You are on page 1of 62

MECHANICS OF

PROGRESSIVE
COLLAPSE:
WHAT DID
AND DID NOT
DOOM WORLD
TRADE CENTER,
AND WHAT CAN
WE LEARN ?
ZDENĚK P. BAŽANT

Presented as a Mechanics Seminar at Georgia Tech,


Atlanta, on April 4 ,2007, and as a Civil Engineering
Seminar at Northwestern University, Evanston, IL,
on May 24, 2007
Collaborators:
Jialiang Le
Mathieu Verdure
Yong Zhou
Frank R. Greening
David B. Benson
Structural
System

- framed
tube
Previous Investigations
• Computer simulations and engrg. analysis at NIST — realistic,
illuminating, meticulous but no study of progressive collapse.
• Mechanics theories of collapse:
1. Northwestern (9/13/2001) — still valid
2. E Kausel (9/24/2001) — good, but limited to no dissipation
3. GC Clifton (2001) — “Pancaking” theory: Floors
collapsed first, an empty framed tube later? — impossible
4. GP Cherepanov (2006) — “fracture wave“ hypothesis — invalid
5. AS Usmani, D Grierson, T Wierzbicki…special fin.el. simulations

• Lay Critics: Fletzer, Jones, Elleyn, Griffin, Henshall,


Morgan, Ross, Ferran, Asprey, Beck, Bouvet, etc.
Movie “Loose Change” (Charlie Sheen), etc.
1
Review of Elementary
Mechanics of Collapse
Initial Impact – only local damage, not overall
Tower designed for impact of Boeing 707-320 (max. takeoff weight is 15% less, fuel capacity
4% less than Boeing 767-200)

Momentum of Boeing 767 ≈ 180 tons × 550 km/h


Momentum of equivalent mass of the interacting
upper half of the tower ≈ 250, 000 tons × v0
Initial velocity of upper half:
v0 ≈ 0.7 km/h (0.4 mph)
Assuming first vibration period T1 = 10 s:
Maximum Deflection = v0T / 2p ≈ 40 cm
(about 40% of max.hurricane effect)
13% of columns
were severed
on impact, some
more deflected
1. 60% of 60 columns of impacted face (16% of

Failure Scenario 287 overall) were severed, more damaged.


2. Stress redistribution ⇒ higher column loads.
3. Insulation stripped ⇒ steel temperatures
up to 600oC→yield strength down -20% at
300oC,-85% at 300oC, creep for > 450oC.
4. Differential thermal expansion +
viscoplasticity ⇒ floor trusses sag, pull
perimeter columns inward (bowing of
columns = buckling imperfection).
5. Collapse trigger: Viscoplastic buckling of hot
columns (multi-floor) → upper part of tower
falls down by at least one floor height.
6. The kinetic energy of upper part can be
neither elastically resisted nor plastically
absorbed by the lower part of tower ⇒
progressive collapse (buckling + connections
sheared.)
a) b) c) d) e) f)

I. Crush-Down Phase II. Crush-Up Phase


Toppling
like
a tree?
Why Didn't the Upper Part Fall Like
a Tree, Pivoting About Base ?
a) c) e)
 MP
F1
m
H1 h1

d x
F1 MP
b) d)  f)
mx  FP
H1
m F
g 3
Possible ? Fmax  mg
8
(The horizontal reaction at pivot)
> 10.3× (Plastic shear capacity of a floor)
South tower
impacted
eccentrically
Plastic Shearing of Floor Caused by Tilting
(Mainly South Tower)

a b c d e
Elastically Calculated Overload
m
Dynamic elastic overload factor calculated for
maximum deflection (loss of gravity potential h
of mass m = strain energy)

 The column response could not be


elastic, but plastic-fracturing
Can Plastic Deformation Dissipate
the Kinetic Energy of Vertical
Impact of Upper Part? 1 2
n = 3 to 4 plastic hinges per column line.
3
Combined rotation angle:
Dissipated energy:
Kinetic energy = released gravitational potential energy:

Only <12% of kinetic energy Collapse could not have


was dissipated by plasticity in  taken much longer than a
1st story, less in further stories free fall
Plastic Buckling
F
Case of P1 P1
single floor u MP P1
buckling
 L/2
L=2Lef h L 
Fc ≥ Fs MP
…can propagate P1
dynamically
F0 Yield limit F0 Yielding
Fc < Fs Wf Shanley
bifurcation Plastic
Load F

