Learning from the past

Failures during
and after construction
Both design and construction errors are identified
BY DOV KAMINETZKY
PARTNER
FELD, KAMINETZKY AND COHEN, P.C.

Large-scale structural failure is a nightmare that haunts Accordingly, failures can be reduced by more compe-
the construction industry. The financial devastation, the tence in design, construction, and construction supervi-
demolished reputations and the loss of life that could re - sion. The possibility of a major error in design actually
sult from a collapse have troubled the sleep of probably getting through the construction phase is indeed re-
e ve ry architect, engineer, contractor or owner at some mote, considering all of the stages of checks and con-
time. trols in design, estimating, detailing, field supervision,
and construction through which a job must go. In con-
This frightening quality of failures almost guarantees
crete construction, fortunately, there is a certain amount
that they will continue to happen. Fear, embarrassment
of informal load testing inherent in the construction
and the gag of interminable lawsuits have kept informa -
process itself and collapses are much more common
tion on failure from traveling quickly enough, what little
during construction than after completion and full oc-
of it ever gets into general circulation at all.
cupancy.
The way to dispel a nightmare is to attack it with hard It is natural, when forms and slabs collapse during
fact, with eyes open wide and the mind alert. . . . avail - concreting, to assume that the formwork was at fault.
ability of complete and accurate information could be the This is not always true; the collapse of one 4-story con-
first step towards shaking the dread of collapse. crete structure was thought at first to be caused by form
—Engineering News-Record, June 4, 1981 failure, but later investigation showed that some of the
exterior wall columns were not on the solid rock as-
These timely remarks underscore some of the prob- sumed in the design plans. A column settled, became in-
lems confronting our firm during more than five o p e ra t i ve, and the slabs collapsed. Other cases have
decades of investigating both major and minor con- been reported in which slabs collapsed due to weakness
struction failures. A few cases are presented here in the caused by duct openings at high-stress points. When
hope of furthering the understanding and awareness lower floor slabs collapse they carry upper floor forms
needed to prevent such disasters. with them, and the situation sometimes looks like a
The term failure indicates not only structural collapse formwork failure until a closer study is made.
but a wide range of nonconformity with design expec-
tations or requirements—such as unwanted settle- Torsional cracking in slabs
ments, deformations, cracks, bulges and misalignments. Unstiffened edge slabs of flat plate floors require top
If one takes time to measure the shape, position, and reinforcement at exterior edges to prevent torsional
condition of completed stru c t u re s, many failures to cracking at the stiff connection to the column. Columns
comply with good design and construction practice should be so located as to avoid reentrant corners at ex-
could be found. terior edges of flat plates that have shallow beams or no
Many recent failures can be traced to: beams at all. The load transfer will induce high torsion-
al stress in the slab edge or the face of the shallow span-
• errors in reading drawings drel and split it near the column. In a 16-story apartment
• design errors where this recommendation was disregarded, bearing
• sloppy construction practices piers were later added to provide direct support at the
• poor communication between designer and contractor reentrant corners and neutralize the torsional moment.
• inadequate construction supervision In a 40-story apartment of similar layout, flat slabs
exhibited cracking at the columns. The use of bearing cracked area rebuilt was subjected to a load test and it
walls as a remedy was out of the question, and structur- failed after 22 hours with a full load. Another load test
al steel brackets were bolted to the column face to pro- made after adding the brackets was quite successful.
vide similar support for the corner. One floor slab with This added support detail was provided at eight columns
on each of the upper levels (Figure 1). Self-anchoring
bolts were set into drilled holes 10 inches deep, and each
bolt was tested for pullout resistance to twice the design
requirement. The proof tests were all satisfactory.

