buildings

• Ancient masonry structures at high


• Simplified indices used to
risk during seismic events due to
study the performance of
improper connections between
buildings under seismicity
components
• 58 Churches selected all
• Overturning collapse of the
over Portugal for this
perimeter walls
exercise
• Lead to extensive study to
prioritize retrofitting works and
recreate damage scenarios

• Churches - open plan, greater height to width ratio,


presence of thrusting horizontal structures from
Index 1 : In-plan area ratio vaulted ceilings and timber roof.
Index 2: Area to weight ratio
• Ample geometry survey drawings and
Index 3 : Base shear ration documentation available
• Portugal is divided into 4 seismic zones (A-D),
higher to lower: 25 from A; 11 from B, C, D each.
In plan ratio: Performance Base shear ratio:
Criteria
 A wi is the area of the  F E = total base shear(Vsd)
earthquake resistant walls  F Rd = Shear strength of
in the direction i the structure( contribution of
 S is the total plan area Area to weight ratio: the earthquake resistant
walls).

 A wi is the area of the


earthquake resistant walls
• Eurocode 8 recommends in the direction i
values up to 5–6% for regular  G is the quasi-permanent • FE = β X G, where β is an
structures with rigid floor vertical action. equivalent seismic static
diaphragms
coefficient related to the design
• In cases of high seismicity, a ground acceleration.
minimum value of 10% seems
to be recommended for • In cases of high seismicity, a • F Rd is related to the cohesion
historical masonry buildings minimum value of 1.2 m2/MN and friction angle
• High seismicity cases - design seems to be recommended for
historical masonry buildings • A minimum value equal to one
ground acceleration larger than
seems acceptable.
.2g

Properties taken
1. Density of masonry = 20kN/m3
2. Weight on roofs = 2kN/m2
3. β = .22
Commentary
• Index 1 and Index 2 are independent of the design ground acceleration
• These indexes should be larger with increasing seismicity, assuming buildings to have identical
safety
• For indexes 1 and 2, the seismicity is taken into account by multiplying the threshold by
the seismicity coefficient a according to the zones A, B, C and D, respectively equal to 1.0, 0.7,
0.5 and 0.3.
• Index 3 should be constant in different seismic zones, as it considers the effect of seismicity

Results
1. Index 1 indicates a unexpected variation for the churches, because the average values exhibit
minor differences according to the seismicity contrarily to the expected dependency (A > B >
C > D).
2. Four churches (16%) in zone A and three churches (27%) in zone B, violate the adopted
criterion.
3. Not a single building violates the criterion of Index 2 and seems to indicate that the threshold
needs revision and is in conflict with Index 1.
4. Index c3, as a direct result from the constant values of indexes 1 and 2, exhibits increasing
values with decreasing seismicity.
5. On average, Index is on the verge of violation for the adopted criterion of zone A, but
adequate for the other zones. Individually, seventeen churches (68%) in zone A and one
church (9%) in zone B, violate the adopted criterion.
Interpretation of results

1. Index 1 is independent from the height, which is considered a major drawback. Therefore, only
Indexes 2 and 3 are further analyzed.

2. The comparison between Index 2 and Index 3 is equivalent to compare (1/h) and (1/b), or
height and seismicity, if cohesion is ignored.

3. Equivalent heights of the churches in the seismic zones are calculated by h equiv = G/ (Aw × γ)

4. With this equivalent heights, the threshold values of Index 2 is recalculated for zones A, B, C,
D as shown in the following table.

5. The simultaneous fulfilment of Indices 2 & 3 is considered as the safety criteria.

6. Considering the above clause, 9 churches of the 22 churches in the previous run and 58 of the
original churches are deemed to be unsafe and needs further study.

7. The threshold value of Index 3 also needs further revision, as 20 of the last 22 churches violate
the minimum criterion of safety.
2. Uncertainty quantification in structural health monitoring : Applications on
cultural heritage buildings
• SHM : A knowledge based
Utility of monitoring system installed
assessment tool to quantify and
in CH:
reduce uncertainties regarding
structural performance
• Validation of adopted structural
models
• Alternative to intrusive repair work
•Identify progressive damage
•Validation of effectiveness of
• Cultural heritage sites –important
strengthening techniques
cultural and economic asset
•Replacement if required, of
strengthening techniques to allow
• Italy – characterized by high
continuous control of response
seismicity

• Paper reports application of monitoring The Roman Arena of Verona


 acquire the vibration characteristics of
strategies to minimize interventions and
the monument by means of
control response when damage detected. accelerometers
 control the surveyed crack pattern
• Statistical model and algorithms through displacement transducers
developed to reduce uncertainties
The “Sala dei Battuti” of the
Conegliano cathedral
• Post processing techniques applied to
 a non-contact optic monitoring system
identify modal parameter and filter out installed to evaluate the crack opening
environmental effects in frescoed surfaces
Case Study 1

