Escolar Documentos
Profissional Documentos
Cultura Documentos
To cite this article: Meng-Hao Tsai & Wen-Shinn Shyu (2015) A proper estimation of inelastic dynamic increase factor in
support-loss experiments, Journal of the Chinese Institute of Engineers, 38:3, 372-381, DOI: 10.1080/02533839.2014.970388
Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the
Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and
are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and
should be independently verified with primary sources of information. Taylor and Francis shall not be liable for
any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever
or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of
the Content.
This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions
Journal of the Chinese Institute of Engineers, 2015
Vol. 38, No. 3, 372–381, http://dx.doi.org/10.1080/02533839.2014.970388
Dynamic increase factor (DIF) is usually approximated as the ratio of maximum to neutral response in dynamic support-
loss experiments. However, this neutral-based calculation may underestimate the real DIF for elasto-plastic response. In
this study, an analytical formulation is thus proposed to correct the neutral-based DIF based on measured dynamic
response. Accuracy of the analytical formulation is verified with the results of a small-scale experiment. An iterative pro-
cedure, which is validated by the experimental results, is suggested for practical application of the correction approach.
Its application to sudden support-loss tests of prototype buildings is demonstrated with numerical simulation for two
model frames. The analysis results indicate that the corrected DIFs may be approximated to the real DIFs in most cases.
Downloaded by [Meng-Hao Tsai] at 08:59 25 March 2015
Due to the coupled vertical oscillation under corner column loss, the correction approach is more favorable for interior
column loss situations that are dominated by a single mode. The ductility demand obtained from the iterative procedure
may serve as an indication of structural damage under column loss.
Keywords: dynamic increase factor; support-loss test; ductility demand; iterative analysis procedure
1. Introduction and Williamson 2004; Kim and Kim 2009; Kim, Kim,
It is well known that dynamic loadings may provoke dif- and An 2009; Ruth, Marchand, and Williamson 2006;
ferent structural responses from static loadings even if Tsai and Lin 2008). In the latest issued UFC 4-023-03
their magnitudes are equal. This difference in structural guidelines (U.S. DoD 2009), two different magnification
response is generally defined as the dynamic amplifica- factors, namely the load increase factor (LIF) and
tion effect. Its basic definition may be found in funda- dynamic increase factor (DIF), are suggested for consid-
mental structural dynamic theory and is usually ering the dynamic effect in the linear and nonlinear static
represented by a dynamic amplification factor (DAF) analysis, respectively. The LIF is intended to account for
(Chopra 1995; Clough and Penzien 1993). The intention both the material nonlinearity and dynamic effect, while
of the DAF is to facilitate the prediction of maximum the DIF is only responsible for the latter. Also, analytical
dynamic structural response using static analysis results. expressions of DAFs for elasto-plastic SDOF models
The DAF of elastic structures is of concern in most have been proposed for progressive collapse analysis of
cases. Numerical and analytical solutions of elastic DAFs building frames under column loss (Tsai 2010; Tsai and
for several loading patterns have been established (Biggs Lin 2009).
1964; Chopra 1995). Nevertheless, in the wake of grow- Recently, several prototype dynamic experiments were
ing attention on structural response under sudden column conducted to investigate the gravity load–redistribution
loss, it is necessary to investigate the effect of structural behavior (Matthews, Elwood, and Hwang 2007; Sasani,
plasticity on the DAFs. The DAF obtained from a single Bazan, and Sagiroglu 2007; Sasani and Sagiroglu 2008,
degree-of-freedom (SDOF) model subjected to a step- 2010; Sasani et al. 2011; Tian and Su 2011). Prototype
loading function has been used for building structures experiments are always costly and valuable. Therefore, it
under sudden column loss. The dynamic effect has been seems infeasible to conduct a series of large-scale
simulated by a constant DAF equal to 2.0 in the static dynamic tests as desired. Due to the nature of failure tests,
analysis procedures as recommended by US General usually only dynamic response under a specific loading
Service Administration (GSA) (U.S. GSA 2003) and the magnitude may be obtained from a prototype specimen.
