Professional Documents
Culture Documents
Abstract: For several decades, the engineering profession has considered techniques to analyze the potential that structures could
Downloaded from ascelibrary.org by The University of Adelaide on 04/28/18. Copyright ASCE. For personal use only; all rights reserved.
experience disproportionate collapse and to design them for greater resistance to such collapse. First interest in such design followed the
partial collapse in 1968 of the Ronan Point building in London, a high rise residential structure that experienced full height collapse of a
portion of the building following a relatively small kitchen-related gas explosion. Interest in collapse phenomena continued to build
following the attack on the Alfred P. Murrah building in 1995 and has been at an apex since the collapses of the twin towers at the World
Trade Center and the nearby World Trade Center 7 building in 2001. Presently researchers and engineers are studying structural perfor-
mance during extreme deformations, systems to resist disproportionate collapse, and methods to analyze collapse potential. The goal is to
develop techniques to accurately and cost efficiently assess collapse potential and to enhance robustness at appropriate cost. Analysis
methods in common use include sophisticated dynamic, nonlinear modeling of structural systems with high-fidelity structural analysis
computer software, and simplified approaches that are intended to capture the essential behaviors during collapse scenarios. Unfortunately,
the sophisticated approaches require software not normally owned by design engineers, substantial experience in the modeling of collapse
phenomena, and time and cost implications that cannot be supported by the present design fees and, indeed, are not warranted for many
situations. Simplified analysis methods in common use are generally empirically based. Hence, they do not capture the essential behaviors
of collapse mechanisms, and are of uncertain applicability for all but structural systems for which they have been calibrated. This paper
presents two energy-based methods that capture the essential physics of collapse phenomena, and have potential to be developed into
simplified procedures for collapse potential assessment.
DOI: 10.1061/共ASCE兲0887-3828共2006兲20:4共336兲
CE Database subject headings: Collapse; Engineering profession; Energy.
Collapse Phenomenon
Downloaded from ascelibrary.org by The University of Adelaide on 04/28/18. Copyright ASCE. For personal use only; all rights reserved.
冉 冊
series of elastic analyses.
Fe1 A numerical model of the structure is constructed with one
KE0−1 = ␦e1 Mg − 共8兲
2 column missing to represent the initiating event. Consider the
general case, where masses are represented as lumped masses at
where Fe1⫽internal force resistance that is induced at deflection all nodes, i. A loading case corresponding to 1 g is applied to
␦e1. simulate the application of an external load to evaluate strain
The energy comparisons can be taken beyond the elastic limit. energy. Under this loading case, each node, i, will deflect verti-
Using the assumption that materials have bilinear strength behav- cally to ␦Ej,i, with j = 1 to signify the first phase. This is the de-
ior, the evaluation can be continued with a subsequent linear formation that would occur if the structure was able to remain
analysis, using a reduced stiffness to represent the effect of elastic under the influence of quasi-static application of gravity
yielding. Hinge locations can be assumed to continue to deform, with a column missing. The strain energy actually absorbed dur-
accumulating additional inelastic strain energy without further ing the first phase while deformation is elastic can be calculated
increase in force. In the second phase of behavior following as follows:
formation of the first plastic hinge in the structure and up to the Compare the forces 共e.g., moments兲 at critical locations in the