Shear failure caused by construction error

A thre e - ye a r-old concrete plaza deck, serving as the
roof of a garage, collapsed in New York without warn-
ing, crashing down on parked cars. The roof consisted
of a 16-inch-deep waffle slab with 3 feet of earth cover.
About half of a symmetrical entrance plaza, an area 45 x
50 feet, failed. The other half remained in place, appar-
ently in sound condition. Failure was a clean punching
shear, with little effect beyond the shear cut.
In spite of the builder’s long experience with this type
Figure 1. Torsional shear failure at column near reentrant
slab edges. Solution shown here involved repair followed by of construction, the 12-inch-deep concrete caps that
installation of supplementary support brackets attached to should have extended 10 inches beyond the column
column to resist the twisting action. faces had been omitted at all of the columns in the fail-
ure area (Figure 2). They had also been omitted at all of
the nine columns in the symmetrical area that did not
fail. The only difference in conditions between these two
areas was a stopped-up drain in the failed area. This re-
sulted in a frozen earth cover on the deck that failed; the
other half of the deck was well drained. The factor of
safety for three years had been 1.05.
The failed slab was reconstructed with column heads
and new columns; the other half of the deck was con-
siderably strengthened by new girders, capitals, and col-
umn jackets.

Shear failure: design or construction error?

At first glance, the tragic failure described next resem-
bled a formwork collapse. Howe ve r, it was determined
that it resulted from excessive punching shear in the flat

BASIC RULES FOR PREVENTING 5. Just as a ship cannot be run by two captains, a con-
CONSTRUCTION FAILURES struction job must be run by one individual—not by a
1. Gravity always works—if you don’t provide perma- committee. That individual must have full authority to
nent support, something will fall. plan, direct, hire and fire; and full responsibility for pro-
duction and safety.
2. Chain reaction will make a small fall into a large fail-
ure, unless you can afford a fail-safe design, where suf- 6. Good craftsmanship is needed on the part of the de-
ficient reserve support is available when one compo- signer, the vendor, and the constructor teams.
nent fails.
7. Some designs are unbuildable. Attempts to produce
3. It requires only a small error or oversight, in design, in architectural gems may stretch the limit of safe build-
detail, in material strength, in assembly, or in protective ability even with our most sophisticated equipment
measures, to cause a large failure. and techniques.

4. Eternal vigilance is necessary to avoid small errors. If 8. There is no foolproof design, there is no foolproof
there are no capable foremen on the job and in the de- construction method—without careful control.
sign office, then supervision must take over the chore of
local control. Inspection service and construction man- 9. The best way to generate a failure on your job is to
agement cannot be relied upon as a secure substitute. d i s re g a rd the lessons to be learned from failures of
others.
 Figure 3. Columns remained standing following collapse of flat plate floors in this
office building. Probable failure cause was excessive punching shear in the flat plates
where reinforcement continuity was interrupted at some of the columns.

Figure 2. The roof of this parking garage in New York City or columns, which of course prevented slab steel from
collapsed suddenly three years after construction. running through the columns. The design called for a
Immediate cause of failure: a plugged drain in the earth complex reinforcement assembly around each interior
cover above. With water unable to drain from the soil, the column within the slab thickness, but how these assem-
weight on the garage roof increased, precipitating the blies could be placed within the zone of high shear and
collapse. Key reason for failure: the contractor failed to
still permit the duct openings was not clear.
construct the called-for concrete cap at the top of each
column. In looking at a plan view of the engineering Inadequate mudsills under shoring
drawings, the contractor mistook the lines representing the
concrete cap for the outline of the spread footing. Had Workmen were placing concrete for the second floor
there been better field inspection of this job, or had the of a building addition when steel shoring supporting
drawings been clearer, this failure might have been the area collapsed and eight workmen fell 20 feet to the
prevented. basement level. The foundations for the shoring were
16-inch-wide pieces of plywood cut at random lengths.
Because of the flexibility of the wood, soil pressure un-
der the shoring towers was approximately 5 tons per
plates at the columns. square foot, or 5 times what the ground could ade-
Reinforced concrete flat plate floors (no column capi- quately support.
tals or drop panels) were 10 inches thick, supported on The soil in the construction area was muddy due to
square columns spaced 24 feet on centers in both direc- heavy rains prior to the accident. This produced differ-
tions. Columns rested on concrete-filled pipe piles dri- ential settlement of adjacent legs of shoring towers (Fig-
ven to bedrock. Typical columns were 25 inches square
at the basement level, decreasing to 20 inches square be-
tween the second and fourth floors.
First and second floors were several weeks old at the
time of the accident, and forms and shores had been re-
moved. The third floor concrete was at least 20 days old;
forms had been removed. The slab was reshored to the
second floor, and was carrying the formwork for the
fourth floor. Concrete had been placed in the fourth
floor forms only a short while when most of the east
wing, an area about 72 x 144 feet, dropped all the way to
the cellar. The other three wings were little damaged ex-
cept where they adjoined the collapsed section.
Significantly, almost all of the columns remained
standing full height after the collapse (Figure 3). Top-sto-
ry column forms remained in place and very little rein-
forcement projected from the free-standing columns at Figure 4. Wet soil conditions, inadequate mudsills and shore
any floor level. Plans indicated 10 x 14-inch duct open- loads five times what the ground could adequately support
ings in the slab along two adjacent faces of some interi- led to shoring collapse.
 Figure 5. Spall
at bearing
surface on
precast beam
supporting
hollow core
slab.

