Professional Documents
Culture Documents
Failure by Design
Failure by Design
by
Granger Meador
Physics Teacher,
Bartlesville High School
Bartlesville, OK
failurebydesign.info
FAILURE BY DESIGN
(Two Case Studies of Bridge Design Failure)
by Granger Meador
Published by the Author
Version 2.0
Contact information:
Granger Meador
Physics Instructor
Bartlesville High School
1700 SE Hillcrest Drive
Bartlesville, OK 74003-7299
1-918-336-3311
FAX 1-918-337-6226
Email: gmeador@bps-ok.org or inquiry@meador.org
WWW: http://failurebydesign.info
Page 2 of 34
PURPOSE
Disasters are inherently intriguing to students, and can motivate them to engage in
high levels of analysis. My goals in having my own 11-12th grade physics students
analyze the Hyatt Regency hotel walkways collapse and the failure of the Tacoma
Narrows bridge include:
develop students analytical and deductive skills in determining the possible and
actual failure modes of a structure
illustrate for students how engineers and architects must utilize physics in
designing safe structures
have students use their knowledge of force loading and vectors to explain the
flaw in the design change of the Hyatt walkways
introduce concepts of wave motion and aerodynamics to explain the vibrations
and failure of the Tacoma Narrows bridge
Page 3 of 34
ONLINE PRESENTATION
In addition to this handout, there is online presentation at http://failurebydesign.info
with graphical images, animations, and movies. The online presentation is adapted
from a PowerPoint presentation, and you can progress through it in your web browser.
VT-20
Twin Views of the Tacoma Narrows Bridge Collapse
Page 4 of 34
The Disaster:
On July 17, 1981 the hotel had been open for a year and a
local radio station was holding a dance competition. By 7 pm the atrium was
crowded with between 1500 and 2000 people, with many spectators observing from
the walkways. At 7:05 pm a loud crack echoed through the building and the second
and fourth floor walkways collapsed, killing 114 people and injuring over 200 others.
Page 5 of 34
Hyatt Puzzle
What part(s) of the walkways failed? Given the overall design as
constructed, what parts of the support structure could have failed to
cause both the second and fourth-floor walkways to collapse? What
evidence would you look for to decide which part actually failed?
The fourth-floor walkway was
suspended from the roof by
three sets of hanger rods. The
rods passed through
longitudinally welded box beams
and were capped with washers
and nuts which held up the
walkway by pressing on the
underside of the box beams.
The second-floor walkway was
suspended from the fourth-floor
walkway by another set of
hanger rods. These rods were
held up by washers and nuts
pressing on the top surface of
the 4th-floor box beams. The rods
passed through box beams for
the second-floor walkway and
were again capped with washers
and nuts pressing against the
underside of the 2nd-floor box
beams.
Long I-beams running down the
sides of each walkway were
suspended from the box beams by
angle brackets. At right is a closeup view of the construction of the
4th-floor walkway and its
connections to the hanger rods.
Page 6 of 34
Walkway floor:
The hanger rods from the roof to the fourth-floor walkway snapped or the roof
connection failed.
Examination of the accident scene would show the upper rods broken off, most likely
near the roof where the load in each rod would be maximized.
2)
3)
The fourth-floor box beams gave way at the ends, allowing the washers and nuts
underneath them to punch through.
The accident scene should reveal nuts stripped from the hanger rods, but few if any
punctures of the nuts through the fourth-floor box beams.
The accident scene should reveal washers and nuts still on the rods hanging from the
roof, with punctures through the box beams.
4)
One would expect a mix of the above physical evidence in the debris.
Page 7 of 34
The fourth-floor box beams gave way, but not at the ends, allowing the washers
and nuts above them to punch through.
The physics of the situation makes this highly unlikely: the washer/nuts under the beam
carry a much greater load than the washer/nuts above the beam. This would,
however, be able to cause the full failure of the fourth-floor walkway if the punctured
box beams separated sufficiently along the welds to allow the washers and nuts under
the beam to escape.
6)
7)
8)
9)
10)
This scenario suffers from the same defect as the one above.
Each of the above three scenarios would likely only lead to the failure of the secondfloor walkway, rather than both the second and fourth-floor walkways. However, if one
side of the second-floor walkway failed, sufficient stress or torque could be placed on
the upper walkway to make it collapse.
