Sei sulla pagina 1di 13

Engineering Structures 40 (2012) 205217

Contents lists available at SciVerse ScienceDirect

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

Static and dynamic analysis of a reinforced concrete at slab frame building


for progressive collapse
Seweryn Kokot , Armelle Anthoine, Paolo Negro, George Solomos
Joint Research Centre Institute for the Protection and Security of the Citizen, European Laboratory for Structural Assessment, Ispra, Italy

a r t i c l e

i n f o

Article history:
Received 1 September 2011
Revised 2 February 2012
Accepted 5 February 2012
Available online 28 March 2012
Keywords:
Progressive collapse
Reinforced concrete frame building
Column removal
Alternate load path method
Dynamic nonlinear analysis

a b s t r a c t
The problem of structural progressive collapse has been investigated using a real-scale reinforced concrete at-slab frame building, which has survived collapse after two of its central columns had been physically destroyed. The numerical study undertaken considers three loading scenarios, in which alternately
three different columns are being instantaneously removed, and in each case the structural response of
the frame is calculated. A nite-element linear static analysis has rst been conducted. To account for
severe dynamic effects occurring during fast dynamic events, such as explosions or impacts, dynamic
linear and nonlinear time history analyses have next been performed. For each scenario the results have
been processed in terms of demand-resistance ratios at critical cross-sections, and thus it has been
assessed whether the building would be susceptible to progressive collapse according to certain allowance criteria prescribed in technical guidelines. In this respect, three denitions of dynamic factors are
introduced and their effective applicability is assessed in view of actually calculated and guidelines-suggested values. Results show overall that the approaches of linear static and dynamic analyses would have
produced progressive collapse conditions. The nonlinear dynamic analysis predicts no mechanism which
might lead to progressive collapse, even though several plastic hinges would be formed. Merits of using
static or dynamic, linear or nonlinear analyses are discussed.
2012 Elsevier Ltd. All rights reserved.

1. Introduction
Progressive collapse of structures occurs when a local failure
triggers successive failures and leads to the total collapse or a collapse disproportionate to the original cause. There have been a few
world-wide known examples of progressive collapses such as that
of the Ronan Point residential apartment building (London, 1968),
of the Alfred P. Murrah Federal Building (Oklahoma, 1995), etc. The
rst progressive collapse regulatory documents followed the
Ronan Point partial collapse and were included into the British
standards. In turn, after the total collapse of the World Trade Center towers, many research activities have led to more detailed
guidelines on designing and preventing progressive collapses
(e.g. [13]).
There are basically two approaches when dealing with the evaluation and prevention of progressive collapses in a given structure.
The rst indirect approach consists in ensuring that the structure
satises prescriptive design rules (such as requirements on structural integrity and ductility or the presence of vertical and horizontal ties). The second direct approach uses two possibilities
depending on whether local failure is allowed or not. If local failure

Corresponding author.
E-mail address: seweryn.kokot@jrc.ec.europa.eu (S. Kokot).
0141-0296/$ - see front matter 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.engstruct.2012.02.026

is allowed, then the structure must be veried using the alternate


load path method in which a load-bearing element is removed
from the structure. If no local failure is allowed, then key elements
must be designed to sustain a notional accidental action. More detailed information on the state-of-the-art in the eld of progressive
collapse can be found in [38].
Following the issue of the guidelines [1,2], there have been published several research papers presenting numerical analyses concerning progressive collapse of steel and reinforced concrete
structures. To give a few examples of recent studies, Marjanishvili
and Agnew [9], analysed a model of a nine-storey steel momentresistant frame building applying four methods of the indirect approach using SAP 2000 nite element software. Similarly Fu [10]
considered a twenty-storey steel composite frame building model
under column removal using ABAQUS. Tsai and Lin [11] investigated progressive collapse resistance of an earthquake-resistant
reinforced concrete building model subjected to column failure
using SAP 2000. Kwasniewski [12] analysed an eight-storey steel
building using LS-DYNA software focusing on 3D detailed
modelling and identifying critical parameters for the potential of
progressive collapse. Recently, Iribarren et al. [13] used a more
sophisticated approach consisted of a detailed modelling of reinforced concrete cross-sections to analyse a ve-storey RC planar
frame model.

206

S. Kokot et al. / Engineering Structures 40 (2012) 205217

the removal of two central columns, but also how challenging


the structural testing against progressive collapse is.
However, buildings can be exposed to fast dynamic abnormal
events, such as bomb explosions or impacts, so the dynamic nature
of the loading must be considered. Therefore, the rst objective of
this work is to re-evaluate the previously mentioned frame building using linear and nonlinear dynamic analyses according to the
alternate load path method. In other words, this study tries to answer the question: what would have happened if the columns had
been destroyed dynamically in an almost instantaneous manner?
The second objective is to learn how much the effects of the fast
dynamic column removal are greater compared to the effects of
the static approach. To this end, three denitions of dynamic factors are discussed. Finally, the third objective is to identify the
advantages when redistribution of bending moments due to plastic
hinges is taken into account.
2. Description of the structure
The structure was a 3-storey 2-bay reinforced concrete frame
building with a 0.24 m thick slab (Fig. 1). The height of each storey
was 2.7 m and the bays were 6 m and 4 m long. The structure contained two main frames connected together with transverse beams
(Fig. 2). The girder beams were 1 m wide and 0.24 m high. The
frames were supported by square columns of dimensions
0.4  0.4 m. In each frame, there existed an eccentricity of 0.2 m
between the axes of beams and columns. Because of the reduced
beam height, they had quite high reinforcement on both sides,
with only some rebars anchored to the column joints. The details
of reinforcement are shown in Figs. 3 and 4.

Fig. 1. Front view.