Unified Facilities Criteria UFC 4-023-03 by the Sometimes, it is challenging to obtain all the required
Department of Defense (DoD) (U.S. DoD 2005). information from the limited test data. Since static
However, several studies revealed that a constant DAF response cannot be obtained from the prototype specimen,
of 2.0 may lead to results inconsistent with those the DIF is thus usually estimated from the measured
obtained from nonlinear dynamic analyses (Kaewkulchai maximum and neutral responses in the dynamic failure
test (Matthews et al. 2008; Tian and Su 2011). This simple the value of W =k is denoted as ub , which represents the
approach is appropriate for elastic DIF, which is equal to spring deformation as the support is released statically, the
2.0. However, as the tested structure is loaded into the DIF is then calculated as umax =ub ¼ 2. Since ub is also the
inelastic phase, the obtained DIF may not really reflect neutral position of the block under the support-loss
the influence of plasticity on dynamic amplification. In vibration, the DIF is conventionally estimated as the ratio
this study, a correction approach is proposed for the of maximum dynamic response to the neutral response.
neutral-based DIF, which is obtained from dividing the However, as the maximum deformation umax goes beyond
maximum displacement by the neutral displacement in the elastic limit, the elastic resilience is dependent on the
dynamic support-loss tests. An analytical relationship nonlinear behavior of the material. The DIF may be no
between the neutral-based DIF and ductility demand is longer equal to 2.0.
developed. The ductility demand is then used to estimate Suppose that the weight of the block, W, is sufficient
the DIF under the support-loss test. An iterative procedure to load the spring into the inelastic range under support
is suggested for the correction approach. Test results of a loss. The yield deformation and post-yield stiffness ratio
small-scale experiment are used to verify the proposed of the spring are assumed as uy and a, respectively.
formulation. Practical application of the correction Then, neglecting inherent damping, the potential loss
approach to structural frames is demonstrated and will be equal to the stored strain energy in the spring at
discussed. the instant of maximum deformation. That is,
Downloaded by [Meng-Hao Tsai] at 08:59 25 March 2015
1
2. Derivation and correction of the neutral-based Wumax ¼ Py uy f1 þ ðl 1Þ½2 þ aðl 1Þg
2 n o
DIF 1
¼ Py uy aðl 1Þ2 þ 2l 1 ; (2)
2.1. Analytical derivation 2
Consider a block with weight W and suspended by a where Py ¼ kuy and l ¼ umax =uy are the static yield
spring with elastic stiffness k, as shown in Figure 1(a). force and ductility of the spring, respectively. Although
Suppose that the block is originally supported such that the sustained external loading is W , the maximum spring
the spring is not stretched at the beginning. If the block force Pmax may be larger than W at the maximum defor-
is not so heavy as to load the spring into the inelastic mation, as shown in Figure 1(b). A positive post-yield
range, an energy equation of the block under support stiffness ratio is adopted herein because it is regarded as
loss may be written as a necessity for stable support-loss vibration. Pmax is, in
fact, in equilibrium with the sustained loading W and the
Wu ¼ ku2 =2 þ mv2 =2; (1) inertia force induced by the block. As the support-loss
vibration ceases at neutral displacement ub , it comes to a
where m, u, and v are the mass, displacement, and velocity static equilibrium state when the spring force is equal to
of the block, respectively. Damping-induced energy loss is W . If the plastic behavior of the spring follows the kine-
neglected in Equation (1). At the maximum deformation matic strain-hardening rule, the difference between the
umax , the kinetic energy is zero. Therefore, Wumax is equal maximum and the neutral deformation of the spring may
to the strain energy ku2max =2, and umax is equal to 2W =k. If be expressed as
Figure 1. (a) Schematic of a spring–mass system and (b) Force–displacement response of an elasto-plastic spring under step
loadings.
374 M.-H. Tsai and W.-S. Shyu
d ¼ umax ub ¼ ðPmax W Þuy =Py ; (3) estimate the corresponding ductility demand for the
support-loss test. The estimated ductility demand is
where Pmax ¼ Py ½1 þ aðl 1Þ. Upon substituting
written as
Equation (2) into the above equation, it may be derived
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
that gð1 aÞ
l¼ for g 2: (7)
d ¼ umax ub ¼ ðaðl2 1Þ þ 1Þuy =2=l for l 1: gð2 aÞ 2
(4)
The values of umax and ub may be obtained from the Then, the ductility l from Equation (7) may be substi-
measured displacement time histories in support-loss tuted into Equation (6) for correction of the neutral-based
tests. Hence, the DIF is calculated as DIFs. In this way, the DIFs corrected from the neutral-
based DIFs will be exactly identical to the real DIFs
umax 2l2 from Equation (6). Thus, both the ductility demand and
g¼ ¼ for l 1:
umax d ð2l2 1Þ aðl2 1Þ real DIF may be obtained from the support-loss experi-
(5) ment. However, it is assumed that the post-yield stiffness
ratio is given beforehand in this correction approach.
