Professional Documents
Culture Documents
MK 11863
MK 11863
by
McGi11 University
_.. _----
@ Gregory John Kemp 1972
- i -
Gregory J. Kemp
ABSTRACT
The object of this investigation was to study the strength and
behaviour characteristics of simply supported, two way post-tensioned,
square slabs subjected to a uniformly distributed load. The uniform load
was applied by a large flexible air filled bag bearing against a stiff
reaction frame.
Three slabs were tested to destruction. The second slab was a
duplicate of the first slab and used for verification purposes. The third
slab had a lower volume of steel. than its predecessors but aIl other dimen-
sions were identical.
The slabs behaved elastically whilst the section remained uncracked.
Due to the fact that the slab warped slightly due to prestressing and creep
and also the fact that the corners were free to deflect no classical elastic
ar
analysis cou Id be used. The slabs were analysed by the finite element
method up to failure and the results were in very close agreement with those
found experimentally.
Yield line theory predicted loads of approximately 50% of the
actual failure loads. This discrepancy was a result of the yield line
theory ~eglecting the large amount of tensile membrane action which occurred.
- il. -
Grégoire J. Kemp
RESUME
ACKNOWLEDGEMENTS
TABLE OF CONTENTS
Page No.
ABSTRACT. i
RESUME • • ii
ACKNOWLEDGMENTS . iii
TABLE OF CONTENTS iv
LIST OF FIGURES . vi
LIST OF PLATES. . viii
NOTATION. • • . ix
1. INTRODUCTION. • • 1
1.1 . General . • • • • • • • . 1
1.2 Object and Scope . . . • • • 1
2. THEORETICAL CONSIDERATIONS . • • 3
2.1 General . • • • . 3
2.2 Limit Analysis . . . . . . 3
2.3 Yield Criteria for Slabs . 4
2.4 Unbonded Prestressing Tendons. 7
2.5 Membrane Action. • 8
2.6 Review of Previous Work on Prestressed Concrete Slabs. 9
3. EKPERIMENTAL SET UP . 10
3.1 General • . • . 10
3.2 Description of Slabs . • • • . 10
3.3 Control Specimens . . 11
3.4 Formwork and Casting . 13
3.5 Concrete. • . • • • 13
3.6 Prestressing • . 19
3.7 Slab Testing Frame . 22
3.8 Instrumentation. . 28
Page No.
4.4 Slab 3 .
....... ............. 46
5. ANALYS!S OF RESULTS 50
5.1 Effective Prestressing at Time of Testing . • • 50
5.2 Yield Line Theory Analysis • • • • . . . • . 50
5.3 Elastic Analysis . . •. . ... 52
5.4 Cracking Load . . • . • • . 56
5.5 Finite Element Analysis. 59
5.6 Strains and Discussion of Resulte .• 74
6. CONCLUSIONS. • • . . . . . . • . . . . . . . • . . . . • . . 81
- vi -
LIST OF FIGURES
5.7 Load v. Strain in Wires in Slab 1. 2nd and 3rd Load .;.
" '. ;64
Cycles
LIST OF PLATES
l Jacking Operation 24
II Jack, Pump and Manometer 24
III Bearing Plates and Anchorages 24
IV Test Frame and Bag 27
NOTATIONS
= mercury
l = Second Moment of Area of a Section
k = ratio of depth of compression zone at ultimate to effect-
ive depth
z = section modu1us
= transverse def1ectibn
}l = Poisson's ratio
INTRODUCTION
1.1 General
Prestressed concrete slabs are becoming increasingly common in civil
engineering structures. They are used in bridges, lift slab building floors,
pressure vessels and generally in areas with large spans and large loads.
Very little research has been carried out on prestressed concrete slabs and to
date most slabs are designed assuming that they behave elastically. Prestressed
concrete beams have received considerable research attention and are designed
by both elastic and ultimate load methods, but the behaviour of their slab
counterparts, particularly near failure, remains almost unknown. This is inspite
of the vast knowledge accumulated on reinforced'concrete slabs.
The fact that two way prestressed concrete slabs are far more difficult
to model mathematically than prestressed concrete beams or reinforced concrete
slabs is readily seen. Prestressed concrete beams can be considered two dimen-
sional and thus it is only necessary to determine moments, strains, stresses,
and crack patterns in one plane. The behaviour of reinforced concrete slabs
can be fairly easily modelled because of its almost perf~ctly elasto-plastic
moment-curvature characteristic. This has led to yield line theory and mathe-
matical programming techniques.
1) it is the type of load for which slabs are designed using code
specifications.
2) it is a load which causes the least complications in actual
loading and has no punching or shear consequences.
3) it lends itself weIl to analysis.
The first slab tested was square, post-tensioned with unbonded tendons.
A duplicate test was made to verify results of the first slab. In addition a
third slab was tested being identical to the first test except for an increase
in the spacing of the prestressing steel. Emphasis is put on a full understand-
ing of the behaviour of simply supported prestressed slabs in general and not
on the effects of different parameters.
- 3 -
2. THEORETICAL CONSIDERATIONS
2.1 General
Prestressed concrete structures are generally designed by assuming
that they behave elastically. Prestressed concrete slabs are designed similarly,
however, there is often difficulty in determining the elastic moments due to
complex boundary conditions. Nevertheless these problems can be overcome for
design purposes by using suitable approximations and assumptions. Examples are
the load balancing method and the use of code design specifications. The
ultimate strength of statically determinate beams can be determined and is
already adopted in most codes. Indeterminate structures are a more difficult
problem since their load capacity is not just a property of the cross-section
but also depends on their rotation ability and the geometry of the structure.
The same factors will also effect prestressed concrete slabs.
For reinforced concrete slabs this problem has been overcome by
assuming that the material behaves as a r~gid plastic material. There has been
a large amount of work done on reinforced concrete slabs and many of the
theories can be said to yield a satisfactory estimate of the ultimate load.
The overall procedure is called Limit Analysis and assumes a rigid-plastic
material.
increases very slowly with rotation thus giving a moment curvature diagram with
an ever increasing moment.
Using rigid plastic assumptions, it is not possible to obtain the
complete load-deflection characteristics of a slab but only the collapse load.
Limit analysis draws attention to the fact that even exact collapse loads are
not always possible and may only provide an upper or lower limit called
"bounds".
