Академический Документы
Профессиональный Документы
Культура Документы
design of slabs
G. S. T. ARMERt
where the moments M,, M y , and twist M,,,, follow Timoshenko's nomen-
clature,ll and p is the distributed load intensity at the point.
4. Normally the stress fieldis complicated.WithaJohansen-type slab
it is hardly ever possible to put M,, = 0, except in very special cases2 How-
ever, the essential feature is that Hillerborg's method is an attempt to design
a slab with variable reinforcement, and not to analyse a slab with given uniform
reinforcement.The intention is to makethe stress fieldcoincidewith the
field of resistance moments due to the reinforcement. That being so, Hiller-
borg can, for the first time, deliberately makeM,, = 0 so as tomake the direc-
tions of reinforcement (assumed to be placed along theX - and y-axes) coincide
withthe principal moment directions. Evidently equation (1) can thenbe
superseded by equations representing twistless beam-strip action:
a=M_
,
_a$ -- -ap . . . . .
and
where 01 is the proportion of load taken in the strips in the x-direction, and
(1 -a) in the y-direction. Thus the load may be divided between the strips,
or, as more often happens in strip theory, the value of a is taken as either
0 or 1. When a = O all the load isdispersed by strips in the y-direction;
when a = 1 the x-strips take all the load. Also the value of P could change
throughout the slab without affecting the validity. Point loads are treated as
local concentrations of p .
5. Lines of stress discontinuity are then introduced, as shown in Fig. 1 for
a rectangular slab supported on edge-beams, the intention being to signify
sudden changes in the direction of load dispersion SO that the strips become
loaded as shown and give rise to the diagrammatic representation of beam
loads shown in Fig. 2. It has always been an attractive feature of the strip
methodthat the beam loads are knowninintensity and in distribution.
Indeed, they must be known for any lower-bound solution, but by contrast
they are not provided by yield-line theory.*
286
THETHEORY O F THE S T R I P METHOD F O R DESIGN OF S L A B S
c
Y
Band
width
$9
Fig. 3. Unsymmetricalslab Fig. 4. Simply supported square slab
6. In Fig. 1, in regions 1 and 2 the value of CL is 1;in regions 3-7 the value
of a is zero.Beyondinsisting thatthe accumulated shearforceineach
imaginary strip is taken care of when theload dispersionchangessign,
Hillerborg does not place any other restrictions on the discontinuity lines.
Thus, although discontinuous stressfields are allowed'sainlimitanalysis,
Hillerborg does not examine rigorouslythe rules for permissible discontinuities,
and this feature will be examined later. For a symmetrical slab the bending
moment in strip a-a (Fig. 1) is uniform over the centre portion, but it is not
uniform in strip a-a of Fig. 3. This shows that the centre length of strip a-a
must be reinforced even though it carries no load locally-an important point
not emphasized in the literature. It should be noted at this stage that the
discontinuity lines are quite arbitrary, provided that continuity of shear is
observed in strips. Line cc in Fig. 1 is not a discontinuity line, it is a line of
zero shear. The load-dispersion lines of Fig. 2 have therefore nothing to do
with imaginary yield lines. It is curious that such load-dispersion diagrams,
often quoted in codes of practice in conjunction with either yield-line theory
or elastic analysis, are not correctin either case but they are compatible only
with strip theory.
3 287
WOODAND ARMER
Reinforcement in strips with varying moments
7. The size of regions 1 and 2 (Fig. 1) affects the amount of reinforcement
but the choice is not critical. With a simply supported square slab (Fig. 4)
the discontinuity lines may be the diagonal lines themselves, although other
systems are possible. It follows that the maximum moment in each strip is
a variable quantity, whereas in Fig. 1, region 7, all the short strips had a con-
stant maximum moment ofpla/8. In region 7, therefore, a band of reinforce-
ment can be stopped off at any convenient length, easily evaluated, as in an
ordinary beam. This is not so in Fig. 4 where the moments are continuously
variable. If triangular load regions are used, Hillerborg considered it
reasonable to place the reinforcement in uniform bands where the average
maximum momentfor strips within that band was taken as thedesign moment.
