Sunteți pe pagina 1din 20

Engineering Structures 217 (2020) 110794

Contents lists available at ScienceDirect

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

Simplified damage models for circular section reinforced concrete bridge T


columns

Camilo Perdomo, Ricardo Monteiro
University School of Advanced Studies (IUSS), Pavia, Italy

A R T I C LE I N FO A B S T R A C T

Keywords: Damage models are fundamental elements within probabilistic frameworks, such as Performance-Based
Bridges Earthquake Engineering (PBEE), for the seismic assessment of RC structures. For bridges, in specific, there are
Circular section column not many available experimental results for the characterization of damage in the RC columns hence the nu-
Damage limit state merical counterpart acquires particular relevance. In this paper, via Monte Carlo simulation and comparison
Component damage model
with complementary experimental results, a damage model consisting of a library of median and logarithmic
Seismic assessment
dispersion values for curvature ductility, displacement ductility and drift, describing the probability of reaching
a particular damage limit state are generated as a function of key parameters of circular RC bridge columns. Four
damage limit states are considered: slight, moderate, extensive and complete. Simplified relationships based on
the full datasets are proposed for easier implementation. The proposed damage models can be used for threshold
identification and economic loss assessment in performance-based earthquake engineering of RC bridges.

1. Introduction These vulnerability measures at the structure level, although applicable


to individual structures, are generally more suitable for large scale or
Current Performance-Based Earthquake Engineering (PBEE) regional assessments, as has been implemented by others in the past
methods aim to communicate the risk associated in existing or new [6].
structures due to seismic hazard in such a way that owners, stake- However, for the PEER-PBEE framework, a key input parameter is
holders and regulatory organizations can easily understand it and use it the probabilistic description of components reaching or exceeding a
to make rational decisions. To accomplish such a task, efforts have been predetermined damage state based on the value of a specific structural
directed in order to communicate this risk in terms of economic loss due response quantity or Engineering Demand Parameter (EDP), since the
to physical damage, economic loss due to closure and interruption of evaluation of damage consequences (e.g. economic losses) is performed
regular operation and the estimation of possible casualties. Among the at the component level and then aggregated to determine the risk state
different approaches available to compute these outputs, the Pacific of the entire structure. Furthermore, current PBEE (e.g. FEMA-P-58,
Earthquake Engineering Research Center framework (PEER-PBEE) [7]) also requires the evaluation of residual deformations after the
[1,2] currently stands as the most thorough. main-shock, of which, residual drifts have been identified as a key
Regarding physical damage, extensive previous research has ad- parameter to assess the repair suitability and the aftershock capacity
dressed the probability of structural systems to reach or exceed a par- [8].
ticular limit state based on geometrical and mechanical parameters. In this regard, Kowalsky [9] developed closed form expressions for
Park and Ang [3] determined the likelihood of reinforced concrete (RC) curvature ductility, displacement ductility and drift ratio for well con-
components to enter a state of collapse using a damage index and fined circular section RC columns based on moment-curvature analysis
subsequently extended this concept to assess the state of “tolerable results and the well-known plastic hinge method [10], as a function of
damage” in RC buildings under a given seismic hazard [4]. In another the column’s diameter, length and axial load. Specific expressions were
implementation of this approach, Hwang et al. [5] used capacity/de- proposed for two different limit states, defined as serviceability and
mand ratios to compute fragility curves for bridges for three damage damage control, which correspond to predefined strain values in either
limit states, defined as no/minor damage, repairable damage and sig- the concrete or reinforcement steel. The outcome of these expressions
nificant damage, as a function of Peak Ground Acceleration (PGA). can be seen as the mean EDP value at which the column reaches the


Corresponding author at: Piazza della Vittoria No. 15, 27100 Pavia, Italy.
E-mail addresses: camilo.perdomo@iusspavia.it (C. Perdomo), ricardo.monteiro@iusspavia.it (R. Monteiro).

https://doi.org/10.1016/j.engstruct.2020.110794
Received 16 October 2019; Received in revised form 4 May 2020; Accepted 11 May 2020
0141-0296/ © 2020 Elsevier Ltd. All rights reserved.
C. Perdomo and R. Monteiro Engineering Structures 217 (2020) 110794

Nomenclature excessive longitudinal reinforcement deformation


ϕDLS − 3, BE section curvature at damage limit state 3 according to
The following symbols are used in this paper: Berry and Eberhard
ϕDLS − 3, MK section curvature at damage limit state 3 according to
a regression coefficient Moyer and Kowalsky
Ag gross section area ϕDLS − 3, S section curvature at damage limit state 3 associated with
Ah transverse reinforcement bar area transverse reinforcement spacing
Ast total longitudinal reinforcement area ϕmax section curvature at core concrete crushing, or long-
ALR column axial load ratio itudinal reinforcement fracture, or at 20% drop of the
α factor for the computation of the shear strength provided maximum moment capacity
by the concrete ϕy section nominal yield curvature
b regression coefficient ϕy, N nominal yield curvature computed using mean material
β fitted logarithmic standard deviation value for a log- parameters
normal distribution ϕ' y section yield curvature
βale aleatory uncertainty around a median γ factor for the computation of the shear strength provided
βcss factor for the computation of the shear strength provided by the concrete
by the concrete H column length
βepis epistemic uncertainty around a median value k plastic hinge length factor
βtot total uncertainty around a median value Lp plastic hinge length
β0 , β1, β2 , β3 , β4 , β5 , β6 ,β7 regression coefficients Lsp strain penetration length
c neutral axis depth M section moment capacity values
c0 concrete cover thickness MN section nominal moment
c. o. v. coefficient of variation. My section yield moment
dbl longitudinal reinforcement bar diameter μ fitted mean value for a normal distribution.
D section diameter μΔ, DLS − i column displacement ductility at each damage limit state
D' section confined core diameter i, i = 1, 2, 3, 4
Δ column lateral displacement capacity values μϕ, DLS − i curvature ductility at each damage limit state i, i = 1, 2, 3,
ΔDLS − i column lateral displacement at damage limit state i, i = 1, 4
2, 3, 4 μΔ, SF column displacement ductility at shear failure
ΔP − Δ column lateral displacement at P-Delta instability μϕ, MK curvature ductility at damage limit state 3 according to
ΔSF column lateral displacement at shear failure Moyer and Kowalsky
Δ y, N column nominal yield displacement computed using mean N number of points used for fitting statistical distribution
material parameters p value at which a fitted distribution becomes statistically
Δ' y column yield displacement significant according to a GOF test
Δ
column drift ratio at each damage limit state i, i = 1, 2, 3, P column external axial load
H DLS − i
ρl longitudinal reinforcement ratio
4
Δ
column drift ratio a shear failure ρυ transverse reinforcement ratio
H SF ρυ, conf transverse reinforcement ratio corresponding to code
Ec concrete elasticity modulus
confinement requirements
EDP engineering demand parameter
ρυ, eff effective transverse reinforcement ratio
Es longitudinal or transverse reinforcement elasticity mod-
s transverse reinforcement spacing
ulus
smax transverse reinforcement reference spacing for computa-
εc concrete strain in the extreme compressed fibre
tion of longitudinal reinforcement buckling
εc, DLS − 3 concrete strain in the extreme compressed fibre at fracture
σ fitted standard deviation value for a normal distribution
of the transverse reinforcement
θ fitted median value for a lognormal distribution
εs strain in the longitudinal reinforcement extreme tensioned
θΔ/ H median drift ratio from regression equation
fibre
θ μΔ median displacement ductility from regression equation
εs, y longitudinal reinforcement yield strain
θ μϕ median curvature ductility from regression equation
εsu longitudinal or transverse reinforcement ultimate strain
θp _bb plastic rotation at the onset of longitudinal reinforcement
f 'c unconfined concrete compressive strength based on stan-
buckling according to Berry and Eberhard
dard cylinder test
θsss inclination of the diagonal shear crack
f 'cc confined concrete maximum compressive strength
V column developed lateral force
fy longitudinal reinforcement yield strength
Vc column shear strength provided by the concrete
f yh transverse reinforcement yield strength
Vp column shear strength provided by the axial load
fu longitudinal reinforcement ultimate strength
VP − Δ shear produced by P-Delta effects.
γ factor for the computation of the shear strength provided
VR column total shear strength
by the concrete
Vs column shear strength provided by the transverse re-
ϕ section curvature response values
inforcement
ϕDLS − i section curvature at damage limit state i, i = 1, 2, 3, 4
Z Moyer and Kowalsky buckling coefficient and
ϕDLS − 3, εc section curvature at damage limit state 3 associated with
transverse reinforcement fracture
ς inclination angle of the compression strut due to axial
ϕDLS − 3, εs section curvature at damage limit state 3 associated with load.

2
C. Perdomo and R. Monteiro Engineering Structures 217 (2020) 110794

corresponding limit state and are intended for use in design; thus, should, in principle, be easily obtained or estimated based on experi-
consideration of the uncertainty on the mean value, another funda- ence or existing statistical data.
mental parameter within the PEER-PBEE framework, was not given.
Furthermore, Panagiotakos and Fardis [11] generated closed form 2. Definition of damage limit states
expressions for drift ratio in rectangular RC components (beams, col-
umns and walls) at yielding and ultimate capacity based on an ex- In this study, four damage limit states are defined, namely, slight,
tensive survey of experimental results and statistical regression ana- moderate, extensive and complete. Qualitatively, the defined limit
lysis. In a similar manner, Berry and Eberhard [12] proposed closed states are associated with the following general physical conditions in
form expressions for displacement ductility and drift ratio for rectan- the columns: i) Slight (DLS-1): related to the yielding of the column,
gular and circular section columns at the onset of concrete spalling and minor damage is expected, however repair actions might be necessary
longitudinal reinforcement buckling, respectively. In both studies, in order to prevent further deterioration due to other effects (e.g. cor-
mean EDP values obtained from the closed form expressions were rosion); ii) Moderate (DLS-2): related to the onset of cover concrete
complemented with the measured dispersion between the fitted equa- spalling, repair actions are necessary in order to replace the affected
tions and the experimental data, therefore, uncertainty was taken into concrete volume and prevent further deterioration; iii) Extensive (DLS-
account. The proposed closed form expressions are a function of the 3): related with the onset of buckling of the longitudinal reinforcement
component geometrical characteristics (e.g. diameter, length, cover or the fracture of the transverse reinforcement (whichever happens
concrete thickness), material mechanical characteristics (e.g. char- first), repair actions might involve the replacement of concrete and
acteristic concrete compression strength, reinforcement yield strength) reinforcement or the reconstruction of the column; and, iv) Complete
and reinforcement quantities. (DLS-4): related with the crushing of the concrete core or the fracture of
The relationships proposed in the two aforementioned studies have the longitudinal reinforcement, the column might be able to carry
been used for the generation of damage models for economic loss as- gravity loads, but resistance to lateral deformations cannot be guar-
sessment in past studies [13]. It is noted that the expressions proposed anteed, this damage state is associated with a condition of collapse.
by Panagiotakos and Fardis [11] and Berry and Eberhard [12] require In line with previous research [9,10], the onset of each damage limit
detailed information (e.g. transverse reinforcement diameter and spa- state is herein defined with a strain-based criterion, as explained in the
cing) about the column under investigation in order to characterize its following subsections.
damage models. This feature may hinder their implementation in si-
tuations when a large stock of existing structures has to be assessed, for 2.1. Curvature at the different damage limit states
which such detailed information is generally not readily available.
Moreover, Billah and Alam [14] developed damage models at sev- For a given column with predetermined geometrical, load, material
eral limit states for shape memory alloy (SMA)-reinforced concrete and reinforcement characteristics, a moment–curvature analysis is
bridge piers based on results from Incremental Dynamic Analysis (IDA) performed and the corresponding output is used to identify the onset of
[15] providing approximate expressions to estimate the residual drift each damage limit state as shown next.
after the mainshock as well. More recently, Stefanidou and Kappos [16]
also developed closed-form expression for the calculation of limit state 2.1.1. Curvature at DLS-1
thresholds in terms of drift ratio based on a series of parametric in- According to the recommendations of Priestley et al. [10], the slight
elastic analyses for circular section RC columns (and other cross-sec- damage limit state can be associated with the curvature computed with
tional shapes), uncertainty in the predictions from the proposed closed- Eq. (1).
form equations was measured directly from the obtained data.
MN
In contrast, other previous research studies dealing either with the ϕDLS − 1 = ϕ' y
My (1)
damage fragility characterization or loss assessment of bridges have
'
implemented damage models limited for a specific column configura- where ϕ y and My are the first yield curvature and moment, respectively,
tion [17,18] and using a priori measures of dispersion in the capacity and are related with a concrete strain of εc = 0.002 or a reinforcement
[19]. Further studies, such as Cardone [20], have employed damage strain of εs = εs, y = f y / Es , whichever occurs first, and f y and Es are the
models constructed with generally accepted formulae to compute mean longitudinal reinforcement yield strength and elasticity modulus, re-
EDP values at the onset of each predefined limit state that are never- spectively. MN is the nominal moment, related with a concrete strain of
theless not directly related with physical phenomena in the investigated εc = 0.004 or reinforcement strain of εs = 0.015, whichever occurs first.
component with a priori values of dispersion in the capacity. Also,
Monteiro et al. [21] and Zelaschi et al. [22] have used, in their bridge 2.1.2. Curvature at DLS-2
fragility assessment studies, predefined mean EDP values generally A conservative estimate of the onset of cover concrete spalling is
accepted to describe the onset of the considered limit state for a specific given by Eq. (2), according to the recommendations of Priestly et al.
column configuration with a priori values for the dispersion in the ca- [10].
pacity. All these studies would benefit with the consideration of more
ϕDLS − 2 = ϕ (εc = 0.004) (2)
refined damage models that somehow take into account the particular
characteristics of each individual column under investigation. where ϕ (εc = 0.004) is the curvature value at a concrete compression
It becomes thus clear that additional tools that allow for the damage strain at the extreme fibre of 0.004.
characterization without such detailed parameters, which generally
would require expensive scanning tasks during reconnaissance cam- 2.1.3. Curvature at DLS-3
paigns or office data collection, are necessary, especially when a large As mentioned above, the extensive damage limit state is associated
stock of structures is to be assessed and a first screening for identifying with onset fracture of the transverse reinforcement or the onset of
critical structures is warranted in order to efficiently allocate resources. buckling of the longitudinal reinforcement. According to Priestley et al.
When such a framework is referred to loss assessment of RC bridges, the [10] a conservative estimate of the concrete compression strain at the
lack of comprehensive approaches becomes even more evident. With extreme fibre for the onset of fracture of the transverse reinforcement is
the above discussion in mind, the main goal of this study is to propose given by Eq. (3), where εc, DLS − 3 is the concrete compression strain, f yh is
approximate, yet accurate, damage models for cantilever circular sec- the transverse reinforcement yield strength, εsu is the transverse re-
tion RC bridge columns, generated on the basis of key column para- inforcement ultimate strain, f 'cc is the confined concrete maximum
meters, such as diameter, length and axial load. These parameters strength and ρυ is the transverse reinforcement ratio, given by Eq. (4).