In the European Laboratory for Structural Assessment (ELSA), a


reinforced concrete at-slab frame building was tested to evaluate
its safety against collapse (see [14,15]). First, static linear and nonlinear analyses of the building under column removals were performed and then two columns of the building were actually
demolished, one after the other, to observe the building behaviour.
This experiment has shown not only that the structure survived

2.1. Materials
The materials of the structure were C25/30 concrete and S500
steel. In addition, laboratory tests were performed on cubes of concrete and on three specimens of each rebar diameter. The mean value of concrete strength f cm is 32.8 MPa. For /20 steel rebars the

Fig. 2. Floor plan.

S. Kokot et al. / Engineering Structures 40 (2012) 205217

207

where As is the area of reinforcement bars in the beam cross-section, fs is the characteristic value of strength of steel (524.6 MPa
for rebars /20), and d is the distance from the centre of reinforcement to the extreme compressed concrete bres of the beam
cross-section. The calculated resistance for the beams are presented
in Table 1 (a longer bay, b shorter bay). Note that for those
beams which will undergo bending reversal after the column removal two values of resistance are listed in Table 1 (positive and
negative moment). The assumption of neglecting the axial forces
is justied because usually the axial forces in beams are relatively
small and they increase/decrease the bending moment resistance
only marginally.
For columns, the pure axial resistance is calculated as

Nr Ac fc As fs

Fig. 3. Elevation and column rebars.

mean yield, ultimate strength and ultimate strain are as follows:


f y 524:6 MPa, f t 642:6 MPa and eu 11:07%.
2.2. Resistance of the frame elements
Assuming that in beams failure is due to bending (neglecting
axial and shear forces), the approximate beam moment resistance
is calculated as

M r 0:85As fs d

where Ac is the area of the concrete cross-section and fc is the


strength of concrete in compression (32.8 MPa). Their approximated pure bending resistance is calculated via Eq. (1), Table 2.
However, for columns, the inuence of axial force on bending moment resistance cannot be neglected. Therefore the actual bending
moment resistance is obtained from the interaction diagrams calculated with the commercial software SAP 2000 for four types of column cross-sections.
The internal forces are obtained from a FE calculation using the
SAP 2000, and in the most loaded cross-sections they are compared
to the corresponding resistance values.
2.3. Summary of the quasi-static experiment for progressive collapse
The structure was rst tested pseudodynamically against a design earthquake. The results reported in [16] showed that the
structure suffered minor damage. Then the structure was devoted

Fig. 4. Beam rebars.

208

S. Kokot et al. / Engineering Structures 40 (2012) 205217


Table 1
Resistance of beams.
Beam

Mr (kNm)

Mr (kNm)

Floors 12
a left
a middle
a right
b left
b middle
b right

197.087
92.349
225.242
225.242
92.349
112.621

176.815
176.815

Floor 3
a left
a middle
a right
b left
b middle
b right

168.932
92.349
197.087
197.087
92.349
112.621

148.660
148.660

Fig. 5. State of the building at the end of Phase 1.


Table 2
Axial and simplied bending resistance of columns.
Column

Nr (kN)

Mr (kNm)

Floor 1
1
2
3

5836.481
6170.202
5836.481

96.821
151.282
96.821

Floor 2
1
2
3

5836.481
5836.481
5697.431

96.821
96.821
74.128

Floor 3
1
2
3

5994.072
5836.481
5697.431

122.539
96.821
74.128

to controlled demolition with the goal of investigating its safety


against collapse.
The experiment consisted in cutting the columns, one after another. In the rst phase, one central column was cut out and, as can
be seen in Fig. 5, the building withstood the lack of this load-bearing member. In the second phase, the other central column was removed and, again, the structure survived (see Fig. 6). Then, there
was concern that the building would collapse in an uncontrolled
manner (after a complete removal of another column), therefore,
for safety reasons, it was decided to progressively destroy two
external columns to provoke a pancake-type collapse (see Fig. 7).
Since the experiment took into account only the static behaviour of the structure, a question arises whether the structure would
have survived if a column/columns had been destroyed dynamically in an instantaneous manner. In the following sections, the results of numerical linear and nonlinear, static and dynamic
analyses are presented to give a preliminary answer to this
question.

Fig. 6. State of the building at the end of Phase 2 (Note that the second central
column is cut at the base).

3. Finite element model


A nite element model of the analysed structure has been created in SAP 2000. It contains 186 frame elements and 171 nodes.
The element and node numbers shown in Fig. 8 are used in the sequel to display the numerical results. The rst longer bay in the xdirection is referred to as a-bay, while the second one as b-bay.
In this work three principal scenarios have been considered:
sudden removal of the central column, of the left corner column
and nally of the right corner column, one at a time.
Although both guidelines [1,2] recommend applying the combination of permanent (dead) and variable (live) loads in structural

Fig. 7. The building during the pancake-type collapse.

analyses, in this paper only the self-weight is considered to be able


to compare the numerical results with the experiment. The permanent load was equal to 3.5 kN/m2 (actual concrete structure
weight) plus 2.0 kN/m2 (representing several permanent xtures
on the structure) and was modelled as a uniformly distributed

S. Kokot et al. / Engineering Structures 40 (2012) 205217

209

Fig. 8. Finite element model of the analysed frame in SAP 2000 element and node numbers.

