Wangetal 2013 Final
Wangetal 2013 Final
net/publication/258406776
CITATIONS READS
4 364
6 authors, including:
Some of the authors of this publication are also working on these related projects:
Modeling and response of civil structures under earthquake excitation View project
All content following this page was uploaded by Xiang Wang on 10 February 2015.
(a) (b)
Figure 1. (a) Building structural skeleton, and (b) complete test building.
MODEL DESCRIPTION: OPENSEES
Finite Elements and Material Models Used
A three-dimensional model of the fixed-base test building is constructed in the
OpenSees platform using force-based fiber beam column (line type) elements to
represent the columns and walls of the test building. Rigid-link constraints are
implemented at the wall-slab connections to ensure compatible deformations between
the floor slabs and walls. The selected material model for the reinforcing steel in the
columns and walls is a uni-axial modified Menegotto-Pinto constitutive model
introduced by Filippou et al. (1983), which has been shown to be computationally
capable of capturing both kinematic and isotropic hardening as well as the
Bauschinger effect. The model utilized for the concrete material is a Kent-Park
constitutive model (Kent and Park, 1971) with multi-linear unloading/reloading rules,
which takes into account the effects of stiffness degradation and energy dissipation
under cyclic loading. To address the issue of localization in force-based beam
elements for strain-softening behavior, the material stress-strain relation is modified
based on the constant fracture energy assumption as proposed by Coleman and
Spacone (2001).
For modeling the nonlinear behavior of the different types of frame beams of the
test building, zero-length fiber section elements are introduced at the interfaces that
connect the frame beams and the joints. It is assumed that the overall nonlinear
response of the beam-column joint is lumped at the connecting interface, and the
remaining portion of the joint is modeled as rigid body elements for simplicity. This
method, originally proposed by Zhao and Sritharan (2007), aimed at modeling the
strain penetration effect of reinforced concrete columns, with the plane section
assumption that associates rigid-body rotation at the zero-length section with the fiber
deformation. In the current model, the zero-length section element has the same
geometry as the adjacent beam fiber section; however, different material constitutive
models are adopted within the fibers. In the zero-length section, deformation of the
reinforcement fiber represents the total bond-slip integrated along the anchored length
in the beam-column joint. Similarly, deformation of the concrete fiber is taken as the
sum of the concrete strain over the half width of the beam-column joint.
Due to the lumped nature of the fiber section at the beam-column joint, the
accuracy of the numerical results depends significantly on the ability of the fiber
section to capture the interface behavior characteristics (e.g. stiffness degradation,
bond deterioration pinching effect, and contact/detachment between the joint and the
beam) of the different types of frame beams. A generic hysteretic material is adopted
since it facilitates both simplicity and versatility, supports implementation of multi-
linear segments, and allows an analyst to take into account a wide range of
component hysteretic behaviors. Within this model, the slip-bar stress relation of the
rod at the interface is required to define the envelope of the hysteretic material model.
The model also includes parameters that control the nonlinear hysteretic response
under cyclic loading. The beam-column joint model demonstrates reliable predictions
when compared with full-scale frame-beam crucifix shaped component tests
conducted under quasi-static reversed cyclic loading. Details of these tests were
similar to those adopted in the test building (Chang et al., 2013).
The slab-beams at the gravity (west) bays and in the transverse direction of the
test building are also modeled by force-based fiber beam-column elements. The
effective width of the slab-beams corresponds to that of the adjacent column joints, as
additional hoops were provided at the edge of the slab at each floor. The selected
concrete and reinforcing steel material models for the slab-beams are identical with
those used for the columns and walls.
Diagonal steel rods at the east side of the test building are modeled using
corotational truss elements that are capable of considering the large deformation of
the members. Since the rods in the test building were only designed to be effective
when they are subjected to tensile forces, the selected material constitutive model is
nonlinear elastic with zero strength in compression.
Mass Distribution
The masses of the structure excluding the foundation and all the nonstructural
components on the test building are considered in the model. The masses of columns,
beams, and walls of the test building are distributed to their corresponding element
end nodes. The masses of the exterior façade, namely the balloon framing with stucco
overlay at levels 1-3 and precast concrete cladding panels at levels 4-5, are distributed
to the perimeter nodes based on individual tributary lengths, and the masses of the
remaining nonstructural components and the structural slabs were uniformly
distributed to the tributary nodes at each floor.
Damping Properties
Rayleigh damping is implemented in the model to reflect the effects of energy
dissipation contributions from various structural and nonstructural components. For
the pre-test simulations, the mass damping coefficient and stiffness coefficient are
defined to achieve 2% damping ratio at the periods of the first and second
longitudinal modes. These periods are determined by conducting modal analysis by
the numerical model with its initial stiffness.