These upper and lower bound analyses are weIl documented (2), (3),
(4) and it will suffice to mention only bri~fly their salient points. To
determine an upper bound on the collapse load it is assumed that the slab
behaves as a mechanism and rotates either about a fracture line at yield
(called a yield line) or a boundary. The equilibrium of the surface between
these hinge lines is examined thus equating the slab load and the plastic
moment. This leads to the collapse load for that particular mechanism. In
the terminology of Limit Analysis this is called a "Kinematically Admissible
Velocity System." The theory does have limitations in that if the wrong
failure mechanism is chosen, too high a collapse load will be predicted. Also
the theory does not provide any information as to the distribution of moments
in the non-yielding segments of the slab.
On the other hand, lower bound solutions require the determination
of the entire moment field over the slab. If a pattern of bending and twisting
moments can be found such that the equilibrium conditions are satisfied and
the yield cr~terion is nowhere exceeded then this constitutes a "Statically
Admissible Stress Field" and is a lower bound. A lower bound solution can
never predict a load greater than the true ultimate load.
Thus • • • • • (1)
· • • . • (2)
M = A s f y (d - ~)
~
• • • . • (3)
= As f y d (1 - • • • • • (4)
A,
where p =~ , the steel percentage.
bd
I _ _-;;"~ C
~T
+y
My
t M
f~ permissable
strains
A
-M M +x
-x
C
-M
-y
Fig.2.2 Square Yield Criterion. Isotropic Reinforcement top and bottom.
m'
-~ M
x
-ID
still rectangular ego assume negative moment is reduced ta m and 'y' moment
reduced ta m' for positive moments. The yield criteria is as shown in
Fig. 2.3.
In Fig. 2.2 the point A is the condition when the moments along
mutually perpendicular axes reach the yield moment. Such is the case at the
centre of a square slab symmetrically loaded. Point C would refer to the
corner of the same slab providing negative reinforcement is provided. Point
A infers no twist and uniform moment in aIl directions whilst point C implies
twist equal ta the yield moment itself.
Square yield criterion also implies that it applies ta a yield line
not in the direction of the reinforcement. This is justified by assuming
across any fracture line not in line with the reinforcement axes that both
sets of bars must yield. Thus Mx = M and My = M, then for axes 'n'and 't'
likewise Mt = M.
in the steel can be distributed a10ng its 1ength. For this reason the steel
stress at the most critica1 section for bending moment is 1ess in an unbonded
section than a bonded section.
As an unbonded beam is 10aded there is 1ittle increase in steel
stress and cracking is not distributed very much. As a resu1t, for the same
moment an unbonded beam has a 1arger rotation at the critica1 section than a
bonded beam. Consequent1y the unbonded beam fai1s at the 10wer moment.
Un1ike a bonded beam the moment-curvature re1ationship (11) of an
unbonded member cannot beana1ytica11y predicted with any assurance since
there is no way of accurate1y knowing the stress in the steel at any particu1ar
10ad. This is further comp1icated by the fact that the stress will vary a10ng
the steel 1ength as a resu1t ·of frictional forces with the concrete. This
is significant at high deformations.
Before the member has cracked the steel stress remains constant
(5) but after cracking it increases at an increasing rate but the steel
stress at fai1ure is a1ways 1ess than maximum strength (10). The tendon stress
at fai1ure de pends on the steel percentage in the section and tests (5), (10)
have shown the increase in stress varies from 20 to 80 k.s.i. This is higher
than the ACI Building Code (ACI 318 - 63) value of 15 k.s.i. Even if it were
possible to estimate the variation of the steel stress between cracking 10ad
and u1timate 10ad the strain distribution over the section wou1d be unknown
(because the steel strain is not equa1 to the strain in the adjacent concrete)
and hence curvature cou1d not be ca1cu1ated.
for the slabs tested herein. It has been found that the relative increase
in strength over the yiêld 1ine solution is greatest for low steel percent-
ages wh en the membrane action is tensile (13).
3• EXPERIMENTAL SETUP
3.1 General
The decision variables in a prestressed concrete slab are:
It was decided to test the slabs under a uniform load using air
pressure contained within a light flexible bag. The edges were stmply supported
with the corners free to deflect. The size of the slab was determined from
the following considerations:
i) The mintmum size of prestressing steel and the associated
anchorages and jacks available.
ii) The pressure which cou Id be developed within the air bag.
iii) The size of the loading frame which could conveniently fabri-
cated in the laboratory.
iii) Draped cab les wou1d have increased the u1timate load which was
not wanted since the u1timate load per square foot with straight tendons
was many times higher than that which slabs are expected to resist in practice.
A1so the reaction frame may have deformed excessive1y if the u1timate load
was significant1y increased.
Slab 1 had the tendons spaced at 8 inch centres. This gave a
steel percentage of 0.75% and the value pf su was 1ess than 0.3 as required
-f-'-
by ACI 318-63. c
The steel running east-west was p1aced at the mid-depth and the
steel running north-south was p1aced a1ternative1y above and be10w the layer
at the mid-depth. This a110wed nine wires to be p1aced in each direction.
The wires in the north-south direction were symmetrica1 but one more wire
was p1aced on the tension side of the neutra1 axis than the compression side.
Refer to Fig. 3.1.
Slab 2 was identica1 to Slab 1 and was used for verification
purposes.
Slab 3 had the wires spaced at 12 inch centres and this resu1ted
in six wires in each direction. In the north-south direction a11 wires
were at the mid-depth of the concrete section. In the east-west,'direction
they were p1aced a1ternative1y above and be10w those in the north-south
direction but for symmetry four were p1aced on the tension side and two on
the compression side. Refer to Fig. 3.2.
•
•
•
•
•
00 •
.u
_Ill
00
•
•
~
.
--{ t- 2"
r-~------~------~------~~----~~.
~ __•__-:.~__•__~.____•__~.~__•__~.~__•__~I tensionload
under face
5 at 12" = 60"
00
•
o
\0
Il
•
: N
.-1
.u
III •
11"\
•
• - tension face
00 under load
-4 t- 2"
• • • • • •
Thus for each slab, there were twe1ve 6 in. x 3 in. and six 12 in.
x 6 in. cy1inders cast, as we11 as the three sma11 beams.