3
C. Perdomo and R. Monteiro Engineering Structures 217 (2020) 110794

ρυ f yh εsu of buckling of the longitudinal reinforcement is given by Eq. (14):


εc, DLS − 3 = 0.004 + 1.4
f 'cc (3) f y dbl ⎞
H
θp _bb = 0.006(1 + 7.19ρυ, eff )(1 + 3.129ALR)−1 ⎜⎛1 + 0.651 + 0.227 ⎟
According to Kowalsky [9], Eq. (3) is a conservative estimate based ⎝ D D ⎠
on the energy equilibrium approach [23] and experimental results show (14)
that the concrete extreme fibre will be able to sustain deformations at
least 1.5 times larger. in which ρυ, eff is the effective transverse reinforcement ratio, defined in
Eq. (15), D is the section diameter, H is the column length and H / D is
4Ah the column aspect ratio.
ρυ =
D 's (4)
fy
' ρυ, eff = ρυ
In Eq. (4) Ah is the transverse reinforcement bar area, D is the
f 'c (15)
confined core diameter and s is the transverse reinforcement spacing.
Accordingly, the curvature at DLS-3 related with concrete strain at The curvature ductility at the onset of longitudinal reinforcement
transverse reinforcement fracture is given by Eq. (5). buckling according to the Berry and Eberhard [25] is thus given by Eq.
ϕDLS − 3, εc = ϕ (εc, DLS − 3) (16):
(5)
θp _bb
For well confined sections, the extensive damage limit state can also ϕDLS − 3, BE = ϕy +
Lp (16)
be associated to excessive deformation in the longitudinal reinforce-
ment [10] as given in Eq. (6), where εsu in Eq. (6) corresponds to the where Lp is the plastic hinge length, defined in Eq. (17), in which Lsp is
longitudinal reinforcement ultimate strain. the strain penetration length, given in Eq. (18), and k is the plastic
ϕDLS − 3, εs = ϕ (0.6εsu ) hinge length factor, defined in Eq. (19).
(6)
Lp = max {kH + Lsp , 2Lsp} (17)
For poorly confined sections, the possibility of longitudinal re-
inforcement buckling has to be assessed, hence, in this study three al- Lsp = 0.022f y dbl (18)
ternatives were implemented for this purpose. Firstly, the simple for-
mulation recommended by Priestley et al. [10] was evaluated, in which
⎧ ⎛f ⎞ ⎫
the curvature at the onset of buckling is given by Eq. (7): k = min 0.2 ⎜ u − 1⎟, 0.08
⎨ fy ⎬
⎩ ⎝ ⎠ ⎭ (19)
16dbl − s ⎞
ϕDLS − 3, S = ϕser + ⎛ ⎜ (12ϕy − ϕser )

⎝ 16dbl − smax ⎠ (7) The curvature at the onset longitudinal reinforcement buckling is
thus the minimum value from those obtained using Eqs. (5)–(7), (9) and
where ϕser is the curvature associated with a longitudinal reinforcement (16), as shown in Eq. (20).
strain of εs = εs, y , dbl is the longitudinal reinforcement bar diameter,
ϕy = ϕDLS − 1 is the nominal yield curvature, s is again the transverse ϕDLS − 3 = min {ϕDLS − 3, εc , ϕDLS − 3, εs , ϕDLS − 3, S , ϕDLS − 3, MK , ϕDLS − 3, BE } (20)
reinforcement spacing and smax is given by Eq. (8):

⎛f ⎞ 2.1.4. Curvature at DLS-4


smax = 3 + 6 ⎜ u − 1⎟ dbl ≤ 6dbl The curvature at the complete damage state is given by Eq. (21),
f (8)
⎝ y ⎠ where ϕmax is the lowest curvature corresponding to concrete core
where fu is the longitudinal reinforcement ultimate strength. crushing, first longitudinal reinforcement fracture or 20% drop in the
Secondly, the Moyer and Kowalsky [24] buckling model was eval- maximum moment capacity.
uated, according to which the curvature at the onset of longitudinal ϕDLS − 4 = ϕmax (21)
reinforcement buckling is given by Eq. (9):
ϕDLS − 3, MK = ϕy μϕ, MK (9)
2.2. Curvature ductility at the different damage limit states
where μϕ, MK is the curvature ductility at the onset of longitudinal re-
inforcement buckling, given by Eq. (10): Curvature ductility at the different damage states is estimated based
−2.5 on a nominal yield curvature [10] computed using mean material
s
μϕ, MK = 2 ⎛ ⎞ ⎜ ⎟ Z parameters, as defined in the next section, according to Eqs. (22) and
⎝ dbl ⎠ (10) (23), where ϕy, N is the reference nominal yield curvature and μϕ, DLS − i is
where the coefficient Z is given in Eq. (11), in which ALR is the axial the curvature ductility at each damage limit state i.
load ratio, given in Eq. (12), and ρl is the longitudinal reinforcement 2.25εy
ϕy, N =
ratio given in Eq. (13). D (22)
Z = (260 + 325ALR) + (20 − 25ALR)(ρl − 0.5) (11) ϕDLS − i
μϕ, DLS − i =
P ϕy, N (23)
ALR = '
f c Ag (12)

Ast 2.3. Force-displacement relationship


ρl =
Ag (13)
The force-displacement relationship of each column was also com-
'
In Eqs. (12) and (13) P is the external axial load, f c is the un- puted in order to determine the lateral displacement at which each
confined concrete compressive strength based on a standard cylinder damage state is reached, as well as to assess shear failure. The lateral
test, Ag is the gross section area and Ast is the total area of longitudinal force developed in the column is calculated using its height H and the
reinforcement. moment values from the moment-curvature analysis. The corre-
Thirdly and finally, the Berry and Eberhard [25] buckling model sponding displacements at each value of lateral force are computed
was also implemented. In this proposal, the plastic rotation at the onset with Eq. (24):

4
C. Perdomo and R. Monteiro Engineering Structures 217 (2020) 110794

'
⎧Δ y '
ϕ
for ϕ ≤ ϕ' y 2.5. Lateral displacement at the different damage limit states
⎪ ϕy
Δ=
⎨ ' M

' M
( ) '
⎪ Δ y My + ϕ − ϕ y My Lp H for ϕ > ϕ y (24)
Each curvature point computed in the moment-curvature analysis
has a corresponding lateral displacement, therefore, lateral displace-
ments at the attainment of each limit state (ΔDLS − i ) can be retrieved by
where ϕ are the curvature response values, Δ' y is the yield displacement
identifying those corresponding to the limit state curvatures computed
given in Eq. (25) and Lp is the plastic hinge, previously defined in Eq.
previously – Δ(ϕDLS − i ) .
(17).
Furthermore, for the complete damage limit state (DLS-4), the
(H + Lsp )2 possibility of P-Delta instability is assessed. The shear produced by the
Δ' y =ϕ' y
3 (25) P-Delta effect is computed with Eq. (34), where the value of 0.5 is an
empirical factor recommended by Priestley et al. [10] to account for the
It is noted that the computed lateral displacement response of the
typical cyclic response of RC members.
column ignores shear deformations, mainly because of two reasons:
firstly, for typical circular sections for which the aspect ratio (H/D) is 0.5P Δ
VP − Δ =
larger than 2, shear deformations tend to be small, and secondly, and H (34)
primarily, because in this fashion conservative estimates of the lateral It is assumed that the lateral displacement at which P-Delta in-
displacement at the onset each damage state are obtained [10]. stability takes place corresponds to the point where the shear developed
by flexural response is equal to the P-Delta shear, as shown in Eq. (35).
2.4. Shear strength
ΔP − Δ = Δ(V = VP − Δ) (35)
The column shear strength is also estimated in order to identify If such point is found for a given column, the displacement corre-
possible shear failure and its corresponding displacement. In this study, sponding to DLS-4 is taken as the minimum of that of P-Delta instability
the Modified UCSD model is implemented because its consideration of and that identified from the strain-based criterion, as shown in Eq. (36).
the loss of shear strength with increasing curvature demand and its
good match with experimental results in this type of elements [10]. ΔDLS − 4 = min {Δ(ϕDLS − 4 ), ΔP − Δ} (36)
Other proposals that predict shear strength in RC elements under
monotonically or cyclically increasing displacements could also be
2.6. Displacement ductility and drift at the different damage limit states
implemented [e.g. 26,27].
For completeness, the Modified UCSD model as implemented is this
Displacement ductility at the different damage states is estimated
study is described in the following paragraphs. The total shear strength
based on the nominal yield displacement [10] computed using mean
of the column at the critical section is given in Eq. (26), where VR is the
material parameters, as defined according to Eqs. (37) and (38), where
total strength, Vc is the shear strength provided by the concrete, Vp is the
Δy, N is the reference nominal yield curvature and μΔ, DLS − i is the dis-
shear strength provided by the effect of external axial load and Vs is the
placement ductility at each damage limit state i.
shear strength provided by the transverse reinforcement, given in Eqs.
(27), (28) and (29), respectively. (H + Lsp )2
Δy, N = ϕy, N
3 (37)
VR = Vc + Vp + Vs (26)
ΔDLS − i
μΔ, DLS − i =
Vc = αβcss γ f 'c 0.8Ag (27) Δ y, N (38)

Vp = Ptanς (28) In the same manner, drift ratios at which each damage limit state is
reached are computed according to Eq. (39).
π Ah f yh (D − c0 − c ) cotθsss
Vs = Δ Δ
2 s (29) = DLS − i
H DLS − i H (39)
The constants α , βcss and γ in Eq. (26) are given in Eqs. (30), (31)
and (32), respectively. In Eq. (28), ς is the inclination of the compres-
sion strut and its tangent is defined in Eq. (33), where c is the neutral 2.7. Shear failure limit state
axis depth. In Eq. (29), c0 is the concrete cover thickness and θsss is the
inclination of the diagonal shear crack, taken constant with a value of As explained in the previous subsections, flexural and shear re-
30° [10]. sponses are treated separately, meaning that the computed flexural
response does not directly account for the possibility of shear failure
M before reaching its ultimate capacity. Because of this feature of the
α=1≤3− ≤ 1.5
VD (30) proposed procedure, shear failure is treated as a separate damage state,
however, it corresponds to the limit damage state of complete damage
βcss = 0.5 + 20ρl ≤ 1 (31)
(DLS-4). It is thus clear that shear failure could take place at a dis-
0.29forμϕ ≤ 1 placement lower than DLS-4 as computed with Eq. (36), or even DLS-3