… cannot Plastic buckling lh


inevitable! buckling

Fc F Service
s
load Expanded scale
0 0
0 0.5h h 0 0.04h
Axial Shortening u
2
Gravity-Driven Propagation
of Crushing Front in
Progressive Collapse
Two Possible Approaches to
Global Continuum Analysis
• Stiffness Approach  homogenized elasto-
plastic strain-softening continuum — must
be NONLOCAL, with characteristic length =
story height … COMPLEX !
• Energy Approach – non-softening
continuum equivalent to snap-through*
— avoids
________________________
irrelevant noise …SIMPLER !
* analogous to crack band theory, or to van der Waals
theory of gas dynamics, with Maxwell line
Crushing of Columns of One Story
One-story equation of motion:: ü = g – F(u) / m(z)
Initial condition: v  velocity of impacting block
Collapse arrest criterion: Kin. energy K < Wc
F0 Crushing Lumped Mass

Rehardening
Resistance F(u)
Crushing force, F

Wc λh
ΔFd Wb
Dynamic Snapthrough 1 2
mg 
ΔFa
Fc Maxwell Line 3

0 u0 u c uf h Lower Fc for
Floor displacement, u multi-floor buckling!
a) Front accelerates b) Front decelerates c) Collapse arrested
Real Crushing λ
Crushing force, F
F0 F0 F0
Resistance F(z) F(z) λh ΔFd W1 = K h
λh ΔFd W1 = W2 Fc
ΔFd W1 = W2
mg Fc mg
ΔFa mg ΔFa
Fc u
u u 0
0 zc h 0 h h
v v v
v2 >v1 for Fc
for Fc v1
λh λh λh
Floor velocity, v

v1 Deceleration
Acceleration
v1 Deceleration
Deceleration u Acceleration
u 0 u
0 h 0 h
Displacement h
v 1 v2 > v1 v 1
v
v1
g-Fc/m g-Fc/m
v1
v1 v2 < v1
t t
0 t zc 0 tzc Time t 0 time
Mean Energy Dissipation by Column Crushing, Fc, and
Compaction Ratio, λ, at Front of Progressive Collapse
Internal energy (adiabatic) potential : W = ∫ F(z)dz Total potential = Πgravity - W
a) Single-story plastic buckling L = h
h h
Floor n n-1 n-2 n-3 n-4
Fpeak
Crushing Force, F

Wc Wc λh
Fc energy-
Fc equivalent
b) Two-story plastic buckling L = 2h snapthrough
Fpeak = mean
2λh
crushing
Fc
Fc force
c) Two-story fracture buckling L = 2h
Fpeak
Fs Service load
Fc Fc
Distance from tower top, z
2 Phases of Crushing Front Propagation
Crush-Down Crush-Up
(Phase I of WTC) (Phase II of WTC
or Demolition)

Mass Collapse
shedding front
Phase II
Collapse
front
1D Continuum Model for Crushing Front Propagation
λ = compaction ratio = Rubble volume within perimeter
Tower volume h)
Crush-Down
a) g) Can 2 fronts propagate . i)
up and down m(z)vCrush-Up
b) ζ .
C z0 ζ simultaneously ? m(z)g
– NO !
m(y)y
z
C z0 . m(y)g
Fc’< Fc if slower zΔt
s0 B Fc F Fc Fc
than free fall c
. s = λs Fc .
z 0
Phase 1
downward yΔt μy. 2
H
Δ c)
t d)
A A C y

y0 = z0
e)
r0B’ C r = λr0 B’ λz0
λ(H-z0) B B λH B

Phase 1. Crush-Down Phase 2. Crush-Up


Diff. Eqs. of Crushing Front Propagation
I. Crush-Down Phase:

z(t)
z0
Resisting force Buckling Comminution Jetting air
Intact
Criterion of Arrest (deceleration): Fc(z) > gm(z)
z0
y(t)
II. Crush-Up Phase: Compacted

Inverse: If functions z(t), m(z),


l(z) are known, the specific
energy dissipation in collapse,
Fc(y), can be determined
Compaction ratio:
fraction of mass ejected outside perimeter
Resistance and Mass Variation along Height

