Professional Documents
Culture Documents
12-2010-A New Method For Progressive Collapse Analysis of RC Frames Under Blast Loading PDF
12-2010-A New Method For Progressive Collapse Analysis of RC Frames Under Blast Loading PDF
Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct
A new method for progressive collapse analysis of RC frames under blast loading
Yanchao Shi a,b , Zhong-Xian Li a,b, , Hong Hao a,b,c
a
Key Laboratory of Coastal Civil Engineering Structure and Safety (Tianjin University), Ministry of Education, Tianjin 300072, China
School of Civil & Resource Engineering, The University of Western Australia, WA 6009, Australia
article
info
Article history:
Received 30 June 2008
Received in revised form
1 February 2010
Accepted 4 February 2010
Available online 29 March 2010
Keywords:
Reinforced concrete (RC) frame
Blast loading
Progressive collapse
Numerical analysis
Initial damage
Non-zero initial condition
Direct numerical simulation
Member-removal procedure
abstract
The progressive collapse of structures under blast loading has attracted great attention all over the
world. Some guidelines give specific procedures to analyse the progressive collapse of building structures.
Numerical analysis and laboratory test results of the progressive collapse of structures have also been
reported in the literature. Because the progressive collapse of a structure induced by blast loading
occurs only after the blast-loading phase, most of these studies and guideline procedures perform
progressive analysis by removing one or a few load-carrying structural members with static and zero
initial conditions. The damage on adjacent structural members that might be induced by blast loads
and the inevitable non-zero initial conditions when progressive collapse initiates are neglected. These
simplifications may lead to inaccurate predictions of the structural collapse process. In this paper, a new
method for progressive collapse analysis of reinforced concrete (RC) frame structures by considering nonzero initial conditions and initial damage to adjacent structural members under blast loading is proposed.
A three-storey two-span RC frame is used as an example to demonstrate the proposed method. Numerical
results are compared with those obtained using the alternative load path method, and with those from
comprehensive numerical simulations by directly applying the blast loads on the frame. It is found that
the proposed method with a minor and straightforward extension of the simplified member-removal
procedure is efficient and reliable in simulating the progressive collapse process of RC frame structures.
It requires substantially less computational effort as compared to direct numerical simulations, and gives
more accurate predictions of the structural progressive collapse process than the member-removal
approach.
2010 Elsevier Ltd. All rights reserved.
1. Introduction
Progressive collapse refers to the failure of one or a group of
key structure load-carrying members that gives rise to a more
widespread failure of the surrounding members and partial or
complete structure collapse. It is defined as the spread of an initial
local failure from element to element resulting in the collapse of an
entire structure or a disproportionately large part of it [1]. Many
accidental and intentional events, such as false construction order,
local failure due to accidental overload, damage of a critical component by earthquake and explosion, could induce the progressive
collapse of a structure. However, because of the high peak, short
duration and negative phase of the blast load, the progressive collapse induced by an explosion is very different from that by earthquake ground excitations. With the recent progressive collapse of
the Alfred P. Murrah Federal Building and World Trade Centre due
to blast and impact, research is focused more than ever to make
buildings safer from progressive collapse induced by blast and impact loading.
For an economic and safe design of structures against progressive collapse to blast loads, a reliable progressive collapse analysis is essential. Because of the catastrophic nature of progressive
collapse and the potentially high cost of retrofitting buildings to
resist it, it is imperative that the progressive analysis methods
be reliable [2]. Engineers need an accurate and concise methodology to produce trustworthy and timely results. Thus, many researchers have been spending lots of effort in developing reliable,
efficient and straightforward progressive collapse analysis methods recently.
Krauthammer et al. [3] developed a procedure for studying
progressive collapse both theoretically and numerically, and established a reliable structural damage assessment procedure to
predict a possible future phase of progressive collapse. Luccioni
et al. [4] carried out an analysis of the structural collapse of a reinforced concrete building caused by a blast load. In the analysis,
1692
1693
1
v(y) = vmax 1
(3)
Mt = mL
(4)
(5)
Ft (t ) = p(t )L
(6)
ke = KL k
(7)
where Fe (t ) is the equivalent load. Ft (t ) is the total load of the system, which is equal to the load per unit length p(t ) multiplied by
the total length of the member L. KL is the ratio of the equivalent
load to the total load, which is also related to the boundary condition and the deflection shape function. Time t is the same in the
two systems.
