Paper: O’Dwyer/O’Brien

Paper

Design and analysis of concrete slabs using a

modified strip method
D. W. O’Dwyer, BE, MEngSc, CEng, MIEI
Trinity College, Dublin

Professor E. J. O’Brien, BE, MEngSc, PhD, CEng, MIStructE

National University of Ireland

Synopsis reinforcement top andbottom, and encastrkat its edges, is illustrated in Fig
This paperdescribes a technique for the ultimate limit state 1. For the mechanism assumed inFig 1 (a), the relationshipbetween thecol-
analysis or design of reinforced concrete slabs. The modified strip lapse load,qc,and the plasticmoment capacity, %, is2:
method presented is a development of the strip technique using
linear and quadratic programing optimisation. The first stage of
the procedure uses linear programing to find the distribution of
loads between the strips which, in the analysis case, leads to the
maximum collapse load factor or, in the design case, the minimum
reinforcement requirement.
The load distributions leading to the largest collapse load
factor or minimum reinforcement requirement are not unique. The
second stage of the procedure uses quadratic programing to
search among the optimal distributions to find the load
distribution which minimises the sum of squares of the differences
between the elastic moments and the strip moments. This results
in a solution which does not deviate excessively from the results of
a linear elastic analysis.
The method is of particular use in the design of slabs which
have irregular patterns of loading, irregular shape, irregular
support conditions or which include openings.

Introduction
The design of reinforced concrete slabs is an everydaystructural engineer-
ing task. Where the slab tobe designed is rectangular and where the sup-
port conditions areregular, a popular approach is to use graphs ortables to
determine the distribution of moments’. Such tables are generally theresult
of linear elasticanalyses for uniformly distributed loads acting on slabs of
constant thicknesswhose supports, whether pinned or encastr6, are uniform
and infinitely stiff. Although these conditions
are not met in practice, design-
ers often assume that irregularities in the support conditions, loadingand
slab stiffness can be accommodated by small plastic deformations in the Fig I (a).Hinge patternfor simple yieldline analysis
slab. Such techniques are useful for the design of standard slabs but are
clearly unsuited for more complex situations.

Plastic assumption and yieldline method

The assumption that reinforcedconcrete slabs behave plasticallyis justified
because they are generally under-reinforced2. Load tests on under-rein-
forced concrete slabs3 have indicated that failure occurs after the formation
of yieldlines, alongwhich the tensile side of the slab cracks and the rein-
forcement yields.
If the shape, loading and support conditions of a slabare regular, it is pos-
sible topredict the mechanism (the pattern of yieldlines), by which it will
faiP4. By examining the virtual work equations associatedwith any given
mechanism, it is possible to calculatethe imposed load required to causethe
mechanism to form. The mechanism will form when therate of change in
potential energyof the loaddue tothe formation of the mechanismexceeds
or equals the rate at which energy is expended in forming it.
The yieldline method has significant disadvantages. Firstly, it generates
an upper bound on the collapse load. For any assumed patternof yieldlines,
the predicted collapse load will exceed or equal the true value. The differ-
ence between the two is a function of how closely the assumed pattern of
yieldlines agreeswith the pattern at failure. If the critical yieldline pattern
has been identified, the calculated collapseload will be correct. A second
disadvantage is that the yieldline technique does not give the supportreac-
tions along the slab’sedge. Thisis a particular disadvantagewhere the slab
is supported by edge beams whose designis dependent onhow the slab
transfers load onto them.
The yieldline analysis of a square, isotropically reinforced slab, with equal Fig I(6).Form of hingepattern of true collapse mechanism

The Structural EngineerVolume 76/No 17 1 September 1998 329

Paper: O’Dwyer/O’Brien

&=4x ....(1)
mP

The relationship between the collapse load and the plastic moment capaci-
ty calculated using the correct hinge pattern, illustrated in Fig 1(b), gives5:

4 L2 = 42.85 1
C ....(2)
mP
For this example, the yieldline analysis based on the simple hinge pattern
overestimates the collapse load by 12%. The critical yieldline patterns can
be readily predicted if the slab’s shape, loading and support conditions are
straightforward.
However, when the slab shape or the support or loading conditions are
complex, identifying the critical pattern becomes more difficult.