formation of the next hinge at deflection ␦e2 关Fig. 2共d兲兴, the structure to corresponding plastic capacities 共e.g., fully plastic
change in potential energy follows the same formulation moment capacities兲 at the same locations, and calculate the ratios
PE1−2 = Mg共␦e2 − ␦e1兲 共9兲
F Pi
The strain energy still is the area under the force-deformation R1,i = 共15兲
curve, but now takes the form FE1,i
冋 1
KE1−2 = 共␦e2 − ␦e1兲 Mg − 共Fe1 + Fe2兲
2
册 共11兲 1 = Min共R1,i兲 共16兲
Combining these calculated energies with the energies associated The deformation at which the first hinge forms 共Fig. 3兲 is
with the first, elastic phase, at the time of formation of the second
hinge in the system the calculated energies are ␦e1,i = 1␦E1,i 共17兲
⌬PE0−2 = Mg␦e2 共12兲
and the effective forces that cause this deformation are
1
⌬SE0−2 = 关Fe1␦e2 + Fe2共␦e2 − ␦e1兲兴 共13兲 Fe1,i = 1mig 共18兲
2
The energies in the system at formation of the first hinge are
1
⌬KE0−2 = Mg␦e2 − 关Fe1␦e2 + Fe2共␦e2 − ␦e1兲兴 共14兲
2 n
冉 冊
n
1
⌬KE0−1 = 兺 mig1␦E1,i 1−
2
共21兲  j−1␦Ej−1,i + ␦0j−1,i − ␦oj,i
i=1 ⬘j−1,i = 共28兲
␦Ej,i
If ⌬KE0−1 is greater than zero, then additional phases of analysis
are required. For each required phase, a hinge is placed at the Of course, when the analytical model is changed to represent the
location or locations where yielding was predicted to occur in the failure of elements of the structure, the conversion from one
prior phases, and equal and opposite forces are applied to the phase to another is not exact unless there is only one degree of
members that connect to the hinges to represent the plastic force freedom that is being tracked. A single value of ⬘j−1 will not scale
capacity at these locations. multiple deflections exactly. Therefore, it is incumbent on the
In general, these plastic hinge forces applied without gravity engineer to select the value that closely represents the behavior of
will cause an upward displacement, which is designated as ␦oj,i, the structure. Normally, this will be the value determined for the
and forces, Foj,i, in the elements of the structure. While these location where the structure is deforming the most.
displacements and forces never occur during a collapse sequence, The selected value of ⬘j−1, is substituted for  j−1, and the
their calculation is useful in the determination of collapse poten- process continues. In phases following a rupture, the energies are
tial, as described below.
n
The next step is to remove the internal force couples related to
formation of prior hinges, and to apply a load case of 1 g to ⌬PE j−1−j = 兺 mig共␦ej,i − ␦ej−1,i兲 共29兲
determine the elastic deformations, ␦Ej,i, and element forces, FEj,i, i=1
冉 冊 冋 册
n
1
 j = min
F Pi − Foj,i
FEj,i
共22兲 ⌬KE j−1−j = 兺
i=1
mig共␦ej,i − ␦ej−1,i兲 1 − 共⬘j−1 +  j兲
2
共31兲
and new effective elastic deformations, ␦ej,i, are calculated as A close examination of the formulas for the components of energy
follows: for the first elastic phase, phases after yield, and phases after
rupture reveals that they all take the same form, and can be rep-
␦ej,i =  j␦Ej,i + ␦oj,i 共23兲 resented by Eqs. 共29兲–共31兲 above, with ␦E0,i, ␦o0,i, ␦o1,i, FE0,i, and
0 all equal zero.
In phases after the formation of hinges, the energies are
Therefore, the determination of collapse potential can be re-
n solved by calculating the sum of the kinetic energy changes in
⌬PE j−1−j = 兺 mig共␦ej,i − ␦ej−1,i兲 共24兲 each phase, and comparing to zero.
冋 册
i=1
k n
1
1
n ⌬KEk = 兺 兺 mig共␦ej,i − ␦ej−1,i兲 1 − 共⬘j−1 +  j兲
2
共32兲
兺
j=1 i=1
⌬SE j−1−j = mig共␦ej,i − ␦ej−1,i兲共 j−1 +  j兲 共25兲
2 i=1 This can be expanded by substituting the relationship for ⬘j−1
k n
⫻ 1− 冋 1
2␦Ej,i
共 j−1␦Ej−1,i + ␦oj−1,i − ␦oj,i +  j␦Ej,i兲 册 Fig. 4. Example framing
Downloaded from ascelibrary.org by The University of Adelaide on 04/28/18. Copyright ASCE. For personal use only; all rights reserved.
共34兲
Recognizing again that
k
1 ␦2ej − ␦ej−1
2
− 2␦oj共␦ej − ␦ej−1兲
␦ej,i =  j␦Ej,i + ␦oj,i 共35兲 ⌽k = ␦ek − 兺
2 j=1 ␦Ej
艋0 共39兲
the expression for kinetic energy can be regrouped and simplified.
for which 0 = ␦o0 = ␦o1 = ␦E0 = Fo1 = 0.