⎪ for some configurations. It is then proposed that for the complete da-
γ = 0.29 +
⎨ (
0.05 − 0.29
13 − 1 )
(μϕ − 1) for 1 < μϕ ≤ 13 mage state, post-processing of the data be carried out to determine the
⎪ 0.05forμϕ > 13 most likely failure mode of a given column. For simplicity, it was as-
⎩ (32)
sumed that the transverse reinforcement has the same mechanical
0.5(D − c ) properties as the longitudinal one.
tanς = The lateral displacement at shear failure is identified as that when
H (33)
the developed shear in the column is equal to its shear strength, as
In Eq. (32), μϕ are the curvature ductility values computed based on shown in Eq. (40). The displacement ductility and drift ratio at shear
the same yield curvature used in Eq. (1). The shear strength at each failure are then defined by Eq. (41) and Eq. (42), respectively.
point of the estimated lateral displacement response can thus be cal-
culated. ΔSF = Δ(V = VR) (40)

5
C. Perdomo and R. Monteiro Engineering Structures 217 (2020) 110794

ΔSF and 5.538% in the estimated statistics for the median value (θ) and
μΔ, SF =
Δ y, N (41) lognormal standard deviation (β), respectively, between the im-
plementation of 60 and 120 simulated sections. These differences are
Δ Δ reduced to 0.038% and 1.408%, for θ and β, between the im-
= SF
H SF H (42) plementation of 120 and 180 simulated sections, which are considered
acceptable. The implementation of 120 simulated sections also re-
3. Case-study and generation of structural response data presented a good compromise between computational time and output
data generation. With this approach, in this study, over approximately
The general approach followed for the generation of the RC bridge 3.5 million data points (per DLS) were generated for subsequent re-
columns damage models was to select the key parameters that define sponse analysis. The general procedure for the generation of structural
the RC columns as structural elements and proceed with a full random response data and development of the damage models is summarized in
generation of several different hypothetical ones, whose response was Fig. 1.
analyzed and statistically post-processed. A set of RC column para- Furthermore, for the mean section, the nominal yield curvature and
meters were therefore defined in order to cover, in a comprehensive yield displacement were evaluated using Eqs. (22) and (37) and mo-
manner, actual columns that can be found in existing bridges. The ment-curvature analyses were performed for each of the generated
parameters adopted in this study, from which the different case-study sections. With respect to the moment-curvature analyses, reference is
columns were generated, are shown in Table 1. All the possible com- made to the Mander et al. [23] and King et al. [34] models for the
binations of parameters from Table 1 were investigated. The values of stress-strain relationships of the concrete and reinforcing steel, re-
f 'c and f y were taken as mean values for each generated column. spectively. The concrete cover thickness was taken constant throughout
Furthermore, the parameter ρυ, conf in Table 1 corresponds to the the analyses with a value 50 mm, considered appropriate for bridge
transverse reinforcement ratio necessary for confinement requirements columns. The reinforcing steel elasticity modulus was also taken con-
according to a design code. In this study, confinement requirements stant with a value of 200GPa, whereas the concrete elasticity modulus
were computed according to the AASHTO [28] provisions, following was computed for each simulated section using Eq. (44), wherein the
the expression described in Eq. (43). input and output are given in MPa.

⎧ Ag f' f' ⎫ Ec = 5000 f 'c (44)


ρυ, conf = max 0.45 ⎛⎜ − 1⎞ c , 0.12 c

⎨ ⎝ Ac ⎠ fy fy ⎬ (43)
⎩ ⎭ All the moment-curvature analyses were carried out using the
software CUMBIA [35], which allows for incorporation of all the
parameters mentioned previously with the usual assumptions for sec-
3.1. Probabilistic description of material parameters
tional analysis, i.e. the consideration that plane sections remain plane
after deformation and a perfect bonding between concrete and re-
In order to generate variability in the structural response, statistical
inforcing steel.
distributions were assigned to the material properties. Based on the
After the moment-curvature analysis was performed, the force-dis-
recommendations made by Ellingwood and Hwang [29], Kappos et al.
placement response and shear strength evaluation for columns con-
[30] and Verderame et al. [31], the material properties and their cor-
structed with each simulated section were carried out. Sample results
responding statistical distributions, as adopted in this study, are sum-
for a column with D = 2.5 m, f’c = 35 MPa, fy = 450 MPa, ρl = 2.5%,
marized in Table 2.
ρν/ρν,conf = 0.5, ALR = 0.2 and H/D = 5 are presented in Fig. 2, where
Moreover, the parameters shown in Table 2 were considered as
the response of the section and column with mean parameters are
statistically independent. It is recognized that there exists a relationship
highlighted.
between the yield strength and the ultimate deformation capacity in the
In Fig. 2b, the upper lines with a general bilinear shape represent
reinforcement steel (i.e. the larger the yield strength, the lower the
the shear strength of the columns according to the Modified UCSD
ultimate deformation capacity). However, for the values investigated
model. Finally, with all the data from the simulated sections and col-
for yield strength in this study, and the rather conservative value for
umns, and the nominal yield curvature and displacement determined,
ultimate deformation, the aforementioned statistical independence was
curvature ductility, displacement ductility and drift ratio at the dif-
deemed acceptable.
ferent predefined damage limit states were identified, according to the
procedure explained previously.
3.2. Computation of the structural response

In specific, for every possible combination of mean parameters in 4. Generation of damage model
Table 1, a section and column characterized by a set of mean para-
meters, was defined. Then, Monte Carlo simulation [32] was used to Once all the structural response data and the different EDP values at
generate different values of the material properties according to the the different DLS were computed, the EDP data could be grouped and
statistical distributions indicated in Table 2, for a total of 120 simulated treated statistically. In order to obtain appropriate statistical distribu-
sections, based on the findings from Monteiro [33], including the one tions with a high level of statistical significance, EDP values were
with the set of mean parameters. This enabled the definition of em-
Table 1
pirical cumulative distribution functions with 2400 data points, once
Column parameters.
the results were grouped according to the controlling parameters of the
EDP damage model at each considered DLS. D [m] f'c [MPa] fy [MPa] ρl [%] ρv/ ρv,conf [ − −] ALR H/D
The size of the sampling for the distribution fitting was considered
0.8 21 400 0.5 1.0 0.1 1.5
sufficient given that no significant changes were found in the statistics 1.0 28 420 1.0 0.8 0.2 3.0
that control fitted statistical distributions when a larger number of 1.5 35 450 1.5 0.6 0.3 5.0
sections was considered in the simulation process. An example of this is 2.0 42 500 2.0 0.5 0.4 7.0
presented in Table 3, where the parameters fitted for a lognormal dis- 2.5 48 2.5 0.4 9.0
3.0 3.0 0.2
tribution (i.e. median (θ) and lognormal standard deviation (β)) with
3.5 3.5
different numbers of simulated sections for a given column are pre- 4.0 4.0
sented. Table 3 shows that, on average, there is a difference of 0.204%

6
C. Perdomo and R. Monteiro Engineering Structures 217 (2020) 110794

Table 2 fitted distributions. Results for other EDPs and columns can be found in
Statistical description of material properties. Perdomo [36].
Parameter Mean c.o.v.b Distribution Fig. 3 also shows the computed control parameters for the normal
and lognormal distributions ( μ and σ ; θ and β , respectively), the p
f'c f'ca 0.20 Normal values for the Kolmogorov-Smirnoff (K-S) test [32] and the number of
fy f ya 0.11 Lognormal
observed data points (N ), for each investigated DLS. It is noted that in
fu/fy 1.40 0.11 Normal
εsu 0.12 0.10 Normal
this case the lognormal distribution fails to appropriately describe the
observed data at all DLS even for a level of statistical significance of 5%,
a
According to Table 1. as shown by the p values, however, visually, both distributions tend to
b
Coefficient of variation. describe well the observed data without large discrepancies. It also
must be pointed out that the same remark does not hold for all columns
Table 3 or EDPs. In other cases, the lognormal distribution fits the observed
Variation in fitted statistics given the number of simulated sections for a column data with a high p value and the normal distribution fails even at a 5%
with D = 2.5 m, ALR = 0.2, ρ l = 2.5%, ρv/ρv,conf = 0.5 and H/D = 5. significance level. Nevertheless, it was generally found, among all fitted
DLS-1 DLS-2 DLS-3 DLS-4 distributions, that both of them visually fit quite well the observed data.
Following current practice, preference will be given to the estimation of
NSS* θ Β θ β θ β θ β the parameters of lognormal distributions, mainly because it avoids the
60 0.0089 0.0973 0.0127 0.0760 0.0286 0.1355 0.0357 0.1496
issue of sampling negative EDP values, reason for which they are fre-
120 0.0089 0.1012 0.0127 0.0795 0.0286 0.1449 0.0356 0.1595 quently implemented over normal or other types of distributions [25].
180 0.0089 0.1000 0.0127 0.0789 0.0285 0.1422 0.0357 0.1568 However, other types of distributions might also be used to char-
acterise, in statistical terms, for the description of earthquake-induced
* Number of simulated sections. damage in structural systems or components; Beta distributions [37]
and Gamma or Weibull distributions [38] have been shown to provide a
grouped according to their column (or section) geometrical parameters good probabilistic description at different damage limit states.
and reinforcement quantities. This means that for a column with a It is also noted the comparatively small, with respect to previous
given diameter, aspect ratio, longitudinal and transverse reinforcement research, values of the logarithmic standard deviation at all DLS. It was
ratio, and axial load ratio the EDP values for different DLSs were re- found that these values present little variability among columns and
trieved. Normal and lognormal statistical distributions were fitted to EDP at their respective DLS, thus, those shown in Fig. 3 can be seen as
the observed data and their distribution parameters computed (i.e. representative. Furthermore, these small values of dispersion are at-
mean and standard deviation, and, median and logarithmic standard tributed to the evident incapacity of the procedure to account for
deviation, respectively) using the method of the moments [32]. The epistemic uncertainty, an effect that has been observed in previous
goodness-of-fit (GOF) of the fitted distributions was assessed using the studies [30]. Indeed, in the way the simulation procedure was set,
chi-squared and Kolmogorov-Smirnoff tests. A sample of this procedure within the approach to generate the damage models, only the aleatory
is shown in Fig. 3 for the same column displayed in Fig. 2 for drift ratio, uncertainty in the mechanical properties of the constituent materials of
where the series “observed” corresponds to the empirical cumulative a given column was accounted for. As such, further efforts to account
distribution function defined using the results of the simulation process, for epistemic uncertainty have been made.
and the series “Normal Dist.” and “Lognormal Dist.” correspond to the

Fig. 1. General procedure for the construction of the damage models.

7
C. Perdomo and R. Monteiro Engineering Structures 217 (2020) 110794

Fig. 2. Sample of structural response calculation: (a) Moment-curvature; (b) Force-displacement and shear strength.

Fig. 3. Fitted statistical distributions for sample column: (a) DLS-1; (b) DLS-2; (c) DLS-3; (d) DLS-4.

4.1. Regression analysis values of the regression coefficients for all EPDs and DLS for the best
performing equations are summarised in Table 4.
In line with previous research [11,12] the approach implemented
β
herein to account for epistemic and total uncertainty in the estimated f y β5 ⎛ f ⎞ 6 H β7
median EDP values requires the estimation of predictive closed form EDP = β0 D β1 ρl β2 (1 + ALR) β3 ρυ, eff β4 ⎜⎛ ⎟⎞ ⎜ u ⎟ ⎛ ⎞
⎝ f 'c ⎠ ⎝ f y ⎠ ⎝ D ⎠ (45)
expressions. Therefore, the EDP values from the simulation procedure
were gathered, for each DLS, and closed form solutions were fitted to
In general, it was found that the performance of all fitted equations
the observed data using nonlinear regression.
is similar and, given their good match with respect to the simulation
The basic form of the closed form solution proposed in this study is
results, as shown in Table 4, they are considered satisfactory. The small
shown in Eq. (45), which keeps the basic form of those implemented in
values of variability, as expressed by the coefficients of variation shown
previous research and contains all the parameters that are deemed re-
in Table 4, are in line with the findings presented in Fig. 3. The equa-
levant for the prediction of EDP values. Several variants of the basic
tions presented in Table 4 are not necessarily the best performing ones
equation are fitted for each EDP (four for curvature ductility and six for
with respect to the simulated data and, specifically for drift ratio, the
displacement ductility and drift ratio). The behaviour of the best per-
latter correspond to other alternatives of Eq. (45) for all DLSs, not
forming equations (with respect to the experimental data, which are
presented in this document due to space limitations. Detailed results
shown later) for drift ratio at each DLS are shown in Fig. 4, whereas the
can be found in Perdomo [36].