•1st Century AD Monitoring Systems


• 1939 – to prevent visible out-of- 1. Since 1990s:
plane tilt and protect from WWII,
Geometric, Material survey, NDT (dynamic
construction of massive buttresses
Damage survey, Visual identification tests) and
• At the end of the war removed and inspections , Deep MDT (mechanical
replaced with post-tensioned steel inspections continuous coring and flat
jack tests)
cables vertically placed along the
entire height of the massive pillars
2. Since 2010 :
Vibration characteristics 16 single-axis
Static Inner gallery of the Arena accelerometers
sensors
Behaviour of the crack 20 displacement transducers
Displacemen Main cracks pattern
t transducers
Correlated with 4 sensors of temperature and
6 Acc Wing of the Arena : two environmental parameters relative humidity.
different levels along the out-
of-plane direction  3 different but interconnected aspects:
7 Acc Top of arches along the (i) variations control of the static
elliptical perimeter of the measurements,
structure radially (ii) daily identification of modal parameters
3 Acc Base of the building to record
(iii) registration and analysis of exceptional
the ground acceleration events (e.g. earthquakes, concerts, etc.).
Uncertainty reduction in modal
parameters estimation of the Arena Validation of the procedure
of Verona
• Proposed algorithm successfully tested for
• OMA ( Operational Modal Analysis) identification of modal parameters
techniques both in time and frequency
domain applied. • Time histories, daily recorded by the
dynamic monitoring system, were used as
• Enhanced Canonical Correlation Analysis input data
(ECCA) - Authors have developed an
automated OMA technique to reduce • In this study total of 460 analyses have
uncertainties arising from numerical analysis been conducted
• Advanced cluster analysis to identify stable
modes of the structure – elimination of noise • Data also analysed not only a single day
record but also a month of consequential
• Efficient in identification of modes of the monitoring (30 setups) before entering in a
wing of the Arena of Verona (slender and long-term monitoring period
flexible structure) and the Amphitheatre
(stiff and massive) • Based on the initial dynamic identification,
• Other techniques fail - the structure being the variations of the fundamental natural
stiff and massive - characterized by high frequencies of the Arena's Wing are
damping values - ambient vibrations not constantly extracted from the measured
able to excite sufficiently well all the vibrations.
frequencies in the range of interest
Results

Fig. 9a shows all


modes that have
been identified
through the 460 per
formed parametric
modal analyses

Fig. 9b represents
the identified poles
after the cleaning
process of the
stabilization
diagram

Fig. 9c shows the


modal parameters
obtained after
cluster analysis
Results

Comparing Fig. 9c and


Fig. 9d, most of the
noisy modes are
eliminated by
inspecting the
consistency between
modes estimated

First structural modes


extracted from the
cluster analysis and
represented in a
frequency vs. damping
plot in Fig. 9e

Fig. 9f shows that


mean values and
confidence intervals
confirm the estimated
modal parameters with
very small
uncertainties
Results
Results
Reliability of the identified
modes

The probability distribution


(red line) of the first eigen
frequencies estimated over the
460 analysis compared with the
results of the normal
distribution, obtained from the
sample mean and sample
variance (blue line).

Frequencies estimated
for 30 days and their
95% confidence level is
displayed. The estimates
are in general accurate
but they vary from one
record to another
Results

Based on the initial dynamic identification, the variations of the fundamental natural frequencies of the Arena's Wing
extracted from the measured vibrations. The frequency development vs. time and vs. temperature gives a sound indication on
the dynamic characteristics of the structure and it is strongly related to the environmental parameters

Summarizes the results of the application of the automated subroutine for the identification of the vibration modes of the
structure during the monitoring period (05/12/2011–30/07/2014).
Results
• The standard deviation of natural frequencies show that environmental conditions have a strong
influence, especially for modes 2–7

• The coefficient of variation (CV) varies from 2.62 % to 5.06 %.

• Mode 1 seems to be less influenced by environmental factors with a CV of about 1.92%.

• Regarding the estimation of modal damping coefficients, the results show higher dispersions.
However, reasonable mean damping values were calculated.

• All structural modes, apart from mode 4, present, on average, damping coefficients around 1%: this
value is certainly underestimated, given that masonry structures are usually characterized by higher
damping ratios in the range of 4–7%.

• It is worth noting that mode 4 is characterized by a higher damping coefficient, confirming that
dissipation mechanisms occur mainly where such mode shape present the maximum bending, i.e. just in
correspondence of the anchorage of post-tensioned steel cables inserted along the massive pillars of the
wing during the strengthening intervention of 1956 .