400 400

300 300

200 200

100 100

0 0

0 200 400 600 800 1000 1 1.2 1.4 1.6


Variation of resisting force due Variation of mass density, m(z),
to column buckling, Fb, (MN) (106 kg/m)
Energy Potential at Variable Mass

Crush-Down
Crush-Up

Note:
Solution by quadratures is possible for constant average
properties, no comminution, no air ejection
Collapse for Different Constant Energy
Dissipations
Tower Top Coordinate (m) (for no comminution, no air)

Wf = 2.4 GNm
fall arrested
2

1.5
free 1
fall 0.5
phase 1
0
phase 2

λ= 0.18 , μ= 7.7E5 kg/m , z0 = 80 m , h = 3.7 m

Time (s)
Collapse for Different Compaction Ratios
Tower Top Coordinate (m) (for no comminution, no air)

transition between
phases 1 and 2

free
fall λ= 0.4
0.3
Wf = 0.5 GNm , 0.18
μ= 7.7E5 kg/m ,
z0 = 80 m , h = 3.7 m
0
Time (s)
Collapse for Various Altitudes of Impact
for impact 2 floors below top
Tower Top Coordinate (m)

mg < F0,heated
5 (≈ 2.5 E7 GNm)

free
fall 20

phase 1 55
phase 2
λ= 0.18 , h = 3.7 m
μ= (6.66+2.08Z)E5 kg/m
Wf = (0.86 + 0.27Z)0.5 GNm

Time (s) (for no comminution, no air)


Crush-up or Demolition for Different
Constant Energy Dissipations
Wf = 11 GNm
Tower Top Coordinate (m)

fall arrested

asymptotically
6 parabolic end
free 5
fall 4
3
2
0.5

λ= 0.18 , μ= 7.7E5 kg/m , z0 = 416 m , h = 3.7 m

Time (s) (for no comminution, no air)


Resisting force as a fraction of total
Impacted Floor Number Impacted Floor Number
96 81 48 5 F 110 81 64 25 F 101 110
100% 100%
Resisting Force /Total Fc

Fb Fb
75% 75%
Crush-down Crush-down
ends ends
50% 50%
Fs Fs
North Tower South Tower
Fb Fb
25% Fs Fa 25%
Fs Fa

Fa Fa
0% 0%
0 4 8 12 0 4 8 12

Time (s) Time (s)


Resisting force / Falling mass weight
Impacted Floor Number Impacted Floor Number
96 81 48 5 F 110 81 64 25 F 101 110
100 100
Fc / m(z)g

10 10

1 Crush-down 1 Crush-down
ends ends

0.1
North Tower 0.1
South Tower
0 4 8 12 0 4 8 12
Time (s) Time (s)
External resisting force and resisting
force due to mass accretion
Impacted Floor Number Impacted Floor Number
Resisting force Fc and Fm (MN)

96 81 48 5 F 81 64 25 F
2500 2500

1250 1250 Fm
Fm

Fc Fc

North Tower South Tower


0 0
0 4 8 12 0 4 8 12
Time (s) Time (s)
3
Critics Outside
Structural Engineering Community:

Why Are They Wrong?


Lay Criticism of Struct. Engrg. Consensus
1) Primitive Thoughts:
 Euler's Pcr too high Shanley
 Buckling possibility denied bifurcation
 Plastic squash load too high, etc.
 Initial tilt indicates toppling like a tree?
— So explosives must been used !
No ! — horizontal reaction is unsustainable
~4º tilt due to
Like a Mass asymmetry of
Tree? Centroid damage

~25º (South Tower)


No ! No ! non-accelerated
Ft
rotation about
vertically moving
mass centroid
2) Collapse was a free fall ! ? Therefore the
steel columns must have been destroyed
beforehand — by planted explosives?
Video Record of Collapse of WTC Towers

North Tower South Tower


Tilting Profile of WTC South Tower
t
2
t
H1
1 m e
 2
H1
s Video c
-recorded
(South
Tower)
Initial tilt
H1
C  t  (1  cos  )
2

East North
Comparison to Video Recorded Motion
(comminution and air ejection are irrelevant for first 2 or 3
seconds)
Tower Top Coordinate (m)

420 420
From crush-down From crush-down
differential eq. differential eq.