Fig. 9. Loads on the frame simulation of the column removal (from SAP 2000).

linear load applied to the girders to account for the one-way


behaviour of the concrete slabs.
In the current dynamic analyses the simulation of the column
removal is performed as follows: The column to be removed is rst
replaced by the corresponding reaction forces at the appropriate

node. In the second step these reaction forces are simultaneously


and abruptly brought to zero. In practice this is accomplished in
SAP 2000 by applying at these points a similar set of forces/moments of increasing magnitude in the opposite direction (see
Fig. 9). The rate of the column removal is specied by a time

210

S. Kokot et al. / Engineering Structures 40 (2012) 205217

Table 3
Bending moments in beams, no column removal, comparison with resistance, frames
1 and 2.
Frames 1 and 2

Ms (kNm)

Beam

a left

a mid

a right

b left

b mid

b right

Floor 3
Floor 2
Floor 1

43.34
49.59
47.22

30.04
27.18
28.26

52.52
51.98
52.18

29.30
23.07
26.90

11.08
11.46
11.28

17.84
23.31
19.84

32.53
29.43
30.60

26.65
23.08
23.17

14.87
10.24
11.94

12.00
12.41
12.21

15.84
20.70
17.62

The results obtained from these static computations are compared with the structural resistances using the so called demandresistance ratios (DRRs), also referred to as demand-capacity ratios.
A local DRR is dened in each section as
8
in beamsbending moment only
>
< M max =M r
DRR N max =N r
in barsaxial force only
>
:
M max =M r N in columnscombined bending moment and axial force

Ms/Mr (%)
Floor 3
Floor 2
Floor 1

25.66
25.16
23.96

Table 4
Axial forces and bending moments in columns, no column removal, comparison with
resistance, frames 1 and 2.
Frame 1

Ns (kN)

Column

Floor
Floor
Floor
Floor
Floor
Floor

3
3
2
2
1
1

top
bot
top
bot
top
bot

50.45
50.45
102.02
102.02
153.17
153.17

91.02
91.02
177.99
177.99
267.20
267.20

31.78
31.78
66.49
66.49
99.38
99.38

Floor
Floor
Floor
Floor
Floor
Floor

3
3
2
2
1
1

top
bot
top
bot
top
bot

23.21
16.89
12.02
14.82
10.46
5.98

17.84
13.45
9.86
11.75
8.09
4.87

Floor
Floor
Floor
Floor
Floor
Floor

3
3
2
2
1
1

top
bot
top
bot
top
bot

126.38
126.38
140.57
140.57
212.03
212.03

91.24
91.24
97.38
122.09
127.84
127.84

Floor
Floor
Floor
Floor
Floor
Floor

3
3
2
2
1
1

top
bot
top
bot
top
bot

18.37
13.36
8.55
10.54
4.93
2.82

19.55
14.74
10.13
9.62
6.33
3.81

Ms (kNm)
43.34
28.06
21.53
26.91
20.31
9.09
Mr(Ns) (kNm)
147.78
147.78
128.30
128.30
136.71
136.71
Ms/Mr(Ns) (%)
29.33
18.99
16.78
20.97
14.86
6.65

where Mmax and Nmax are the maximum moment and axial force
acting on the section while Mr and Nr are the bending moment
and axial resistances of the section, respectively. The global DRR
is taken as the maximum local DRR over the entire structure i.e.
DRRmax. For reinforced concrete structures, both [1,2] specify that
the value of 200% for the demand-resistance ratio should not be exceeded, otherwise the structure is deemed as prone to progressive
collapse.

4.1. Before column demolition


This phase concerns the frames in the intact state, i.e. all elements are present, as compared to the subsequent phases where
one or more columns are destroyed.
The results, being exactly the same for both frames, are displayed only once. The values of internal forces (bending moments
and axial forces) in the most representative/critical cross-sections
are given in Table 3 for beams and in Table 4 for columns. The loading corresponds to the aforementioned self-weight of (3.5 + 2) kN/
m2. In the Tables, the resultant internal forces are given at the different cross-sections (l left, mid - midspan, r right) of each bay
(a longer bay, b shorter bay) together with the ratios between
the resultant internal forces and the element resistances (demand/
resistance ratio DRR). Note that the Mr values in these Tables are
obtained from the corresponding interaction diagrams. As an
example, Fig. 10 shows how the value of Mr is obtained for the
rst-oor central columns (with rebars /20) under the axial force
Ns = 267.20 kN. The maximum values of demand/resistance ratios
are: at the midspan of the a-beams on the third oor
(DRR = 32.53%) and at the top of the left column on the third oor
(DRR = 29.33%) and clearly these values are relatively small.

4.2. One central column removed


function, which is chosen as a linear ramp. For actual bomb explosions, the time in which a structural member is destroyed is very
short (some milliseconds). In the presented FE calculations, the removal time is selected to be 5 ms, which implies a quasi instantaneous removal. The dynamic effects of the removal rate on the
dynamic response of the structure have already been investigated
in [7] and the results have showed that the most unfavourable dynamic effects occur when the column is destroyed within a time
close to zero (below 5 ms). The dynamic computations are performed starting from the equilibrium position of the intact structure under gravity loads (zero initial velocities) and assuming a
5% viscous damping.
4. Linear static analysis
These analyses have already been performed and the results reported in [14]. However, to make this paper self-contained, they
have been reproduced to facilitate comparison with the dynamic
analyses.

Many current progressive collapse provisions in codes, standards and guidelines (e.g [1,2,17]) require that the load-bearing
elements are removed anywhere in the structure, one at a time,
and check if progressive collapse could occur. As the rst damage
scenario a central column in the rst frame is removed. The bending moment and axial force distributions for both frames in the
most critical cross-sections are given in Tables 57.
The linear static analysis shows that the most loaded cross-sections are in the rst frame, namely the right-end of the b-beam on
the second oor (DRR = 123.72%) and the top of the right column
on the third oor (DRR = 107.77%). The vertical displacement at
node 48 is equal to 0.0167 m.
These results indicate only minor yielding, so the structure
would not be susceptible to collapse, statically. However, according to the guidelines in [1,2], a structure would be susceptible to
progressive collapse, dynamically. This is because its demandresistance ratio exceeds 200% when the permanent loads are multiplied by a factor of 2 (to account for dynamic effects) in the computation of internal forces.

211

S. Kokot et al. / Engineering Structures 40 (2012) 205217

Finding Mr(Ns) from the interaction diagram (rebars 20)

Table 6
Axial forces and bending moments in columns, central column removed statically,
comparison with resistance, Frame 1.