This decision was taken in the belief that the method was generally basedon a
safe lower-bound solution with a reserve of strength. Moreover, since one
band width is not likely to sufiice for the whole slab, it is necessary to study
a typical loaded area such as abcd (Fig. 4) with a band width ad=w.
8. A typical strip, with loaded length l at a distance z from ab and span L,
has a maximum bending moment of
la 1
pl.1-p- = -.PP = m
2 2
say, where
l la - 11
= (-++ll.
,, \-.-
/;P \
1
= 2 . p .(average loaded length)”.(K)
= (maximum
moment
in mid-strip). K . . (3)
where K is shown in Fig. 5.
9. Hillerborg refers to the average loaded length as the ‘mean span’, but
this is likely to be confused with the actual span of the strips. Also, equation
(2) is only valid for avalue m = + P P , implying symmetricallyloaded strips and
zero shear along the discontinuity line. However, these values of K are used
whether or not there is zero sheararising from symmetry;also, these valuesof
K are used for the mostgeneralirregularly shaped loaded area shownin
Fig. 6,again no doubt because of the alleged reserveof strength expected from
alower-boundapproach.Thus, when there is astrongshear at the dis-
continuity line, Hillerborg evaluates the maximum moment correctly for the
mid-strip and then applies the factor K.
V = O*0836pL4
12. The solution by yield-line theory1 is, for uniformly placed reinforce-
ment without top steel in the corners, M=pLa/22. Hence
V = 2 .La.pLa/22 = 0.091pL4,
showingthat yield-line theory uses,in this treatment, slightly more steel,
Disregarding the bands, with continuously variable reinforcement the absolute
minimum moment volumea for this Hillerborg-typeload dispersion is
Vmin.=0 . 0 6 2 5 ~ 1 ; ~showing
. that Hillerborg’s treatment is rcasonably econ-
omical. It should be pointed out that the reinforcement required for bands
similar to Band 2 above is often below the minimum code requirement for
slabs.
Alternative treatment of discontinuity lines
13. In limitanalysis there is no intrinsic reason why the discontinuity
lines should be straight, and it would appear to the Authors that Hillerborg
has sometimesneedlessly complicated the calculations by employingtriangular
and trapezoidal shapes of loaded regions. The discontinuity linesmayin
fact be chosento fit the bands of reinforcement identically, and to give uniform
treatment of the strips, as shown in Fig. 8. In addition to avoiding the use of
equation (3) and Fig. 5, the solution now happens to be exact and strictly in
accordance with limit analysis, which is certainly not true of any averaging-
of-moments process. It should be noticed that the outer longitudinal strips
theoretically carry no load. In reality they will carry very little load, so that
only nominal reinforcement is required. This remark islikely to apply to
most edge bands (e.g. region 2, Fig. 7).
Treatment of holes
14. Holes present no problemwith the strip method, so long as it is feasible
to provide strong bands round the hole acting as beams (Fig. 9). Thus strips
a-a are supportedin turn by the strong bandb-b, the interreactive forces being
suitably distributed. It is easy to see that the distribution of reaction on the
290
THE T H E O R Y O F T H E S T R I P M E T H O D FOR DESIGN OF SLABS
c o f slab
1
I
X X A 3 5 A
......l...
Load dispersion
Key + -+- Contraflexureline
Intensity of load on edge beam -..-.- Stress discontinuity
edge beam is therefore known at some stage of the calculations, even though
this loading diagram appears to be crudely discontinuous. In marked con-
trast to yield-line theory, no part of the structure can be left without its accom-
panying statementof loading, and this feature of the strip method is of instant
appeal to designers.
Fig. 11. Triangular slab with free edge Fig. 12. Load dispersionin an irregular slab
17. Hillerborg was forced to adopt a different treatment when the strips
were not supported at their far ends by a symmetrical, or nearly symmetrical,
system. Thus, in Fig. 11 showing a triangular slab with a free edge, the strips
are cantilevered out from the fixed edges, and carry both the distributed load
and the reaction from the other strips in the positive-moment region, which
latter strips may have to change direction as shown. On the contraflexure
line itself the total vertical reactions must be balanced which means, in effect,
that the strips change in width. If Qland bl refer to the shear per unit width
in region 1, and the corresponding width of strip, when Ql .bl = Q2. b2, i.e.