This DIF, g, is defined as the neutral-based DIF in this
This is usually not the case in practice. Therefore, an
study since it is determined from the neutral response.
iterative procedure is proposed in a later section to deter-
As revealed from Equation (5), the neutral-based DIF
Downloaded by [Meng-Hao Tsai] at 08:59 25 March 2015
2.2. Correction
3. Experimental verification
A general definition of the DIF is usually given as the
3.1. Test setup
ratio of maximum dynamic displacement to static
displacement under the same load demand. Based on a A small-scale test setup with manual support-loss mecha-
similar philosophy, it may be defined as the ratio of nism was devised for the DIF experiment (Tsai and You
static to maximum dynamic force response under the 2012). Figure 3(a) shows a schematic drawing for the
same displacement demand (Tsai 2010). For better elevation view of the test setup. The test specimen is
distinction, the former and the latter are, respectively, made of structural steel with design yield stress of
defined as displacement-based DIF and force-based DIF. 400 MPa and has a cross section of 30 mm wide and
Therefore, if the same displacement demand umax is 3 mm deep. From material tests of the specimen, the
selected as the basis, the DIF will be expressed as yield stress and strain are equal to 422 MPa and 0.002,
respectively. Details of the test setup may be gathered
Pmax 2l½1 þ aðl 1Þ from Tsai and You (2012). Iron blocks with average
DIFp ¼ ¼ for l 1: (6)
W faðl 1Þ2 þ 2l 1g weight of 6.30 N and lead plates with average weight of
12.37 N each are used as the imposed loadings. A rect-
This force-based DIF is regarded as real DIF for compar-
angular steel basket is used to accommodate the imposed
ison with the neutral-based DIF. Analytical formulations
loadings. A laser displacement meter is installed on a
of displacement-based DIF may be derived if the same
reference frame to measure the displacement 225 mm
force demand is used as the basis.
away from the presumed fixed end. A hanger is fixed to
Figure 2(a) and (b) presents the variation of
the reference frame with a pinned connection. It is used
numerical neutral-based and real DIFs with ductility for
to support the imposed loading before sudden loss. The
selected post-yield stiffness ratios, respectively. It is seen
hanger may be manually knocked off by a hammer to
that the real DIF may increase when the ductility is lar-
simulate the sudden-loss scenario in the dynamic tests.
ger than a certain value for positive post-yield stiffness.
As for the static tests, the applied loading is increased by
However, the neutral-based DIF shows an asymptotically
sequentially placing the iron blocks into the basket and
decreasing trend with ductility. Relative errors between
the displacement response of each loading step is mea-
the neutral-based and real DIFs are shown in Figure 2(c).
sured. The picture of a deformed specimen under static
Except for the case with zero post-yield stiffness, the
loading is shown in Figure 3(b).
errors monotonically increase with ductility. The neutral-
based approach may significantly underestimate the real
DIF for large ductility. Also, the relative error appears
insensitive to the variation of post-yield stiffness ratio 3.2. Test results
when the stiffness ratio is larger than 0.1. As observed Figure 4(a) shows the fundamental nonlinear static load–
from Equation (5), the neutral-based DIF may be used to displacement response of the specimens. The ordinate is
Journal of the Chinese Institute of Engineers 375
3.3. Verification
Idealized approximation of the nonlinear static response
is established for verification of the proposed correction
approach. Because of the strain-hardening behavior, it is
more appropriate to idealize the static load–displacement
response of the test specimen as a trilinear curve. The
elastic, post-yield, and hardening stiffness are, respec-
tively, estimated as 20.65, 5.19, and 11.07 N/cm. The
post-yield and hardening stiffness ratios and the thresh-
olds of load and displacement response for the trilinear
approximation are summarized in Table 1. In the table,
Py , Ps , and Pu are, respectively, the yield loading, the
loading at the onset of strain-hardening, and the maxi-
mum applied loading in the static test. Their correspond-
ing displacements are denoted as uy , us , and uu . The
post-yield and hardening stiffness ratios are, respectively,
calculated as
Figure 3. (a) A schematic drawing of the test setup and (b) A photo of a deformed specimen.