7000
6000

axial force [kN]

5000
4000

Frame 1

Ns (kN)

Column

Floor 3
Floor 2
Floor 1

81.30
166.99
250.39

3
10.37
1.68

81.58
175.84
267.37

30.24
18.54
27.84
39.55

107.81
77.40
60.40
60.30
74.43
25.19

112.29
110.78

100.04
116.69
154.49

26.93
16.51
25.13
35.70

107.77
77.37
51.76
43.00
48.18
16.31

Ms (kNm)

3000
Floor
Floor
Floor
Floor
Floor
Floor

2000
1000
Ns = 267.20kN

0
1000
0

3
3
2
2
1
1

top
bot
top
bot
top
bot

Mr(Ns) (kNm)

Mr=212.03kNm

50

100

129.79
85.39
65.89
96.46
41.52
31.79

150
200
250
300
bending moment [kNm]

350

400

Floor 3
Floor 2
Floor 1

152.61
138.86
151.84
Ms/Mr(Ns) (%)

Fig. 10. Interaction diagram for a column with rebars /20.

Table 5
Bending moments in beams, central column removed statically, comparison with
resistance, frames 1 and 2.

Floor
Floor
Floor
Floor
Floor
Floor

3
3
2
2
1
1

top
bot
top
bot
top
bot

85.05
55.95
47.45
69.47
27.34
20.94

Frame 1

Ms (kNm)

Beam

a left

a mid

a right

b left

b mid

b right

Floor 3
Floor 2
Floor 1

130.01
150.60
138.13

35.93
28.51
34.10

45.94
51.69
50.41

77.46
99.12
91.30

18.26
14.54
12.20

110.25
139.33
136.21

0.0142 m. In this case, the linear static calculation indicates that


the structure would not be prone to progressive collapse neither
statically (DRR < 200%) nor dynamically (DRR < 200%)

38.91
30.87
36.93

30.90
29.23
28.51

52.11
56.06
51.64

19.77
15.74
13.21

97.89
123.72
120.95

5. Linear dynamic analysis

Ms/Mr (%)
Floor 3
Floor 2
Floor 1
Frame 2
Floor 3
Floor 2
Floor 1

76.96
76.41
70.09
Ms (kNm)
46.52
54.43
51.71

29.92
27.04
28.14

49.56
47.42
47.94

33.69
29.47
33.07

11.38
11.59
11.46

12.85
16.65
13.32

32.40
29.28
30.47

25.15
21.05
21.28

17.09
13.08
14.68

12.32
12.55
12.41

11.41
14.78
11.83

Ms/Mr (%)
Floor 3
Floor 2
Floor 1

27.54
27.62
26.24

4.3. One left corner column removed


In the second damage scenario a left corner column is removed
from the rst frame.
The maximum demand-resistance ratios are reached on the
third oor at the right-end of the a-beam (DRR = 132.64%) and at
the top of the right column (DRR = 92.31%). The vertical displacement at node 25 equals 0.0552 m. Therefore, according to the rules
of thumb mentioned earlier, a progressive collapse is unlikely under static conditions (DRR < 200%) but is possible under dynamic
conditions (DRR > 200%).

4.4. One right corner column removed


The last case deals with the removal of a right corner column
from the rst frame. This case is similar to the previous one and
is more favourable because the span of the right bay is shorter.
The demand-resistance ratios are far below 100% in all members.
The maximum DRR values are 66.23% for beams and 39.2% for columns and the vertical displacement at node 69 is equal to

This section presents the results of the three scenarios of column removal using linear dynamic analysis. The advantage of this
kind of calculations is that dynamic effects are inherently incorporated in the analysis as opposed to an a priori assumed dynamic
factor to be applied on the results of the static analysis. Since it
provides a more realistic distribution of the internal forces over
the structure, the linear dynamic analysis is expected to give a
more reliable estimate of the actual maximum demand-resistance
ratio (DRRmax) characterising the structural robustness against progressive collapse. Furthermore, the actual dynamic factor that
should be applied to the static analysis results, can be computed
a posteriori.
It is however worth mentioning that the notion of dynamic factor is well-dened only for a single-degree-of-freedom system
where all quantities (force, displacement, DRR, etc.) lead to the
same dynamic/static ratio. In a multi-degree-of-freedom system,
different denitions can be adopted which lead to different values
of the dynamic factor, namely:
 The ratio of the dynamic and static maximum deection at the
top of the removed column (Denition 1),
 the maximum ratio of the dynamic and static local DRR (Denition 2),
 the ratio of the dynamic and static DRRmax (Denition 3).
Despite the apparent soundness of the rst two denitions, the
third denition may be preferable because it provides a weighted
global dynamic factor, as it will be conrmed by the results of
the linear and nonlinear dynamic analyses. If this dynamic factor
is applied to the static results, the output of the dynamic analysis
is recovered in terms of robustness (value of DRRmax).

212

S. Kokot et al. / Engineering Structures 40 (2012) 205217

Table 7
Axial forces and bending moments in columns, central column removed statically,
comparison with resistance, frame 2.
Frame 2

Ns (kN)

Column

Floor 3
Floor 2
Floor 1

51.47
104.61
157.22

92.34
183.00
273.94

29.44
60.88
90.60

17.15
12.79
6.20
8.66
7.56
3.42

15.29
12.01
6.18
7.97
6.82
3.41

126.61
141.35
155.51

90.83
96.38
126.30

13.55
10.10
4.39
6.13
4.86
2.20

16.83
13.22
6.41
6.58
5.40
2.70

Table 9
Maximum axial forces and bending moments in columns, central column removed
dynamically, comparison with resistance values, Frame 1, linear analysis.
Frame 1