18. The conditions in Fig. 11 seem to imply that the strips in region 1 must
of necessity be parallel to the free edge. This is not so, for the slab near the
free edge can always be heavily reinforcedso as toform a strong band almost
indistinguishable from a 'beam'. The remaining strips may then spandirectly
across to this beam band. This simpler treatment is preferred by the Authors.
Superimposition of strips not at right angles
19. The strips shown in Fig. 11 are equivalent to curved beams, and this
kind of strip placement can result in strips not crossing at right angles, as
shown in Fig. 12. In one direction the strips are like those in Fig. 11 ; in the
other direction thetreatment issimilar tothe strips in Fig. 1. Fig. 12 is
typical of many produced by Hillerborg. The designer must always remember
that strips must continue right across the slab until some suitable reaction
point is encountered, and continuity of shears must be preserved, otherwise
the diagrams can become bewildering. Skew slabs may be treated by a series
of skew strips, with discontinuity lines similar to those in Fig. 1, each set of
strips dealing withthe appropriate loaddispersion, the reinforcement following
the strips. In hismanyworkedexamples, Hillerborge-' clearlyexpects the
designer to treat every case on its own merits.
Critical examination in the lightof limit analysis
General comments
20. It is obvious that the strip methodis at one and the same time a power-
ful method of design and a daring, perhaps startling, use of limit analysis.
292
THE T H E O R Y O F T H E S T R I P M E T H O D FOR DESIGN O F SLABS
and insisting upon this being continuous, leads to the condition that where the
strips join
. . . . (S)
This is the same result as given by Hillerborg’s intuitive treatment given by
equation (4), showing that the strip method rigorously satisfies the rules of
limit analysis on such discontinuities.
and on substituting from equations (7) and (S), it is eventually found that
. . . . .(9)
This is the same expression as in equation (6), which means that any value of
X willprovide the exact collapse load p for which the slab wasdesigned.
There are therefore aninfinite number of simultaneous modesof collapse. This
+v’*
Fig. 15. Trial mode of collapse
295
WOOD AND A R M E R
moment I
Fig. 16. Illustratingthe'plastic
potential' or theory of 'normality
Princlpal moment 2 of strains'
result alters the whole attitude given to this design method; clearly a more
generalized statement is required. To do this the formal proofs of upper and
lower bounds are briefly re-examined. The key to such proofs is the state-
ment relating maximum dissipation of energy to normal plastic strains (Fig. 16),
thus :
If the principal (plastic) strains are normal to the yield locus at the point
of yield, then the work done is a maximum.
In simpler terms, if the principal moments in the stress field in an isotropic
slab coincide with the direction of yielding, then more energy is dissipated
than with any other possible stress state on yield locus.
28. The limit theoremscan be put veryconcisely as follows.Let {ac}
represent the correct [ =exact] stress field at collapse, and { K ~ the
} corresponding
exactfieldof strains. So that D{U~}.{K~} represents the internal dissipation
of energy at collapse. Let the external correct collapse loads be denoted by
{Po},and the field of deflexions be {Sc}, being compatible with { K ~ } . Then by
the theorem of virtual work, since the structure is everywhere in equilibrium,
E being the external work,
D{U0}.{Kc) = E{Po1.{8J.. . . . . . (10)
Suppose there exists another different stress field{ U } , everywhere in equilibrium
with loads {P},where {P}=A{Po},A being a constant multiplier for all the
various loads. The theorem of virtual work then allows any arbitrary test
mechanism, so choose the correct mechanism, giving
' D{U}.{KC} = E{P}.{S,}. . . . . . * (11)
If the correct stress field follows the 'normality rule', then this new stress field
probably does not, SO that
DIU}. {Kc} d D(uc}.{K,) . . . . . * (12)
whence
E { P ) .{&l d E{Po}.{&l
or
A G l . . . . . . . * (13)
which is the lower-bound theorem.