The strip method

The strip method, introduced by Hillerborg6 and developed by Wood &
Armer’, is also based on the assumption that reinforced concrete slabs
behave plastically. However, the strip method, unlike yieldline analysis,
provides a lower bound on the collapse load. The strip technique is based Fig 3 . Hillerborg’s strip method
on the lower bound or ‘safe’ theorem of plasticity. It involves findinga set
of moments which are in equilibrium with the loads on the structure and
which do not exceed the plastic moment capacity of the slab at any point. other direction. Further, the valueof a is not restricted to being between zero
The ‘safe’ theorem ensures that the collapse load factor associated with any and unity. However, when Hillerborg’s method is applied by hand, the value
statically admissible set of moments will be less than, or equal to, the true of a for each sectionof slab is often set to either zero or one.
collapse load factor. Once the load has been allocated between the notional slabs spanning in
Consider the segment of slab and the associated moments and shear the two orthogonal directions, the one-way-spanning slabs are subdivided
forces illustrated in Fig2. The condition which must be satisfied for the ele- into strips and each strip is analysed as an independent beam. As they are
ment to be in vertical equilibrium is: assumed to behave independently, the strip with the lowest collapse load fac-
tor dictates the overall collapse load - e.g. symmetry would suggest shar-
d2my
d2mx +-+2--=-4 d2mxy ...(3) ing the load on the slab of Fig I equally between the two orthogonal
ax2 ayLv2 &ay directions, i.e. a = X. This assumption leads to the following relationship
where m, and my are moments/unit lengthon the X andY faces, respectively, between the collapse load and the plastic moment:
and mXyrepresents the torsional momenthnit length. Hillerborg suggested
setting the torsional term to zero and choosing m, and my accordingly. &=32 ....(6)
Setting the torsional term in eqn (3) to zero gives: mP
which is 75% of the true collapse load. In contrast,
a linear elastic analysis*
... gives the relationship between the maximum moment and the applied uni-
form load as:
which can be rewritten as:
qL2 - 19.34
-- ....(7 )
mmax
...(5 ) which is 45% ofthe true collapse load and only 59% of the Hillerborgstrip
d=-(l-a)q
ay2 1 result. If the objective is design rather than analysis, eachstrip is designed
as an independent beam and,this in manner, the reinforcement requirements
where a is a factor reflecting the degree of loadsharing between the two in the X and Y directions in the real two-way spanning slab are calculated.
orthogonal directions. Hillerborg’s technique involves sharing the load at Both the yieldline method and the Hillerborg strip technique are listed as
each point on the slabbetween two notional one-way spanning slabs which acceptable design methods inBS 8 1 10 and the draft Eurocode EC2.
span in the reinforcement directions which are usually orthogonal, as illus-
trated in Fig 3. The value of a determines how the loadis divided between Analysis using the modifiedstrip method
the X andY directions. Hillerborg’s approach does not require that the load If a slab is subdivided into a grid of rectangular elements as illustrated in
on a given element be carried by spanning exclusively in either one or the Fig 4,then, depending on how the load is shared between theX andY strips
intersecting at each element (i.e. the value of a at each element), a differ-

6myx 6mxy
myx + - 6Y mxy + -6x
6V 6X

Fig 2. Slab element Fig 4. Subdivision of slab into a series of orthogonal strips

330 TheStructuralEngineerVolume 76/No 17 1 September 1998

 Paper: O’DwyedO’Brien

ent allowable load will be obtained. Assuming infinite ductility, any man- TABLE l - Form of additional constraints for each type of edge condition
ner of sharing the load will generate an acceptable lower-bound solution to
the collapse load. The objective in analysisis to find the pattern
of load dis-
tribution which givesthe maximum collapse load factor,h.
Eqn (4), the basis of Hillerborg’s analysis, is the vertical equilibrium - MP= < Edge moments = < MP
equation for a grillage model in which the torsional capacity is ignored.
Applying numerical approximations for the derivatives in eqn 4 at(ij) node
and including the load factor gives: Simply supported edge
Edge moments = 0