N max
(kN)
d

Column

Floor 3
Floor 2
Floor 1

105.59
221.10
328.27

91.02
177.99

118.76
262.91
406.82

67.52
41.98
71.50
91.72

170.27
118.30
109.33
90.09
150.73
102.47

12.24
16.54
11.37
22.18

118.33
117.71
262.00
261.21
399.56
403.76

113.37
114.36

106.42
130.93
175.72

59.95
37.03
63.57
80.20

159.83
111.16
83.50
58.68
86.10
58.31

M max
(kNm)
d

Ms (kNm)
Floor
Floor
Floor
Floor
Floor
Floor

3
3
2
2
1
1

top
bot
top
bot
top
bot

46.74
29.31
24.44
30.31
21.54
15.78

Floor
Floor
Floor
Floor
Floor
Floor

3
3
2
2
1
1

top
bot
top
bot
top
bot

Mr(Ns) (kNm)
Floor 3
Floor 2
Floor 1

147.94
128.75
137.34

Nd (kN) for M max


d

Ms/Mr(Ns) (%)
Floor
Floor
Floor
Floor
Floor
Floor

3
3
2
2
1
1

top
bot
top
bot
top
bot

31.59
19.81
18.98
23.54
15.68
11.49

189.97
134.29
87.40
157.38
66.94
92.68

Floor
Floor
Floor
Floor
Floor
Floor

3
3
2
2
1
1

top
bot
top
bot
top
bot

105.20
105.44
220.88
220.74
229.94
221.17
Mr(Nd) (kNm)

Floor 3
Floor 2
Floor 1

156.38
147.23
147.29
M max
/Mr(Nd) (%)
d

Table 8
Maximum bending moments in beams, central column removed dynamically,
comparison with resistance, frames 1 and 2, linear analysis.
Frame 1

Md (kNm)

Beam

a left

a mid

a right

b left

b mid

b right

Floor 3
Floor 2
Floor 1

190.21
220.48
186.13

52.12
40.21
49.57

102.04
98.50
96.26

136.70
180.72
176.65

25.93
21.16
15.92

174.17
228.94
239.37

56.44
43.54
53.68

68.64
55.71
54.44

91.95
102.21
99.91

28.08
22.91
17.24

154.65
203.28
212.54

Md/Mr (%)
Floor 3
Floor 2
Floor 1

112.60
111.87
94.44

Md/Ms local dynamic factor


Floor 3
Floor 2
Floor 1

1.46
1.46
1.35

Frame 2

Md (kNm)

Floor 3
Floor 2
Floor 1

56.25
68.53
63.92

1.45
1.41
1.45

2.22
1.91
1.91

1.76
1.82
1.93

1.42
1.46
1.30

1.58
1.64
1.76

30.38
27.41
28.70

54.08
55.26
58.63

46.62
49.40
49.62

12.01
11.99
12.01

18.79
27.71
29.51

Md/Mr (%)
Floor 3
Floor 2
Floor 1

33.30
34.77
32.43

32.90
29.68
31.08

27.44
24.53
26.03

23.65
21.93
22.03

13.01
12.98
13.01

16.68
24.60
26.20

1.38
1.68
1.50

1.06
1.03
1.05

1.46
1.66
2.22

Md/Ms local dynamic factor


Floor 3
Floor 2
Floor 1

1.21
1.26
1.24

1.02
1.01
1.02

1.09
1.17
1.22

The procedure used in the calculations has been presented in


Section 3. In summary, the following steps were carried out in
SAP 2000:
 Build a FE model,
 nd the internal forces at the top of a column to be removed
under the self-weight loading,

Floor
Floor
Floor
Floor
Floor
Floor

3
3
2
2
1
1

top
bot
top
bot
top
bot

121.51
85.87
59.35
106.89
45.03
62.92

M max
=Mr N d
d
(local dyn. factor)
M s =M r N s
Floor
Floor
Floor
Floor
Floor
Floor

3
3
2
2
1
1

top
bot
top
bot
top
bot

1.43
1.53
1.25
1.54
1.65
3.00

2.23
2.24
2.53
2.25

1.48
1.44
1.61
1.36
1.79
3.58

 remove this column from the FE model and apply these internal
forces as reactions in its place,
 apply these reaction forces again but in the opposite direction
using a linear ramp function,
 perform linear time history analysis with initial conditions (static deformation and zero velocities) and 5% critical damping.
The results obtained from these dynamic computations (time
histories of internal forces) are compared with the resistances
using Eq. (3) and with the corresponding static responses obtained
in the previous section.
5.1. One central column removed
The response of the structure to the sudden removal of the central column in the rst frame is determined. Based on the envelopes of the internal forces (bending moments and axial forces)
in both frames, the corresponding maximum values for beams
and columns are presented in Tables 8 and 9. For columns of Frame
2 (results not reported in full due to space limitation), the maximum value of DRR is 38.08% at the top of the left third column.
As could be expected, Frame 2 is signicantly less affected than
Frame 1 where the column was removed, a fact that can be explained by the one-way behaviour of the at-slab frame. The most
critical sections in terms of demand-resistance ratio are the right-

213

S. Kokot et al. / Engineering Structures 40 (2012) 205217

Deflections at nodes 48, 25 and 69

0
0.01
0.02

deflection [m]

0.03
0.04
0.05
0.06
central column removed node 48
left column removed node 25
right column removed node 69

0.07
0.08
0.09
0.1

0.2

0.4

0.6
t [s]

0.8

1.2

Fig. 11. Vertical deections for three column removal scenarios at nodes of maximum deections for linear dynamic analysis.