3.5 Concrete
Type III high ear1y strength Portland Cement manufactured by the
Canada Cement Co. was used for aIl slabs. The coarse aggregate consisted of
a type of crushed limestone and had a maximum size of 1/4 in. The fine
- 14 -
aggregate consisted of crushed silica sand. Two sizes were used in equal
quantities, these corresponded to the A.S.T.M. sieve sizes no. 24 and no. 40.
The mix proportions for·each slab were the same. The ratio of
cement: coarse aggregate; fine aggregate was 1: 1.61: 1.65. The water-
cement ratio was 0.50. The aggregate was thoroughly dry prior to mixing.
A representative stress-strain curve is shown for each slab in
Figs. 3.3, 3.4, 3.5. It was found that the envelope of the stress-strain
curves at the time of testing encompassed the stress-stnain curve at the time
of prestressing. Thus it was assumed that there was no difference in the
modulus of elasticity between the time at prestressing and the time at testing.
The strains on the 12 in. x 6 in. control cylinders were measured
using 20 mm. electrical resistance strain gauges. Two gauges were connected
in series in both vertical and horizontal directions on opposite sides of the
cylinder. The horizontal strains allowed the Poisson's ratio to be determined.
Poisson's ratio varied considerably for each cylinder.as has been reported
by many previous investigators. AlI tests indicated a constant ~ over the
first 80% of the ultimate load and increasingsharply at failure. This
constant value of p was always very close to 0.20. Table 3a lists the
measured physical properties of the concrete for each slab. For aIl slabs
the slump was 1.5 inches and the initial tangent modulus of elasticity was
6
5.0 x 10 p.s.i.
The tensile strength f~ was ob~ained by testing the small concrete
beams in an Instron testing machine under a symmetrical two point load. The
total load P was measured to the nearest pound weight. The associated
moment in the shear free middle zone is thus 1/2 P x 1/3i = P ~ /6 The
b d2
section modulus Z = ---6--- and thus the stress in the outer fibre is ~~
~
For the size of the beams tested herein the stress in the outer fibre equalled
3/2 P p.s.i. The above calculations assume that the beam behaves linearly
elastic until rupture. The author placed strain gauges on the top and bottom
faces of. the beams during testing and could detect no non-linearities.
AlI slabs and control specimens were cured under moist burlap for
seven days.
- 15 -
··
..-f
e1ement ana1ysis
finite
II)
·
p..
Ul
Ul
~
.E-c
Ul
STRAIN x 10 6
7000
6000
"'
5000
""..
fi)
"p.. 4000
11)
11)
~
~
11)
3000
2000
1000
STRAIN x 10 6
7000
6000
5000
~
.
.
fil
0..
CI)
CI)
4000
~
~
CI)
3000
2000
1000
STRAIN x 106
Table 3a
CONCRETE PROPERTIES
Age at Age at
Mean Std. Dev. Post Testing
fI
c f'c ft:. Tensioning (days)
3.6 Prestressing
The smallest commercially available plain wire prestressing system
was chosen. This was the Freyssinet mono-wire system using 0.276 inch (7 mm.)
diameter wires anchored by cone anchorages. The mono~wire jack allowed the
wire to pass through the centre of the piston. Each wire was stressed by
the jack and then anchored off by hammering in the radial friction type wedges.
See Plates I, II, III.
The mild steel bearing plates were 4 in. x 2 in. x 1/4 in. In order
to reduce the tensile strains under the bearing plates two 1/8 in. diameter
mild steel deformed bars were placed behind the plates. These were 1-1/2 inches
vertically apart and ran along the whole length of each edge. This also pre-
vented spalling of the concrete edges due to handling and yet because of its
size and position would have no effect on the load-deflection characteristics
of the slab.
The 0.276 inch diameter wire had a guaranteed ultimate strength of
235 k.s.i. ie. 14.1 kips and a proportional limit of 196 k.s.i. ie. Il.8 kips.
The 0.2% offset occurred at 224 k.s.i. ie. 13.4 kips. The manufacturers
figures were verified by testing three 15 inch long specimens in the Instron
testing machine. Two strain gauges were placed on each specimen and the
stress-strain curve found to correspond to the one given by the manufacturer.
The ultimate strength of each specimen was always ab ove the guaranteed value
and the average value of these tensile tests was 245 k.s.i. ie. 14.7 kips.
The final elongation was 5%. The stress-strain curve is shown in Fig. 3.6.
The manometer for the jack was calibrated especially for this size
of jack and had a scale graduated in 200 pound increments. In order to cal-
culate the coefficient of friction in the jack a wire was tested in a 22 ft.
long pre-tensioning bed. Four strain gauges were placed on the wire and the
strains were recorded for each one kip increment of load as shown on the
manometer. The coefficient of jack friction was found to be 3% which was
in exact agreement with that quoted by the manufacturer.
The complete prestressing operation consisted of the following
stages:
a) The wires were weIl greased and placed in the formwork together with the
bearing plates and the mild steel reinforcement at the edges. The wires
were supported on specially made concrete blocks. The concrete was poured
and allowed to cure before removing from the formwork.
- 20 -
U.T.S.--'
240
200
.
~
.
11) 160
~
tJl
tJl
~
~
E-t
tJl
120
80
40
b) Prior to post tensioning a11 strain gauges on both the steel and concrete
were read. For the wires in Slabs 1 and 2 which did not have strain
gauges attached the wires were pu11ed in order to obtain the residua1
frictiona1 resistance. The bond broke at an average 10ad of 1200 pounds
and the residua1 friction had an average value of 600 pounds. There was
1itt1e variation for the 26 wires tested.
c) The protruding wires and bearing plates were c1eaned and the anchorages
p1aced and secured at the non-jacking end. The anchorages were 100se1y
p1aced at the jacking end and the jack p1aced over the wire and anchorage
so that it reacted against the bearing plate. The jack was 10aded to
1400 pounds to draw the anchorage at the non-jacking end against its bear-
ing plate. The 10ad was removed to 600 pounds and the protruding piston
1ength measured. This gave a value eO' The jack was hand pumped unti1
the manometer read the desired 10ad and the piston 1ength was again
measured, el' The wedge in the anchorage at the jacking end was hammered
in and the 10ad in the jack removed to 600 pounds and the piston 1ength
again measured e2' By taking these piston measurements the wire e10ngation
cou1d be measured and used to verify the manometer readin~ and the strain-
gauge readings.