where mXijand myij are the moments/m in theX and Y directions, respec- Free edge
tively, evaluated at node (ij) and ijy and 6y are the corresponding intervals Moments perpendicular to the
between nodes. Wij is the portion of vertical load applied to node (i,j). This edge at 1st internal node = 0
equation must be satisfiedfor node i,j to be in vertical equilibrium. A sim-
ilar equilibrium equation can be formulated for each node in the grillage,
i.e. for each rectangular element in the slab. In matrix form these equations ject to the assumptions of plasticity, and ensures the safety of the slab.
can be expressed as: However, the moments which the linear programing algorithm finds may
differ greatly from the moments that actually occur. This leads to two poten
....(9) tial difficulties - first, the designer will be scepticalof the results and, sec-
ondly, the real reinforced concrete slab may not be able to accommodate th
where plastic rotations requiredto achieve the optimised moments. It is clearly of
interest to find a solution that is similar to what will actually occur and that
[mX]and [my] are the matrices of moments/m in the X and Y requires a minimum amountof plastic hinge rotation.
directions, respectively, evaluated at the nodes This information cannot be obtained directly from the linear programing
[wl the
matrix
is of applied
loads
and algorithm because the number of optimal solutions may be very large or infi-
matrix
theis[Cl of coefficients: nite. However, once the optimum collapse load factoris known, an appro-
priate distribution of moments can be foundby reformatting the problem.
-2 1 0 . . 0-
One approach to minimising the need for plastic hinge rotation and which
1-2 1 0 0 should find a distribution of moments similar to the moments obtained from
0 1 - 2 . a linear elastic analysis, and FzYi,j,is to minimise the magnitude of the
[c]=0 0 1
0
.
0 . . 1 0
-2
0

. 1
greatest difference between the linear programing moments and the elastic
moments. This could be formulated as a linear programing problem. An
alternative and more traditional approach which is akin to curvefitting is to
minimise the sumof the squares of the differences between the linear pro-
0 0 0 1- 2 gramming and the elastic moments. Thus the objective functionthis in sec-
ond stage of the optimisation is to minimise:
Clearly, at the boundaries, the relevant equations will be replaced by the
boundary conditions.
Any statically admissible setof moments must satisfy all of these equi-
librium equations. In addition, for a set of moments to be acceptable, the
magnitude of moment at each node, in each direction, must not exceed the subject to eqn(9) and inequalities (10) and ( l 1) and subject to the equality
plastic moment capacity of the slab at that point. Therefore, in addition to that h equals its collapse load value.
the equilibrium equations, there are two additionalofsets constraints at each This is a quadratic programing problem and can be solvedby formulat-
node: ing the Kuhnnucker’ conditions and using a modified linear programing
technique to find a feasible solution’.’’.“.
,.(lo) The value of the collapse load factor h may have to be reduced if the
amount of redistribution is excessive; BS 8 110 requires that the maximum
...(11) elastic moments are not reducedby more than 30% through redistribution.

where [mxIhog, [myIhog and [mXlsag, [mYISagare the moment capacitiesat the Analysis example using the modified
strip method
nodes for hogging and sagging moment, respectively. Any set of moments The modifiedstrip method was used to calculate the collapse load factor for
which satisfies eqn (9) and does not violate constraints(10) and (1 1) at all a uniformly distributed load applied to the slab shown in Fig 5. The slab is
the nodes in the slab is a statically admissible solution and guarantees a 5m square and simply supported on three sides with the fourth sidefree. The
lower bound. The objectis to find the maximum value of the load factor,h, uniform load was modelled as a series of point loads appliedat the centres
for which an acceptable solution exists, i.e. as high a lower bound on the of one 100 0.5m x 0.5m square elements.The point loads were250N, rep-
collapse load factoras possible. resenting a uniformly distributed loadof 1kN/m2.
An equilibrium equation with the form of eqn (8) was formulatedfor each
Linear and quadratic optimisation loaded node. The moments at all nodes, both loaded and support, were con-
Only the moments and the collapse load factor in eqn (9) are unknown. strained to be less than the known local moment capacity of the slab; con-
Therefore, the problem h, sub-
is that of maximising the collapse load factor, straints with the form of eqns (10) and (1 1). The slab was assumed to be
ject to thelinear eqns (9) and inequalities (10) and (11). This can bereadi- isotropically reinforced with a plastic moment capacity of 20kNm/m in
ly formulated as a standard linear programing problem’ and, once the both hogging and sagging.
problem is expressed in this form, standard packagescan beused to find the The free edge was incorporatedby setting the mXijmoments to zero at the
optimum collapse load. Additional constraints must be added to the formu- slab’s free right-hand edge andat the loaded node nearest this edge, since
lation to take account of boundary conditions at the slab’s edges and sup- no vertical reactionis applied at the edge nodes.
ports. The form of the additional constraints for each typeof edge condition The initial phase of the optimisation found the maximum collapse load
is summarised in Table 1. factor to beh = 14.4, i.e. a maximum UDLof 14.4kN/m2.
Linear programing will yield the maximum collapse load factor h sub- The second stage in the procedure found, from among the optimal solu-
ject to the given constraints. However, the set of moments which the linear tions, the setof moments for which the sumof squares of differences from
programing algorithm finds correspondingto the optimum collapse loadh the elastic distribution of moment was a minimum. The elastic distribution
is often not unique, i.e. there may be a large number of possible sets of used was for a UDL of 14.4kN/m2 and included an allowance for the twist-
moments which satisfy the equilibrium equations and moment constraints ing moments, mXYijin accordance with theWood & Armer equations7.The
and which result in the maximumh. Each of these solutions is valid, sub- elastic moment distribution is illustrated in Fig 6 and the optimal solution

