You are on page 1of 21

Finite Element Specialists and Engineering Consultants

Article 4: A Review of some Published Ultimate Load Capacities for Steel Plates (Part I)
By

Angus Ramsay, Husain Jubran Al-Gahtani & Edward Maunder

Synopsis
In the design or assessment of a standard plate configuration, such as the common case of a simply
supported rectangular plate with a uniformly distributed load to be considered in this article, the
engineer may resort to published data to establish the ultimate load capacity of his/her plate with an
expectation that the data is accurate. However, the engineer will find that various publications
provide different results, and that these differences can be very significant. For example, the values
given in Roark’s “Formulas for Stress and Strain” and the SCI’s “Steel Designers’ Manual” for a plate
with aspect ratio 0.3 are 176kPa and 103kPa respectively and without recourse to further work the
engineer might be unsure as to which, if any, is correct. The verified lower-bound limit analysis
[Article 1] shows that, whilst conservative, neither are in fact correct and that the collapse load is
actually slightly greater than 231kPa. Whilst an engineer must work on a conservative basis, the
imposition of such overt conservatism as exhibited in this example, which is not in the control of the
engineer, can lead to a design using more material than is necessary or, in the case of assessment,
unnecessary remedial work. This article reviews some of the published strength data for the simply
supported rectangular plate configuration in the light of more modern approaches for evaluating the
elastic moments and the plastic collapse load. The updated data is presented in a form suitable for
use by practising engineers and an example of its use is provided.

Introduction
The observation of significant differences in published results for the simple plate configuration
considered in this article, see figure 1, was used as the background to a recent NAFEMS Benchmark
Challenge, (NBC02), [2], where readers of the magazine were asked to use their finite element
systems to try and uncover the truth.


 Poisson’s ratio only influences the
elastic solution, the plastic solution
UDL () over full plate
being independent of this (elastic)
A  parameter. For the elastic solutions
= in this article a value of  = 0.3 has



been used which is consistent for
the ductile steels being considered.
simply supported edge

Figure 1: Simply supported rectangular plate with a UDL

Copyright © Ramsay Maunder Associates Limited (2004 – 2016). All Rights Reserved.
The responses to this challenge were interesting in that those who had modelled the problem using
plate elements did not manage to capture the true collapse load with any accuracy, whereas two
respondents who used solid elements were able to get extremely close to the load obtained from
the authors lower-bound limit analysis software tool, EFE, (this value not being disclosed to readers
of the magazine). In addition to verifying EFE, the challenge gave credence to the idea that the
published data is overtly conservative. In exploring where this conservatism has come from, the
authors have gone back to the original references used to generate this data and the findings are
rather interesting.

Definitions
The shape of the plate is defined by the aspect ratio and depending on which dimension is greatest,
this parameter can be either greater than or less than unity. As some of the equations used later in
this article require  ≥  then of the shapes shown in figure 2, only those on or above the diagonal
(shown in green) are allowed. Clearly, though, the results for those shapes below the diagonal are
simply obtained through the symmetrical equivalent above the diagonal.

Figure 2: Admissible plate shapes

As the aspect ratio decreases, the plate will ‘span’ across the shortest dimension and so the span of
the plate is defined as the dimension .

In considering the elastic solution for plates then, as the stresses vary linearly through the thickness,
it is sufficient to consider the stresses on the surface or skin of the plate. The corresponding elastic
moments are simply expressed in terms of the surface stress as =   ⁄6. In order to judge when
first yield occurs, it is common to adopt the terminology of an equivalent stress where this will
depending on the yield criterion and might be, for example, the maximum principal stress or the von
Mises stress. The equivalent stress can be given the symbol  with the maximum value over the
plate surface  so that, for a material with a yield stress  , first yield occurs when  =  . It
is common for plates to express the load in terms of a non-dimensional load parameter  as shown
in equation 1.

Copyright © Ramsay Maunder Associates Limited (2004 – 2016). All Rights Reserved.
∙ ∙  (1)
=
 

As  varies over the plate so does  and where the equivalent stress is a maximum, the load
parameter is a minimum – see equation 2.