• In general as light trend of decrement of all natural frequencies is noted with the approach of the hot
season, demonstrating the correlation between variation of natural frequencies and environmental
parameters.
Case Study 2
• The Sala dei Battuti Hall -The Conegliano Statistical modeling of environmental
Cathedral – a masterpiece of the Renaissance effects
building
• Once the environmental effects are filtered
• Frescoes of the Battuti Hall are subject to serious out from recorded data, it is possible to
cracking accurately decompose the measurements
into their reversible and irreversible
• Differential settlements between the facade of the components
cathedral and the surrounding structures, namely the
massive bell-tower – leads to cracking • In this case ARX model used - an Auto-
Regressive output (randomness) and an
Monitoring Systems exogenous input (independent)
• Preliminarily a correlation analysis
• A network of displacement transducers + a
performed with the objective of identifying
temperature / relative humidity sensor – significant
the time series of environmental parameters
cracks
presenting higher correlation coefficients
• A contactless structural damage control method with the time series containing the measured
with high definition optical camera - to control the output (i.e. crack opening).
behaviour of the more visible lesion in the frescos
• Correlation between T and H
• The equipment detects the relative displacement and y given in the table:
of four small optical targets placed on the two
sides of the crack and applied by means of a T-H y- T y- H
special glue, completely removable, that does not r -.8038 -.9409 .776
damage the painted surface
Statistical model of environmental parameters

• The selection of the ARX model ( orders of the exogenous input (na), autoregressive output (nb) and
the delays between input and output (nk)) based on quality criteria proposed by Peeters & De Roeck:
loss function(V), the final prediction error (FPE) and the coefficient of determination (r2) given by:

N = number of measurements, d = number of estimated parameters, k = time step, yk = measured


output, xk = input parameter.
The parameters of the model, equal to (na + nb) were calculated using a least-squares procedure
minimizing the residuals ek

• Once the best ARX model is obtained from the first half-year of monitoring (estimation phase), new
measured environmental parameters (fresh data), collected during the second half-year (validation
period), are used to simulate the response of the studied crack

• From the comparison between the simulated crack behaviour and its recorded counterparts, changes
caused by structural damage can be distinguished from those caused by varying environmental
conditions.
• The Fig. shows the temporal evolution of the
displacement of the frescos crack during a
period of 13 months before and after the
elimination of the environmental effects
performed through the described dynamic
regression models.

• It can be observed that the variation of filtered


value is reduced to a small range, which proves
that the selected ARX model properly describes
the factors with greater influence on the output.
Any other event on the cracks, besides
temperature, would represent a potential
damage
Conclusions

• For the static data case, a new contactless technique was presented for the frescoes monitoring,
revealing a good correlation with standard measurements.
• Modal parameter identification via advanced cluster analyses integrated into system identification
techniques was developed.
• Due to the analysis with different combinations of the model order (n) and the number of rows of the
output block matrices (i), most of noisy modes are eliminated.
• The new procedure, based on cluster analysis and statistical data treatment, proves to be very effective
in the identification of the structural modes. The reliability of the algorithm has been demonstrated on
dynamic data recorded not only from the slender and flexible wing of the Arena but also from the
massive and rigid amphitheatre, difficult to excite uniformly.
• The uncertainty on the modal parameters varies from one record to another, and variability during the
long-term monitoring process is taken into account with mean values and confidence intervals.
• Finally the outcomes of almost three years monitoring are presented, exploiting the developed
automated SHM algorithms running on the central server of university where data are continuously
transmitted and processed on arrival
• Autoregressive models were successfully implemented to filter out the environmental effects from
static records. Damage detection algorithms were applied and confidence intervals calculated around the
predicted values to distinguish changes caused by structural damage from those induced by varying
environmental conditions.
3. Probabilistic reliability assessment of a heritage structure under horizontal
loads
• Methodology for probabilistic reliability assessment of heritage buildings – a historical aqueduct

• Survey of the aqueduct demonstrate that when out-of-plane rotation of pillars > 6 degree, cracks so
significant that temporary structures required to support it.

• Objective - To assess the reliability of the structure under horizontal load , w.r.t the inclination of
the pillars, considering Bayesian updating of the wind speed.

Probabilistic reliability assessment

• Computation of • Verification of
• Identification – structural reliability
• Definition of the probability of failure
ultimate and
limit state  (1- pf) > (1- pd)
serviceability limits
underlying the pf= P[G<0]=P[Z<1]  β > βt
– failure modes –
theory of reliability where target
basic variables to
On the basis of the reliability is
govern the failure E<R;
reliability index β represented by
mode G = R-E=0 ;
Z = R/E=1 ; (1- pd)/ βt
β = Φ-1(pf)
• Definition of the • Bayesian approach
where E = action, R
probability density - defining
= resistance (both where Φ-1 is the
function fi(Xi) for probabilistic models
random variables) standard normal
basic variable Xi for existing building
G = safety factor; Z cumulative
that govern the as the nature of
= safety margin. distribution function
limit state information obtained
.
being indirect.
Methodology of planning tests and inspections on historic structures (ISO13822)
Global model and analysis, local
model
• Structural analyses to reveal static and
dynamic properties, stress patterns and
state of damage.
• Original design and dimensioning
replicated in the analysis model for
better understanding of the defects
• Ad hoc models + Numerical models
• Study of local mechanisms for safety
evaluation due to poor connections
between components