Through Eqs. (3)(7), all the parameters of the equivalent SDOF
system can be derived. Suppose that the equivalent blast load is
Fe (t ); since the blast load is of very short duration, the equation of
motion can be approximately written as
Fe (t ) = Me a.
R td
v td =
(1)
L/ 2
y
L/2
(2)
(8)
Then
0
Fe (t )dt
Me
Z
s td =
0
td
Ie
R td R t
v dt =
(9)
Me
0
Fe (t )dtdt
Me
(10)
in which td is the loading duration. vtd and std are the velocity and
displacement at time td , respectively. If the blast load is assumed
to be triangular, the displacement can be finally derived as
2
vt td .
(11)
3 d
Therefore, both the initial velocity and displacement can be
obtained based on the above equations. It is worth noting that if
s td =
1694
the blast load duration is so small that the member has no time
to deform during the blast-loading phase, the initial displacement
will be very small compared with the length of the member.
In this case, for the purpose of simplification, this small initial
displacement is ignored in the progressive collapse analysis of the
RC frame. However, the velocity is not necessarily small, depending
on the blast-loading impulse, and it is always important to include
it in the analysis.
3.2. Initial damage
Initial damage is another very important parameter that should
be considered in the progressive collapse analysis of RC frames. The
damage severity can be estimated by using the pressureimpulse
(PI) diagram for RC members [2023]. The damage degree of the
member is obtained by using the following proposed procedure.
Suppose that the pressureimpulse diagram of a column developed by the authors [20], as shown in Fig. 2, is available, the procedure is as follows.
a. Estimate the pressure and impulse acting on the member and
locate it in the pressureimpulse diagram in the PI space.
b. If the point is in the small damage range, for example, the
damage index D is smaller than 0.2, the initial damage of this RC
member is ignored.
c. If the point is in the range of collapse (D > 0.8), it means that
the RC column is totally damaged and does not have or only has
minimum load-carrying capacity. In this case, the corresponding
column(s) are removed at the beginning of the progressive collapse
analysis.
d. If the point is in other ranges, for example point A in Fig. 2,
the damage degree of this column is obtained by interpolation between the two adjacent damage degrees DA and DB . It should be
noted that the proper interpolation is done by deriving an intermediate PI curve that has point A on it because of the nonlinearity of the PI curve. In this study, a program is written in MATLAB
to derive this PI curve. Fig. 2 shows the pressureimpulse curve
corresponding to the damage degree D.
In order to model the initial damage of an RC member, one
should relate the above damage degree to the member material
strength and stiffness degradation. In order to do this, a few
assumptions related to the member damage are made. First, it is
assumed that damage only occurs in the concrete material. This is
reasonable because, for an RC member that still maintains a certain
level of load-carrying capacity at the end of the blast-loading
phase, the steel bar is normally in the elastic range. This is because
the blast-loading duration is very short, and damage to the RC
member in this loading phase is usually brittle failure. Therefore,
the reinforcement is unlikely to enter the plastic deformation
stage. Moreover, plastic deformation of steel bars is associated with
large cracks in the concrete, which dramatically reduces the loadcarrying capacity of the RC member. Second, it is assumed that the
(12)
Edmg = KE E (1 D)
(13)
where fc0 and E are the yield compressive strength and the Youngs
modulus of the undamaged concrete; fc0,dmg and Edmg are the yield
compressive strength and the Youngs modulus of the damaged
concrete, respectively. KY and KE are the modification factors used
to reduce the errors arising from the simplifications related to the
assumptions of uniform loading and damage only to the concrete.
In this study, however, both KE and KY are set to be 1, which, as will
be demonstrated, gives a very good representation of the effect of
structural member damage on the progressive collapse. However,
more examples with different blast-loading and structural damage
scenarios need be analysed to derive more appropriate KY and KE
values.