8
C. Perdomo and R. Monteiro Engineering Structures 217 (2020) 110794

Fig. 4. Comparison of fitted equations to simulated data: (a) DLS-1; (b) DLS-2; (c) DLS-3; (d) DLS-4.

Table 4
Regression coefficients for best performing predictive equations.
EDP Equation DLS β 0 β 1 β 2 β 3 β 4 β 5 β 6 β 7 Meana c.o.v.b

μϕ 1 1 1.758 −0.008 0.081 −0.545 0.042 0.000 0.000 0.000 0.996 0.096
3 2 2.463 0.000 −0.087 −2.283 0.011 0.000 0.000 0.000 0.990 0.142
1 3 80.074 0.000 0.002 −1.879 0.673 −0.049 0.000 0.000 0.974 0.265
4 4 74.706 0.000 −0.099 −2.220 0.667 0.000 −0.087 0.000 0.988 0.158
μΔ 2 1 1.572 −0.027 0.094 −0.560 0.040 0.053 0.000 −0.030 0.994 0.113
2 2 3.964 −0.059 0.050 −1.503 0.038 −0.060 0.000 −0.146 0.994 0.113
6 3 61.778 0.000 0.164 −1.643 0.570 0.000 0.523 −0.346 0.982 0.200
6 4 81.224 0.000 0.099 −2.024 0.610 0.000 0.583 −0.389 0.974 0.200
Δ/H 2 1 0.004 −0.094 0.147 −0.560 0.038 0.105 0.000 0.839 0.992 0.125
2 2 0.010 −0.126 0.103 −1.503 0.036 −0.007 0.000 0.723 0.992 0.123
2 3 0.179 −0.121 0.218 −1.643 0.564 0.081 0.000 0.524 0.977 0.224
4 4 0.280 0.000 0.152 −2.024 0.614 0.000 0.000 0.481 0.971 0.242

a
Mean of observed-to-predicted values.
b
Coefficient of variation of observed-to-predicted values.

4.2. Accounting for epistemic uncertainty simulation process for reinforcement content, material strength and
axial load ratio, were retrieved from the Panagiotakos and Fardis [11]
To account for the total and epistemic uncertainty in the predicted study; whereas 25 experimental results for DLS-2, 39 for DLS-3 and 21
mean and median values, the closed form expressions computed pre- for DLS-4 were retrieved from the PEER structural performance data-
viously were compared with experimental results. The experimental base [39], also keeping in mind that the characteristics of the specimens
data were obtained from the study of Panagiotakos and Fardis [11] and were well within (or very close) to the parameters of the simulation
from the PEER structural performance database (SPD) [39]. A total of process.
247 experimental results for DLS-1 and DLS-4, selected for consistency It must be noted that results from the Panagiotakos and Fardis [11]
with the parameters of the simulation, i.e. columns with symmetric study are given directly in terms of drift ratio and correspond to rec-
reinforcement (without diagonal reinforcement) without bar slip in the tangular sections, hence, for the implementation of the predictive
support and ductile steel, with values within the ranges of the equations, the depth of the section in the direction of flexure was

9
C. Perdomo and R. Monteiro Engineering Structures 217 (2020) 110794

assumed as the column diameter. Due to space limitations, only the Table 5
behaviour of the best performing equations with respect to the ex- Performance of fitted equations with respect to experimental data.
perimental results for drift ratio at each DLS are shown in Fig. 5. In Measure DLS-1 DLS-2 DLS-3 DLS-4
Fig. 5 the experimental results come from Panagiotakos and Fardis
a
(2001) for DLS-1 whereas for the other DLS, the experimental results Mean 1.580 1.458 1.045 1.220
Mediana 1.497 1.256 0.909 0.930
come from the PEER SPD database. At DLS-4, when compared with
c.o.v. 0.383 0.419 0.399 0.614
experimental results from Panagiotakos and Fardis [11], there is a
slightly lower performance of the fitted equations, however, the dif- a
of experimental-to-predicted values.
ferences are not large. Further details can be found in Perdomo [36].
The accuracy of the best performing fitted equations for drift ratio, Table 6
measured in terms of mean, median and coefficient of variation of the Estimation of epistemic uncertainty.
experimental-to-predicted values is shown in Table 5. The performance
Measure DLS-1 DLS-2 DLS-3 DLS-4
for displacement ductility is similar at DLS-2, DLS-3 and DLS-4, where
data for comparison were available. β ale 0.12 0.12 0.22 0.24
It is seen that the worst performance is obtained at DLS-1, where the β tot 0.37 0.40 0.38 0.57
best performing equation offers a median value of experimental-to- β epis 0.35 0.38 0.31 0.51

predicted drift of 1.5, this is attributed to the fact that most of the ex-
perimental specimens have a low aspect ratio, meaning that the con-
Table 7
tributions of shear deformation might be important, as well as due to
Sample of database with damage models for drift ratio at DLS-3.
the incongruence of comparing rectangular sections with circular sec-
tions. Still, the presented results are in line with the findings from D [m] ALR ρ l [–] ρv/ρv,conf [–] H/D θ β
Panagiotakos and Fardis [11], however, with less dispersion, probably 0.8 0.2 0.025 0.5 5 0.0368 0.34
due to the use of members where the plastic hinge method is more 1.0 0.2 0.025 0.5 5 0.0330 0.33
adequate (e.g. members without diagonal reinforcement). It is thus not 1.5 0.2 0.025 0.5 5 0.0306 0.33
surprising that the predictive equation behaves better for the simulated 2.0 0.2 0.025 0.5 5 0.0293 0.33
2.5 0.2 0.025 0.5 5 0.0286 0.33
data (Fig. 4a) with respect to the experimental data (Fig. 5a). The
3.0 0.2 0.025 0.5 5 0.0279 0.33
proposed equation captures in a good manner the response obtained 3.5 0.2 0.025 0.5 5 0.0274 0.33
from the simulated data by accounting for most of the controlling 4.0 0.2 0.025 0.5 5 0.0270 0.34
parameters that drive the predictions using the plastic hinge method
(e.g. D, fy, H⁄D), whereas experimental drifts at yielding are affected by
a number of factors that are not accounted for with the plastic hinge drift underestimation with a median value of experimental-to-predicted
method formulation [11]. drift of 1.2, this is attributed to the rather conservative value of con-
Moreover, at DLS-2 the best performing equation also offers a slight crete strain at which the limit state was deemed to be reached (i.e.

Fig. 5. Comparison of fitted equations to experimental data: (a) DLS-1; (b) DLS-2; (c) DLS-3; (d) DLS-4.

10
C. Perdomo and R. Monteiro Engineering Structures 217 (2020) 110794

Fig. 6. Median curvature ductility and dispersion as function of the different parameters: (a) DLS-2; (b) DLS-3.

Fig. 7. Median displacement ductility and dispersion as function of the different parameters: (a) DLS-2; (b) DLS-3.

Fig. 8. Median drift ratio and dispersion as function of the different parameters: (a) DLS-2; (b) DLS-3.

εc = 0.004 ). In this regard, Berry and Eberhard [12] report from a series predictive equation captures well the response of the simulated data by
of experimental results that the mean strains at which concrete spalling taking into account most of the variables that control the reaching of a
was observed have values of εc = 0.005 and εc = 0.008 for rectangular concrete strain of εc = 0.004 in a given column (e.g. D , ALR , H / D ).
section columns and circular section columns transversally reinforced Given this conservative value of concrete strain at which the simulated
with spirals, respectively. The coefficients of variation of the previous data and, by extension, the predictive equations are set, it is reasonable
mean data have values of 0.551 and 0.563, for rectangular section to expect that the predictive equation underpredicts experimental drifts
columns and circular section columns as well, showing a large varia- at concrete spalling. Experimental drifts would be generally related
bility of concrete compressive strains at spalling. It is, therefore, not with larger compressive strains at spalling, therefore being larger than
surprising that the predictive equation behaves better for the simulated the ones predicted by Eq. (45). Fig. 5b confirms this expectation.
data (Fig. 4b) with respect to the experimental data (Fig. 5b). The For DLS-3 and DLS-4 the best performing equations offer a median

11
C. Perdomo and R. Monteiro Engineering Structures 217 (2020) 110794

Fig. 9. Simplified relationships for curvature ductility: (a) Median; (b) dispersion.

Fig. 10. Simplified relationships for displacement ductility: (a) Median; (b) dispersion.

Fig. 11. Simplified relationships for drift ratio: (a) Median; (b) Dispersion.

value of experimental-to-predicted drifts of 0.9 and 0.93, respectively, The total uncertainty is related to the aleatory and epistemic un-
showing an overall good agreement. These median values of experi- certainties, βale and βepis , respectively, according to Eq. (47).
mental-to-predicted EDP reassure the use of the median values of the
fitted distributions obtained previously in this section. βtot 2 = βale 2 + βepis 2 (47)
In order to assess the effect of the epistemic uncertainty, the general
properties of a lognormal distribution were used to estimate its value. Furthermore, aleatory uncertainties at each DLS can be estimated
The logarithmic standard deviation is related to the coefficient of var- using Eq. (46) with c. o. v. values contained in Table 4. With the total
iation according to Eq. (46) [32]. Thus, the total uncertainty around the and aleatory uncertainties estimated, the epistemic uncertainty at each
median values can be computed with Eq. (46) using the c. o. v. values DLS was computed using Eq. (47) and the results are presented in
presented in Table 5. Table 6.
Finally, the values of dispersion computed previously (e.g. Fig. 3)
β = log (1 + c. o. v.2 ) (46) for the fitted distributions are updated according to Eq. (47) with the

12
C. Perdomo and R. Monteiro Engineering Structures 217 (2020) 110794

Table 8 the damage model datasets are investigated in this section in order to
Regression coefficients and dispersion of simplified relationships for curvature seek for further simplifications. A preliminary scrutiny of the datasets
ductility. proved that the transverse reinforcement content (ρυ / ρυ, conf ) has a no-
DLS-1 DLS-2 ticeable effect in the prediction of the median EDP values at the at-
tainment of each DLS, which is clearly a result of the sectional analysis
a b β a b β implemented in this study.
As a consequence, the trends for other parameters will be in-
ρv/ρv,conf = 1.0 1.118 −0.408 0.36 3.118 −3.961 0.42
ρv/ρv,conf = 0.5 1.113 −0.438 0.36 3.084 −3.939 0.42 vestigated for fixed values of ρυ / ρυ, conf , namely, 1.0, 0.5 and 0.2. These
ρv/ρv,conf = 0.2 1.106 −0.477 0.36 3.057 −3.928 0.42 values are associated with: (i) columns designed according to modern
DLS-3 DLS-4 seismic design provisions; (ii) columns designed for shear but without
consideration of confinement requirements or capacity design princi-
a b β a b β
ples [40]; and (iii) columns for which only minimum shear reinforce-
ρv/ρv,conf = 1.0 16.172 −17.856 0.33 24.371 −29.231 0.54 ment is provided, respectively. It is thought that bridge columns with
ρv/ρv,conf = 0.5 9.710 −10.440 0.33 14.918 −18.621 0.54 these likely configurations can be fairly identified based on the year of
ρv/ρv,conf = 0.2 5.660 −6.879 0.34 8.801 −11.837 0.54 construction and/or location of the bridge, which makes this char-
acterization convenient.

values of epistemic uncertainty estimated in Table 6. The same values


are used at each DLS for each EDP investigated. It is noted that this 5.1. Curvature ductility
assumption is not particularly strict, since the values of epistemic un-
certainty for curvature ductility might be different. However, computed Via scrutiny of the datasets for curvature ductility, for all DLS, the
values of epistemic uncertainty for displacement ductility were quite following trends were found: the median curvature ductility decreases
similar to those shown in Table 5 for DLS-2 to DLS-4, therefore, it was almost linearly with increments in ALR ; the influence of the column
considered appropriate to use the values proposed in Table 6 for the diameter and longitudinal reinforcement ratio is small. The variation of
three EDPs investigated. the dispersion values is small across sections. A sample of the observed
At the end of the updating procedure, a set of databases per DLS variation of the curvature ductility with the aforementioned para-
were constructed per each EDP and an example sample, for drift ratio at meters, for two DLS, is shown in Fig. 6.
DLS-3, is presented in Table 7.
5.2. Displacement ductility and drift ratio
5. Investigation of EDP trends
Regarding the displacement ductility and drift ratio, for all DLS, the
The constructed datasets, as shown in Table 7, are quite compre- median displacement ductility decreases nonlinearly with increasing
hensive and cumbersome to handle. In fact, the databases for the three H / D , whereas the median drift ratio increases almost linearly with
analysed EDPs (curvature ductility, displacement ductility and drift increasing H / D . For the computation of the median value, the influence
ratio) contain multiple row entries per DLS. It is also noted that in order of the column diameter is noticeable and ALR also plays a significant
to use the proposed databases, specific information about reinforcement role, clearly decreasing EDP values with its increment. On the other
content is required, a feature that is intended to be avoided in this hand, the influence of the longitudinal reinforcement ratio, although
study. not negligible, is small in comparison with the impacts of the column
The trends and relative relevance of the parameters required within diameter and axial load. The trends and relative importance of the