Downloaded by [Meng-Hao Tsai] at 08:59 25 March 2015
Figure 4. (a) Nonlinear load–displacement response of the specimens and (b) Typical displacement time histories under three
different loading demands.
are compared with the neutral-based and real DIFs in accuracy may decrease if the nonlinear response presents
Figure 5(a). Apparently, the correction leads to better significant strain-hardening behavior. This is observed
approximation to the real DIFs. Relative errors with from the pronounced error under the largest ductility
respect to the real DIFs are shown in Figure 5(b) for the demand in Figure 5(b). In fact, a similar formulation
neutral-based and corrected DIFs. Except for the DIF may be derived for trilinear approximation if necessary.
under the largest ductility demand, the relative errors are In such a case, more complicated expressions are
significantly reduced through the correction. expected for the idealized tri-linear response. For
example, the relationship between the neutral-based DIF
and the ductility will become
4. Limitations of the analytical formulation
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
4.1. Idealized elasto-plastic response g½ð1 a1 Þ þ l2s ða1 a2 Þ
l¼ (9)
Since the proposed analytical formulation is derived 2g 2 a2 g
based on an assumption of elasto-plastic response, its
Journal of the Chinese Institute of Engineers 377
Figure 5. (a) Comparison of DIFs and (b) Relative errors with respect to the real DIFs.
Downloaded by [Meng-Hao Tsai] at 08:59 25 March 2015
2l½1 þ a1 ðls 1Þ þ a2 ðl ls Þ
DIFp ¼ h i:
1 þ a1 ðls 1Þ þ 2ðls 1Þ þ 2a1 ðls 1Þðl ls Þ þ a2 ðl ls Þ2 þ 2ðl ls Þ
2
(10)
the value of lmax will increase toward infinity as the
These two equations, Equations (9) and (10), are post-yield stiffness ratio approaches zero, an engineer-
applicable to the strain-hardening range with l ls , ing-based maximum ductility ratio may be provided in
where ls stands for the ductility threshold of strain hard- practice. Moreover, Equation (7) may have a meaningful
ening. Moderate improvement in the corrected DIF for ductility value only if gð2 aÞ 2 [ 0. This brings
the strain-hardening vibration may be obtained with the about a\2ðg 1Þ=g. Meanwhile, Equation (7) reveals
trilinear approximation. Nevertheless, since elasto-plastic that the ductility demand l increases with the post-yield
response is usually the dominant behavior in support-loss stiffness ratio for a fixed neutral-based DIF, g. Thus, a
tests, the proposed correction approach is sufficient for lower p bound of ductility may be obtained as
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
most practical applications. lmin ¼ g=ð2g 2Þ with a ¼ 0. To sum up, the post-
yield stiffness ratio and corresponding ductility demand
are bounded as
4.2. Bounded estimation of post-yield stiffness ratio
and ductility 0 a\min:f2ðg 1Þ=g; 0:5g; (12a)
Usually, the post-yield stiffness ratio cannot be deter- pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
mined from a single dynamic support-loss test. However, g=ð2g 2Þ l ð1 aÞ=a: (12b)
a predetermined post-yield stiffness ratio is needed for From the above investigation, an iterative procedure is
utilization of Equations (6) and (7). It is noted that from proposed for the determination of the post-yield stiffness
Equation (2) and Pmax ¼ Py ½1 þ aðl 1Þ, Equation (3) ratio and ductility demand. At first, an initial guess for a
may be rewritten as
378 M.-H. Tsai and W.-S. Shyu
is made according to Equation (12a). Then, Equation (7) average of lmax and lmin . Detailed steps of the procedure
is used to calculate the resulting ductility l. Since the are shown in Figure 6. Although the resulting ductility
value of lmax is dependent on a, the mean value of lmax may not really equal the true value, it can still be used
and lmin may vary accordingly. The post-yield stiffness as an index of damage. Figure 7(a) compares the
ratio is adjusted until the ductility is approximated to the ductility estimate by using the iterative procedure to that
obtained from the static and dynamic test results under
various applied loadings. It is seen that except for the
strain-hardening region, approximate ductility may be
obtained. In the figure, the experimental ductility is
calculated as the maximum dynamic displacement
divided by yield displacement, so it is less than 1.0 in
the elastic region. Figure 7(b) shows the corresponding
experimental and predicted DIFs. Accuracy of the itera-
tive procedure is comparable to the previous estimation
with a known post-yield stiffness ratio. Therefore, its
application to general elasto-plastic dynamic response
may be confirmed.