This would have allowed one section of a walkway to collapse, but the remaining
sections would likely have survived.
Page 8 of 34
Original design
As-built
The size of the washers and nuts and the design of the
box beam were NOT changed when this change in the
4th-floor connection was made.
Page 9 of 34
The Kansas City Building Code required a minimum support value of 151,000 newtons for each
hanger rod/box beam connection in the original design.
A test showed that each connection could, on average, support only 90,000 newtons. This was
only 60% of the required minimum for the original design.
The design change doubled the load on the fourth floor box beam, so the code would have
called for a support value of 302,000 newtons at the fourth floor connections, yet as-built those
connections could still only support 90,000 newtons.
Q:
Assume that the two walkways themselves each weighed 220,000 N. Each of the six 4th-floor
hanger rod/box beam connections could support 90,000 N, as could the six 2nd-floor
connections. How many 150 lb (667 N) people could the walkway connections support
a) as built? b) as originally designed? c) if built with the original design, but meeting code?
A.
a)
AS-BUILT:
(6 * 90,000 N) (2 * 220,000 N) = 100,000 N remaining capacity
(we have to subtract the weight of BOTH walkways from ONLY the 4th floor total
connection strength, since as-built those connections carried both walkways)
100,000 N / 667 N = 150 total people on both walkways
(video at time of collapse showed about 80 people on the 4th-floor walkway)
b)
AS ORIGINALLY DESIGNED:
(6 * 90,000 N) + (6 * 90,000 N) (2 * 220,000 N) = 640,000 N remaining capacity
(as designed the 2nd and 4th-floor connections were independent of each other)
640,000 N / 667 N = 960 total people on both walkways
(which is over SIX TIMES more than the as-built design)
c)
Thus one can see that if the design change had not been made, the walkways would likely still be intact
to this day, giving no indication that they did not meet the conservative building code.
It also turned out that the hanger rods did not meet specifications: the preliminary design sketches for
the walkways called for hanger rods with a strength of 413 megapascals. This was omitted from the final
structural drawings and the contractor specified rods with a strength of only 248 megapascals.
However, their weakness played no role in the catastrophe.
Page 11 of 34
As-built analogy:
Now shift one of the lower scales so that it hangs from the other lower
scale. Now the lower box beam scale reads the same amount as
before, but the upper box beam scale reads twice that, almost as
much as the top roof connection scale.
Page 12 of 34
Hyatt Consequences
There were a number of warnings during its construction that the Hyatt hotel project
was troubled. The atrium roof collapsed during construction, fortunately during a
weekend so that no one was injured. After the walkways were up, there were reports
that construction workers found them unsteady under heavy wheelbarrows, but the
construction traffic was merely rerouted and the walkway design not subjected to
intense scrutiny.
After the terrible tragedy, the Kansas City Star newspaper hired consulting engineers to
review the evidence and identify the cause of the accident. Within four days of the
failure, the front page of the Star showed drawings pinpointing the cause. The
investigative reporting of the accident won a Pulitzer Prize that year.
Within days of the tragedy, the remaining walkway for the third floor was dismantled
and removed in the middle of the night, despite protests from the city mayor. The
hotel owners argued it was a hazard, but attorneys for the victims objected. Engineers
studying the tragedy were disappointed they did not have the chance to use the
remaining walkway to test theories that the tempo of dancers on the walkways may
have contributed to the collapse (similar to the vibration problems of the Tacoma
Narrows Bridge).
After a twenty-month investigation, the U.S. attorney and county prosecutor
announced they had no evidence of a federal or state crime in connection with the
collapse. But two months later the state attorney general charged the design
engineers with gross negligence and the U.S. Department of Commerce eventually
concluded that the steel fabrication contractor (Havens Steel Company) was not at
fault. The design engineers had approved the original design that did not meet code
as well as the design change that weakened the structure. During a 26-week
administrative trial the detailer, architect, fabricator, and technician on the project all
testified that during construction they had contacted the project engineer regarding
the structural integrity of the connection detail. Each time he assured them that the
connection was sound, claiming to have checked the detail when in reality he had
never performed any calculations for this design at all. The Missouri Board of Architects,
Professional Engineers, and Land Surveyors eventually convicted engineer of record
Jack D. Gillum and project engineer Daniel M. Duncan of gross negligence,
misconduct, and unprofessional conduct in the practice of engineering. Their Missouri
(and later Texas) professional engineering licenses were revoked, and their company
(Gillum-Colaco or G.C.E. Inc.) lost its certificate of authority as an engineering firm. As
of September, 2000 Gillum was an Adjunct Professor of Civil Engineering at Washington
University in St. Louis, Missouri.