end of the b-beam on the rst oor (DRR = 212.54%) and the top of
the right column on the third oor (DRR = 159.83%). Since the demand-resistance ratio for beams exceeded the 200% threshold, the
building is susceptible to progressive collapse.
As for local dynamic factors (Denition 2) in beams, the maximum values are reached at the right-end of the a-beam at the third
oor in the rst frame but also at the right-end of the b-beam at
the rst oor in the second frame (2.22), while in columns, the
maximum dynamic factors are much larger and reach values of
3.58 and 8.71 in the rst and second frame, respectively. This fact
demonstrates that it is difcult to draw any conclusion from the local dynamic factors because they are highly heterogeneous
throughout the structure, especially in columns where the static
and dynamic forces are quite different. On the other hand, much
more representative is the global dynamic factor according to Definition 3 and here for beams, it is equal to 1.72, while for columns
1.48.
The time history of the maximum displacement of the structure
at node 48 is plotted in Fig. 11. It is seen that the dynamic curve
practically oscillates about the corresponding static deection value, and the maximum is obtained for t 0.09 s.
From the maximum displacement of the dynamic (0.0268 m)
and static (0.0167 m) responses at node 48, a ratio of 1.60 is found,
which can be interpreted as another global dynamic factor (Denition 1).
5.2. One left corner column removed
For the left corner column removal scenario, the maximum demand-resistance ratios are 199.54% for beams (right-end of the abeam of the third oor in Frame 1) and 172.79% for columns (top of
the right column of the third oor in Frame 1). In this scenario, we
can notice even larger local dynamic factors (up to 29.43). For this
scenario, the maximum demand-resistance ratio (almost 200%) is
on the verge of treating the building as acceptable/unacceptable
against progressive collapse.
The global dynamic factor (Denition 3), for beams, equals 1.50,
while for columns 1.87.
Fig. 11 shows how the vertical displacement at node 25 varies
in time. The maximum value is 0.091 m at time t 0.2 s. The ratio
of the maximum linear dynamic deection and the deection for
the linear static analysis is 0.091 m/0.0552 m = 1.65.

Fig. 12. Denition of a plastic hinge for beam elements.

5.3. One right corner column removed


For the right corner column removal scenario, the maximum
demand-resistance ratios are 84.54% for beams (right-end of the
a-beam on the third oor of Frame 1) and 69.41% for columns
(top of the right column on the second oor of Frame 1). In this
case, the maximum local dynamic factor is the largest observed
so far (59.34), which can be explained by the small static bending
moment of 0.32 kNm. On the other hand, the global dynamic factor, for beams, equals 1.28, while for columns 1.77.
Fig. 11 shows the vertical displacement at node 69 in time. The
maximum value is 0.0207 m at time t 0.31 s. The ratio of the
maximum linear dynamic deection and the deection of the static
analysis is 0.0207 m/0.0142 m = 1.46.
These results demonstrate that this is the most favourable nonfailure scenario and that the structure bridges over the lacking column very efciently. In fact, as will be seen below, the structure
remains always in the elastic range.
6. Nonlinear dynamic analysis
The nonlinear dynamic analysis is the most advanced method
for predicting the response of a structure when a load-bearing

214

S. Kokot et al. / Engineering Structures 40 (2012) 205217

(
DRR

nlin

if no yielding occurred;
100  Mmax =Mr


max plastic rotation
100 1 ultimate
if
yielding occurred:
plastic rotation
5

This nonlinear DRR coincides with the linear DRR in the absence of
yielding (DRR < 100%). In the presence of yielding ( DRR > 100%), the
nonlinear DRR measures the distance to the ultimate plastic rotation (point C of the moment-curvature relationship, Fig. 12). As
for the linear DRR, the value of 200% is marking the threshold not
to be exceeded (failure of the section) although this does not necessarily implies the collapse of the structure.
In the current analysis three plastic hinges are introduced in
each beam (left, mid and right) and two in each column (bottom
and top), thus resulting in 36 plastic hinges for each frame.
6.1. One central column removed

Fig. 13. Final locations of plastic hinges for the central column removal.

element is removed quasi-instantaneously. The fundamental difference with the linear dynamic analysis is that inelastic behaviour
and/or geometric nonlinearities are taken into account. Therefore
in the SAP 2000 modelling it requires the denition and assignment of plastic hinges at specied positions of selected members.
The nonlinear time history analysis with the same initial conditions is performed next.
The stressstrain (r  e) relationship for concrete is assumed
parabolic in the rst part and constant in the second part according
to the following equation

rc f c



e 2

0:002


;
0:002

0:002 < e < 0;

where fc is the compressive strength, e is the strain in concrete,


0.002 is the strain value at which the parabola ends. For strains
between 0.002 and 0.0035 (which is considered as the failure
strain) the stress remains constant.
For steel, the stressstrain (r  e) relationship is assumed
bilinear. The rst part is linear elastic with a yield stress of
524.56 MPa and a modulus of elasticity of 206 GPa, while the
second part is plastic with linear hardening and an ultimate
stress of 642.56 MPa.
In SAP 2000, the plastic hinge behaviour is dened by a piecewise linear moment-plastic rotation relationship, the characteristics of which are identied from the idealised moment-curvature
relationship of the section. An example is given in Fig. 12 for a
beam section: point B is dened by the yielding moment and point
C by the ultimate moment and the corresponding plastic rotation.
The curve is usually extended by a softening and residual branch
which has however no importance in the present study since the
plastic hinges never reach their ultimate capacity. For columns,
this moment-plastic rotation relationship depends also on the normal force and this interaction may be activated in SAP 2000. The
values of points B, C, D and E are taken from Ref. [18].
To be able to directly compare the nonlinear results to the linear
ones, the following nonlinear demand-resistance ratio (DRRnlin) is
dened as