Downloaded by [Meng-Hao Tsai] at 08:59 25 March 2015
Figure 7. (a) Comparison of the estimated ductility under various applied loadings and (b) Comparison of the iterative DIFs under
various applied loadings.
Journal of the Chinese Institute of Engineers 379
Figure 8. (a) Elevation view and section details of the 3-story frame model and (b) Elevation view and section details of the 10-story
frame model.
nonlinear static analysis results, are used to determine It is assumed that the selected column is lost suddenly
Downloaded by [Meng-Hao Tsai] at 08:59 25 March 2015
the loading magnitudes for nonlinear dynamic analyses in the dynamic loading procedure (Kim, Kim, and Park
(Tsai 2010, 2012). Figures 9(a) and (b) shows the non- 2009; U.S. DoD 2009). In the dynamic loading proce-
linear static and pseudo-static response curves of the dure, the selected column is at first replaced with its sec-
three-story and ten-story frames, respectively. The ordi- tional resultants applied to the column-removed joint to
nate is expressed as a multiplier of the total vertical reac- simulate the intact condition of the frame. Then, a set of
tion divided by an assigned downward loading, W , equal-but-opposite step loadings is applied to the column-
which is the sum of a uniformly distributed loading removed joint to simulate the sudden column loss sce-
equal to 10 kN/m on each span adjacent to the failed nario. Contribution of inherent damping is neglected
column. The abscissa is expressed as the ductility herein. Maximum and neutral displacements of the
demands obtained from dividing the displacement of the column loss joint are measured in the nonlinear dynamic
column loss joint by the yield displacement. The yield analyses. Tables 2(a) and 2(b) summarize the dynamic
displacement is determined from the commencement of response and associated analysis results using the iterative
the first hinge event. For varied extent of plasticity, three procedure. In the table, the real DIFs are calculated as the
different loading magnitudes for the corner and interior ratios of static to dynamic loading magnitudes under the
column loss scenarios are selected for the nonlinear same displacement demands. It is seen that except for the
dynamic analyses, as shown in Tables 2(a) and 2(b), corner column loss scenarios under the lowest loading
respectively. The loading magnitudes are expressed as demand, the corrected DIFs, DIFp , may have better
multipliers of the assigned downward loading. approximation to the real values than the neutral-based
Figure 9. (a) Nonlinear static and pseudo-static response curves of the 3-story frame and (b) Nonlinear static and pseudo-static
response curves of the 10-story frame.
380 M.-H. Tsai and W.-S. Shyu
DIFs, g. In general, the correction approach presents use- not very close to the exact value, the estimated ductility
ful predictions for the interior column loss scenarios. This may serve as an index of damage induced by the support
may be caused by the fact that lateral vibration of the loss. Application of the proposed approach to estimation
structural frame is coupled with the vertical oscillation of structural DIFs under sudden column loss is numeri-
under corner column loss. Also, the proposed approach cally validated with two moment-resisting steel frames.
leads to better prediction under larger ductility demands. Approximated DIFs are obtained from correction of the
Therefore, the proposed correction approach is more neutral-based DIFs that are calculated from the maxi-
accurate for dynamic support-loss tests dominated by sin- mum and neutral displacements. Accuracy of the ductil-
gle-mode vibration. Besides, although in some cases the ity demands which resulted from the iterative procedure
predicted ductility demands are not really close to the true is acceptable for evaluation of structural damage.
values, they may still be used to roughly evaluate the Because the vertical oscillation caused by corner column
extent of structural damage without the need of sophisti- loss may be coupled with lateral vibration mode, the cor-
cated nonlinear dynamic simulation. Since the variation rection approach is more favorable for interior column
of DIF decreases as the ductility increases, the proposed loss scenarios.
correction approach may result in more accurate predic-
tion of DIF than ductility demand.
Nomenclature