of residua1 friction. This amount of over stressing being ha1f the residua1
friction ie. 0.3 kips. Prior to transfer the force in the wire at the jack
was 12.1 kips and thus the force in the wire after transfer was 9.6 kips
ie. 67% of the U.T.S. This resu1ted in a concrete stress of 600 p.s.i. for
Slabs 1 and 2 ans 400 p.s.i. in Slab 3. The associated strain in the concrete
6
was 120 x 10- in. lin. for Slabs 1 and 2 and 80 in. lin. for Slab 3. The
associated 10ss in prestress force in the first wire stressed due to subsequent
stressing was 0.21 kips for Slabs 1 and 2. Thus the first wires were over
stressed by 0.2 kips. This is on1y 1.6% of the total app1ied 10ad. Slab 3
was a1so over stressed by this amount.
The first wires stressed will be further stressed by subsequent
stressing of wires perpendicu1ar to the wires under consideration as a resu1t
of the Poisson effect. Thus for Slabs 1 and 2 where the strain in the concrete
is 120 x 10- 6 the strain in the perpendicu1ar direction is p ttmes this
amount. For poisson's ratio of 0.2 this amounts to a strain of 24 x 10 -6
and hence the first wires wou1d be further stressed by an amount of 0.04 kips.
This represents 0.3% of the desired 10ad and was thus ignored. A1so it wou1d
have been beyond the range of the manometer sensitivity.
Upon complet ion of the stressing operation the wires protruding
beyond their anchorages were mechanica11y cut off and a11 strain gauges were
read. The slab was then 1ifted into the testing frame and positioned for
testing.
a) Prior to p1acing the slab the channel sections were removed and 1/8 in.
amber gum rubber strips p1aced on top of the box section in order to
cushion the slab as it was positioned.
b) The slab was then 1ifted in and positioned.
c) The channel sections were rep1aced and 3/8 in. clearance 1eft between
the slab and the 1 in. diameter support bar. A 1/8 in. thick strip of
neoprene rubber was inserted between the slab and the support. The
remaining gap was required to a110w for the lifting of the corners.
d) The channel section was then accurate1y 1eve11ed, with the use of a
surveyors 1eve1 and rod by adjusting the nuts ab ove and be10w the channel.
This ensures that the slab bears against supports which border one plane.
The channel section was then tight1y bo1ted.
e) In order to prevent the wire or anchorages from shooting out from the
frame in the event of a wire breaking during testing a 1/8 in. thick steel
plate was bo1ted to the f1ange of the channel section as shown in Plate XIV.
The pressure on the slab was measured using a U tube manometer with
mercury. The mercury 1eve1 cou1d be read to 0.02 in. accurate1y. This
represents a pressure of 1.4 1b./sq. ft. The air was supp1ied by bott1ed
compressed air. The inf10w and outf10w of the air was contro11ed by a standard
y valve regu1ator. Refer to Fig. 3.8.
Plates IV and V show the reaction frame prior to p1acing the
channel supports.
24 -
Plate I.
Jacking Operation
Plate II.
Jack, Pump, and
Manometer
Plate III
Bearing Plates
and Anchorages
- 25 -
slab
25 lb.
PLAN
1 rubber bag
7'-0" x 10" x
5" box _ _ _... r3/4" p1ywood base
JI)-~t----------------~
L---1 . L-.J 8' -0" x 6" x' 2" x 8.2 lb. Channel at
~'~'~'~~~~~~~~~~~~'~".==.=="="='='=='="='="=~'''~='.~.'.~.~~~~~
r- .r
X
10" centre '11
regul.ators
valve
Mercury
manometer
compressed
air
N'
0\
- 27 -
3.8 Instrumentation
A symmetrical square slab possesses the property that each octant is
identical. However, for the slabs tested, it was impossible to have the pre-
stressing steel at the same height in both directions and thus they can only be
considered to have identical quadrants. Hence only one quadrant of the slab
was elaborately instrumentated. A check for symmetry was made by taking some
deflections and strains in other quadrants.
The deflections of aIl slabs were measured using 0.001 in. dial
indicators mounted on a steel frame supported by the channel sections. This
gave the deflections of the slab relative to the supports. The deflections
were measured at positions shown in Figs. 3.9, 3.11, 3.12 for Slabs 1, 2, and 3
respectively. In addition it was desired to know the differential deflection
of the supports and so dial gauges were placed along one edge of one channel
section for Slab 1. The deflections were measured relative to the floor but
only the difference in deflections along the edge was of importance. The
positions of these dial gauges is also shown in Fig. 3.9.
Electrical resistance strain gauges were used in Slabs 1 and 3
on both the steel and concrete. The gauges placed on the prestressing wires
were of special significance since to the author's knowledge straingauges
have notbeenplaced on post tensioning wires which use no ducting. The concrete
will tend to scrape the concrete off the wires in such a system because of
there being only a very thin layer of grease between the steel and the concrete.
The problem was overcome by first placing aIl gauges at the non-jacking end
of the wires where relative movement between steel and concrete is a minimum.
Small 5 mm. long gauges were used. The gauges were glued to the surface, water-
proofed and then their wires bent back at their middle. The strain gauges were
then placed between sheets of teflon with their ends left protruding. These
ends were soldered to the electrical lead wires. The whole lot was taped with
a plastic backed tape and coated with a special sponge rubber compound. The
whole lot was then re-taped. Thus any movement of up to 3/8 in. between the
steel and concrete'could occur without failure of the gauge. The principle
being that the strain gauge wires were free to slide between the teflon sheets.
In Slab 1 only four out of the ten wires behaved satisfactorily but this was
- 29 -
improved for Slab 3 where nine out of the twe1ve gauges were satisfactory.
The most 1ike1y causes of trouble are:
channel support
-- -,
......--support
-..• ."
w
S iE ~N
+-r +-1 . ..