The nonlinear dynamic analysis for one central column removed shows that, at time 0.039 s, two plastic hinges are activated
almost simultaneously in the rst frame, one at the top of the right
column on the third oor and the other at the right-end of the bbeam on the second oor. Shortly after, at 0.040 s, another plastic
hinge is activated at the right-end of the b-beam on the rst oor.
The nal spatial conguration of the plastic hinges activated after
the sudden column removal is shown in Fig. 13.
The time history of the maximum displacement at node 48 is
plotted in Fig. 14. Two solid lines show the comparison between
the linear and nonlinear time histories of the displacement at node
48. The divergence of the two curves at time 0.04 s is caused by the
formation of the rst plastic hinges mentioned above. Clearly larger deections (up to 0.0315 m), always overshooting the static
ones, are observed during the nonlinear dynamic analysis owning
by the formation of the plastic hinges. The ratio of maximum
deections at point 48 for nonlinear and linear dynamic analyses
is 1.18 (0.0315 m/0.0268 m), thus the global dynamic factor (maximum dynamic nonlinear displacement divided by static linear displacement) is slightly higher than that of the linear case, that is
1.89 (0.0315 m/0.0167 m) instead of 1.60 (see subsection 5.1). On
the contrary, the bending moments are much lower than in the linear case especially, where the plastic hinges are activated, as is to
be expected.
With regards to the properties of the hinges, Fig. 15 presents the
results for a beam plastic hinge (element No. 96), while Fig. 16
shows the case of a column plastic hinge (element No. 160). There
is the possibility to follow the detailed behaviour of the hinge and
the following information can be produced:
 The skeleton path of the plastic hinge (thin black line) including
the threshold points (yield, ultimate, residual) and the associated levels of damage (pink1 rst yielding, blue immediate
occupancy, cyan life safety and green collapse prevention).
The skeleton path is the moment-plastic rotation relationship
without normal force,
 the actual path followed (thick black line),
 the current time step,
 the values of the plastic moment and rotation at that current
time step.
For a beam plastic hinge, the actual path follows exactly the
skeleton path, while for a column plastic hinge, the actual path
usually deviates from the skeleton path because of the inuence
of the normal force on the moment-plastic rotation relationship.
1
For interpretation of colour in Figs. 1, 57, 911, 1318, the reader is referred to
the web version of this article.

215

S. Kokot et al. / Engineering Structures 40 (2012) 205217

Deflections at nodes 48, 25 and 69


0
NL=L

0.02
L
NL

deflection [m]

0.04

0.06

0.08

central column removed node 48


left column removed node 25
right column removed node 69

0.1
NL

0.12

0.2

0.4

0.6

0.8

1.2

t [s]
Fig. 14. Vertical deections for three column removal scenarios at nodes of maximum deections for nonlinear dynamic analysis.

Fig. 15. Plastic hinge at the right-end of the b-beam on the rst oor.

Fig. 17. Final locations of plastic hinges for the left column removal.

In the present case, the demands in the plastic hinges are all below their ultimate capacity. In fact, according to Eq. (5), the maximum DRRnlin value is 140% in beams and 125% in columns. The
global dynamic factor (Denition 3) for beams is equal to 1.13,
whereas for columns 1.16. The nonlinear dynamic analysis thus
demonstrates that the structure would have survived a sudden removal of the central column.
6.2. One left corner column removed

Fig. 16. Plastic hinge at the top of the right column on the third oor.

In the nonlinear analysis of a left corner column removal, two


hinges are activated almost simultaneously at time 0.094 s, one
at the right-end of the a-beam on the rst oor, the other on the

216

S. Kokot et al. / Engineering Structures 40 (2012) 205217

Deflection at node 48
0
1 column linear dynamic
1 column nonlinear dynamic
2 columns linear dynamic
2 columns nonlinear dynamic

0.005

deflection [m]

0.01

0.015

0.02

0.025

0.03

0.035

0.2

0.4

0.6

0.8

1.2

t [s]
Fig. 18. Comparison of the displacement at node 48 for the cases of one central column and two central columns removal.

bottom of the left column on the second oor. The nal spatial distribution of activated plastic hinges is shown in Fig. 17.
The ratio of the maximum deections at node 25 for nonlinear
and linear dynamic analyses is 1.29 (0.117 m/0.091 m), leading to a
global dynamic factor (Denition 1) of 2.12 (0.117 m/0.0552 m)
with respect to the linear static analysis.
The dashed lines in Fig. 14 compare nonlinear and linear time
histories of displacement at node 25. The maximum DRRnlin values
are 149% for beams and 134% for columns. Similarly, the global dynamic factor according to Denition 3 is 1.12 for beams and 1.45
for columns.
Table 10
Summary of results.

This nonlinear dynamic analysis shows that the structure would


have survived a sudden removal of the left corner column. Again,
the total or partial collapse would not have happened thanks to
an appropriate activation of plastic hinges and redistribution of
bending moments.

6.3. One right corner column removed


When it comes to the case where a right corner column is removed, the linear dynamic analysis has shown that the structure