The tragedy eventually led to what has been reported as over $140 million in damages
awarded in civil cases brought by the victims and their families. These amounts
dwarfed the half million dollar cost of the building.
Failure By Design by Granger Meador (failurebydesign.info)
Page 13 of 34
Hyatt Prevention
This accident could have been prevented by a better design, better design
procedures, and ethical engineer behavior.
Better Design:
There are a bewildering number of design details in a massive project like the Hyatt
Regency, and it is too easy when one presented with the accident as a puzzle to think
that the problems in the design were obvious. But there were some severe
deficiencies in the walkway design. The original design, with its extremely long hanger
rods and awkward nut threading, was difficult to implement, and it did not meet
building codes.
Engineers have suggested possible improvements:
Objective:
Design and engineer a bridge using 1/8 inch x 1/8 inch balsa wood to hold as much
mass as possible over a 12-inch span.
Construction:
1. The overall width of the bridge may not exceed 3 inches.
2. The overall length of the bridge may not exceed 15 inches.
3. The bridge shall allow a 1-inch wide x 3/4-inch thick board to pass through it over
the roadbed.
4. The bridge shall have no structures below the abutments that support the bridge.
5. The bridge shall allow a 3/8-inch bolt to pass through the center of the bottom of
the bridge unobstructed (for testing).
6. The bridge shall be constructed entirely of 1/8-inch x 1/8-inch balsa wood.
7. Any common adhesive may be used at the joints of the wood members.
8. Adhesives may be used only at joints.
9. Wood joints may be notched if desired.
10. The mass of the bridge may not exceed 40 grams.
Competition:
1. The mass of the bridge will be determined before testing.
2. All bridges will use the same test device.
3. The bridge will be tested using a 1-1/2-inch wide x 10-inch long x 3/4-inch thick
wood block with a hole in its center for the testing mechanism.
4. The bridge is placed on the bridge tester, with the span set at 12 inches.
5. The teacher will attach the testing mechanism to the bridge.
6. When the testing mechanism is ready, the student will begin to add sand to the
bucket at the rate he or she chooses.
7. There will be a 10-minute time limit on adding sand to the bucket.
8. The student will continue to add sand to the bucket until the bridge collapses, and
the bucket falls.
9. The mass of the sand will be measured.
10. The bridge supporting the greatest load is the winner.
11. In the event that the bucket is completely filled without breaking for more than one
bridge, then, of those bridges, the one that has the least mass wins.
Variations:
Vary the length of the span of the bridge from 8 inches to 24 inches.
Assign dollar amounts to supplies (wood and glue), and judge the bridges on cost
efficiency (cost per pound of load held).
Products for this competitive event can be found in the Pitsco Competitive Events Theme Catalog.
Page 15 of 34
Objective:
Design and engineer a bridge using toothpicks (or Blunt End Structure Sticks) to hold as
much mass as possible over a span of 6 inches.
Construction:
1. The overall width of the bridge may not exceed 3 inches.
2. The overall length of the bridge may not exceed 9 inches.
3. The bridge shall allow a 1-inch wide x 3/4-inch thick board to pass through it over
the roadbed.
4. The bridge shall have no structures below the abutments that support the bridge.
5. The bridge shall allow a 3/8-inch bolt to pass through the center of the bottom of
the bridge unobstructed (for testing).
6. The bridge shall be constructed entirely of toothpicks (or Blunt End Structure Sticks).
7. Any common adhesive may be used at the joints of the wood members.
8. Adhesives may be used only at joints.
9. The mass of the bridge may not exceed 20 grams.
Competition:
1. The mass of the bridge will be determined before testing.
2. The bridge will be tested using a 1-1/2-inch wide x 4-inch long x 3/4-inch thick
wood block with a hole in its center for the testing mechanism.