Table 9
Regression coefficients and dispersion of simplified relationships for displacement ductility at DLS-1.
ALR = 0.1 ALR = 0.2 ALR = 0.3 ALR = 0.4

D [m] a b β a b β a b β a b β

ρv/ρv,conf = 1.0 0.8 1.073 −0.045 0.37 1.098 −0.040 0.36 1.016 −0.045 0.36 0.917 −0.054 0.37
1.0 1.053 −0.042 0.37 1.081 −0.038 0.36 1.002 −0.041 0.36 0.908 −0.049 0.37
1.5 1.018 −0.037 0.37 1.050 −0.032 0.36 0.986 −0.034 0.36 0.898 −0.041 0.37
2.0 0.994 −0.032 0.37 1.033 −0.027 0.36 0.975 −0.029 0.36 0.888 −0.035 0.37
2.5 0.979 −0.029 0.37 1.023 −0.025 0.36 0.967 −0.026 0.36 0.882 −0.031 0.36
3.0 0.970 −0.026 0.37 1.015 −0.022 0.36 0.960 −0.023 0.36 0.877 −0.027 0.36
3.5 0.963 −0.025 0.37 1.007 −0.020 0.36 0.955 −0.021 0.36 0.873 −0.025 0.36
4.0 0.954 −0.022 0.37 1.002 −0.019 0.36 0.953 −0.020 0.36 0.869 −0.023 0.36
ρv/ρv,conf = 0.5 0.8 1.072 −0.044 0.37 1.088 −0.038 0.36 1.007 −0.045 0.37 0.900 −0.051 0.37
1.0 1.046 −0.042 0.37 1.071 −0.036 0.36 0.993 −0.040 0.36 0.894 −0.047 0.37
1.5 1.011 −0.036 0.37 1.040 −0.029 0.36 0.976 −0.033 0.36 0.883 −0.039 0.37
2.0 0.988 −0.031 0.37 1.025 −0.027 0.36 0.964 −0.028 0.36 0.873 −0.033 0.37
2.5 0.973 −0.028 0.37 1.012 −0.023 0.36 0.957 −0.025 0.36 0.867 −0.030 0.37
3.0 0.962 −0.026 0.37 1.006 −0.022 0.36 0.950 −0.023 0.36 0.862 −0.027 0.37
3.5 0.955 −0.024 0.37 0.999 −0.020 0.36 0.946 −0.021 0.36 0.857 −0.024 0.36
4.0 0.949 −0.022 0.37 0.994 −0.018 0.36 0.942 −0.019 0.36 0.855 −0.022 0.36
ρv/ρv,conf = 0.2 0.8 1.064 −0.044 0.37 1.077 −0.037 0.36 0.989 −0.041 0.37 0.881 −0.048 0.37
1.0 1.040 −0.041 0.37 1.058 −0.034 0.36 0.977 −0.037 0.37 0.873 −0.044 0.37
1.5 0.998 −0.034 0.37 1.029 −0.028 0.36 0.960 −0.030 0.36 0.862 −0.036 0.37
2.0 0.979 −0.031 0.37 1.011 −0.025 0.36 0.948 −0.026 0.36 0.854 −0.031 0.37
2.5 0.964 −0.027 0.37 1.001 −0.022 0.36 0.942 −0.023 0.36 0.847 −0.028 0.37
3.0 0.951 −0.024 0.37 0.994 −0.021 0.36 0.936 −0.021 0.36 0.843 −0.025 0.37
3.5 0.946 −0.023 0.37 0.986 −0.019 0.36 0.930 −0.019 0.37 0.839 −0.023 0.37
4.0 0.940 −0.021 0.37 0.979 −0.017 0.36 0.928 −0.018 0.36 0.835 −0.021 0.37

13
C. Perdomo and R. Monteiro Engineering Structures 217 (2020) 110794

Table 10
Regression coefficients and dispersion of simplified relationships for displacement ductility at DLS-2.
ALR = 0.1 ALR = 0.2 ALR = 0.3 ALR = 0.4

D [m] a b β a b β a b β a b β

ρv/ρv,conf = 1.0 0.8 2.658 −0.258 0.41 2.141 −0.202 0.41 1.809 −0.178 0.41 1.577 −0.169 0.415
1.0 2.556 −0.246 0.41 2.071 −0.191 0.41 1.750 −0.166 0.41 1.525 −0.155 0.415
1.5 2.374 −0.219 0.41 1.948 −0.168 0.41 1.660 −0.144 0.41 1.447 −0.130 0.414
2.0 2.237 −0.195 0.41 1.858 −0.149 0.41 1.593 −0.125 0.41 1.393 −0.113 0.414
2.5 2.146 −0.178 0.41 1.800 −0.135 0.41 1.553 −0.114 0.41 1.363 −0.103 0.414
3.0 2.074 −0.164 0.41 1.752 −0.123 0.41 1.513 −0.102 0.41 1.332 −0.092 0.413
3.5 2.019 −0.152 0.41 1.716 −0.115 0.41 1.489 −0.095 0.41 1.313 −0.085 0.413
4.0 1.968 −0.141 0.41 1.680 −0.106 0.41 1.467 −0.088 0.41 1.295 −0.079 0.413
ρv/ρv,conf = 0.5 0.8 2.637 −0.258 0.41 2.119 −0.202 0.41 1.791 −0.179 0.41 1.554 −0.170 0.416
1.0 2.539 −0.246 0.41 2.054 −0.192 0.41 1.733 −0.166 0.41 1.506 −0.156 0.416
1.5 2.352 −0.218 0.41 1.930 −0.168 0.41 1.637 −0.142 0.41 1.428 −0.133 0.415
2.0 2.213 −0.195 0.41 1.835 −0.149 0.41 1.571 −0.125 0.41 1.373 −0.114 0.415
2.5 2.124 −0.178 0.41 1.774 −0.134 0.41 1.530 −0.113 0.41 1.339 −0.103 0.414
3.0 2.046 −0.163 0.41 1.725 −0.122 0.41 1.492 −0.103 0.41 1.306 −0.091 0.414
3.5 1.990 −0.151 0.41 1.691 −0.113 0.41 1.467 −0.095 0.41 1.289 −0.086 0.414
4.0 1.944 −0.141 0.41 1.658 −0.106 0.41 1.442 −0.088 0.41 1.271 −0.079 0.414
ρv/ρv,conf = 0.2 0.8 2.630 −0.259 0.41 2.105 −0.203 0.41 1.774 −0.180 0.41 1.538 −0.172 0.417
1.0 2.525 −0.248 0.41 2.035 −0.192 0.41 1.714 −0.168 0.41 1.490 −0.161 0.417
1.5 2.331 −0.219 0.41 1.903 −0.168 0.41 1.620 −0.145 0.41 1.405 −0.135 0.416
2.0 2.189 −0.195 0.41 1.814 −0.149 0.41 1.548 −0.126 0.41 1.351 −0.117 0.415
2.5 2.100 −0.178 0.41 1.752 −0.135 0.41 1.505 −0.114 0.41 1.312 −0.105 0.415
3.0 2.025 −0.163 0.41 1.703 −0.124 0.41 1.469 −0.103 0.41 1.283 −0.095 0.415
3.5 1.973 −0.153 0.41 1.664 −0.113 0.41 1.443 −0.097 0.42 1.263 −0.087 0.415
4.0 1.925 −0.142 0.41 1.634 −0.106 0.41 1.420 −0.089 0.41 1.243 −0.080 0.415

Table 11
Regression coefficients and dispersion of simplified relationships for displacement ductility at DLS-3.
ALR = 0.1 ALR = 0.2 ALR = 0.3 ALR = 0.4

D [m] a b β a b β a b β a b β

ρv/ρv,conf = 1.0 0.8 14.523 −0.517 0.35 12.802 −0.508 0.35 11.179 −0.504 0.35 10.089 −0.505 0.35
1.0 12.894 −0.492 0.35 11.330 −0.480 0.35 9.757 −0.475 0.35 8.633 −0.473 0.35
1.5 12.027 −0.447 0.35 10.090 −0.432 0.35 8.685 −0.426 0.35 7.673 −0.423 0.35
2.0 10.975 −0.408 0.35 9.166 −0.392 0.35 7.885 −0.385 0.35 6.936 −0.381 0.35
2.5 10.220 −0.378 0.35 8.578 −0.364 0.35 7.386 −0.356 0.35 6.492 −0.352 0.35
3.0 9.624 −0.354 0.35 8.094 −0.338 0.35 6.962 −0.332 0.35 6.115 −0.327 0.35
3.5 9.173 −0.334 0.35 7.722 −0.319 0.35 6.637 −0.312 0.35 5.831 −0.307 0.35
4.0 8.751 −0.312 0.35 7.410 −0.300 0.35 6.376 −0.293 0.35 5.587 −0.289 0.35
ρv/ρv,conf = 0.5 0.8 8.094 −0.460 0.35 7.478 −0.455 0.34 6.363 −0.446 0.36 5.316 −0.436 0.34
1.0 7.671 −0.443 0.35 6.475 −0.421 0.36 5.525 −0.413 0.34 4.922 −0.415 0.34
1.5 7.110 −0.402 0.35 5.962 −0.382 0.34 5.226 −0.375 0.34 4.635 −0.375 0.34
2.0 6.449 −0.364 0.34 5.621 −0.349 0.34 4.930 −0.341 0.34 4.353 −0.338 0.34
2.5 6.201 −0.340 0.34 5.417 −0.324 0.34 4.731 −0.315 0.34 4.138 −0.312 0.34
3.0 5.970 −0.316 0.35 5.210 −0.300 0.34 4.492 −0.292 0.34 3.916 −0.288 0.34
3.5 5.838 −0.299 0.35 5.029 −0.283 0.34 4.307 −0.273 0.34 3.756 −0.270 0.34
4.0 5.689 −0.283 0.35 4.839 −0.267 0.34 4.142 −0.257 0.34 3.624 −0.254 0.34
ρv/ρv,conf = 0.2 0.8 4.825 −0.384 0.35 3.745 −0.343 0.34 3.086 −0.326 0.34 2.589 −0.316 0.34
1.0 4.349 −0.358 0.34 3.634 −0.331 0.33 3.148 −0.324 0.33 2.635 −0.314 0.34
1.5 4.046 −0.326 0.34 3.390 −0.298 0.33 2.951 −0.290 0.33 2.504 −0.280 0.33
2.0 3.858 −0.299 0.34 3.251 −0.272 0.33 2.825 −0.264 0.33 2.478 −0.262 0.33
2.5 3.770 −0.279 0.34 3.154 −0.254 0.33 2.596 −0.237 0.34 2.203 −0.227 0.33
3.0 3.579 −0.258 0.34 2.904 −0.230 0.33 2.508 −0.218 0.33 2.224 −0.216 0.33
3.5 3.405 −0.244 0.34 2.856 −0.217 0.33 2.518 −0.210 0.33 2.219 −0.206 0.33
4.0 3.300 −0.229 0.34 2.823 −0.205 0.33 2.483 −0.198 0.33 2.178 −0.194 0.33

different column parameters on the drift ratio at the different damage 6. Simplified relationships
limit states are found to be consistent with previous research developed
by Reza et al. [41], were closed-form expressions for the prediction of Based on the observed trends, simplified relationships, obtained
the lateral displacement of bridge piers at the cracking of concrete, through statistical regression, are proposed for the computation of the
longitudinal reinforcement first yield and crushing of the concrete core median EDP as function of the most relevant column parameters.
or shear failure were developed. When it comes to the dispersion va- Several functional forms for these simplified relationships were tested,
lues, their variability across different columns is small. A sample of the concluding that for curvature ductility and drift ratio a simple linear
observed behaviour is shown in Figs. 7 and 8 for displacement ductility relationship was adequate whereas for displacement ductility a power
and drift ratio, respectively. law was instead better. The functional forms of the simplified re-
lationships are shown in Eqs. (48)–(50) for curvature ductility, dis-
placement ductility and drift ratio, respectively, where a and b are the

14
C. Perdomo and R. Monteiro Engineering Structures 217 (2020) 110794

Table 12
Regression coefficients and dispersion of simplified relationships for displacement ductility at DLS-4.
ALR = 0.1 ALR = 0.2 ALR = 0.3 ALR = 0.4