冋 册
k n For analyses that do not consider incremental failure of ele-
1
⌬KEk = 兺 兺 mig共␦ej,i − ␦ej−1,i兲
j=1 i=1
1−
2␦Ej,i
共␦ej,i + ␦ej−1,i − 2␦oj,i兲 ments, acceptance criterion can be derived directly from Eq. 共32兲
k
1
共36兲 ⌽k = ␦ek − 兺
2 j=1
共␦ej,i − ␦ej−1,i兲共 j−1 +  j兲 艋 0 共40兲
n
Column survival can be checked when a phase of analysis indi-
⌬KEk = 兺
i=1
mig␦ej,i cates that collapse is arrested. The column forces determined in
that phase can be used to verify the adequacy of the critical col-
冋 册 umns.
k n
1 m ig
− 兺兺
2 j=1 ␦Ej,i
共␦ej,i − ␦ej−1,i兲共␦ej,i + ␦ej−1,i − 2␦oj,i兲 This approach can be adapted for complicated frames by track-
i=1 ing forces and capacities in the elements in the immediate vicinity
共37兲 of the column that is assumed to be removed as the initiating
event.
n
⌬KEk = 兺
i=1
mig␦ej,i Example Application of Push-Down Procedure
In this simple example, the application of push-down analyses
k n coupled with energy balance concepts is used to assess the capa-
1 mg
− 兺兺
2 j=1
i
i=1 ␦Ej,i
关␦2ej,i − ␦ej−1,i
2
− 2␦oj,i共␦ej,i − ␦ej−1,i兲兴 bility of a structure to arrest collapse. The structure used to illus-
trate these concepts consists of the simple grid frame illustrated in
共38兲 Fig. 4 with nine supports on a 6.1 m 共20-ft兲 grid pattern. W410
⫻ 53 共W16⫻ 36兲 girders frame across the supports in one direc-
Generalized, the determination of resistance to collapse can be tion and W360⫻ 32.9 共W14⫻ 22兲 beams in the second direction.
assessed through a series of linear elastic analyses that yield arti- The girders and beams framing to supports are assumed to be
ficial upward deflections, ␦oj,i, and forces Foj,i, associated with continuous over the central support and fixed against rotation at
applied internal plastic force limits; the effective elastic vertical the edge supports. Beams framing at the midspan of girders are
deflection, ␦Ej,i, and forces, FEj,i, that would have occurred in assumed to have pinned support conditions. Fig. 5 is a stick figure
each phase if the structure remained elastic under gravity load representing the framing only at the support lines, and the repre-
with pins at the locations where elastic strengths have been ex- sentation of the weight supported by this framework idealized as
ceeded; and factors,  j, that are the scale factors indicating the a 178 kN 共40 kip兲 load at the center support. This system is ana-
actual fraction 共or multiple兲 of gravity load that the structure can lyzed for ability to arrest collapse in the event of removal of the
support elastically in the various configurations with pins and central support.
eliminated elements.
While this approach clearly is complicated for multiple degree
of freedom 共MDOF兲 systems, most analyses procedures assume
that a single vertical load-carrying element is notionally removed.
For this assumption, the structural system usually can be repre-
sented reasonably with a single degree of freedom defining verti-
cal motion 共i.e., by assuming that all floors deflect by the same
amount at the location of the removed column兲, hence with only
one deflection tracked. In these cases, the acceptance criterion can
be simplified and the mass terms drop out, to create the following
expression: Fig. 5. Line diagram of key framing and supported weight
冉 冊
⌬SE1−2 = ⫻ 178 ⫻ 共0.0714 − 0.0538兲 ⫻ 共1.64 + 1.82兲
1.64 2
⌬KE0−1 = 178 ⫻ 1.64 ⫻ 0.0328 ⫻ 1 −
2 = 5.42 kN-m 共47.9 kip-in.兲 共47兲
= 1.72 kN-m 共15.2 kip-in.兲 共44兲
The calculated kinetic energy is positive at the end of the first
phase, indicating that the downward fall of the structure is not
1
冋
⌬KE1−2 = 178 ⫻ 共0.0714 − 0.0538兲 1 − 共1.64 + 1.82兲
2
册
arrested at the formation of the first set of hinges. It is necessary = − 2.29 kN-m 共− 20.2 kip-in.兲 共48兲
to carry the analysis another step.