29. It should be noted that alternative correct stress fields can exist. These
allow changes in the rigid regions only, where the dissipation of energy is not
taking place. Moreover there is nothing in the proof to prevent the use of
variable yield moments, made to fit the stress fields; nor would discontimous
stress fields affect the dissipation of energy, provided there are no unbalanced
forces.
296
THETHEORY OF THESTRIPMETHOD FOR D E S I G N O F S L A B S
30. For an upper-bound solution consider an alternative mechanism {a,}.
This imposes a definite strain field { K ~ } . Then, using the normality rule
(Fig. 16), the corresponding stress states are defined, on the yield lines only.
This enables a work equation
D{u~}.{K~}= E{P,} .{an} . . . . . . (14)
to be set up, noting that {Pn}
is not everywhere in equilibrium with {au}.
31. If this (incorrect) mechanism is now used as a valid virtual displace-
ment with the correct stress field {U,}, and since ‘normality’ will not prevail,
then, by virtual work, .
and
Hence
{Pu}3 {P,}, or h 3 1 . . . . . . (15)
providing an upper-bound solution.
32. This far the proof will only be upheld if there is isotropic (square mesh)
reinforcement, although this may be varied in intensity. This arises because
the conditions in Fig. 16 presuppose that the yield locus is the same whatever
the principal momentdirections. With orthotropic or skewreinforcement,
because of anisotropy, the slab can yield in directions other than theprincipal
moment directions. Consequently equation (12) requires redefining.
33. Let M,, be the normal moment on any yield line of the correct collapse
mode, of length L,, and rotation 4,. Then D{U,).{K~} is actually Z(M,,, .L,.+,),
due to the reinforcementwhichisprovided.Now let M,,, be thenormal
moment on thesame yieldline L,, due to another stress field {U}. Then
D{u}.{K,} is actually I(Mn,.L, .4,).
34. To establish a lower-bound solution, the inequality (12) must hold.
Since there can be no restriction in direction of yield lines, in general
M,, M,,, . . . . . . . (16)
i.e. the normal moment in any direction due to the trial stress field must be
less than the normal yield moment which the reinforcement could sustain in
that direction atthat point. This proof independently reaches thesame
conclusion as Kernpl‘ in a recent paper defining the criterion of yield for
orthotropic reinforcement. In addition Kemp shows that this test-of-normal-
moment-in-every-direction criterion obeys thenormality of strains. It is
therefore important to note that, without this recent extension of the yield
criterion the basis of Hillerborg’s method with variable reinforcement remained
intuitive.
35. It can now be seen that, if {uH} is a Hillerborgstress field in equilibrium
with loads {P}then the virtual work theorem allowsanytestmechanism
{ K ~ } from
, which
D{u~}.{K,} = E{P}.{8,}. . . . . . . (17)
However, if, and only if, the field of resistance moments due to the reinforce
ment coincides identically with this stress field, then the normal moments at
297
WOOD AND ARMER
any point will coincidewith the yield criterion in any direction, so that
equation (17) then is preciscly the work equation for any mcchanism of col-
lapse. Hence basically Hillerborg’s method provides an exact solution with
an unlimited number oJsimultaneous modes. Hillerborg’s ‘averaging’ process,
equation (2), by failing to satisfy the coincidence mentioned above, does not
provide a lower-boundsolution. Control ofanalysisislost. Indeed in
many cases it can be proved that an upper-bound solution for thecollapse load
results. The designer who instead places the reinforcement in a conservative
manner, carefully accounting for any lack of symmetry in the slab(cf. $8 7-g),
would then achieve a lower-bound solution.
36. On test, slabs designed by the strip method may be expected to yield
in nearly all directions at failure, and so they do, somewhatlike a plastic
hammock. At working loads, however,there are hardly any cracks to be seen.
Whereas with a Johansen-slab dcsigned by yield-line theory there are rigid
regions of the slabstill remaining at collapse, with a Hillerborg-slab such rigid
regions tend to disappear. This is the direct result of the strip method having
produced a design with more efficient and economical use of reinforcement,
but there is a corresponding price to pay in terms of increaseddeflexions.