D [m] a b β a b β a b β a b β

ρv/ρv,conf = 1.0 0.8 26.285 −0.561 0.56 21.101 −0.552 0.55 17.749 −0.552 0.56 15.954 −0.559 0.56
1.0 21.392 −0.530 0.56 17.515 −0.518 0.55 14.665 −0.515 0.55 12.931 −0.517 0.56
1.5 18.015 −0.477 0.56 14.709 −0.465 0.55 12.498 −0.460 0.55 10.993 −0.460 0.56
2.0 15.845 −0.433 0.56 12.995 −0.421 0.55 11.071 −0.416 0.55 9.689 −0.414 0.56
2.5 14.501 −0.403 0.56 11.949 −0.390 0.56 10.212 −0.385 0.56 8.934 −0.383 0.56
3.0 13.477 −0.376 0.56 11.145 −0.363 0.56 9.510 −0.359 0.56 8.317 −0.355 0.56
3.5 12.749 −0.355 0.56 10.541 −0.343 0.56 8.994 −0.337 0.56 7.875 −0.334 0.56
4.0 12.067 −0.332 0.56 10.041 −0.323 0.56 8.585 −0.317 0.56 7.490 −0.314 0.56
ρv/ρv,conf = 0.5 0.8 15.727 −0.529 0.55 12.338 −0.512 0.55 10.548 −0.511 0.55 8.826 −0.508 0.56
1.0 12.981 −0.497 0.55 10.426 −0.481 0.55 8.767 −0.477 0.55 7.498 −0.474 0.56
1.5 11.008 −0.445 0.55 8.892 −0.429 0.55 7.518 −0.421 0.55 6.515 −0.421 0.55
2.0 9.723 −0.403 0.55 7.911 −0.387 0.55 6.710 −0.379 0.55 5.800 −0.375 0.55
2.5 8.943 −0.373 0.55 7.298 −0.356 0.55 6.201 −0.348 0.55 5.375 −0.346 0.55
3.0 8.311 −0.345 0.55 6.816 −0.329 0.55 5.795 −0.323 0.55 5.019 −0.319 0.55
3.5 7.902 −0.325 0.55 6.474 −0.309 0.55 5.507 −0.301 0.55 4.776 −0.299 0.55
4.0 7.524 −0.306 0.55 6.189 −0.292 0.55 5.257 −0.283 0.55 4.578 −0.282 0.55
ρv/ρv,conf = 0.2 0.8 8.746 −0.481 0.54 6.801 −0.460 0.54 5.301 −0.444 0.55 4.099 −0.427 0.56
1.0 7.555 −0.450 0.54 5.959 −0.427 0.54 4.790 −0.415 0.54 3.849 −0.404 0.55
1.5 6.422 −0.396 0.54 5.155 −0.375 0.54 4.293 −0.365 0.54 3.574 −0.359 0.55
2.0 5.747 −0.356 0.54 4.636 −0.333 0.54 3.899 −0.325 0.54 3.302 −0.322 0.54
2.5 5.334 −0.327 0.54 4.326 −0.306 0.54 3.637 −0.296 0.54 3.098 −0.293 0.54
3.0 4.986 −0.301 0.54 4.069 −0.281 0.54 3.432 −0.271 0.54 2.934 −0.268 0.54
3.5 4.762 −0.284 0.54 3.889 −0.263 0.54 3.281 −0.253 0.54 2.816 −0.249 0.54
4.0 4.563 −0.266 0.54 3.730 −0.245 0.54 3.158 −0.237 0.54 2.707 −0.233 0.54

Table 13
Regresssion coefficients and dispersion of simplified relationships for drift ratio at DLS-1.
ALR = 0.1 ALR = 0.2 ALR = 0.3 ALR = 0.4

D [m] a b β a b β a b β a b β

ρv/ρv,conf = 1.0 0.8 0.0017 0.0016 0.38 0.0017 0.0016 0.37 0.0016 0.0015 0.36 0.0015 0.0013 0.37
1.0 0.0015 0.0016 0.38 0.0015 0.0016 0.37 0.0014 0.0015 0.36 0.0013 0.0013 0.37
1.5 0.0012 0.0015 0.38 0.0012 0.0016 0.37 0.0011 0.0015 0.36 0.0010 0.0013 0.36
2.0 0.0009 0.0015 0.38 0.0010 0.0016 0.37 0.0009 0.0015 0.36 0.0009 0.0013 0.36
2.5 0.0008 0.0015 0.38 0.0008 0.0016 0.37 0.0008 0.0015 0.36 0.0008 0.0014 0.36
3.0 0.0007 0.0015 0.38 0.0008 0.0016 0.37 0.0007 0.0015 0.36 0.0007 0.0014 0.36
3.5 0.0007 0.0015 0.38 0.0007 0.0016 0.37 0.0007 0.0015 0.36 0.0006 0.0014 0.36
4.0 0.0006 0.0015 0.38 0.0006 0.0016 0.37 0.0006 0.0015 0.36 0.0006 0.0014 0.36
ρv/ρv,conf = 0.5 0.8 0.0017 0.0016 0.38 0.0017 0.0016 0.37 0.0016 0.0015 0.36 0.0015 0.0013 0.37
1.0 0.0015 0.0015 0.38 0.0015 0.0016 0.37 0.0014 0.0015 0.36 0.0013 0.0013 0.37
1.5 0.0011 0.0015 0.38 0.0011 0.0016 0.37 0.0011 0.0015 0.36 0.0010 0.0013 0.36
2.0 0.0009 0.0015 0.38 0.0010 0.0016 0.37 0.0009 0.0015 0.36 0.0008 0.0013 0.36
2.5 0.0008 0.0015 0.38 0.0008 0.0016 0.37 0.0008 0.0015 0.36 0.0007 0.0013 0.36
3.0 0.0007 0.0015 0.38 0.0007 0.0016 0.37 0.0007 0.0015 0.36 0.0007 0.0013 0.36
3.5 0.0007 0.0015 0.38 0.0007 0.0016 0.37 0.0007 0.0015 0.36 0.0006 0.0013 0.36
4.0 0.0006 0.0015 0.38 0.0006 0.0016 0.37 0.0006 0.0015 0.36 0.0006 0.0013 0.36
ρv/ρv,conf = 0.2 0.8 0.0017 0.0016 0.38 0.0017 0.0016 0.37 0.0016 0.0015 0.37 0.0014 0.0013 0.37
1.0 0.0015 0.0015 0.38 0.0014 0.0016 0.37 0.0014 0.0015 0.36 0.0013 0.0013 0.37
1.5 0.0011 0.0015 0.38 0.0011 0.0016 0.37 0.0011 0.0015 0.36 0.0010 0.0013 0.37
2.0 0.0009 0.0015 0.38 0.0009 0.0016 0.37 0.0009 0.0015 0.36 0.0008 0.0013 0.37
2.5 0.0008 0.0015 0.38 0.0008 0.0016 0.37 0.0008 0.0015 0.36 0.0007 0.0013 0.36
3.0 0.0007 0.0015 0.38 0.0007 0.0016 0.37 0.0007 0.0015 0.36 0.0006 0.0013 0.36
3.5 0.0007 0.0015 0.38 0.0007 0.0016 0.37 0.0006 0.0015 0.37 0.0006 0.0013 0.36
4.0 0.0006 0.0015 0.38 0.0006 0.0016 0.37 0.0006 0.0015 0.36 0.0005 0.0013 0.36

regression coefficients. regression coefficients are determined depending on the DLS, axial load
ratio, diameter and quantity of transverse reinforcement. With respect
θ μϕ = a + ALR·b (48) to the dispersion, given its low variability across sections or columns, a
simple average was taken as representative. A sample of the fitted
H b
θ μΔ = a ⎛ ⎞ equations for DLS-3 is presented in Figs. 9–11 for curvature ductility,
⎝D⎠ (49)
displacement ductility and drift ratio, respectively, along with the data
H used for their regression. It is generally seen that the fitted equations
θΔ/ H = a + ⎛ ⎞·b bound the data within ±30% for the median and ±5% for the disper-
⎝D⎠ (50)
sion. Complete results for the regression coefficients and estimated
Moreover, for curvature ductility, regression coefficients are de- dispersion are presented in Tables 8–16.
termined depending on the DLS and the quantity of transverse re- In Eq. (48), ALR is the independent variable, whereas in Eqs. (49)
inforcement; whereas for displacement ductility and drift ratio, the

15
C. Perdomo and R. Monteiro Engineering Structures 217 (2020) 110794

Table 14
Regression coefficients and dispersion of simplified relationships for drift ratio at DLS-2.
ALR = 0.1 ALR = 0.2 ALR = 0.3 ALR = 0.4

D [m] a b β a b β a b β a b β

ρv/ρv,conf = 1.0 0.8 0.0060 0.0021 0.41 0.0046 0.0020 0.41 0.0038 0.0018 0.41 0.0033 0.0016 0.41
1.0 0.0053 0.0021 0.41 0.0040 0.0020 0.41 0.0033 0.0018 0.41 0.0028 0.0017 0.41
1.5 0.0042 0.0022 0.41 0.0031 0.0021 0.41 0.0026 0.0019 0.41 0.0022 0.0017 0.41
2.0 0.0034 0.0022 0.41 0.0026 0.0021 0.41 0.0021 0.0019 0.41 0.0018 0.0017 0.41
2.5 0.0030 0.0022 0.41 0.0023 0.0021 0.41 0.0019 0.0019 0.41 0.0016 0.0017 0.41
3.0 0.0027 0.0023 0.41 0.0020 0.0021 0.41 0.0016 0.0019 0.41 0.0014 0.0017 0.41
3.5 0.0024 0.0023 0.41 0.0019 0.0021 0.41 0.0015 0.0019 0.41 0.0013 0.0018 0.41
4.0 0.0022 0.0023 0.41 0.0017 0.0021 0.41 0.0014 0.0019 0.41 0.0012 0.0018 0.41
ρv/ρv,conf = 0.5 0.8 0.0059 0.0021 0.41 0.0045 0.0020 0.41 0.0037 0.0018 0.41 0.0032 0.0016 0.41
1.0 0.0052 0.0021 0.41 0.0040 0.0020 0.41 0.0032 0.0018 0.41 0.0028 0.0016 0.41
1.5 0.0041 0.0022 0.41 0.0031 0.0020 0.41 0.0025 0.0019 0.41 0.0022 0.0017 0.41
2.0 0.0034 0.0022 0.41 0.0026 0.0021 0.41 0.0021 0.0019 0.41 0.0018 0.0017 0.41
2.5 0.0030 0.0022 0.41 0.0023 0.0021 0.41 0.0018 0.0019 0.41 0.0015 0.0017 0.41
3.0 0.0026 0.0022 0.41 0.0020 0.0021 0.41 0.0016 0.0019 0.41 0.0014 0.0017 0.41
3.5 0.0024 0.0022 0.41 0.0018 0.0021 0.41 0.0015 0.0019 0.41 0.0013 0.0017 0.41
4.0 0.0022 0.0023 0.41 0.0017 0.0021 0.41 0.0014 0.0019 0.41 0.0012 0.0017 0.41
ρv/ρv,conf = 0.2 0.8 0.0059 0.0021 0.41 0.0045 0.0020 0.41 0.0037 0.0018 0.41 0.0032 0.0016 0.41
1.0 0.0052 0.0021 0.41 0.0039 0.0020 0.41 0.0032 0.0018 0.41 0.0028 0.0016 0.41
1.5 0.0041 0.0021 0.41 0.0031 0.0020 0.41 0.0025 0.0018 0.41 0.0022 0.0016 0.41
2.0 0.0034 0.0022 0.41 0.0025 0.0020 0.41 0.0021 0.0018 0.41 0.0018 0.0017 0.41
2.5 0.0029 0.0022 0.41 0.0022 0.0020 0.41 0.0018 0.0019 0.41 0.0015 0.0017 0.41
3.0 0.0026 0.0022 0.41 0.0020 0.0021 0.41 0.0016 0.0019 0.41 0.0014 0.0017 0.41
3.5 0.0024 0.0022 0.41 0.0018 0.0021 0.41 0.0015 0.0019 0.41 0.0012 0.0017 0.41
4.0 0.0022 0.0022 0.41 0.0016 0.0021 0.41 0.0013 0.0019 0.41 0.0011 0.0017 0.41

Table 15
Regression coefficients and dispersion of simplified relationships for drift ratio at DLS-3.
ALR = 0.1 ALR = 0.2 ALR = 0.3 ALR = 0.4