 ∙ ∙  (2)
 =
  

The minimum load parameter may be scaled with the ratio of the maximum equivalent stress and
the yield stress to determine the ultimate (elastic) load parameter   as shown in equation 3.

 (3)
 = 
 

Thus, the elastic limiting load  is given by equation 4.

 (4)
 =   
∙ 

When considering plastic collapse of a plate, the ultimate (plastic) collapse load is written in terms of
the load factor , which is generally obtained from limit analysis, as shown in equation 5.

 =  (5)

In this manner the ultimate (plastic) collapse parameter may be expressed in terms of the ultimate
(elastic) collapse parameter as shown in equation 6.

  (6)
  =   =
  

Based on a rectangular plate cross section it is common practise to adopt the idea that in order to
develop full section yield the load to cause plastic collapse is 3/2 times the load to cause first yield:

3 (7)
 ≥ 
2

For the case where the moment field across the plate is constant then this approximation is exact
but for all other forms of moment field the approximation produces a lower-bound, hence, the
inequality in equation 7.

In this article the results will be presented in terms of the ultimate collapse parameter either elastic
or plastic depending on the context. These parameters are functions of the aspect ratio, the plate
size, the thickness and the yield stress.

Copyright © Ramsay Maunder Associates Limited (2004 – 2016). All Rights Reserved.
The article will include contour plots of elastic moments and these can be non-dimensionalised using
equation 1 and recognising the relationship between elastic surface stress and elastic moment as
shown in equation 8 for the ! moment component.

1 6 !
"! = =
(8)
  ∙ ∙ 

Navier’s Exact Elastic Solution


Navier’s solution for the transverse displacement is the double infinite series given in equation (9).
The curvatures and moments may be derived from the displacement field in the normal manner.

1 1 & -&
16% %  sin sin (9)
$= ))  
&'( -.   +  - 0
2! 2!

The Cartesian moment fields from the Navier solution are shown in figure 3. In interpreting the
moments it is helpful to recall that, for example, the moment  causes a direct stress  .

Figure 3: Cartesian moments from Navier’s solution ( = 1)

Copyright © Ramsay Maunder Associates Limited (2004 – 2016). All Rights Reserved.
The series converges rapidly, with the first term alone giving a reasonable approximation of the
moments – compare rows 1 and 4 of figure 3. The second row of the figure shows the moments for
the second term and the third row is the total solution (summing to = - = 49) minus the solution
for the first term.

It is interesting to note that maximum bending moments occur at the plate centre and that the first
term of the series over-estimates these by some 10%, i.e., taking the first term would provide a
conservative approximation to the moments. As the torsional moment is zero at the centre of the
plate then the principal moments are equal to the bending moments.

Two yield criteria are considered in this article these being the maximum principal moment and the
von Mises moment. The two yield criteria being considered can be written in terms of principal
moments ! and  , which are determined in the usual fashion, and the moment capacity of the
plate  as shown in table 1. It should be noted that the maximum principal moment criterion is
also called the square yield criterion and is identical to that which would be used for an isotropically
reinforced concrete slab.

Maximum Principal Moment von Mises Moment

−  ≤ ≤+ 
  
Principal Moments  ≥ ! +  − ! 
−  ≤ ≤+ 

8 8
These expressions can be normalised by letting 7! = 89 and 7 = 8;
: :

−1 ≤ 7! ≤ +1
Normalised Principal Moments 1 ≥ 7! + 7 − 7! 7
−1 ≤ 7 ≤ +1

Table 1: Maximum principal moment and von Mises yield criteria

The yield surfaces for these criteria, together with that for the Tresca criterion, will be shown later in
the article (see figure 10). As it is the principal moments or the von Mises moments that will govern
collapse, it is of interest to see how these vary over the plate. These (invariant) moment fields are
plotted for two plate aspect ratios in figure 4.