3. The bridge will be placed on the bridge tester, with the span set at 6 inches.
4. The teacher will attach the testing mechanism to the bridge.
5. When the testing mechanism is ready, the student will begin to add sand to the
bucket at the rate he or she chooses.
6. There will be a 5-minute time limit on adding sand to the bucket.
7. The student will continue to add sand to the bucket until the bridge collapses and
the bucket falls.
8. The mass of the sand will be measured.
9. The bridge supporting the greatest load is the winner.
Variations:
Vary the length of the span of the bridge.
Assign dollar amounts to supplies (toothpicks and glue) and judge the bridges on
cost efficiency (cost per pound of load held).
Model the bridge after a local or famous bridge.
Products for this competitive event can be found in the Pitsco Competitive Events Theme Catalog.
Page 16 of 34
Page 17 of 34
Overhead, side, and cross-sectional views of the bridge and its deck bracing
Page 18 of 34
1.
2.
3.
After proving successful on the model, 1 9/16 -in. steel cables attached a point on
each side span to 50-yd concrete anchors in the ground. Unfortunately these
cables snapped a few weeks later, proving to be an ineffective solution, although
they were reinstalled in a matter of days.
In addition to these cables, center stays and inclined
cables, which connected the main cables to the
stiffening girder at the middle of the bridge, were
installed.
Finally, untuned dynamic dampers, similar to one
that had helped curtail torsional vibrations of
Moisseiffs Bronx-Whitestone Bridge in New York, failed
immediately after their installation in the Tacoma
Narrows Bridge. It was discovered that the leather
Center cable stays
used in the devices was destroyed during the
sandblasting of the steel girders before they were
painted, rendering them useless.
Page 19 of 34
Galloping Gertie
Prior to the day of its collapse, the bridge only exhibited vertical oscillations. The
transverse modes of vibration had nodes at the main towers, with from 0 to 8 nodes
between the towers. The maximum double amplitude (crest to trough) was about 5 ft
in a mode with 2 nodes between the towers, with a frequency of 12 vibrations/min. This
mode would startle motorists as cars ahead of them would disappear from view and
then reappear later. The bridge bounce led a wit to call it Galloping Girdie for its
motion and its side girders. This was corrupted into the sobriquet Galloping Gertie.
Some motorists would drive across the bridge and even stop midway to enjoy the
bouncing sensation, while others made long detours to avoid it.
The most frequently observed vibration was one with no nodes between the towers, a
double amplitude of up to 2 feet, and a frequency of 8 vibrations/min. Measurements
made before the bridge failed indicated that higher wind speeds correlated to higherfrequency vibration modes. But there was no significant correlation between wind
speed and vibration amplitude: winds of 3 or 4 mph could create motions of several
feet, while at other times the bridge remained motionless in winds as high as 35 mph.
The Collapse
A midnight storm on November 7, 1940 probably weakened the K-bracing under the
bridge deck, since a lone observer reported the bridges amplitude of vertical vibration
increased. By early morning the wind speed was 40 to 45 mph, and the bridge was
undergoing large vertical oscillations. By 9:30 am the span was vibrating in 8 or 9
segments at a frequency of 36 vibrations/min and a double amplitude of about 3 feet.
Traffic was shut down, but two cars with three passengers were trapped on the bridge
when it suddenly began to vibrate torsionally (twist) around 10 am. The passengers,
gripping the concrete curbs, crawled to
safety. Unfortunately a frightened dog in
one car refused to budge and had to
be left behind. Farquharson was on the
scene studying the bridge. At one point,
he walked along the torsional nodal line
along the center of the roadway to study
the center stays and, incidentally,
unsuccessfully attempt to retrieve the
dog, which perished in the collapse.
Page 20 of 34
Page 21 of 34
Tacoma Puzzles:
Why did this bridge oscillate so much?
The Tacoma Narrows bridge was a victim of poor aerodynamics. The long, narrow,
and shallow deck with its solid I-beam sides was too flexible to survive the winds of the
canyon in which it was situated. Most suspension bridges used deep, open trusses
which allowed wind to flow through relatively unimpeded. Designers were used to
making sure that a bridge could handle static loads, but the dynamic forces of winds
were seldom considered at the time of this bridges construction. In fact, subsequent
to this failure large bridges and buildings were routinely checked for aerodynamic
stability, often incorporating wind tunnel testing of scale models.