410
Note
400 Free fall
Free fall
uncertainty range

400 First 20m of fall


First 30m of fall South Tower
North Tower (Top part  large falling mass)
380
0 1 2 3 0 1 2

Time (s) Time (s)


Not fitted but predicted! Video analyzed by Greening
Seismic and video records rule out the free fall!

417 m
H From seismic data:
crush-down T ≈ 12.59s ± 0.5s
North Tower
Impact of Free fall with pulverization
compacted with expelling air
rubble layer impeded by single-
on rock base story buckling only 12.81s
of bathtub
8.08s 12.29s 12.62s
Seismic
0m Most likely time rumble
-20 m from seismic record T
Calculated crush-down duration vs. seismic record
Ground Velocity Tower Top Coordinate (m)
450 450
North Tower South Tower
with air ejection with air ejection
& comminution & comminution
300 300
Seismic
Seismic error
Free fall error Free fall
Calculation
Calculation error
Crush-down ends
150 error 150
Crush-down ends

with buckling only with buckling only


0 0
(m/s)

Free fall Free fall


a a b
b c
c
0 4 8 12 16 0 4 8 12
Time (s) Time (s)
3) Pulverizing as much as 50% of concrete
to 0.01 to 0.13 mm required explosives!
NO. — only 10% of kinetic energy sufficed.
How much explosive would be needed to
pulverize 73,000 tons of lightweight concrete of
one tower to particles of sizes 0.01— 0.1mm ?
• 237 tons of TNT per tower, put into
small drilled holes (the energy required is
95,000 MJ; 30 J per m2 of particle surface,
and 4 MJ per kg of TNT, assuming 10% efficiency at
best).
Comminution (Fragmentation and
Pulverization) of Concrete Slabs
Schuhmann's law: M ( D)  M t ( D / Dmax ) k

D
mass of particles < D total particle size
Energy dissipated
= kinetic energy D 3 G f ( D)
loss ΔK
K  W f ( D )   Dmin D
dM ( D)
Cumulative Mass of Particles

0.12 mm 16 mm
1 density of
particle size
(M / Mt)

0.012 mm k
= Dmin 1
0.16mm = Dmin

0.01 0.1 1 10 Particle Size (mm)


Kinetic Energy Loss ΔK due to Slab Impact
Momentum balance: Fragments
m Compacted mv1  mv2  i mi vi
layer
max K for vi  v2 (all i)
v1 Comminuted = ms concrete
slabs Kinetic energy loss:
1 2 1 2
v2 K    mv1  (m   mi )v2 
2 2 
Kinetic energy to
 ms
pulverize concrete K  
z 2

2h [1  ms / m( z )]
slabs & core walls
Total: U  K total  K  Wb  Wa (energy conservation)
Gravitational Concrete
Buckling Air
energy loss fragments
Fragment size of concrete at crush front
Fragment Size at Crush Front (mm)

Impacted Floor Number Impacted Floor Number


96 81 48 5 F 110 81 64 25 F 101 110
Maximum and Minimum

10 10

Crush-down Crush-down
1 ends 1
ends
Dmax
0.1
Dmax
0.1

0.01 Dmin 0.01 Dmin

North Tower 0.001


South Tower
0.001
0 4 8 12 0 4 8 12

Time (s) Time (s)


Comminution energy / Kinetic energy of
falling mass
Impacted Floor Number Impacted Floor Number
96 81 48 5 F 110 81 64 25 F 101 110
100% 100%
Wf / К

10% 10%

1% 1%
Crush-down Crush-down
ends ends

North Tower South Tower


0.1% 0.1%
0 4 8 12 0 4 8 12

Time (s) Time (s)


Dust mass (< 0.1 mm) / Slab mass
Impacted Floor Number Impacted Floor Number
96 81 48 5 F 110 81 64 25 F 101 110
1
1

Crush-down Crush-down
Md / Ms

ends ends
0.5
0.5

North Tower South Tower


0 0
0 4 8 12 0 4 8 12

Time (s) Time (s)


Loss of gravitational potential vs.
comminution energy
1000 1000
North Tower South Tower
Energy Variation (GJ)

Loss of gravitational Loss of gravitational


potential potential
500 500

Ground impact Ground impact

Comminution Comminution
energy energy
0 0
0 4 8 12 0 4 8 12
Time (s) Time (s)
a
4) Booms During Collapse!
Air squeezed out
—hence, planted explosives?
of 1 story in 0.07 s
If air escapes story-by-story, its
mean velocity at base is
Air Jets
va = 461 mph (0.6 Mach), but
h
locally can reach speed of sound

200 m of concrete
dust or fragments
(va < 49.2 m/s, Fa < 0.24 Fc,  pa < 0.3 atm)

5) Dust cloud expanded too rapidly?