MODEL DESCRIPTION: DIANA
While the OpenSees model, with simplified elements and slab modeling may
execute a nonlinear history analysis rapidly, relative to a model constructed using 3D
elements, the later approach facilitates more rigorous account of mechanisms
anticipated to be important in the behavior of the test structure. Importantly,
considering the specific configuration of the test specimen and its plan dimensions,
the slabs are expected to have a dominant influence on the earthquake resistance of
the structure. As seen in Figure 1, the test building was provided with large stairwell
and elevator shaft openings, complicating the slab-frame interaction and load transfer
mechanism in the diaphragm. Therefore, detailed modeling of the RC floor slabs is an
important aspect that needs special consideration in the FE modeling of the test
specimen. To satisfy these requirements, the DIANA finite element analysis software
is selected as a FE-simulation platform. This software has dedicated nonlinear
constitutive material models for modeling of RC structures. Moreover, it supports the
structural type finite elements needed such as 3D beam-column elements and shell
elements with embedded steel reinforcement.
Finite Elements and Material Models Used
The beam and column members of the test specimen are modeled using 3-node
Mindlin–Reissner displacement-based 3D beam-column elements with 18 DOFs per
element (TNO DIANA, 2010a). In usual displacement-based FE structural models
using beam-column elements, the softening-type nonlinearity is localized in one
element. Therefore, the plastic hinge regions of beam and column members are
discretized using a single element, which is referred to as a “plastic hinge” (PH-)
element. The length of the PH-element is assumed as half the height of the associated
member. To maintain computational economy, different integration schemes for the
finite elements (along their length and over their cross-section) are defined based on
their expected level of nonlinearity. The beam and column longitudinal steel
reinforcement are defined using embedded steel bars with a full bond assumption.
Considering the large stiffness of the foundation, the columns were restrained with
fixed boundary conditions at the base during the fixed base test phase.
The slabs and shear walls are modeled using 8-node Mindlin–Reissner
quadrilateral serendipity shell elements available in DIANA. This element has 6
DOFs per node including mechanics-based drilling degrees of freedom. The
reinforcing mesh in each direction of the slab is modeled using uni-axial membrane
elements with an equivalent thickness embedded in the concrete shell elements.
Figure 2 shows the FE mesh of the test specimen as developed in DIANA.
(b)
(c)
(a) (d)
Figure 2. FE mesh of test specimen: (a) 3D view, floor view of (b) level 2, (c) level
4, and (d) roof. (units in m)
The concrete constitutive material model employed in this FE model is a total strain
rotating smeared crack model. The uniaxial compressive behavior of the concrete
material consists of an initial linear and two subsequent parabolic parts. The
unloading and reloading in compression and tension occur along a secant linear
stress-strain path passing through the origin (TNO DIANA, 2010b). The “Model B”
as proposed by Vecchio and Collins (1993) is used as a built-in procedure in DIANA
to incorporate the modified compression field theory (MCFT) in the material model
of the concrete.
To alleviate the spurious numerical localization effects associated with
compression and tension failure of the concrete material, and to eliminate the
resulting mesh sensitivity, the fracture energy concept is used for the softening
branch of the stress-strain curve of the concrete material in both tension and
compression. The fracture energy (both in tension and compression) is usually
considered a constant material parameter for plain concrete. In this study, the tensile
fracture energy is selected as a constant material property, and the tension softening
behavior is defined using Reinhardt tension softening model (Reinhardt, 1984). The
value of the compressive fracture energy for the RC beam and column elements,
however, are selected to reflect the effects of mechanical confinement provided by
the transverse reinforcement in the plastic hinge regions of the beam and column
members. These confinement effects are incorporated by modifying the compressive
uni-axial stress-strain behavior of the unconfined concrete material. To achieve this, a
reverse two-step approach is used in this study. In the first step, the peak compressive
strength and corresponding strain of the concrete material in the plastic hinge regions
of each beam and column members are calculated using Mander’s model (Mander et
al., 1988). The ultimate strain of the concrete material in compression is determined
following the procedure suggested by Scott et al. (1982). As a result, a uni-axial
stress-strain curve can be defined to represent the uniaxial compressive behavior of
the confined concrete in the plastic hinge regions of the beam and column members.
In the second step, the fracture energy in compression for each member is computed
using the stress-strain curve for confined concrete derived in the previous step. For
this purpose, the effective length is defined as the length of the plastic hinge region of
the beam or column of interest. To guarantee mesh objectivity, DIANA internally
imposes a constant tensile and compressive fracture energy per element (i.e., by
scaling the uni-axial stress-strain curve of concrete based on the length of the beam-
column element).
The modified Giuffré-Menegotto-Pinto model as proposed by Filippou et al.
(1983), which is implemented in the DIANA material library, is used to describe the
hysteresis stress-strain behavior of the reinforcing steel. The parameters of the
reinforcing steel constitutive model (e.g., modulus of elasticity, yield strength, strain
hardening ratio) are obtained from the results of the tensile tests performed on
samples of reinforcement obtained from those used in the test building.
Modeling of Inertia and Damping Properties
The self-mass of the beams, columns and shear walls is modeled by assigning the
proper material mass density to the corresponding elements. Since the individual
mass contributions of the interior partition walls, ceilings, balloon frame facade at
levels 1 to 3, and the installed contents on each floor are not significant; their inertial
effects are modeled as an added uniform distributed mass over each floor slab. On the
other hand, the masses of the precast concrete claddings, stairs, elevator, roof
penthouse and cooling tower (at the roof level) are significant; therefore, they are
each are modeled with lumped masses.