The Federal Works Agency investigated the failure after the collapse, and included on
its commission Theodore von Karman, a noted aeronautical engineer. The review
stressed that the bridge had met accepted engineering criteria of the time, having
been built to accepted safety factors for static loading. It did not cast blame on Leon
Moisseiff, but accepted that the bridge failure was the result of design limits being
stretched into previously unexplored areas.
The commission focused on three possible sources for the destructive dynamics of the
bridge: aerodynamic instability producing self-induced vibrations, periodic eddy
formations, and random turbulence. Many textbooks today attribute the collapse to
resonance. Resonance is when a system begins to oscillate with a large amplitude as
it is acted upon by periodic impulses of a frequency approximately equal to one of
the natural frequencies of oscillation of that system. However, an analysis by Billah and
Scanlan (1991) showed that the bridges behavior was not due to simple resonance.
Mathematical models of its motion are better explained by more complex selfexcitation mechanisms.
Commission member von Karman had proposed that the wind blowing across the
bridge deck created turbulent vortices. This process of vortex shedding would have
created alternating high and low pressure regions on the lee side of the bridge,
causing it to oscillate.
Page 22 of 34
Why did this bridge fail when other, longer bridges didnt?
The extreme flexibility of the Tacoma Narrows bridge arose from its shallow deck and
narrow width in comparison to its long span. A comparison to other bridges of the time
is revealing:
Bridge
(Location; Designer; Year Opened)
Deck Depth /
Span Length
Ratio of
Depth to
Span
Deck Width /
Span Length
Ratio of
Width to
Span
Tacoma Narrows
(Puget Sound, WA; Moisseiff; 1940)
8 ft / 2800 ft
1 : 350
39 ft / 2800 ft
1 : 72
Bronx-Whitestone
(Long Island Sound, NY; Moisseiff; 1939)
11 ft / 2300 ft
1 : 209
74 ft / 2300 ft
1 : 31
Golden Gate
(San Francisco, CA; Strauss; 1937)
25 ft / 2400 ft
1 : 168
89 ft / 2400 ft
1 : 47
George Washington
(New York; Ammann; 1931)
36 ft / 3500 ft
1 : 97
106 ft / 3500 ft
1 : 33
The lower depth-to-span and width-to-span ratios of the other bridges made them
much less vulnerable to both vertical and torsional deflections, as shown by the graphs
on the following page.
Page 23 of 34
Use open stiffening trusses which would allow the wind free passage through the
bridge
Increase the width to span ratio
Increase the weight of the bridge
Use an untuned dynamic damper to limit the motions of the bridge (the dampers
on the bridge did not work)
Increase the stiffness and depth of the trusses or girders
Streamline the deck of the bridge
The concept of deep stiffening trusses was applied in retrofitting several other bridges.
Moisseiffs Bronx-Whitestone bridge had a truss added above its side I-beams in 1946.
The famous Golden Gate Bridge in San Francisco had a $3.5 million retrofit to stiffen it
as well.
Page 24 of 34
Several design changes were made when a new Tacoma Narrows Bridge was
constructed, using the original bridges piers. Scale models of the new $18 million
bridge were tested in wind tunnels at the University of Washington during its design, and
it was a four-lane structure with a 60-ft wide deck and 25-ft deep stiffening open
trusses. It has been a successful design for over 50 years.
Page 25 of 34
Analysis of the specifications on this page reveals that the new bridge had a
depth/span ratio of 1:112, versus the 1:350 ratio of the failed design. The new design
also boasted a width/span ratio of 1:47, versus the 1:72 ratio of Galloping Gertie.
Page 26 of 34
Page 27 of 34
Wave Demonstrations
Slinky Wave Demos
Simple but effective demonstrations of wave motion
can be performed with a Slinky, two people, and a
short length of string. While a regular metal or plastic
Slinky will work, you will obtain better results using an
extra-long Slinky, such as the one shown in the
photograph. You can obtain such springs from
science education supply houses.