D [m] a b β a b β a b β a b β

ρv/ρv,conf = 1.0 0.8 0.0407 0.0041 0.35 0.0348 0.0039 0.35 0.0304 0.0034 0.35 0.0276 0.0031 0.35
1.0 0.0328 0.0044 0.35 0.0284 0.0041 0.35 0.0245 0.0036 0.35 0.0219 0.0032 0.35
1.5 0.0266 0.0052 0.35 0.0223 0.0046 0.35 0.0193 0.0040 0.35 0.0172 0.0036 0.35
2.0 0.0220 0.0057 0.35 0.0184 0.0050 0.35 0.0159 0.0044 0.35 0.0140 0.0039 0.35
2.5 0.0192 0.0059 0.35 0.0161 0.0052 0.35 0.0139 0.0045 0.35 0.0122 0.0040 0.35
3.0 0.0171 0.0060 0.35 0.0142 0.0053 0.35 0.0122 0.0046 0.35 0.0108 0.0041 0.35
3.5 0.0155 0.0062 0.35 0.0129 0.0054 0.35 0.0111 0.0047 0.35 0.0098 0.0042 0.35
4.0 0.0141 0.0063 0.35 0.0118 0.0055 0.35 0.0102 0.0048 0.35 0.0089 0.0043 0.35
ρv/ρv,conf = 0.5 0.8 0.0235 0.0029 0.35 0.0207 0.0028 0.34 0.0182 0.0025 0.36 0.0159 0.0021 0.34
1.0 0.0199 0.0032 0.35 0.0169 0.0029 0.36 0.0147 0.0025 0.34 0.0131 0.0023 0.34
1.5 0.0161 0.0036 0.35 0.0135 0.0032 0.34 0.0117 0.0029 0.34 0.0104 0.0026 0.34
2.0 0.0134 0.0038 0.34 0.0112 0.0035 0.34 0.0097 0.0031 0.34 0.0085 0.0028 0.34
2.5 0.0118 0.0040 0.34 0.0099 0.0037 0.34 0.0085 0.0033 0.34 0.0075 0.0029 0.34
3.0 0.0105 0.0042 0.34 0.0088 0.0038 0.34 0.0075 0.0034 0.34 0.0066 0.0030 0.34
3.5 0.0096 0.0043 0.34 0.0080 0.0039 0.34 0.0068 0.0034 0.34 0.0060 0.0030 0.34
4.0 0.0088 0.0044 0.34 0.0074 0.0040 0.34 0.0063 0.0035 0.34 0.0055 0.0031 0.34
ρv/ρv,conf = 0.2 0.8 0.0120 0.0025 0.35 0.0092 0.0022 0.34 0.0077 0.0019 0.34 0.0066 0.0017 0.34
1.0 0.0100 0.0025 0.34 0.0084 0.0023 0.33 0.0073 0.0020 0.33 0.0062 0.0017 0.34
1.5 0.0082 0.0027 0.34 0.0069 0.0024 0.33 0.0060 0.0022 0.33 0.0051 0.0019 0.33
2.0 0.0071 0.0028 0.34 0.0059 0.0026 0.33 0.0051 0.0023 0.33 0.0045 0.0020 0.33
2.5 0.0064 0.0030 0.34 0.0053 0.0026 0.33 0.0044 0.0023 0.33 0.0037 0.0020 0.33
3.0 0.0057 0.0030 0.34 0.0046 0.0026 0.33 0.0039 0.0023 0.32 0.0034 0.0021 0.33
3.5 0.0052 0.0030 0.34 0.0042 0.0027 0.33 0.0036 0.0024 0.33 0.0032 0.0021 0.33
4.0 0.0047 0.0030 0.34 0.0039 0.0027 0.33 0.0033 0.0025 0.32 0.0029 0.0022 0.33

and (50) H / D is the independent variable. Values of the coefficients a fully defined.
and b, and the dispersion value of the damage model β are retrieved
from Tables 8–16 (depending on the desired EDP and DLS) as a function
of the amount of transverse reinforcement, diameter and axial load (as 6.1. Comparison with previous studies
represented by the axial load ratio - ALR) of the actual column. Once
the coefficients a and b are appropriately determined, the median value In order to assess the performance of the fitted simplified relation-
of the damage model (θ ) is then computed with Eqs. (48), (49) or (50) ships with respect to mechanics-based expressions, results are com-
with the actual axial load ratio of the column ( ALR ) for Eq. (48), or the pared with the proposal of Kowalsky [9]. By definition, DLS-2 and DLS-
aspect ratio (H / D ) of the actual column for Eqs. (49) and (50), as long 3 correspond to the limit states defined in Kowalsky’s study as servi-
as these values are within the simulation process range (e.g. H / D = 1.5 ceability and damage control. Since such a study was limited to well
and H / D = 9 ). With θ and β determined, the damage model can then be confined sections, using regression coefficients for ρυ / ρυ, conf = 1, pre-
dicted values for drift ratio are compared for columns with D = 3.0 m

16
C. Perdomo and R. Monteiro Engineering Structures 217 (2020) 110794

Table 16
Regression coefficients and dispersion of simplified relationships for drift ratio at DLS-4.
ALR = 0.1 ALR = 0.2 ALR = 0.3 ALR = 0.4

D [m] a b β a b β a b β a b β

ρv/ρv,conf = 1.0 0.8 0.0703 0.0062 0.55 0.0570 0.0051 0.55 0.0486 0.0043 0.56 0.0445 0.0036 0.56
1.0 0.0533 0.0063 0.55 0.0444 0.0053 0.55 0.0375 0.0045 0.56 0.0334 0.0039 0.56
1.5 0.0401 0.0070 0.55 0.0331 0.0060 0.55 0.0283 0.0051 0.55 0.0252 0.0045 0.56
2.0 0.0323 0.0075 0.55 0.0266 0.0063 0.55 0.0229 0.0055 0.55 0.0202 0.0048 0.55
2.5 0.0277 0.0077 0.55 0.0229 0.0066 0.55 0.0197 0.0057 0.55 0.0174 0.0050 0.55
3.0 0.0243 0.0079 0.55 0.0201 0.0068 0.55 0.0173 0.0058 0.55 0.0152 0.0051 0.55
3.5 0.0220 0.0080 0.55 0.0182 0.0068 0.55 0.0156 0.0059 0.55 0.0137 0.0052 0.55
4.0 0.0198 0.0082 0.56 0.0165 0.0070 0.55 0.0142 0.0060 0.55 0.0124 0.0053 0.56
ρv/ρv,conf = 0.5 0.8 0.0417 0.0044 0.55 0.0331 0.0037 0.55 0.0288 0.0031 0.56 0.0247 0.0026 0.56
1.0 0.0322 0.0044 0.55 0.0261 0.0037 0.55 0.0222 0.0032 0.55 0.0193 0.0027 0.56
1.5 0.0242 0.0049 0.55 0.0196 0.0041 0.55 0.0167 0.0036 0.55 0.0146 0.0031 0.55
2.0 0.0194 0.0051 0.55 0.0159 0.0044 0.55 0.0135 0.0038 0.55 0.0118 0.0033 0.55
2.5 0.0167 0.0053 0.55 0.0136 0.0045 0.55 0.0116 0.0039 0.55 0.0101 0.0034 0.55
3.0 0.0145 0.0054 0.55 0.0119 0.0046 0.55 0.0101 0.0040 0.55 0.0088 0.0035 0.55
3.5 0.0132 0.0055 0.55 0.0107 0.0047 0.55 0.0091 0.0040 0.55 0.0080 0.0035 0.55
4.0 0.0119 0.0055 0.55 0.0098 0.0047 0.55 0.0083 0.0041 0.55 0.0072 0.0036 0.55
ρv/ρv,conf = 0.2 0.8 0.0229 0.0030 0.54 0.0181 0.0025 0.54 0.0146 0.0021 0.55 0.0115 0.0017 0.57
1.0 0.0184 0.0031 0.54 0.0146 0.0027 0.54 0.0120 0.0022 0.55 0.0098 0.0019 0.56
1.5 0.0137 0.0034 0.54 0.0110 0.0029 0.54 0.0092 0.0025 0.54 0.0078 0.0021 0.55
2.0 0.0111 0.0035 0.54 0.0089 0.0030 0.54 0.0075 0.0026 0.54 0.0064 0.0022 0.54
2.5 0.0095 0.0036 0.54 0.0076 0.0031 0.54 0.0064 0.0027 0.54 0.0055 0.0023 0.54
3.0 0.0083 0.0037 0.54 0.0067 0.0032 0.54 0.0056 0.0028 0.54 0.0048 0.0024 0.54
3.5 0.0075 0.0037 0.54 0.0060 0.0032 0.54 0.0051 0.0028 0.54 0.0044 0.0024 0.54
4.0 0.0068 0.0038 0.54 0.0055 0.0032 0.54 0.0046 0.0028 0.54 0.0040 0.0024 0.54

Fig. 12. Comparison with mechanics-based proposal: (a) Serviceability; (b) damage control.

Fig. 13. Computed damage models: (a) ρυ / ρυ, conf = 1; (b) ρυ / ρυ, conf = 0.2 .

17
C. Perdomo and R. Monteiro Engineering Structures 217 (2020) 110794

Fig. 14. Behaviour of drift ratio at shear failure: (a) Median; (b) dispersion.

and the results are shown in Fig. 12. (geometry, materials, reinforcement quantities, axial load) were found
In Fig. 12 the solid lines correspond to Kowalsky’s proposal whereas to highly influence the prediction of shear failure, (iii) an additional
the dashed lines correspond to the proposed relationships. It is seen that parameter, f y / f 'c , was found to be determinant in the prediction of
the models broadly correspond with each other, although the match is shear failure, i.e. a column with the same geometrical and reinforce-
better for the serviceability limit state. Nevertheless, it is noted that ment layout characteristics is more likely to present shear failure if the
Kowalsky’s proposal tends to compute slightly lower EDP values in both value of f y / f 'c is higher, which is a direct effect of the Modified UCSD
DLS for the investigated column. Keeping in mind that Kowalsky’s model formulation; (iv) median EDP values at shear failure are not a
findings were based on a reference section with fixed values of concrete continuous function (i.e. not defined for all values) of the diameter nor
and reinforcement yield strength, the comparison with the mechanics- the aspect ratio, given that, e.g. columns with large aspect ratio are very
based model was taken as supportive of the proposed relationships. unlikely to present shear failure; and (v) the constructed datasets con-
Similar observations can be made when curvature ductility or dis- tain data points that are unlikely to be found in practice; for example,
placement ductility are compared. there are reported values of EDP at shear failure for columns with a
high quantity of longitudinal reinforcement in tandem with very low
6.2. Summary description of obtained damage models quantity of transverse reinforcement. Since the moment capacity is
correlated with the moment demand and the moment demand is cor-
By defining the characteristics of a given column, e.g. diameter, related with the shear demand, which in turn would control the pro-
aspect ratio, axial load ratio and reinforcement quantities, the median vided shear capacity, these points are thought to be unrealistic. A
and dispersion values for such column at each predefined limit state can sample of the observed behaviour of drift ratio at shear failure is shown
be retrieved from the produced datasets or computed with the simpli- in Fig. 14 and similar observations were found for displacement duc-
fied relationships. With both values, a lognormal cumulative distribu- tility.
tion function can be computed. An example of such damage models, The observations made before render the formulation of simplified
with drift ratio as EDP, is presented in Fig. 13 for the same column equations for EDP at shear failure difficult, therefore, at the moment,
investigated in Fig. 2, however, for values of ρυ / ρυ, conf = 1 and only the constructed dataset is deemed useful for the construction of
ρυ / ρυ, conf = 0.2 . damage models for the probabilistic characterization of shear failure.
In Fig. 13 the solid lines correspond to damage models created using
the values from the developed datasets, whereas the dashed lines cor- 7. Discussion of results and generated damage models
respond to those obtained using the simplified equations. For those
limit states where the dashed lines are not visible it is because they In this study, a number of assumptions were made in order to
overlap with the solid ones, showing, in this case, a very good agree- construct the damage models associated with each predefined limit
ment between the comprehensive and simplified alternatives. With the state. As such, these assumptions limit the applicability of the proposed
defined damage models, consequence functions (e.g. repair cost, dis- datasets or simplified equations to members with the following char-
ruption time) can be defined for each DLS and PBEE for different de- acteristics: i) cantilever columns; ii) columns where bar slip at the
cision variables can be carried out. critical section is not expected or not relevant; iii) columns where
corrugated steel is used for the longitudinal reinforcement (e.g. no
6.3. Displacement ductility and drift ratio at shear failure smooth bars); and, iv) columns where reinforcement splices at the cri-
tical section are not present.
Datasets similar to those presented in Table 7 were also produced In addition, pre-existing detrimental conditions due to lack of
for displacement ductility and drift ratio at shear failure following the maintenance and exposition to the environment, such as loss of re-
same procedure outlined before, however, with a value of logarithmic inforcement cross section due to corrosion, were not taken into account.
dispersion accounting for epistemic uncertainty of 0.54 (maximum Consequently, less suitability should be expected for old structures,
found in the comparison of experimental-to-predicted values for the where, for example, smooth bars were used in construction and re-
flexural response). inforcement splices were located at the column’s base. Nevertheless, for
Several observations can be made with respect to the behaviour at preliminary assessment, the proposed tools are deemed to be useful.
shear failure: (i) the value of dispersion adopted for accounting for Moreover, for what concerns the characteristics of the studied col-
epistemic uncertainty can be debatable, since no direct comparison umns and the parameters for the simulation process, it should be con-
with experimental results was made, and the median values predicted sidered that the proposed relationships should not be used for columns
from the simulations were obtained using the Modified UCSD model with diameters, aspect ratios, axial load ratios and reinforcement
only; (ii) all the parameters defining the characteristics of the column quantities outside of the values presented in Table 1. When it comes to