IIIII~~ Prestressing
---- ---- ---- wires
LI -1-1-+ +-1
,I-t at 5~"
LI~ lB%"
~
t-I-trl-i 1"
S--+j-- E
N
18~"
, ..
edge of slab
101--- support
18~"
1"
2"
74"
- - - --,
9~"!' \ 18\"
9~ .. 1 ~ 2"
o 0./.;
9~" 1 ~
6. a
--~~~!- ~~'~ --
1
1 1"
1
~-------
1 1 l , 1
1-'-1-'·-1
, 1-1 ~I-I-I
1-1-1-1-1-1--
1-1-1-1-1 1
1"
4.1 General
The density of the concrete was found by weighing the control cylinders
and it was estÜRated that the unit weight of the concrete with the stressing
steel was 145 pounds per cubic foot. This corresponds to 0.34 inches of mercury
(24 p.s.f.) for a two inch thick slab. Thus a load reading of 0.34 on the mercury
manometer meant that the slab was floating freely on the air-filled bag. A
load of 0.68 inches of mercury was equivalent to the slab being loaded by its
own weight. This condition was used as datum for aIl measurements of deflections
and strain. A pressure of 1 in. of mercury corresponds to 70.7 p.s.f or
346 kg./sq.m.
The test program consisted of three distinct loading cycles. In
each cycle the load was incrementally brought up to a certain level and then
completely removed incrementally. The first cycle tested the slab before
cracking; the second cycle checked the first cycle and the slab was then brought
pa st the cracking load; cycle three checked the second cycle and brought the
slab to failure.
4.2 Slab 1
Prior to testing Slab 1 it was noticed that some warping had occured.
This was thought to be a result of the eccentric prestressing in one direction
~ the fact that creep had taken place in the 22 days between post tensioning
and testing. The slab profile was determined by means of a surveyor's level
and rod. It found that a fIat plane could be drawn between four corners.
Fig. 4.1 shows the deflection of the slab relative to this plane.
The first load cycle took the load up to 6 inches of mercury (425 p.s.f.)
in half inch (35.4 p.s.f~ increments and unloaded in 1 inch (70.7 p.s.f.)
increments. This upper level was below the cracking load as calculated in the
next chapter. No cracking was observed at this load level nor was any indicated
by the strain gauges on the tension face nor from the load-deflection curves.
The second cycle took the load to a mercury level of 12 inches (850 p.s.f.
in one inch (70.7 p.s.f.) increments. Cracking was visually observed at 10 inches
of mercury (707 p.s.f.) but indicated by the strain and dial gauges at 8 inches
(565 p.s.f.).
The third loading cycle loaded the slab to failure in one inch of
mercury increments. Strain and dial gauges were last read at a load level
0.02 - 37 -
0.01
~~_____________________________________
~ ________~edge
~North
0.00
0.03
0.02
0.01
~ ______________________________________ ~L/4
0.00
0.04
0.03
0.02
0.01
L/2
o.oo-----------------------------------------J
0.03
0.02
0.01
3L/4
o.oo--------------------------------------------------J
0.01 ~
~--------------
/ South
0.00 '--_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _~_..::::~ edge
J 6.16 ft.
"'
N
L/4
W 6.16ft . L/2 E
3L/4
~1Il"
of 23 inches (1625 p.s.f.) and failure occurred Just prior to a load of 24 inches
of mercury (1700 p.s.f.) being reached. At high load levels much creep occurred
and thus once the load level had been reached gauges had to be read quickly
since the pressure would drop off due to an increase in deflections. The
deflections at these high loads were very large and it was quite obvious that
the slab was being subjected to tensile membrane action.
The visual crack pattern started at a load of 10 inches of mercury
(707 p.s.f.). The cracks originated near the centre and roughly followed the
diagonal but usually fanned out at the corners. With subsequent increases in
load the cracks along the diagonal widened and many new ones formed. Thus the
slab had a wide zone of cracking emanating near the centre and going to the
corners. The cracks fanned out to the edges rather than pass exactly through
the corner.
At loads around 18 in. of Hg. (1270 p.s.f.) deflections became large
and cracks parallel to the boundary supports formed about 16 ins. frorn the centre.
The deflected shape appeared conical and 'the cracks opened to over 1/4 inch
in parts. The largest crack"-- -; widths occurred for those cracks parallel to the
bounda'ries and these were caused as a result of tensile membrane action. The
slab exhibited a jig-saw puzzle type of cracked pattern with distinct differences
in the level of concrete sections adjacent to a crack. In other words sorne
parts of the concrete looked as though they were being punched out from the
adjoining concrete sections. Figs. 4.2a to 4.2c give the sequence of crack
patterns.
The final rupture was explosive in nature and occurred through com-
pression failure of the concrete. The failure zone was not Just a narrow line
but rather a yield region consisting of a wide area of crushed concrete.
The yield zone occurred along the diagonals and along a central crack paraI leI
to the boundary as shown in Fig. 4.2d. The square yield pattern at the interior
of the slab covered a finite width but was centred over the second prestressing
tendon from the centre. That is, the yield line was approximateay 16 ins.
from the centre and thus the square yield pattern had a side length of approxi-
'o. mately 32 ins.
This square type of yield pattern has been obtained in most investiga-
tions where tensile membrane action is present. Taylor et al (13) found this
central square yield pattern in aIl of their tests on simply supported, square
reinforced concrete slabs. They also noted that near failure the deflections
- 39 -
(a) (b)
( c) (d)
r '
Fig.4.2 Schematic Representation of Crack Patterns of Slab 1.
- 40 -
were large and tensile membrane action was present. Final failure also occurred
by compression of the concrete along the diagonals even though it did not occur
explosively.
Plates VIII to XII show both faces of the slab after failure.
After the test the following points were observed about the destroyed
slab.
a) The explosive type rupture caused myriads of new cracks to open. Along·
the failure lines there were so many cracks that much of the concrete in
this region fell away from the slab upon lifting out of the test frame.
b) AlI corners were cracked at right angles to their diagonals as seen in
Plate X.
c) Across the failure lines the prestressing wires had been permanently bent.