With a Hillerborg-slab,keeping the deflexionswithin reasonable limitsis
more dependent upon membrane action.la
Fig. 17. Examples takenfrom ‘Strimlemetoden’: (a) slab with column;(b) L-shaped
slab
Column
Fig. 18. Type-3 element (new treatment) ;these elements are subjected to a uni-
formly distributed load p
SECTION BB
Fig. 19. Hillerborg's treatment of a square type-3 element : (a) primary load action ;
(b) secondary load action
quarter of a uniformly loaded rectangular slab, size L X aL, with free edges
on corner columns only (Fig. 22). Then the stress field
M* = p gLa
(l-4$) . . . . . (19a)
Mm
P
- ~ X Y . . . . . . . . . (19c)
- L
(4 v, = =0 on edges X = f.Lj2; likewise on f a -,
2
301
W O O D AND ARMER
+
Fig. 20. Calculated moment field for type-3 fieldmoment m,=O. This case is for
R = / , the full line representing the greatest negative moment and the brokenline
the greatest positive moment
(e) +
principal moments at the corners of (Mxu);, f = + p aL218, and
elsewhere of smaller value except approaching the centre lines,
(f) a corner reaction R = - 2Mx, =paL2/4.
45. With allnecessary conditions satisfied, if constant negative moments
- K,(pLa)/8 and - K y .p(aaL2/8)are added all over the quarter slab (Fig. 22)
in the respectivedirections then, without interfering with the equilibrium
equation or the zero shear conditions, a natural type-3 rectangular element is
obtained where, compared with equation (1 8),
/M Y pL2/8
Y V P
J
Y
Fig. 22. Type-3 element obtained from lower-bound solution for rectangular slab
with free edges
of the worst values to be found in each column of mesh points within the sub-
regions 1, 2, 3 and 4. The Appendix thus gives tabulated values forthe
required moment of resistance M X 1 , M Y 1, for the top reinforcement of
region 1, and similarly for the other regions.
47. These tables account for the following parameters:
a = hI1xs K x , Ky,
where the type-3element has constant field moments on the boundary of
pCfa1,a
(1 - K x ) , 7(1 -K,,),
and constant support moments of
Hence
Thus in this case K, = 0.64 and Ky= 0. Then by referenceto therelevant table
in the Appendix-here, that for a = O.4-the design moments M, and M,,
both positive and negative, can be obtained readily by interpolation. The
parts between the zero shear lines and the supports can be reinforced all over
to resist the moment values given by the table or the reinforcement can be
curtailed to fit the field more closely.
50. The loading and support conditions for sections B-B and D-D are
shown in Fig. 23. The lengths X and y are obviously variable and the design
moments in these strips are calculated in exactly the same way as has been
described above in $0 7-9 using Fig. 5. A slab based on this example has been
tested'" and behaved satisfactorily both at working and at ultimate load.
Conclusions-final comments
52. Hillerborg's 'simple' stripmethodprovidesan exact (not a lower-
bound) solution for the collapse load of a slab carrying distributed load if the
reinforcement were made to fit ideally. Point loads may be treated as local
concentrations of distributed load. It is a powerfuldesign method giving
excellent results for the designer who has a good background knowledge of
elastic design. Although there is almost unlimited freedomof choice in plac-
ing the reinforcement, thedesign chosen should not be too far removed from
that expected in elastic design. Thus, in effect, a great simplification of elastic
design is achieved.
304
-~
SECTION DD
PY
Fig. 23. Assumed moment field for a rectangular slab with a central prop
305
W O O D A N D ARMER
Acknowledgements
54. This Paper dealswithwork forming part of the programme of the
Building Research Station, and is published by permission of the Director.
References
1. PRAGER W. An introduction to plasticity. Addison-Wesley. New York, 1959.
2. WOODR. H. Plastic and elastic design of slabs andplates. Thames and Hudson,
London, 1961.