18
C. Perdomo and R. Monteiro Engineering Structures 217 (2020) 110794

material characteristics, columns constructed with concrete strengths simplicity of the proposed damage models makes them suitable and
outside the range of 16–52 MPa, or steel yield strengths outside the appealing for preliminary assessments in bridge portfolios with a large
range of 360–540 MPa, are not recommended to be characterized with stock of structures, where a quick, yet accurate, screening allows to
the proposed tools. Finally, since the reinforcement elasticity modulus efficiently allocate resources, by identifying the most vulnerable
was taken as constant in this study, care should also be placed in structures that require a more comprehensive evaluation.
characterizing columns with reinforcement yield strain in the range of Some limitations can be kept in mind for the applicability of the
0.0018 to 0.0027, which correspond to the considered yield stress proposed datasets or the simplified equations for the development of
range. damage models. Firstly, the proposed damage models are valid for
members with the following characteristics: i) cantilever columns; ii)
8. Conclusions columns where bar slip at the critical section is not expected or not
relevant; iii) columns where corrugated steel is used for the long-
A study for the construction of damage models for cantilever cir- itudinal reinforcement; and, iv) columns where reinforcement splices at
cular section RC bridge columns was carried out and detailed in this the critical section are not present. Secondly, members for which pre-
manuscript. Four damage limit states were defined and were associated existing detrimental conditions due to lack of maintenance and en-
to specific physical events in the members. Three engineering demand vironmental exposure are deemed not to be critical. Thirdly, members
parameters were evaluated for their construction, namely, curvature whose characteristics (e.g. diameters, axial load, reinforcement quan-
ductility, displacement ductility and drift ratio. General characteristics tities, aspect ratio) are within the ranges adopted during the simulation
for the studied columns were defined and by adopting statistical dis- process. Finally, similar outputs for the characterisation of columns in
tributions for material properties, simulated columns were created and double-bending are required in order to know the bounding values of
their structural response was determined using sectional analysis along columns with other end-fixity conditions. Further studies in these di-
with the plastic hinge method, thus, accounting for aleatory uncertainty rections are necessary to provide a more comprehensive characterisa-
in the column’s material parameters. Regression analysis was per- tion for a wider range of columns found in practice.
formed in order to fit closed form expressions for the prediction of
EDPs. These closed form solutions were subsequently used for com- CRediT authorship contribution statement
parison with experimental results to assess the effect of the epistemic
uncertainty, ignored in the simulation process. Camilo Perdomo: Methodology, Formal analysis, Writing - original
As a result of the procedure, a set of databases containing the draft. Ricardo Monteiro: Conceptualization, Methodology, Writing -
median EDP and dispersion values at each of the predefined limit states review & editing.
were created. Moreover, scrutiny of the databases demonstrated that
the estimation of median EDP and dispersion values could be further Declaration of Competing Interest
simplified. Investigation of the trends and relevance of the controlling
parameters showed that curvature ductility could be expressed as a The authors declare that they have no known competing financial
function of the axial load ratio, whereas displacement ductility and drift interests or personal relationships that could have appeared to influ-
ratio could be expressed as a function of the aspect ratio with de- ence the work reported in this paper.
pendences on the column diameter and axial load ratio, for fixed values
of the amount of transverse reinforcement. Acknowledgments
As a consequence, three values of the amount of transverse re-
inforcement with respect to that required for confinement according to The authors greatly acknowledge the support provided by the re-
modern seismic design codes were considered to fit simplified re- search project INFRA-NAT (783298-UCPM-2017-PP-AG), co-funded by
lationships for median EDP and dispersion values calculation. These the European Commission ECHO – Humanitarian Aid and Civil
values of transverse reinforcement can be related with columns de- Protection, and the project “Dipartimenti di Eccellenza,” funded by the
signed according to modern seismic provisions, columns designed for Italian Ministry of Education, University and Research at IUSS Pavia.
shear but without confinement considerations or capacity design prin-
ciples, and columns provided only with minimum shear reinforcement. References
Such characteristics can be assumed based on the year of construction
and/or location of the bridge. A linear relationship was found to be [1] Porter KA. An overview of PEER’s performance-based earthquake engineering
satisfactory for the construction of the simplified relationships for methodology. In: Proc., 9th international conference on applications of statistics
curvature ductility and drift ratio, whereas a power law was suitable for and probability in engineering, San Francisco, California; 2003.
[2] Deierlein GG. Overview of a comprehensive framework for earthquake performance
the prediction of the displacement ductility. The dispersion values were assessment. In: Proceedings of an international workshop; 2004. p. 15–26.
found to be well described by a simple average of the source data, given [3] Park YJ, Ang AH-S. Mechanistic seismic damage model for reinforced concrete. J
its low variability within each limit state. Damage models using a Struct Eng 1985;111(4):722–39.
[4] Park YJ, Ang AH-S, Wen YK. Damage limiting-aseismic design of buildings. Earthq
lognormal cumulative distribution function were subsequently defined Spectra 1987;3(1):1–26.
with the output of the databases or simplified relationships. [5] Hwang H, Jernigan JB, Lin Y-W. Evaluation of seismic damage to Memphis bridges
Consideration was also given to the displacement ductility and drift and highway systems. J Bridge Eng 2000;5(4):322–30.
[6] Kircher CA, Whitman RV, Holmes WT. HAZUS earthquake loss estimation methods.
ratio at shear failure, thus, databases containing its probabilistic de-
Nat Hazards Rev 2006;7(2):45–59.
scription were created. Scrutiny of the databases showed that all con- [7] Federal Emergency Management Agency (FEMA). Seismic Performance Assessment
trolling parameters, including the ratio of reinforcement yield to con- of Buildings, Volume 1 – Methodology. FEMA P-58-1¸ Washington D.C.; 2012.
[8] Ruiz-García J, Aguilar JD. Aftershock seismic assessment taking into account
crete strength, are highly influencing for the estimation of the damage
postmainshock residual drifts. Earthq Eng Struct D 2014;44(9):1391–407.
models; this feature did not enable the development of simplified re- [9] Kowalsky MJ. Deformation limit states for circular reinforced concrete bridge col-
lationships for this limit state. umns. J Struct Eng 2000;126(8):869–78.
The damage models proposed herein can be used in performance- [10] Priestley MJN, Calvi GM, Kowalsky MJ. Direct displacement-based design of
structures. Pavia, Italy: IUSS Press; 2007.
based earthquake engineering, for the definition of consequence func- [11] Panagiotakos TB, Fardis MN. Deformation of reinforced concrete members at
tions and, subsequently, the seismic risk of a given bridge due to yielding and ultimate. ACI Struct J 2001;98(2):135–48.
column vulnerability. Furthermore, based on the presented results, one [12] Berry MP, Eberhard MO. Performance models for flexural damage in reinforced
concrete columns. PEER Report 2003/18. Berkeley, U.S.: University of California;
can assess which EDP is more suitable for this purpose depending on the 2003.
available information and the output from the structural analysis. The

19
C. Perdomo and R. Monteiro Engineering Structures 217 (2020) 110794

[13] Kameshwar S, Padgett JE. Characterizing and predicting seismic repair costs for AASTHO LRFD bridge design specifications, Washington, DC; 2007.
bridges. J Bridge Eng 2017;22(11):040117083. [29] Ellingwood B, Hwang H. Probabilistic descriptions of resistance of safety-related
[14] Billah AHMM, Alam MS. Performance based seismic design of shape memory alloy- structures in nuclear plants. Nucl Eng Des 1985;88:169–78.
reinforced concrete bridge piers. I: Development of performance-based damage [30] Kappos AJ, Chryssanthopoulos MK, Dymiotis C. Uncertainty analysis of strength
states. J Struct Eng 2016;142(12):04016140. and ductility of confined reinforced concrete members. Eng Struct
[15] Vamvatsikos D, Cornell CA. Incremental dynamic analysis. Earthq Eng Struct D 1999;21:195–208.
2001;46:73–93. [31] Verderame GM, Stella A, Cosenza E. Le propietà meccaniche degli acciai impiegati
[16] Stefanidou SP, Kappos AJ. Bridge specific fragility analysis: when is it really ne- nelle strutture in c.a. relizzate negli anni ’60. In: Proc. X Congresso Nazionale
cessary? B Earth Eng 2018;17(4):2245–80. “L’ingegneria Sismica in Italia”., Assocciazione Nazionale Italiana di Ingegneria
[17] Nielson BG, DesRoches R. Analytical seismic fragility curves for typical bridges in Sismica, Potenza-Matera (in Italian); 2001.
the Central and Southeastern United States. Earthq Spectra 2007;23(3):615–33. [32] Ang AH-S, Tang WS. Probability concepts in engineering. River Street, U.S.: John
[18] Ghosh J, Padgett JE. Probabilistic seismic loss assessment of aging bridges using a Wiley & Sons; 2007.
component-level cost estimation approach. Earthquake Eng Struct [33] Monteiro R. Sampling based numerical seismic assessment of continuous span RC
2011;40:1743–61. bridges. Eng Struct 2016;118:407–20.
[19] Choi E, DesRoches R, Nielson B. Seismic fragility of typical bridges in moderate [34] King DJ, Priestley MJN, Park R. Computer programs for concrete column design.
seismic zones. Eng Struct 2004;26:187–99. Research Report 86/12. New Zealand: Department of Civil Engineering, University
[20] Cardone D. Displacement limits and performance displacement profiles in support of Canterbury; 1986.
of direct displacement-based seismic assessment of bridges. Earthquake Eng Struct [35] Montejo LA, Kowalsky MJ. CUMBIA – A set of codes for the analysis of reinforced
2014;43:1239–63. concrete members, theory and user guide. Technical Report No. IS-07-01. U.S.:
[21] Monteiro R, Zelaschi C, Pinho R. Derivation of fragility functions for seismic as- Constructed Facilities Laboratory, Department of Civil, Construction and
sessment of RC bridge portfolios using different intensity measures. J Earthquake Environmental Engineering, NC State University; 2007.
Eng 2017. https://doi.org/10.1080/13632469.2016.1264334. [36] Perdomo C. Direct economic loss assessment of multi-span continuous RC bridges
[22] Zelaschi C, Monteiro R, Pinho R. Critical assessment of intensity measures for under seismic hazard. Ph.D. Dissertation. Pavia, Italy: University Institute for
seismic response of Italian RC bridge portfolios. J Earthquake Eng Advanced Studies (IUSS); 2020.
2017;23(6):980–1000. [37] Lallemant D, Kiremidjian A. A beta distribution model for characterizing earth-
[23] Mander JB, Priestley MJN, Park R. Theoretical stress-strain model for confined quake damage state distribution. Earthq Spectra 2015;31(3):1337–52.
concrete. J Struct Eng 1988;114(8):1804–26. [38] Billah AHMM, Alam MS. Statistical distribution of seismic performance criteria of
[24] Moyer MJ, Kowalsky MJ. Influence of tension strain on buckling of reinforcement in retrofitted multi-column bridge bents using incremental dynamic analysis: a case
concrete columns. ACI Struct J 2003;100(1):75–85. study. B Earthq Eng 2013;11:2333–62.
[25] Berry MP, Eberhard MO. Practical performance model for bar buckling. J Struct Eng [39] Berry M, Parrish M, Eberhard M. PEER structural performance database user’s
2005;131(7):1060–70. manual. Berkeley: Pacific Earthquake Engineering Research Center, University of
[26] Palermo D, Vecchio FJ. Compression field modelling of reinforced concrete sub- California; 2004.
jected to reverse loading: formulation. ACI Struct J 2003;100(5):616–25. [40] Priestley MJN, Seible F, Calvi GM. Seismic design and retrofit of bridges. New York,
[27] Bentz EC, Collins MP, Vecchio FJ. Simplified modified compression field theory for U.S.: John Wiley and Sons; 1996.
calculating shear strength of reinforced concrete elements. ACI Struct J [41] Reza SM, Alam MS, Tesfamariam S. Lateral load resistance of bridge piers under
2006;103(4):614–24. flexure and shear using factorial analysis. Eng Struct 2014;59:821–35.
[28] American Association of State Highways and Transportation Officials (AASHTO).

20

S-ar putea să vă placă și