Thus the sudden rupture caused yielding of the wires in this area.
d) The width of the failure lines on the compression face was also wide.
e) The fact that the concrete could be removed along the failure lines permitted
the slab thickness to be measured in these areas. It was found that the
slab, in a11 places measured, was 2 ins. to within :!" 1/64 in. It was ob-
served, however, that the wires in some areas had floated up during casting
giving an increase in the effective depth of the slab. As far as cou Id be
determined this increase in effective depth was no more than 1/16 in.
f) The thin layer of grease on the wires had not deteriorated the physical
appearance of the surrounding concrete.
g) After the concrete had been removed from around the strain gauges it was
found that aIl gauges were in good condition, none having been scraped from
the tensioning wire. The fact that sorne gauges could not be used during
the test seemed to be due to
i) Faulty soldering of the lead wires.
ii) Lead wires broken during casting.
iii) Strain gauge wites shorting during stressing.
h) Whilst the interior of the slab suffered appreciable permanent deformation
the edgesremained fairly straight.
i) The central square shape crack pattern did not form simultaneously. Each
side formed at a slightly different load. The actual rupture line occurred
on the crack:,;; Une which was last to forme
j) The anchorages on the prestressing wires were no longer flush against their
respective bearing plates indicating that there was no load in the wires
- 41 -
Plate VIII
Slab 1 after
Fai1ure
Plate IX
Tension Face
Slab 1 after
Failure
Plate X
Corner of Slab 1
after Failure
(Top Face)
- 42 -
Plate XI
Bottom Face of Slab 1
after Failure
Plate XII
Crushing of Compression
Face of Slab 1
- ~, -
after failure.
k) None of the wires broke during testing.
4.3 Slab 2
Slab 2 was manufactured to be the same as Slab 1 for verification
purposes. The slab was tested after 9 days compared with 63 days for Slab 1.
Both the compressive and tensile strength of Slab 2 concrete was less than that
in Slab 1, as shown in Table 3a. It was considered that verification could
be provided by a comparison of the load-deflection curves and by comparison of
ultimate load. Consequently Slab 2 was only instrumented with 17 dial gauges.
The loading cycles and increments were exactly the same as in Slab 1.
The behaviour of Slab 2 was almost identical to Slab 1. Visual cracks
commenced at the same load of 10 ins. Hg. (707 p.s.f.). With increase in load
a band of cracks formed approxima~ely parallel with the diagonals. Further
load resulted in large deflections and the formation of cracks parallel to the
boundary. This square pattern formed in the same place as Slab 1 and at a load
of 17 ins. Hg. (1202 p.s.f.) compared with 18 ins. Hg. (1272 p.s.f.) in Slab 1.
The square shaped crack pattern was even more obvious than in Slab 1 as seen
in Plates XIII to XVII.
Prior to cracking it was quite common for cracks to cross over other
cracks indicating a re-distribution in the direction of the principal stresses.
Towards failure the cracks were again very wide (ie. over 1/4 in.), especially
the cracks parallel to the boundary.
Failure occurred explosively and under a constant load of 22 ins. Hg.
(1560 p.s.f.). Slab 1 failed at less than 24 ins. Hg., estimated at 23.7 ins. Hg.
(1675·p.s.f.\., Thus Slab 1 and 2 varied in ultimate load by 7%. The discrepancy
was thought to be mainly due to a small difference in the effective depth.
Assuming at failure that the ultimate capacity is proportional'to the square
of the effective depth then a difference in effective depths of 3·1/2% or
1/30 in. would cause this discrepancy. This shows the extreme care which must
be taken in placing steel in such shallow sections. Another important reason
for the difference in ultimate capacities was the difference in concrete strengths.
After failure the same observations were made as in the previous section
except that two of the edges no longer remained straight but bowed in at the,
middle. This was a result of the large central deflection. The slab had large
permanent deflections but when the slab was pushed back into a fIat plane the
edges also straightened out considerably.
- 44 -
Plate XIII
Slab 2 after
Failure (Top Face)
Plate XIV
Slab 2 after
Failure (Top Face)
Plate XV
Slab 2 Crushed
Concrete Removed
- 45 -
Plate XVI
Compression Face of
Slab 2
Plate XVII
Detail of
Cru shed Area
on Compression
Face of Slab 2
- 46 -
4.4 Slab 3
The first cycle of load went to 5 in. Hg. (354 p.s.f.) in 1/2 in. Hg.
increments and decreased to the dead load (01.68 ins. Hg.) in 1/2 in. Hg. increments.
No cracking occurred during this cycle.
The second cycle brought the load to 10 ins. Hg. (707 p.s.f.) in 1 in.
-Hg. (70.7 p.s.f.' increments and was removed in the same increments. The first
visual cracking occurred at a load of 8 ins. Hg. (565 p.s.f.). Cracking along
the diagonals did not originate from the centre of the slab but rather at the
corners of a 12 inch square central crack pattern as shown in Fig. 4.3a. This
square crack pattern formed between loads of 8 and 10 ins. Hg.
The third cycle loaded the slab in 1 in. Hg. increments until the slab
failed explosively at a load just less than 20 ins. Hg. (1414 p.s.f.). The slab
behaviour was similar to the previous ones. At a load of 15 ins. Hg. (1060 p.s.f.)
two sides of a square formed over the second wires from the centre ie. 18 - 20 ins.
from the centre. The other two sides of the square crack pattern did not forro
clearly as seen in Plate XVIII. The failure lines formed along parts of the
diagonals and on the two cracks parallel to the boundaries and 18 inches from the
centre of the slab as seen in Fig. 4.3c and Plates XVIII to XX.
In comparison to the previous slabs this slab did not crack so exten-
sively on the tension side but the cracks were wider, especially the ones on the
central 12 inch square which measured over 3/8 in. in width. On the compression
face there were many more cracks than in Slabs 1 and 2. Large segments of the
concrete had completely crushed along the failure lines and fell away from the
slab as it was lifted from the test frame as seen in Plate XX. After failure
of the slab similar observations were made as for the previous slabs. The most
important being the following:
a) The slab thickness was 2 inches at aIl the places that could be measured in
the interior of the slab.
b) The distance from the compression surface to the centroid of the steel was
often 1/16 in. to 1/10 in. greater than intended. This was thought to be due
to the steel floating upwards slightly because of vibrating during the pour-
ing of the slab.
c) Failure lines went directly to the corners. AlI corners had cracked perpendi-
cularly to the diagonals.
d) The edges had bowed in at the middle noticeably due to the large vertical
deflections.
- 47; -
e) Wires were permanently bent and most had no load in them after failure.
f) All strain gauges were in good condition after the test.