3. JoHANseN K. W. Yield-line theory. Cement and Concrete Association, London,
1962.
4. JONESL. L. and WOOD R. H. Yield-lineanalysis of slabs. Thames and
Hudson, Chatto and Windus, London, 1967.
5. HILLERBORG A. A plastic theoryforthe design of reinforced concrete slabs.
Proc. 6th Congr. Int. Ass. Brit. struct. Engng, Stockholm, 1960.
6. WLLeRBoRc A. Jamviktsteori forarmerade betongplattor. Betong, 1956, 41
(4) 171-182.
7. HILLERBORG A. Strimlernetoden. Svenska Riksbyggen, Stockholm, 1959. See
also Strip method for slabs on columns, L-shaped plates etc. Translated by F. A.
Blakey, Commonwealth Scientific and Industrial Research Organisation, Mel-
bourne. 1964.
8. WOODR. H. The reinforcement of slabs in accordance with a predetermined
fieldof moments. Concr. Mag., 1968, 2 (2), 69-76.
9. CRAWFORD R. E. Limit design of reinforced concrete slabs. PhD thesis,
University of Illinois, 1962.
10. SWEDISH STATECONCRETE COMWWEE. Massive concreteslabs; specifications
for dcsign methods, etc. Stockholm, 1958.
306
THE THEORY OF THE STRIP METHOD FOR DESIGN O F SLABS
11. TIMOSWENKO S. and WOWOWSKY-KRIECIER S. Theory of plates and shells.
McGraw-Hill, Ncw York, 1959.
12. ARMER G. S. T. Ultimateloadtests of slabs designcd by the strip method.
Proc. lnstn civ. Engrs, 1968. 41 (Oct) 315-334.
13. Recent developments in yicld-line theory. Mug. Concr. Res. Spec. Publ., 1965
(May) 31-62.
14. KE” K. 0. The yield criterion for orthotropically reinforced concrete slabs.
Int. J. meclr. Sci., 1965, 7, 737-746.
307
W O O D AND A R M E R
/1z"--+l
VALUESOFREINFORCEMENTCOEFFICIENTSFOR ALPHAm.4
THE THEORY O F THE STRIP METHOD F O R DESIGN O F S L A B S
309
WOOD AND ARMER
NEGATIVE(TOP'REINF0RCEMENT)
'MXf MX2 M X 3 MX4 MY1 M Y 2 M Y 3 MY4
-280 -000 0039 -000 -256-019 -000 0000
0280 m000 0050 0000 -305 -068 -000 -000
6280 -000 -059 -000 -354 -011 m000
0280 -000 -065 -OOO -403 01 0127
6 -031 -000
-280 0000 -000 04520215 -067 -000
-380 0000 O ; ? ~ ~ O O O O-277. -025 0000 -000
' 0 80 0000 -1 0 -000 0074 m000 0000
. -380 -000 0139 0000 *%'.l23 -012 0000
-380 -000 ,165 .OOO 0424 m172m045 -000
-380 -000 m170 0000 -473e221-085 m000
-480 -000 -239 -000 e280 0036 ~ O O Om000
. 0480 m000 m250 -000 m329 -085 -000 -000
0480 -000 0259 -000 e3780134-012 m000
-480 -000 -265 -000 m427 0183 m045 -000
m480 m 0 0 0 -270 -000 -476 m232 0094 -001
05800007 0339 *OOO -280 -056 m000 o O 0 0
0580 -007 e350 -000 -329 m105 0000 -000
0580 -007 m359 0000 m378 m154 0012 0000
-580 -007 ,365 mOOO 0203 m053 -000
-580 m007 0970 0000 --422 4 7 -252 m102 -005
-680 m079 0439 0000 -280-067 m000 0000
-680. -086 0450- 0000 m329 0116 0000 m000
0680. -092 0459 0003 0378 0165 -012 ~ o O O
06800092-465 0011 m427 - 2 1 4 m053 0000
-680 m092 .e470 m012 -476 0263 -102-015
310
THE THEORY O F T H E S T R I P M E T H O D FOR DESIGN OF S L A B S
31 1