Combining the energy calculations for the two phases shows the
following:
F01,i
that are not near the roof, building corners, or grade levels have
inflection points at midheight and beams in adjacent spans are
fixed at their far ends, leading to
Ib
Rb = 共55兲 Fig. 14. Strain energy absorbed by flexural behavior and change in
Ls
Ib + 4Iab + 12Ic potential energy
Hc
where Iab⫽moment of inertia of the beams in adjacent spans; To develop the appropriate force–displacement relationship
Ic⫽moment of inertia of columns; and Hc⫽column height. over the relevant range of response, it is necessary to determine
Similar relationships for Rb can be developed for other column whether the first hinge forms at the positive or negative moment
locations in the building frame, and for locations where two-way locations, so that the appropriate relationship, either Pfyn共␦兲 or
action is important. Pfyp共␦兲, is used in the analysis. This can be accomplished by
In the inelastic range, if the first hinge forms in the negative calculating the following values and finding which is lower:
moment location
冋冉 冊 册
4ZnFy
3 2EsIb Pen = 共58兲
Pfyn共␦兲 = ␦ + Z nF y 共56兲 Ls共1 − Rb兲
Ls Ls2
where Pfyn共␦兲⫽force necessary to deflect the beam in flexure with 4Z pFy
Pep = 共59兲
yielding at negative moment locations; Zn⫽plastic section modu- Ls共1 + Rb兲
lus in negative bending; and Fy⫽yield stress.
In the inelastic range, if the first hinge forms in the positive where Pen⫽force that would induce first yielding at negative mo-
moment location ment locations; and Pep⫽force that would induce first yielding at
冋冉 冊 册
the positive moment location.
3 2EsIb Once plastic hinges have formed in both locations, the force
Pfyp共␦兲 = 共1 − Rb兲␦ + Z pFy共1 + Rb兲
Ls共1 + 2Rb兲 Ls2 that is sustained while the beam experiences additional displace-
ment is
共57兲
2共Zn + Z p兲Fy
where Pfyp共␦兲⫽force necessary to deflect the beam in flexure with Pfy = 共60兲
yielding at positive moment locations; and Z p⫽plastic section Ls
modulus in positive bending. The next step is to set a deflection limit, ␦ f lim 共perhaps deter-
mined by limiting support rotations兲, that determines when the
flexural system is assumed to lose ability to accumulate additional
energy. For deformations beyond ␦ f lim, the beam has failed and
ceases to support load.
With this formulation, the minimum value of Pfe共␦兲, Pfyn共␦兲 or
Pfyp共␦兲, and Pfy共␦兲 over the range of deflections between 0 and
␦ f lim defines the force–displacement relationship,P f 共␦兲, for the
beam.
To illustrate the approach, representative plots are shown in
Figs. 13–18, based on the following assumed member properties
and structural geometry:
Fig. 13. Force–displacement relationship for flexural behavior Fig. 15. Model for catenary analysis
Fy = 345 MPa 共50,000 psi兲 with the beam axial strain a function of deformation at the loca-
tion of the missing column, axial flexibility, and horizontal stiff-
Ib = Iab = 3.9 ⫻ 10−4 m4 共941 in.4兲 nesses derived from the adjacent structure.