- 48 -
Plate XVIII
Slab 3 after
Failure
(Top Face)
Plate XIX
Slab 3 after
Failure
(Top Face)
Plate XX
Slab 3 after
Failure
(Bottom Face)
- 50 -
5. ANALYSIS OF RESULTS
the value of fsu is not known with any certainty, un1ess experimenta11y measured.
- SI -
24Mu1t
qu1t'= i2
For L = 74 in. = 6.16 ft. the u1tfmàte load for each slab is:
Slab 1: = 1,300 x 24
6.16 2
= 820 Ib./sq.ft.
It shou1d be noted that even when the dead and live load are removed the manometer
will still record the dead weight of the slab because the slab has been loaded
from underneath. Therefore, with the dead load effect the slab shou1d fai1 at
a manometer reading of 11.6 + 0.34 = 12.0 inches of mercury (844 p.s.f.).
The slab actua11y fai1ed at a reading of 24 inches Hg.
Therefore with dead weight, fai1ure reading shou1d have been 12.0 in.Hg. com-
pared to an actua1 fai1ure pressure of 22 in.Hg. (1560 p.s.f.).
Therefore, with dead weight effect fai1ure reading shou1d have been 8.7 in.Hg.
rather than the actual fai1ure 10ad of 20 in.Hg.
Thus, it is seen that the conventiona1 yie1d 1ine theory (an upper
bound) predicted 10ads which were around 50% of the actua1 fai1ure 10ad. The
difference being due to three important factors:
v 4 w = q/D • . . . . 5. 1
where W = transverse
def1ections
3
D = Et /12(1 - p2) = plate stiffness per unit 1ength.
A simp1y supported plate has zero def1ection and zero curvature at the supports.
A uniform 10ad q can he put into a Fourrier series form for the purpose of
ana1ysis.
- 53 -
. . . . . 5.2
Substituting into equation 5.1 for the centre of the plate where x = L/2,
y= L/2 gives
. . • • . 5.3
The summation was computer ca1cu1ated for m and n up to the 15th order and was
found to be equa1 to 0.244.
Therefore, central def1ection W- = 169 L4 x 0.244
DTT 6
0.00406 ~ . . . • . 5.4
D
6 lb. 2/. .
D = 3.47 x 10 .1n. 1n.
The re1ationship between central def1ection and load is shown in F'ig. 5.1
together with the measured ;Load-def1ection curve for Slab 1 (load cycle 1)
and a1so the finite e1ement curve for Slab 1.
The large discrepancy is attributed main1y to the fact that the slab
was not initia11y f1at. As a resu1t of this warping the slab was actua11y sup-
ported on four corner points and thus had a far sma11er stiffness unti1 the edges
came into contact with the supports. From the edge def1ection characteristics
and a1so from a finite e1ement program of a slab supported on four corner points
it was seen that the edges of the slab came fu11y into contact with the supports
at a live load of 2 in.Hg. (141 p.s.f.). Consequent1yonce the edges of the slab
came fu11y into contact with the supports the stiffness increased as seen by
the measured load-def1ection curve in Fig. 5.1. Another major reason for the
discrepancy being the fact that corners were free in the tests and this a1so
gave a lower stiffness than the e1astic mode1.
)C
r-
6.0
I
5.01--
/ 2
,,~ ~lastic theory
/ /~ x
"
~ 4.0
=
..-1
~
3 3.0
1
./ -/''/ ~
/ /_./' . /
2.0t-- . Je . . . . . Finite Element Resu1ts
The greatest bending moments occur at the centre of the slab and
by c1assica1 e1astic theory are re1ated to the curvatures at the centre by
) . 5.7
Substitution into equations 5.1 and 5.2 for the centre of the
~z..w ~2.w
S 1ab where ~ =~ yie 1ds
0;)(:. "'~
m+n-2
. .lo.(m==--2_+:.......!::p-:....:n:...2-::-)...:Io(_-..=.1)=-_-=-2_ _ . • • . 5.8
m n ( m2 + n 2 ) 2
Mx = My = 16q L 2 x O. 269
rr 4
= 0.0442 qL 2
-6
= 28.8 x 10 q • • • • • 5.11
Mc = (f p + f~) l . • • . • 5.12
y
The load causing the cracking moment is obtained from equation 5.9.
)t
/
6.0 1
/
. elastic theory ~;'
5.0 /
" tension face
;'
bO 4.0
=s::
..-l
~ 3.0
r- .1
I
YY· compression face
~ ,
1 ././
)1;
~
/
2.0
1 /
STRAIN x 10 6 VI
.....
Fig.5.2 Slab 1, Load v. ~rincipal Strain at Centre lst Load Cycle
6.0
5.0
elast:l.c theory
tens:l.on
. face Il
1
....= 4.0
~
o~ 1 . / / . . - compression face
VI
00
STRAIN x 106
Fig.5.3. Slab 3, Load v. Principal Strains at Centre, lst Load Cycle
- 59 -
Slab 3: q
c
= 6.6 inches Hg. (467 p.s.f.)
10.0
~ 6.0
4.0
)C . . . Finite Element Results
2.0
DEFLECTION in. 0\
o
Fig.5.4 Slab 1, Load v. Central Def1ection, 2nd Loading Cycle
Fig.S.S Slab 2. Load v. Central Deflection. lst. and 2nd. Loading Cycle.
/-
/"
12.0 ~~ 3rd loading cycle
,/
10.0 ~
.. - X
bO
.
::t: 8 0
~
s::
~
3 . 6 0
40
2.0
8.0
.
bD
::t: /
,.,
s::
. ;1"
~ 6.0 /
.~ /
0
,..:: /
4.0 /
/
1 Jt . . . Finite Element Results .
2.0
0\
N
CENTRAL DEFLECTION in.
Fig.5.6 Slab 3, Load v. Central Deflection. lst. and 2nd. Loading Cycle
- 63 -
supported a10ng a11 four edges. The corners were a110wed to def1ect.
By examining the def1ections a10ng the boundary it was possible to estimate
the point at which the e1ement was no longer in contact with the edge
because of corner lift. Refer Fig. 5.10. This point was estimated to be
8 inches from the corner. Thus in the ana1ysis, after a load 'of 2 in.Hg.
(141 p.s.f.) the slab was supported a10ng a11 of the edges except for a
free hanging 1ength of 8 inches from the corner.