Figure 7. Parapet section, weakened by cracking,

blew off in high wind.

Figure 6. Precast
parapet sections
cracked because of
restraint at welded
connections.

Figure 8. Cracking in stem of tee beam where weld to

supporting steel girder restrained movement.

ure 4). When settling occurred, it caused stress changes

in the entire framework and led to collapse, the investi-
gation concluded.

Cracking and failures of precast

concrete elements
The first instance of cracking and distress of precast
and/or prestressed concrete structures due to end re-
straint appeared on the U.S. scene some 30 years ago.
The same type of failure has occurred repeatedly in
many forms and shapes, but with one consistent result:
serious damage to concrete stru c t u re s, often causing
millions in financial losses. There have also been sever-
al cases of total collapse where the seriousness of the ini- Figure 9. Laboratory test of full-scale notched beam
tial distress was not recognized in time. Here, the crack- resulted in failure at load much below that anticipated by
ing developed further and increased in such magnitude designer.
to cause total loss of the shear resistance at the support-
ing ends. Why this constant recurrence and the failure to
given to alert the construction industry to the serious
heed repeated warnings?
dangers inherent in providing end restraint to precast el-
Partly because of legal restrictions imposed by some
ements. The damage is often magnified when this re-
clients to “bury” the facts in case of embarrassing cir-
straint is coupled with the introduction of notches and
cumstances of loss of taxpayer’s money, but more im-
brackets. A few case histories will be presented to fill the
portant is the fact that insufficient publicity has been
existing void of knowledge on this subject.
 Case 1: A complex in the New York area was built in the
late 1970s of hollow-core precast prestressed slabs bear-
ing on precast concrete walls and beams. At the typical
floors, spans at the bearing surfaces appeared as edge
loading occurred (Figure 5). At the lowe r- g a rage level
precast girders cracked as a result of restraint provided
by end welding plates. Tensile stresses developed at the
nonconfined edges. Edge spalling occurred here, too. On
the roof, precast parapet sections were welded at their
ends thereby causing restraint, limiting their movement,
and generating tensile stresses as a result of contractions
due to shrinkage and temperature. These tensile stress- Figure 10. Collapse of part of a school roof structure in
es exceeded the strength of the panels, which cracked at which precast prestressed joists were too tightly restrained
their ends (Figure 6). One parapet unit was so weakened to allow for the long-term shortening which occurs in such
that a heavy wind totally blew it off the roof (Figure 7). members. Also shown is a typical crack observed in double
tee joists in adjacent spans.
Case 2: A school structure in the East was constructed
in the late 1970s of precast double tee panels supported
on a structural steel frame. The tees were supported on
either the top or bottom flanges of the steel girders. Al-
ternate stems of the tees were welded at their ends to the
supporting girders. For economic reasons the construc-
tion stopped and the partially completed structure was
exposed to environmental effects for a great length of
time. As a result, many of the welded stems cracked, with
predominant cracking at the stems having reduced sec-
tions resulting from bottom notches or top flange block-
ing (Figure 8). Here again the welding at the ends re-
strained the panels from movement and rotation,
resulting in serious damage that had to be corrected by The design also required that each stem of each dou-
p re s s u re-injected epoxy supplemented by steel shear ble tee be welded at both ends at the steel-to-steel seat
plates. level, and a plate in the center of each end of the flange
was welded to an insert plate in the support beams.
Case 3: A hospital structure constructed in the early
Thus, the double tees were tightly restrained at each end.
1960s of precast concrete elements had beams notched
Hardly any supplementary reinforcement was provided