Transverse Waves:
A transverse wave vibrates perpendicularly
(at right angles) to the wave travel (water waves
are a good example). To demonstrate, have
two students each take one end of a Slinky and
stretch it out along the floor (the waves will be
more apparent this way). Have one student move his or her end of a plastic or metal
Slinky back and forth (left and right, like a snake crawling), perpendicular to its stretched
length. The other student must hold his or her end of the Slinky still. A series of
transverse waves will be generated.
Transverse Wave Reflection:
A wave striking the boundary of a more dense medium will partially reflect, and the
reflection will be inverted. This is seen in the above demonstration, where one end of
the Slinky is held by a student, whose grip creates a more dense medium for the wave
energy. However, a wave striking the boundary of a less dense medium will have an
erect reflection. This can be demonstrated by tying some string onto one end of the
Slinky and having the student hold the string rather than the Slinky. The string is a less
dense medium for wave travel, so waves sent toward the student holding the string will
reflect right-side-up.
Standing Waves:
When a series of wave pulses are reflected off a
more dense medium, standing waves can be
generated. These distinctive waveforms have places
where the medium does not vibrate at all, called
nodes, and other places where the medium vibrates
the most, called antinodes. When the students are
demonstrating transverse waves (without using a
string), standing waves with varying numbers of nodes
and antinodes can be generated by having the
student moving the Slinky vary the rate at which he or
she continually moves it back and forth.
Page 28 of 34
Longitudinal Waves:
A longitudinal wave vibrates parallel to (in the
same direction of) wave travel (sound waves
are a good example). This kind of wave was
NOT exhibited in the movies of the Tacoma Narrows Bridge. To demonstrate, have one
student grasp and draw toward himself or herself several coils of a stretched metal
Slinky and then release the coils. The other student must hold his or her end of the
Slinky still. A longitudinal wave pulse will be generated and travel down the length of
the Slinky.
Longitudinal Compressions and Rarefactions:
Longitudinal waves can be composed of compressions, where the parts of the
medium (coils of the Slinky) are closer together than normal, or rarefactions, where the
parts of the medium are farther apart than normal. In the above demonstration, the
students created compressional longitudinal waves. A rarefactional longitudinal wave
can be produced by stretching a segment of the Slinky and then releasing it. The
stretched area (rarefaction) will then travel along the length of the Slinky.
Page 29 of 34
The Wave Machine (from Ronald Edges String and Sticky Tape Experiments)
A fancy Shive wave machine like the
one in the photograph will exhibit
torsional waves. But you can build
your own cheap wave machine with
straws and sticky tape, as described
below.
You will need about half a meter of sticky
tape, and twenty or so straws. Turn each end
of the tape over about one centimeter, and
stick them on the desk, or other flat surface
as shown in the upper figure. Now place the
straws about 1 cm apart crosswise with their
centres on the tape as shown in the lower
figure. Pick up the ends of the tape, and attach one
end to the lintel of a door, or other suitably high place
where the chain of straws is free to oscillate. Now
sharply tap the bottom straw, and a torsional pulse will
travel up the machine and be reflected at the top.
Since the top is fixed, the pulse will be of opposite
sign on reflection, descending. This device can be
used to display almost all the properties of one
dimensional transverse traveling and stationary waves.
For example, increasing the tension by hanging a
weight at the bottom will speed up the waves,
loading the ends of the bottom half of the straws
(which can easily be done by inserting paper clips
into each end of a straw) will be like a "dense
medium"- with a lower velocity and reflection at the
interface with the light straws. Moving the bottom
straw to and fro with the correct period will produce
standing waves.
Page 30 of 34
Page
1
1
Description
Hyatt walkway collapse
Tacoma Narrows bridge
2
4
5
Author photo
CD-ROM clipart
Hyatt hotel exterior
Architects walkway
rendering
Crowd in Hyatt atrium
5
5
5
6
6
9
9
10
10
10
10
12
14
14
17
17
17
18
18
18
18
19
19
20
Page 31 of 34
Description
Twisting bridge
Bridge collapse
Sagging side span
Frayed cable
21
Buckled tower
22
23
24
24
25
25
25
25
25
25
26
26
26
27
27
28
28
28
29
30
Transverse waves
Nodes and antinodes
Longitudinal waves
Shive wave machine
30
30
Page 32 of 34
REFERENCE LIST
used for non-profit educational purposes in accordance with the fair use provisions of copyright law
Page 33 of 34
Page 34 of 34