Bayesian updating Definition of prior probability


models
• NDTs useful for refining the knowledge on prior resistance
parameters • Inadequate information - prior
knowledge required for defining
• High non-homogeneity – reliability increased by increasing the
probability models
number of NDT tests though results are mostly qualitative.
• Bayesian updating of environmental action parameters. Eg - new • From literature and vernacular
Bayesian estimate of the seismic hazard based on limited earthquake architecture information, relevant
observations , prior seismic hazard studies, probabilistic model of the information may be derived by
sea level etc similitude.

Verification of structural reliability can be achieved by solving a multidimensional integration


Case study
• Many collapse mechanisms studied - analyzing the
• Medicean aqueduct in Pisa – 16th century in-plane and out-of-plane equilibrium of the arcades
masonry water work - state of decay.
• In-plane forces - Equilibrium was satisfied when the
• 954 spans – masonry arch structure – 6 km long – centering of the line of thrust was within the
each span 4-7 m tall, 7 m width – 1.2 m depth. transversal section of the arches.
• Affected by vertical settlements and out of plane • Out-of-plane forces – a simple model of an isolated
overturning – inclination up to 9 ° - cracks localized arch analyzed, assuming a rigid body on elastic soil
in key section of arches and upper section of pillars. and several load conditions
• Undressed stone masonry – compressive strength • Verification w.r.t wind action revealed low tensile
2.6 Mpa, Elastic Modulus – 1740 Mpa. stresses and admissible compressions for the masonry.
• Built on swampland lower than surrounding plain, • Three cases of seismic safety indices : perfect
alluvial deposits over gravel deposits uprightness of the arcade, initial overhanging and
• Pillars have deep foundation with timber piles presence of transverse buttress - results reveal
having length 6-10m structure vulnerable to earthquake even in the vertical
position

Summary E 1st 2nd 3rd


(Mpa) (Hz) (Hz) (Hz)
• A global FE static and dynamic analysis ( SAP 2000) with respect to out of
plane actions carried out – original and damaged state
• Verification under seismic action : ultimate and serviceability requirements Unda 1740 .847 1.136 1.563
considered mage
d
• None of the examined sections satisfies the serviceability limit state
Dama 870 .599 .806 1.099
• Earthquakes with a return period of only 50 years already produce flexural
ged
cracking at the bottom of the pillars – key section of the arches affected by
hidden cracks
Summary
• Then a local analysis considering only a pillar and the adjacent arches carried out - implementing a
deformable structural model interacting with soil and considering the second order effects;
• Limit states considered : 1. Out of plane rotation affects the whole series of 11 arches bound by two
consecutive buttresses 2. Out of plane rotation affects the central element

• Stabilizing horizontal force is a function of the displacement of the central arch from mean position .
• A kinematic analysis has been performed on the same macro element, considering as a possible collapse
mechanism :the rigid rotation of the structure around a cylindrical hinge located at the pillar’s base.
• The element has been evaluated both in the vertical and inclined position, also considering the presence of a
stabilizing force transmitted by the adjacent elements.
The results of the FORM analysis are shown. The highest reliability
index is obtained when α ≈ 1.7°, when the stabilizing force assumes
the maximum value V ≈ 80.4 kN. For α > 1.7°, the stabilizing force
decreases, and thus the reliability of the structure. Results reveal that
Extreme values of wind speed follows Gumbel the structure is not safe any more for α > 6.5°
distribution – updating of the distribution is
done according to annual wind speed
observations collected at Pisa.

Target reliability Index defined by :

The target reliability index βt= 3.7 The reliability of the element assessed considering the seismic
force and the Ultimate Limit State (ULS). For α < 3° the
structure can be considered safe.
Conclusions

• In the deterministic analysis carried out in SAP2000, it was seen that the structural
element can be considered safe when it stands in a vertical position, while it is near to
collapse for an inclination of 9°; the limit inclination at which the structure can withstand
seismic action is assessed around 4°.

• In the probabilistic study, results confirm the outcomes of deterministic analysis


previously carried out, according to which the structure is safe when it stands in a vertical
position while it is not safe any more when the inclination becomes greater than 3° if
subjected to seismic action and 6° in case of wind force.

• The opinion of historical buildings‘ experts fulfils a special role in the proposed
methodology, not only for establishing prior probability models and identifying most
relevant failure modes, but also in order to confirm the obtained results.

You might also like