The damping characteristics of the building specimen are modeled using the
Rayleigh proportional damping model by defining 2% damping at the first mode and
at 20Hz. The Rayleigh damping parameters are derived based on the initial/uncracked
linear elastic stiffness matrix of the FE building model and are held constant during
the time history analyses but with the stiffness coefficient applied to the tangent
stiffness matrix.
COMPARISON OF NUMERICAL RESULTS
The modal characteristics of the two numerical models are evaluated using the
initial stiffness and the total mass of the test building including the nonstructural
masses. The modal characteristics of the two models are nearly identical in terms of
their first four modal periods and the deformation mechanism of the mode shapes
(Table 1). Importantly, the fundamental mode is similarly identified by both models
as the longitudinal mode with a period of 0.71 seconds. Modes two through four are
also quite similar or exactly comparable (as in mode four, the second longitudinal
mode with a period estimate of 0.20 seconds. These estimates are reasonably
consistent with modal periods estimated by Astroza et al. (2013) using white noise
and ambient motion based system identification.
Table 1. Modal characteristics estimated by the DIANA and OpenSees models.
Model Characteristic 1st mode 2nd mode 3rd mode 4th mode
Period (sec) 0.71 0.68 0.47 0.20
OpenSees Transverse Torsional and
Shape Longitudinal Longitudinal
and torsional transverse
Period (sec) 0.71 0.58 0.45 0.20
DIANA Transverse Torsional and
Shape Longitudinal Longitudinal
and torsional transverse
The peak floor acceleration responses from the nonlinear time history analysis of
the two models are provided in Figures 2 and 3. Figure 3 compares the predicted
results for peak floor accelerations (PFAs) and peak inter-story drift ratios (PIDRs),
respectively, in the direction of motion input (east-west direction) for motion FB-
1:CNP100, which is scaled to target a service level event for the test building
(Hutchinson et al., 2013). Figure 4 compares the same results for motion FB-
5:DEN67, which is scaled to target the design level event for the test building. The
PFAs and PIDRs are the average of peak responses of the six column nodes at each
floor. In both cases, the predicted PIDRs are comparable between the two models,
while the deviations between the predicted PFAs of the two models are more evident,
and the results from OpenSees are constantly larger when compared with that of
DIANA at almost all the floors. It is noted that both models predict the largest PIDRs
concentrated at the lower levels of the building (levels 2-3 and 3-4) and furthermore
the shapes of the PFAs are consistent amongst the models.
(a) (b)
Figure 3. Comparison of the predicted (a) peak floor accelerations and (b) peak
inter-story drift ratios between DIANA and OpenSees (FB-1:CNP100).
(a) (b)
Figure 4. Comparison of the predicted (a) peak floor acceleration and (b) peak
inter-story drift ratio between DIANA and OpenSees (FB-5:DEN67).
DEVELOPMENT OF NCS DESIGN CRITERIA
NCS product suppliers installing their components within the test building require
criteria to estimate design forces and deformations imposed during the target motions.
Using the numerical models previously described, a recommendation is developed for
product suppliers specific to the design event earthquake (FB-5:DEN67). The input
motion used for these simulations is the original (target) motion prior to shake table
iterations.
Recommended Design Inter-story Drift Ratios
As opposed to the ASCE 7 (2010) code provisions that adopted a constant value,
which is normally 2~2.5% and depends on the type of the structure, as the maximum
drift level for all floors, the recommended design inter-story drift ratios for drift-
sensitive components considers the variation of the inter-story drift responses at
different levels. In this project, the recommended design values are determined by
taking the arithmetic average of the peak inter-story drift results as estimated with the
two numerical models (Table 2).
Table 2. Recommended design inter-story drift ratios (%).
Floor Level 1–2 2–3 3–4 4–5 5 – Roof
OpenSees 2.3 2.9 2.6 2.0 1.0
DIANA 1.2 2.2 2.4 2.0 1.1
Recommended 1.8 2.6 2.5 2.0 1.1
Recommended Design Floor Spectral Accelerations
The recommended design floor spectral accelerations, one the other hand, are
calculated based on the floor acceleration spectra of the two models. Figure 5 briefly
illustrates the proposed procedures for determination of the design floor spectral
acceleration. The floor acceleration spectra are evaluated first using the floor
acceleration time histories from each model. Figure 5(a) shows the calculated 5%
damping elastic floor acceleration spectra at level 2 of the test building. These
spectral values from the two models within the period range from 0.06 to 0.6 seconds
are subsequently averaged at an increment of 0.02 second, which is shown in Figure
5(b). It is noted that 0.06 seconds is suggested as the period defining a rigid
nonstructural component in design codes (e.g. ASCE 7, 2010). The mean and
standard deviation of the averaged spectral values within the specified period range
are then statistically determined. Finally, the design floor spectral accelerations are
determined as the mean plus one standard deviation over the selected period range.
Figure 5(c) provides the mean, the mean plus one standard deviation, and the mean
minus one standard deviation from the given period range.