S. Kokot et al. / Engineering Structures 40 (2012) 205217

remains elastic, so the nonlinear analysis gives exactly the same results as in subsection 5.3.
7. Two central columns removed
In all three scenarios considered, the structure has experienced
limited or no damage. In order to assess the robustness of the
structure, the case of two central columns removal has also been
studied through linear and nonlinear dynamic analyses. It is reminded that the real structure had survived that event.
The results have been summarised in Fig. 18 where the time
history of the displacement at node 48 is plotted for the linear/
nonlinear analysis of one/two column(s) removal. It can be seen
that no matter whether one or two central columns are removed
from the structure, the response does not change drastically. The
period of vibration becomes slightly longer in the second case because the remaining structure is less rigid. This behaviour can be
explained by the one-way action of the at-slab frame. In other
words, each frame appears to be damaged essentially by the removal of its own central column.
8. Conclusions
This work presents the results of an extended study of the atslab frame building which has been analysed and tested quasi-statically at the European Laboratory for Structural Assessment. The
scope of a previous study was limited to the investigation of the
general safety against collapse, and thus it did not consider a possible abrupt removal of columns as it may take place in the event of
a bomb explosion, impact, or other accidental action.
The current investigation includes linear and nonlinear dynamic
time history analyses using alternate load path methods. Three
main scenarios of column removal have been considered: a central
column, a left corner column and a right corner column. In addition, the scenario of two central columns being removed simultaneously has been investigated.
The results of the analyses are summarised in Table 10. This table presents the maximum values of the demand-resistance ratios
(in the most critical cross-sections) and the maximum displacements obtained through linear static, linear dynamic and nonlinear
dynamic analyses. The colours highlight the conclusion drawn
from each analysis in terms of three possible structural states: no
damage, limited damage and extensive damage.
The simplest linear static analysis indicates that the structure
would exhibit limited or no damage if the column is removed statically. However, if the column is removed dynamically, the same
static analysis (with the loading multiplied by 2 to account for
the dynamic nature of the loading) indicates that the structure
would be susceptible to progressive collapse in two scenarios
whereas it would suffer limited damage in the third one.
The linear dynamic analysis indicates a slightly more favourable
situation: the structure would still be susceptible to progressive
collapse for the central column scenario but not necessarily for
the left column scenario as the DRR is slightly below 200%. Furthermore, the structure would remain fully elastic for the right column
scenario. The value 2 of the dynamic factor is therefore conservative. In fact, the actual value of the global dynamic factor found
in the three scenarios ranges from 1.72 to 1.87 (maximum of the
two values reached in beams and columns). Interestingly, the dynamic factor computed from the displacement ranges from 1.46
to 1.6 and thus it underestimates the dynamic effect on the DRR
(nonconservative estimate).
The linear dynamic analysis has revealed that the local dynamic
factor dened in each section as the ratio of the dynamic and static

217

demand-resistance ratios, is insignicant and misleading because


huge dynamic factors may be found in columns for instance, but
they result from the relatively small value of the internal forces
in the static analysis. The global dynamic factor dened from the
displacement of the node above the removed element, does not
present such a drawback but remains quite different from the true
dynamic factor computed as the ratio of the dynamic/static maximum DRR.
The nonlinear dynamic analysis (taking into account the capability of redistribution of internal forces) indicates that the progressive collapse of the building would not have happened, that
is, the propagating failure would have been arrested. For both
the central and left corner column removals several plastic hinges
would have formed in the structure, yet all of them would have
been far below their ultimate capacity (two yellow areas in the
summary Table). For the right corner column removal, no yielding
would have occurred, as already foreseen by the linear dynamic
analysis.
Certainly more sophisticated models could have been adopted
in the analysis. In particular it is noted that the structural member
strengths as calculated here correspond to the static properties of
their constituent materials, concrete and steel. For the current
loading conditions considerations of high strain-rate effects should
be taken into account, and this is the subject of an investigation in
course.
References
[1] GSA Guidelines. GSA progressive collapse analysis and design guidelines for
new federal ofce buildings and major modernizations projects. General
Services Administration (GSA); 2003.
[2] DoD UFC Guidelines. Design of buildings to resist progressive collapse, Unied
Facilities Criteria (UFC) 4-023-03. Department of Defence (DoD); 2005.
[3] NIST Best Practices. Best practices for reducing the potential for progressive
collapse in buildings. US National Institute of Standards and Technology
(NIST), Washington, DC; 2007.
[4] Nair RS. Preventing disproportionate collapse. J Perform Construct Facilit
2006;20(4):30914.
[5] Mohamed OA. Progressive collapse of structures: annotated bibliography and
comparison of codes and standards. J Perform Construct Facilit
2006;20(4):41825.
[6] Starossek U. Progressive collapse of structures. Thomas Telford Ltd; 2009.
[7] Kokot S. Literature survey on current methodologies of assessment of building
robustness and avoidance of progressive collapse. JRC Scientic and Technical
Reports JRC 5598, European Commission, Joint Research Centre; 2009.
[8] Starossek U. Typology of progressive collapse. Eng Struct 2007;29(9):23017.
[9] Marjanishvili S, Agnew E. Comparison of various procedures for progressive
collapse analysis. J Perform Construct Facilit 2006;20(4):36574.
[10] Fu F. 3-d nonlinear dynamic progressive collapse analysis of multi-storey steel
composite frame buildings parametric study. Eng Struct 2010;32:397480.
[11] Tsai MH, Lin BH. Investigation of progressive collapse resistance and inelastic
response for an earthquake-resistant RC building subjected to column failure.
Eng Struct 2008;30:361928.
[12] Kwasniewski L. Nonlinear dynamic simulations of progressive collapse for a
multistory building. Eng Struct 2010;32(5):122335.
[13] Iribarren BS, Berke P, Bouillard P, Vantomme J, Massart T. Investigation of the
inuence of design and material parameters in the progressive collapse
analysis of RC structures. Engineering Structures, in press. doi:10.1016/
j.engstruct.2011.06.005.
[14] Gemelli M, Negro P, Castellani A, Bianchi R, Salandi M. Experimental
evaluation of the safety against the collapse of buildings. Tech. Rep. I.03.102;
European Commission, Joint Research Centre; 2003.
[15] Kokot S, Anthoine A, Negro P, Solomos G. Static and dynamic analysis of a
reinforced concrete at slab frame building for progressive collapse. JRC
Scientic and Technical Reports JRC 62663; European Commission, Joint
Research Centre; 2010.
[16] Negro P, Mola E. Current assessment procedures: application to regular and
irregular structures compared to experimental results. In: Third European
workshop on the seismic behaviour of irregular and complex structures.
Florence; 2002.
[17] EN 1991-1-7 . Eurocode 1 - EN 1991-1-7: Actions on structures - Part 1-7:
General actions - Accidental actions; 2006.
[18] ATC-40. Seismic evaluation and retrot of concrete buildings. ATC-40 Report,
Applied Technology Council, Redwood City, California; 1996.

Potrebbero piacerti anche