3.3. Validation
Numerical simulations are carried out to validate the above
procedures in deriving the initial condition and initial damage of
RC members. A typical RC column, which is extracted from the RC
frame in Section 5, is analysed using LS-DYNA. This column is 3 m
long with a cross-section of 300 mm 300 mm. It has four vertical
steel bars, each having a diameter of 24 mm with the yield stress
335 MPa. The stirrup is D10@200 with the yield stress 235 MPa.
The finite elements and material model for the studied column
are exactly the same as those used in modelling the RC frame in
Section 5, and will be described in detail there. In order to relate
the studied column to the actual member within the frame, column
constraints with higher fidelity are employed. As shown in Fig. 3(a),
a footing and a head are included in the numerical model. The outer
vertical faces of the footing and the head were constrained against
horizontal motions (i.e., in the x-direction and the y-direction)
and the bottom face of the footing is constrained against vertical
motion (i.e., in the z-direction).
The peak overpressure applied on the column is 2495 kPa, while
the reflected impulse is 3642 kPa ms. This is the same as the
blast load acting on column C1 in the RC frame in Section 5. Since
the critical standoff distance considered is 10 m, the blast load
is applied uniformly on one side of the RC column. Fig. 3(b) and
(c) give the contours of the transverse velocity and the effective
plastic strain of the column at the end of the positive phase of
the blast load. As can be seen, the distribution of the transverse
velocity along the column is approximately triangular, as assumed
in Section 3.1. The damage zones are also at both ends. Since the
blast load applied is in the impulsive range, this indicts that the
assumption made in Section 3.2 is reasonable.
In order to derive the initial velocity of the RC column using
the method discussed in Section 3.1, the RC column is simplified
into an SDOF system. The equivalent mass of the column is 369 kg;
the equivalent stiffness is 1.13 108 N/m, the equivalent blast
load is derived as 1437 kN, and the equivalent impulse is 2.10 kN s.
1695
a
Head
S
S
H
Rebar and ties
Concrete
Footing
Fig. 3. RC column model and its transverse velocity and effective plastic strain contours at the end of the blast-loading phase. (a) boundary conditions, (b) transverse velocity
contours, (c) effective plastic strain contours.
1696
Table 1
Material properties of concrete.
Compressive strength
Youngs modulus
Poissons ratio
Density
24 MPa
23 000 MPa
0.2
2500 kg/m3
Table 2
Material properties of steel.
Strength
Youngs modulus
Poissons ratio
Steel ratio
335 MPa
0.3
2%
d
for d 1 s1
fts
ts
1/3
ftd
d
TDIF =
=
for d > 1 s1
fts
ts
TDIF =
ftd
(14)
(15)
where ftd is the dynamic tensile strength at the strain rate d , fts is
the static tensile strength at the strain rate ts ( ts = 106 s1 ), and
log = 6 2, in which = 1/(1 + 8fc0 /fco0 ), fco0 = 10 MPa, and fc0
is the static uniaxial compressive strength in MPa. In compression,
the empirical formulae are given as
CDIF =
CDIF =
fcd
fcs
fcd
fcs
=
d
cs
1.026
for d 30 s1
(16)
(17)
1697
fcs is the static compressive strength, and fcu is the static cube
compressive strength in MPa.
For steel, the dynamic increase factor (DIF ) is given as [28]
DIF =
(18)
104
= 0.074 0.040
fy
414
(19)
where is the strain rate of the steel bar in s1 and fy is the steel bar
yield strength in MPa. This formulation is valid for steel bars with
yield stress between 290 and 710 MPa and for strain rate between
104 s1 and 225 s1 .
In order to simulate the progressive collapse process of the RC
frame, the so-called erosion algorithm is used. This algorithm is
employed to capture the physical fracture process of the material if no significant reverse loading occurs to the fractured elements [29]. There may be a variety of criteria governing the
erosion of the material, such as principal stress, principal strain,
shear strain, pressure, and so on. Xu and Lu [29] used the principal tensile strain as the erosion criterion for reinforced concrete.
The maximum principal tensile strain at failure is assumed as 0.01.
Their simulation results for concrete spallation show a consistent
comparison with the relevant experimental observations. Unosson [30] adopted numerical erosion based on a shear strain criterion to simulate the penetration and perforation of three types
of high-performance concrete (HPC) targets. The maximum shear
strain at failure they used is 0.8 to 0.9.