TheStructuralEngineerVolume 76INo 17 1 September 1998 331

Paper: O’DwyerlO’Brien

closest tothis elastic distribution is shown graphically in Fig7. The elastic R

l
moment distribution had a peak moment of 31.2kNm/m which is 56% l
greater than the allowable moment. An elastic analysiswould limit the max-
imum allowable load to 9.2kN/m2.
-4
Ad
!
The limits onmoment redistribution in BS 8110 would limit the maxi-
mum allowable peak elastic moment to 20kNm/m + 30% or 26kNm/m
which corresponds to h = 12 and a maximum UDL of 12 kN/mz.

Design usingthe modified strip method

This modified strip technique can be used to greatestadvantage in design.
The design formulation is similar to the analysisformulation except that the
design objective is to minimise the amount of reinforcement required. When
a slab thicknessis adopted, the moment capacity at any point is a function
of the areaof reinforcement at that location. Hence minimising the amount
of reinforcement required is approximately equivalent to minimising the
sum of absolute values of the moments. The sum of themoments through-
out a slab is sometimesreferred to as the moment volume7.
The equilibrium condition must be satisfied at eachnode but the design
load is known.Therefore theload factor isremoved from eqn (9) togive:

...
Fig 6(a).Elastic distributionof moments inX direction
Constraint eqns (10)and ( l 1) remain unchanged, with the important dif-
ference that the plastic hogging and sagging moment capacities [mxIhog,
[myIhog and [mx]%,[my]% are now the unknowns. The objective function,
for the most general case, is thesum over allnodes of the absolutevalues
of the four moment capacities at eachnode.

Allowing the plastic moment capacities to fluctuate from onenode to the

next may not be feasible in practice. It is up to the designerto dictate the
level of variation.A sensible approach to adopt is to allow the reinforcement
in each strip tovary independently of adjacent strips and to allow the rein-
forcement within the strip to vary according to thereinforcement details. The
pattern of reinforcement will often be obvious before bar sizes and spacings
are calculated. Controlling thelevel of variation is achieved by limiting the
number of unknown moment capacity variables in eqns (10)and (1 l), i.e.
a single variable may be used to describe themoment capacity at allnodes
in a strip. By providing reinforcement such that themoment capacity of the
slab equalsor exceeds the equilibrium moments, the safetheorem guaran-
tees the safety of the slab. Wood & h e r 7 examined the strip method and
found that, if the reinforcement in the slab, which dictates the ultimate
moment field, is provided to correspond precisely with the equilibrium