5. Fai1ure of the slab occurred when a11 e1ements a10ng the
diagonal had comp1ete1y crushed on the compression side and cracked on
the tension face.
20.0
.
bO
16.0
::r:
=
..-1
12.0
@
. ...:.
1
4.0
41
--1,.-- ...
INCREMENTAI. STRAINS x 10 6 0\
~
Fig.5.7 Load v. Strain in Wires in Slab 1. 2nd and 3rd I,oading Cycles
Fig.5.8 Load v. Strain in Wires in Slab 3. 2nd and 3rd Lo~ding Cycle
24 0
20.0
.
tlO
16.0
::r:
.
c::
o,.f
~
0
12:0
~
8.0 1
2
3 76"
4
4.0 5
6
7 8 9- 10 Il 12
C'\
\J1
INCREMENTAL STRAINS x 10 6
Fig.5.9 Load v. strain in Wires in slab 3. 2nd and 3rd Loading Cycle.
20.0
16.0
bD
.
~ 12 ..0
s::
1. / / / /. ~ ~osition of Wire in slab
....
~.8.0[ A ~ fiJ//1~
1
2
3
4
5
(;
4.01- IJ/ 1 Il . \1 / J , / \1 / • •
7 8 9 10 Il 12
400 800 1200 1600 2000 2400 2800 3200 3600 4000 4400
INCREMENTAL ST RAINS x 10 6 0\
0\
0.4
1 ~.12 in. Hg.
•
0.3 ~ ~ 6 in. Hg.
=
-..1
CIl
z0 0.2
t-I
E-4
e,)
~
r- // ~ 3 in. Hg.
rz..
r:J
I=l 0.1
~
ct.
-0.1
0\
Fig.S.10 Slab 1, Def1ections parallel to, and 2 inches from, the Edge Supports.
"
Fig.5.ll Load v. Central Deflections of Slabs 1 and 2. 3rd Loading Cycle.
24.0
20.0
-
16.0
bD
.
==
=
opof
~
12.0
<
8 ~~
•• • Slab 1
• • • Slab 2
8.0 1- U" ". • Finite Element Results
4.0
0\
0)
3.0
2.S
z
S
~
u
~ 1.5
~
~
1.0
0.5
6 in. Hg.
0\
\0
3.0
~
)t • • • Finite Element Results
'T 20 in. HIZ. --............
....=
~
.~2.0
~
~
12 in. Hg.
1.0,"-
8 in. Hg. 'C' ~~"'- ec!ge
If
Ji' ~
"'"
5 10 15 20 25 30 35
.....
DISTANCE FROM CENTRE in. 0
centre
Fig.5.l4 5lab 3. Load v. Central Deflection. 3rd Load Cycle.
x·
"
16.0
.
tIO
=t:
.
oS
A
~
...:1 8.0
'-
3.5 a..--_ _
3.0
2.5
~
H
E-I
tJ
~
~
~ -1.5
8 in. Hg._
6 in. Hg.
....a
N
4.0
3.0
c::
~
~. • . Finite Element Results
al
Z
0
H
E-I
CJ 2.0
~
~
~
Q
1.0
in. Hg.
in. Hg.
5 10 15 20 25 30 35
"w
DISTANCE FROM CENTRE in.
centre
- 74 -
type fai1ure a10ng the diagona1s towards the centre. The cDushing did not
continue to the centre but stopped at the square crack pattern around the
centre (which had been caused by tensi1e membrane action). Fina11y the
concrete fai1ed'.in compression a10ng one or more of the sides of this
square crack pattern resu1ting in a mechanism and complete destruction of
the slab.
The magnitude of the strains of Slab 1 near fai1ure were about
10% 1arger than those recorded by the uniaxia1 compression test at u1timate.
For Slab 3 the recorded strains were 50% 1arger. This was to be expected
since the compressive tests on the concrete cy1inders cou1d not measure
strains past the u1timate load. Thus no descending portion of the stress-
strain curve cou1d be found. The maximum strains recorded were of the
same order of magnitude as those connnonly pub1ished for uniaxia1 compression
ie. u1timate strains around 0.003. Thus the biaxia1 state of stress did
not significant1y affect the u1timate strain capacity of the slab. However,
the stresses at these 1eve1s of strain cou1d have been significant1y ine
creased due to the biaxia1 state of compression.
The synnnetry of the loading system was checked by p1acing def,lec-
tion gauges a10ng the diagonal on both sidesoof the centre as we11 as
a10ng two sides of the centre 1ines. A11 slabs showed very good synnnetry
espeêiàlly.' âfter ·loads of 4 in. Hg. (283 p. s. f .). Prior to t,his load the
warping a10ng the edges of the slab caused def1ections to be unsynnnetrical.
llast a load of 6 in. Hg. (424 p.s.f.) the synnnetry of def1ection was within
5% and this value became sma11er with increasing load 1eve1s.
( Strains on Top Surface
300
\0
L ~ 6 in. Hg.
....0
~
~ 200
~
~
100
25 30
1400
1200
1000
\0
0
.-1
zH 800
~
E-I
tJ)
600
12 il\ Hg.
400
Hg.
Hg.
200
corner
....,
....,
40' 45
3000
2500
\0
2000
0
~
Z
H
~ 1500
~
1000
500
.....
co
5 15 20 25 30 35 40 45 50
400
300
\D
0
.....
~
z
1-1 200
~ --.......... 5.0 _in. Hg.
E-4
CIl
100 .
2.5 in. iIg.
---- corner
5 10 15 20 25 30 35 40 45 50
3500
3000
1 / 19 in. Hg.
2500
1 /
\0
0
.-4 ".
~
z 2000
H
;1
E-I
CI)
1500
1 // ~
15 in. Hg.
1000
r- I/ /
500 E ~>
~ 12 in Hg.
5 10 15 20 25 30 35 40 45 50
00
o
centre DISTANCE ALONG DIAGONAL in.
- 81 -
6. CONCLUSIONS
REFERENCES
2. Wood, R.H, "Plastic and E1astic Design of Slabs and Plates" 'rhames
and Hudson, London 1961.
14. Magura, D.D., Sozen, M.A., and Siess, C.P. "A Study of Stress Relaxa-
tion in "Prestressing Reinforcement" Journal P.C. I. April 1964 Vol. 9
No.2.