It must be emphasized here that the erosion technique is introduced to overcome the large distortion problem in numerical
simulations. It has no solid physical background. The erosion criteria must be used with caution, as early and premature erosion of
material can lead to incorrect model predictions, and significantly
increase the mesh-size dependency of the calculation [31,32].
Therefore, the limiting value for the erosion criteria, i.e., the maximum value of each damage criterion at failure, cannot be too small.
Otherwise, incorrect model predictions might occur.
In this study, two erosion criteria, i.e., principal strain and
shear strain, are adopted. The element will be deleted if either
one of the two erosion criteria is met. Limiting values for both
the principal strain and shear strain erosion criterion are carefully
selected. First, the range of the limiting value is decided according
to the available references in the literature. For the principal strain
criterion, the limiting value is set to be 0.10 initially, which is
ten times the limiting value for principal tensile strain criterion
in [29]. The limiting value for the shear strain criterion is set
to be 0.8 according to [30]. In order to get reasonable values of
these two erosion criteria, several calculations are carried out; on
gradually increasing the limiting values of these two criteria from
the numerical results, the limiting value of the principal strain
criterion for erosion is decided to be 0.15 and the shear strain
criterion to be 0.9. Further increasing these values will lead to large
distortion of the numerical elements, while decreasing the values
may result in premature eroding of the materials in the structural
model.
5.2. Benchmark analysis
In order to verify the proposed method of progressive collapse
analysis of RC frames, a benchmark progressive analysis of the
example RC frame is carried out by using the direct simulation
method. In the analysis, blast loads acting on the front face of the
RC frame are directly applied to the structure.
The blast scenario considered is a detonation on the ground surface at a distance of 10 m from the centre column in front of the RC
frame. The blast load estimation formulae and pressureimpulse
o
+ 1.980 ln (b)]e0.65Z Ps0F (0.5 Z 10)
n
IrF (0) = 2.154 + 0.291 ln (b) + [136.554
o
+ 65.001 ln (b)]e6Z IsF (0.5 Z < 1)
n
IrF (0) = 1.452 + 0.287 ln (b) + [3.221
o
+ 1.577 ln (b)]e0.65Z IsF (1 Z 10)
(20)
(21)
(22)
where PrF (0) and IrF (0) are the reflected pressure and impulse
respectively at the base of the column, b is the width of the column,
Z is the scaled distance defined by the charge weight and standoff
distance in m/kg1/3 . Ps0F and IsF are the incident pressure and
impulse at the same point as PrF (0), respectively. They could be
easily obtained from the design charts in TM5-1300 [18].
It should be mentioned here that Eqs. (20)(22) can only be used
to predict the reflected pressure and impulse at the base of the RC
column that is exactly in front of the explosive charge center with
a zero degree incident angle. For other columns, the equivalent
standoff distance is used to take into account the incident angle
effect [18].
In this study, the blast load is assumed to be uniform on each
column and all equal to the blast load at the base of the respective
column. However, as the RC frame is of three storeys, the top floor
column is up to 69 m above the ground. Therefore, for the columns
in a different storey, the height effect is considered. The formulae
to estimate the reflected pressure and impulse at height hp are [33]
PrF (hp ) = PrF (0) 31.53Z 2.64 h2p
PrF (hp ) 0 (kPa)
(0.5 Z 10)
(0.5 Z 10)
(23)
(24)
1698
Table 3
Pressure and impulsive asymptotes for PI curves obtained from the formulae in [20].
D = 0.2
Numerical results
D = 0.5
D = 0.8
P0 (kPa)
I0 (kPa ms)
P0 (kPa)
I0 (kPa ms)
P0 (kPa)
I0 (kPa ms)
750
1690
1000
2190
1300
3450
Table 4
Calculated blast loads acting on the columns.
Column
C1 & C3
C2
C4 & C6
C5
C7 & C9
C8
2495
3642
3639
4062
2117
3062
3072
3164
1362
1900
1936
1369
(25)
(26)
in which DL is the self-weight and LL is the live load of the structure. In this paper, the live load considered is 4 kN/m2 . The weight
of the infill walls, which are not modelled, is also applied on the
beams. The value is 80 kN/m2 .