Fig 6(b).Elastic distributionof moments inY direction

moment field, the collapse load will equal thedesign load. Hence this opti-
misation technique can lead tovery efficient designs.
The requirement that the amount of redistribution is limited can be
achieved by adding constraints which ensure that the plastic moment of
resistance provided in any strip is not more than 30% less than the maxi-
mum elastic moment in the strip.
The second, quadratic, stageof the optimisation is of particular impor-
tance where the method is used in design. In the design formulation, the
initial linear programing analysis calculates theminimum amount of rein-
forcement required. However, as was the case in the analysisformulation,
there is morethan one way to reinforce the slab with the minimum amount
of reinforcement while achieving the design load. Of the many possible
ways of reinforcing the slab using the minimum amount of reinforcement,
the designer must findthat pattern which will give satisfactory performance
at serviceloads. As with the analysis case,this can be achieved by finding
the setof moments for which the sum of the squares of differences from the
elastic distribution of moment is a minimum, i.e. eqn (12).
. Consider applying the modified strip method todesign the perforated slab
shown in Fig 8. The positions of the holes in the slab dictate the way in
which this slab can be reinforced. This is clear without evenconsidering how
the load islikely to be distributed. Fig 8(a)shows the reinforcement bands
in the X direction, while Fig 8(b) shows those for theY direction. Within
Fig 5. Square slab with uniform reinforcement top and bottom encastre’ on each band the curtailment details will dictatewhich sections have similar
three sides moment capacities. Using this procedure, the practical reinforcement details

332 TheStructuralEngineerVolume 76/No 17 1 September 1998

 Paper: O’DwyedO’Brien

__1(
X direction Fig 8(a).Design strips spanning in theX direction
Fig 7(a). Optimum distributionof moments in X direction

Fig 8(b).Design strips spanning in theY direction

of the holes, and using the techniqueto size the reinforcement. The modi-
fied strip method can also generate the reactions applied to thesupports,
Fig 7(b).Optimum distributionof moments inY direction which is important for designing edge beams, internal columns, and support
structures.

dictate thereinforcement pattern, and the lineadquadratic programing mod- References

ified strip method provides the optimum area of reinforcement, A,, for each 1. Reynolds, C. E.,Steadman,J. C.: Reinforced concrete designer’shand-
section of each band. book, loth ed., Spon, 1988
It is not essential that the mesh be very dense forthis procedure to give 2. Moy, Stuart S . J.: Plastic methods for steel and concrete structures,
good results. The improvement gained by increasing the number of nodes/ Macmillan Publishers Ltd, 1981
strip is similar to the improvement in the accuracy of a bending moment dia- 3. Armer, G . S . T.: ‘Ultimate load tests of slabs designed by the strip
gram for a uniformly distributed loading as thenumber of point loads used method’, Proc. ICE, 41, October 1968, pp 3 13-331
to model the loading isincreased. 4. Jones, L. L., Wood,R.H.: Yieldlineanalysis of slabs, London, Thames
& Hudson, 1967
Conclusions 5 . Fox, E.N.: ‘Limit analysis for plates: The exact solution for a clamped
The modified strip method presented here isa development of the standard square plate of isotropic homogeneous material obeying the square
strip technique. The method uses standard mathematical programing tech- yield criterion and loaded by uniform pressure’, Proc. Royal Society
niques to calculate theoptimum collapse load factorwhen used to analyse A, 277, August 1974, pp 121-155
a slab and the minimum reinforcement requirement when used to design a 6. Hillerborg, A.: Strip method of design, Viewpoint Publications, 1975
slab. The modified strip method is based on the sameassumptions of plas- 7. Wood, R. H.,Armer, G. S. T.: ‘The theory of the stripmethod for de-
ticity as the stripmethod and thus generates a lower-bound estimate on the sign of slabs’, Proc. ICE, 41, October 1968,pp 285-31 l
collapse load factor. Since it makes no new assumptions and is simply an 8. Timoshenko, S. P., Woinowsky-Krieger, S.: Theory of plates and
optimisation of the standard strip method, it is allowed by many design shells, 2nd ed., New York,McGraw-Hill Kogakusha Ltd, 1959
Codes including,the draft Eurocode, EC2, subject to limits on theamount 9. Taha, H. A.: OperationsResearch, 4th ed., London, Collier MacMillan
of redistribution. Publishers, 1987
The technique is ideally suited to thedesign of slabs with holes or with 10. Saaty, L.,and Bram,J.: Non-linear mathematics,New York, McGraw
complex support conditions or shape. It does not require the designer to Hill Book Company Ltd, 1964
assume either thepattern of load distribution or the moment capacities of 11. Schrage, Li.: UNDO - An optimisation modelling system, 4th ed.,
the slab. It can be used most efficiently by having the designerinput the pat- Scientific Press. 1991
tern of reinforcement, which in a perforated slab isdictated by the positions

The Structural EngineerVolume 76/NO 17 1 September 1998 333