In this formulation, the relationship between the force neces-
Ic = 5.3 ⫻ 10−4 m4 共1,270 in.4兲 sary to deform the beam while it is elastic and deflection is fourth
order, as follows:
冋 册 冉 冊 冋 册 冉 冊冉 冊
Zn = Z p = 2.16 ⫻ 10−3 m3 共132 in.3兲
Pce共␦兲 4
1 2
Pce共␦兲 3
1 1 1
− +
Ab = 0.0135m 共20.9 in. 兲
2 2 ␦ 4klkh ␦ 4klkh kl kh
sumed to be limited by support rotations of 6°兲 is established, the where Pce共␦兲⫽force necessary to deflect the catenary while the
energy accumulated through flexural action can be calculated by beam is elastic; kl⫽axial stiffness of the beam, AbEs / Ls; Ab⫽area
integration of the beam; and kh⫽horizontal stiffness of the surrounding struc-
冕
y
EF共y兲 = P f 共␦兲d␦ 共61兲
0
冫冤冢 冦冋 冧冣 冥
2 1/2
A bF y 1
冉 冊册
Pcy共␦兲 = 2AbFy␦ Ls − + ␦2
kh ␦ 2 1/2
1+
Ls
共64兲
where Pcy共␦兲⫽force necessary to deflect the catenary after the
beam yields in tension.
Downloaded from ascelibrary.org by The University of Adelaide on 04/28/18. Copyright ASCE. For personal use only; all rights reserved.
冕
y
To accomplish this evaluation, we constructed equivalent dia-
EC共y兲 = Pc共␦兲d␦ 共65兲 phragms on three floors. Then we removed the diaphragm ele-
0
ments from two adjacent bays on the floor level of interest at the
where EC共y兲⫽strain energy absorbed by catenary action through perimeter of the building and four bays in a cluster at an interior
deflection y 共Fig. 17兲. location. We applied opposing horizontal forces at the beam/
Fig. 17 suggests that the analyzed condition will not stabilize column connections and calculated horizontal deflections for
through energy accumulated through catenary action alone. equivalent 0.102 and 0.051 m 共4 and 2 in.兲 slabs 共to account for
In the simplified approach described herein, flexural action and cracking兲, with and without corresponding openings in the bays
catenary action are uncoupled 共however, formulations can be de- above and below the floor level of interest. These analyses
veloped to represent coupled phenomena兲. Hence, the combined showed that the effective diaphragm stiffness is influenced by less
strain energy absorbed through deflection of a beam that resists than 10% by the presence of openings on adjacent floors.
deflection through both mechanisms is EF共y兲 + EC共y兲. This com- Hence, for many common conditions, stiffnesses that must be
bined accumulated strain energy can be compared to the change evaluated for catenary action 共which clearly relies on good dia-
in potential energy, Wy, at any value of y less than limiting values phragm action and the ability to transfer the tensile forces from
共e.g., values corresponding to beam failure in flexure or connec- the beam to capable resisting elements兲 often can be estimated on
tion failure in axial load兲. The beam can be judged to fail if for all a floor-by-floor basis. For flexural and catenary action, consider-
deflections the change in potential energy exceeds the accumu- ation needs to be given to the potential that the sudden removal of
lated strain energy, or to survive if the accumulated strain energy a column releases strain energy contained in the compressed col-
at least equals the change in potential energy 共Fig. 18兲. For the umns above, magnifying the effective potential energy that must
analyzed condition, accounting for energy accumulated through be absorbed by lower floors.
both flexural and catenary behavior the structure stabilizes at a The analysis described above can be programmed relatively
deformation of approximately 0.64 m 共25 in.兲. easily with commonly available software such as Mathcad, as was
Adjacent columns can be verified by distributing the end mo- done by the authors to create the example in this paper.
ment of the beams under study in accordance with the relative
stiffnesses of the connecting columns and beams, and augmenting
the column axial service load by the sum of the increases in the Research Needs
vertical components of the end reactions of the beams above.
Similarly, the surrounding structure must also be evaluated for its The methods presented here can be adapted to be practical im-
ability to develop and sustain the catenary tensile forces. provements over current simplified methods to assess dispropor-
This simplified approach can be extended to structures with tionate collapse potential, because they are based on the physical
multiple floors. A parametric study using a representative framing phenomena associated with building failure, rather than a force-
system for a multistory building 共Fig. 19兲 shows that the horizon- based approach that is calibrated for certain structures. Further,
tal stiffnesses on adjacent floors exhibit little coupling when fram- these approaches do not rely on an assumed multiple on applied