At t = 100 ms, after applying the static load to the structure,
all the blast loads acting on the column of the RC frame are applied
to the structure. Fig. 6 shows the collapse process of the RC frame
simulated in the benchmark analysis through the direct simulation
method. From the figure one can see that, at t = 150 ms, the
key column collapses. At this moment, it is obvious to see that the
adjacent columns also suffer a certain level of damage. At about
t = 400 ms, the combination of the vertical load and transverse
load damages the other two first-floor columns heavily and the
two columns begin to collapse. At t = 500 ms, the second-floor
columns begin to fail due to the pulling force of the connected
beams. The whole frame goes down rapidly and collapses to the
ground at about t = 800 ms.
1699
Fig. 6. Collapse process of the RC frame from the benchmark analysis (direct simulation).
Fig. 7. Response process of the RC frame from the GSA nonlinear dynamic analysis.
they are also neglected in the analysis. Therefore, only damage and
the initial velocity of columns are considered.
The blast loads acting on the columns listed in Table 4
are plotted in the pressureimpulse diagram of the RC column
generated in Section 5.2, as shown in Fig. 8. The damage degrees of
1700
Table 5
Initial damage degree and maximum initial velocity and displacement of RC columns.
Column
C1 & C3
C4 & C6
C5
C7 & C9
C8
0.65
5.85
8.06
0.53
4.91
6.70
0.61
5.07
4.93
0.00
3.05
4.03
0.00
2.20
1.47
force of the connected beams, and the whole frame collapses to the
ground at about t = 800 ms.
5.5. Comparison and discussion
a.
b.
c.
d.
1701
Fig. 12. Comparison of the vertical velocities at node N1 from different analyses.
load and live load, negative moment occurs at the joint section, and
the steel reinforcement that is at the bottom side of the beam will
be in compression. After the collapse of the center column, the two
beams above the center column will work as a longer one. The
section of the joint, which is at the middle of the long beam, will
experience a positive moment. Thus, the stress in element E1 is in
tension. In the benchmark analysis and the proposed method, the
tension stress in the element reduces suddenly when the structural
member reaches the ground, whereas that from the GSA analysis
remains almost a constant after t = 400 ms because the structure
stabilizes. The figure also shows that the proposed method and the
benchmark analysis give similar predictions of stress in element
E1.
1702
analysis [34]. All these examples demonstrate the accuracy of using the proposed method in the prediction of structural progressive
collapse induced by explosive loadings.
6. Conclusion
Fig. 14. Comparison of the stress in element E1 (steel reinforcement) from different
analyses.
1703
[24] LS-DYNA. Keyword users manual. Livermore Software Technology Corporation. Livermore (California); 2006.
[25] Tu ZG, Lu Y. Evaluation of typical concrete material models used in hydrocodes
for high dynamic response simulations. Int J Impact Eng 2009;36(1):13246.
[26] Malvar LJ, Ross CA. Review of strain rate effects for concrete in tension. ACI
Mater J 1999;96(5):6146.
[27] Pandey AK, Kumar R, Paul DK, Trikha DN. Strain rate model for dynamic
analysis of reinforced concrete structures. J Strut Eng 2006;132(9):1393401.
[28] Malvar L. Review of static and dynamic properties of steel reinforcing bars. ACI
Mater J 1998;95(6):60916.
[29] Xu K, Lu Y. Numerical simulation study of spallation in reinforced concrete
plates subjected to blast loading. Comput Struct 2006;84(56):4318.
[30] Unosson M. Numerical simulations of penetration and perforation of high
performance concrete with 75 mm steel projectile. FOA-R00-01634-311SE.
Technical report. Tumba (Sweden): Defence Research Establishment Weapons
and Protection Division; 2000.
[31] Tasdemirci A, Hall IW. Numerical and experimental studies of damage
generation in multi-layer composite materials at high strain rates. Int J Impact
Eng 2007;34(2):189204.
[32] AUTODYN. Theory manual. Century Dynamics; 2006.
[33] Shi Y, Hao H, Li Z-X. Numerical simulation of blast wave interaction with
structure columns. Shock Waves 2007;17(1):11333.
[34] Shi YC. Dynamic response and damage mechanism of reinforced concrete
structures under blast loading. Ph.D. thesis, P.R. China: Tianjin University. 2009
[In Chinese].