Expected.
North Tower Collapse in Sequence

Note:
• Dust-laden air jetting out
• Moment of impact cannot be detected visually
Moment of ground impact cannot be seen, but from seismic
record: Collapse duration = 12.59 s (± 0.5 s of rumble)

Note
jets
of
dust-
laden
air
6) Pulverized concrete dust (0.01 to 0.12 mm)
deposited as far as 200 m away? — Logical.

7) Lower dust cloud margin = crush front?


— air would have to escape through a rocket nozzle!
8) Temperature of steel not high enough
to lower yield strength fy of structural steel,
to cause creep buckling?
 fy reduced by 20% at 300ºC, by 85% at 600ºC (NIST).
 Creep begins above 450ºC.
 Steel temperature up to 600ºC confirmed by annealing
studies at NIST.
9) Thermite cutter charges planted?
— evidenced by residues of S, Cu, Zi found in dust?
But these must have come from gypsum wallboard,
electrical wiring, galvanized sheet steel, etc.

10) “Fracture wave” allegedly propagated in a material


pre-damaged, e.g., by explosives, led to free-fall collapse
— unrealistic hypothesis, because:
 A uniform state on the verge of material failure cannot exist
in a stable manner, because of localization instability.
 Wave propagation analysis would have to be nonlocal, but wasn't
 “Fracture wave” cannot deliver energy sufficient for comminution.
4
How the findings can be
exploited by tracking demolitions
Proposal: In demolitions,
measure and compare energy
dissipation per kg of structure.
Use:
1) High-Speed Camera
2) Real-time radio-monitored
accelerometers:
Note: Top part of WTC dissipated 33 kJ/m3
Collapse of 2000
Commonwealth
Avenue in Boston
under construction,
1971
(4 people killed)
The collapse was
initiated by slab
punching)
Murrah Federal
Building in
Oklahoma City,
1995
(168 killed)
Ronan Point
Collapse
U.K. 1968
Weak Joints, Precast Members

Floor
slab

Reinforcing
Bar
Hotel New World

Singapore 1986
Generalization of Progressive Collapse

1) 1D Translational-Rotational 25th floor

--- "Ronan Point" type Gas


Angular momentum and shear exploded
not negligible on 18th floor

2) 3D Compaction Front
Propagation
— will require finite
strain simulation
Gravity-Driven Progressive Collapse
Triggered by Earthquake
MAIN
RESULTS
• All WTC
observations
are explained.

• All lay
criticisms
are refuted.
Download 466.pdf & 405.pdf from Bazant’s website:
www.civil.northwestern.edu/people/bazant.html
References
• Bažant, Z.P. (2001). “Why did the • Kausel, E. (2001). “Inferno at
World Trade Center collapse?” the World Trade Center”, Tech
SIAM News (Society for Industrial
and Applied Mathematics) Vol. 34, Talk (Sept. 23), M.I.T.,
No. 8 (October), pp. 1 and 3 Cambridge.
(submitted Sept. 13, 2001)
(download 404.pdf).
• Bažant, Z.P., and Verdure, M.
(2007). “Mechanics of Progressive • NIST (2005). Final Report on
Collapse: Learning from World
Trade Center and Building the Collapse of the World
Demolitions.” J. of Engrg. Trade Center Towers. S.
Mechanics ASCE 133, pp. 308— Shyam Sunder, Lead
319 (download 466.pdf). Investigator. NIST (National
• Bažant, Z.P., and Zhou, Y. (2002). Institute of Standards and
“Why did the World Trade Center
collapse?—Simple analysis.” J. of Technology), Gaithersburg,
Engrg. Mechanics ASCE 128 (No. MD (248 pgs.)
1), 2--6; with Addendum, March
(No. 3), 369—370 (submitted
Sept. 13, 2001, revised Oct. 5,
2001) (download 405.pdf).

Download 466.pdf & 405.pdf from Bažant’s website:


www.civil.northwestern.edu/people/bazant.html

You might also like