at each end, bearing on concrete brackets cast as part of
in the notched ends, and neither the amount nor loca-
the precast columns. Shortly after construction serious
tion would conform to current recommendations on
cracks developed in both the notched beams and the
connection design.
column brackets. The cracks were so severe that a full-
A state of California Advisory Bulletin (March 16, 1981)
scale notched beam was tested in a laboratory and failed
describes the problem: “A potentially hazardous condi-
at rather a lower load than expected by the design (Fig-
tion may exist in certain buildings which were con-
ure 9). The structure was repaired by adding steel “cradle
structed using precast pretensioned prestressed con-
plates” at a very high cost.
crete framing members. This condition may exist where
Case 4: Cracks developed in the stems of a precast roof inadequate provision was made to allow for the effects of
structure of a school built in the West in the late 1950s. In long-term shortening which occurs in such members.
1980, a section of the roof of the auditorium consisting of The result of this hazardous condition can be structural
approximately 18 prestressed concrete double tee joists, failure and collapse.”
(approximately 38 x 70 feet in area) fell from the center of Summary: After more than 30 years of misuse and
the roof to the floor below (Figure 10). The collapsed misunderstanding of the behavior of precast prestressed
portion of the roof framing consisted of factory fabricat- concrete elements, we believe the following should now
ed pretensioned prestressed lightweight concrete dou- be clear:
ble tee joists spanning 40 feet between prestressed con-
crete girders. The joists were notched to fit on girder • Brackets and notched beams could be designed and
ledges with tops of both elements at roughly the same el- constructed properly and eventually perform well in
evation. To meet seismic requirements for a roof di- s e rv i c e. These elements may be reinforced by either
aphragm, flange shear connectors were welded, typical- post-tensioning and inducing compression in the di-
ly every 6 feet. Similar connections were provided from rection of the expected tensile stresses or by placing
tee flanges to side walls. mild reinforcement to close tolerances at all surfaces
and reentrant corners. In any event, in order to avoid the in this article represent only the tip of a dangerous ice-
damage described above, the concrete elements should berg—one that is all too often concealed as litigation to
be allowed to move sufficiently to reduce the possibility assign liability proceeds. Facts that would be beneficial
of the development of tensile stresses and the resulting may be mothballed for years, while the same errors are
cracks. repeated. Although this learning process may be de-
layed, it must not be stopped. All members of the con-
• Most important, horizontal precast concrete elements struction team—designers, builders, and materials sup-
should not be welded at both ends, but rather allowed pliers—must become involved in learning from these
to move and rotate, so as to avoid considerable damage experiences.
by cracking. Rigidity for lateral loads such as wind and
earthquake may be provided by reinforcing bars grout-
ed or cast in concrete. Acknowledgement
The basic rules for preventing construction failures and the
• Elastomeric bearing pads should permit hori zo n t a l case of shear failure due to construction error were adapted
and vertical movement and rotations and should not from Dov Kaminetzky’s article, “Structural Failures and How
be placed directly at the edges of the bearing surfaces. to Prevent Them,” in the August 1976 issue of Civil Engi-
neering. These basic rules were originally formulated by the
To simplify the message: (1) Avoid brackets and notch- late Jacob Feld, founder of Feld, Kaminetzky & Cohen.
es where possible; (2) Do not weld both ends of precast
elements; and (3) Avoid bearing on unconfined edges.

Conclusions
Some of the general conclusions we have reached PUBLICATION #C810641
through many first-hand failure investigations are pre- Copyright © 1981, The Aberdeen Group
sented here in the “basic rules.” The few examples given All rights reserved