It is seen that the maximum principal moment occurs at the plate centre for both aspect ratios
shown. In terms of the von Mises moment, however, whereas the maximum value for the
rectangular plate is at the plate centre, for the square plate the values at the corners exceed the
centre value by about 10%.

It can be seen that the concentration in the von Mises moment that occurs at the corners of the
plate, is due entirely to the presence of high torsional moments at these points; the normal
moments are, by adherence to the static boundary conditions, zero at the corners. It is noted that
the Kirchhoff plate formulation results in non-zero distributions of torsional moments around the

Copyright © Ramsay Maunder Associates Limited (2004 – 2016). All Rights Reserved.
boundary. If the more complicated Reissner-Mindlin plate formulation had been used, then the
boundary conditions would have included twisting rotations and, therefore, the supports could have
been specified either as soft simple or hard simple supports. Numerical experimentation has shown
that if this formulation were used then the soft simple support condition, whilst moving the peak in
the torsional moments from the corner into the interior of the plate, does not significantly influence
the peak value of moment.

"<8 =0.287

Figure 4: Moment invariants from Navier’s solution

SCI’s Steel Designers’ Manual – Elastic Solutions


The Steel Designers’ Manual (SDM), [3], provides tables which present the Ultimate Load Capacity
for steel plates. These tables are also used by Tata Steel on their website detailing the strengths for
Durbar floor plates. An extract from these tables, for the configuration being considered, is
reproduced in figure 5.

Figure 5: Extract from SDM

Presenting the data in the manner adopted by the SDM disguises the actual relationship between
the collapse load and the various parameters involved. This relationship can be simply stated, as in
other publications, in the form of the non-dimensional ultimate load parameter   as defined in
equation 3.

Copyright © Ramsay Maunder Associates Limited (2004 – 2016). All Rights Reserved.
Instead of using the non-uniform aspect ratio intervals of the SDM, and in order to be compatible
with the Roark aspect ratios used later on in the article, then aspect ratios will be spaced uniformly
at 0.1 intervals in the range 0.2 ≤ ≤ 1. Table 2 presents the ultimate load parameters for these
aspect ratios and these are based on Pounder’s approximate equation 13 for the maximum principal
moment at the plate centre.

1.00 0.90 0.80 0.70 0.60 0.50 0.40 0.30 0.20


 2.26 2.17 2.17 2.25 2.45 2.80 3.40 4.47 6.68
Table 2: Ultimate load parameter from SDM

In NBC02 the plate considered had an aspect ratio of =0.3, a thickness of =10mm, a yield stress of
 =275MPa and a dimension of =2m. Using the ultimate load parameter from table 2, and
equation (4), the collapse pressure or ultimate collapse load can be evaluated:

0.01 (10)
 = 4.47 ∙ 275? ' ∙ = 102@A
0.3 ∙ 2

This value is in agreement with the SDM table of figure 5 (given small rounding differences) but
needs to be scaled up by a factor of almost 2.25 in order to match the actual plastic collapse load
presented in NBC02 (231kPa). To establish why there is such a large difference one needs to look at
the source of the values provided in the SDM. These values come from Pounder’s formula and were
presented in a 1949-50 publication which is the fourth edition of his booklet “The Design of Flat
Plates” [4].

From the preface of his publication, Cuthbert Coulson Pounder (1891-1982) appears to have been a
highly accomplished engineer. He had a 50 year career at Harland & Wolff in Belfast finishing up as
chief technical engineer and was, amongst others, Vice-President of the Institution of Marine
Engineers. He produced scientific papers and textbooks for nearly 60 years and his magnum opus
“Marine Diesel Engines”, which first appeared in 1950, is today being published in its ninth edition.

Whilst the version of Pounder’s publication used in this article dates from 1949-50, it was first
published in 1919, and he tells us in his introduction that after being driven by the absences of
rigorous solutions to the (Kirchhoff) plate differential equations, he set out to formulate ‘… a rational
system of calculations and design which could be relied upon to produce safe, economical and
consistent results for the more distinctive forms of flat plat.’

A thorough reading of Pounder’s booklet reveals that:

1) it is an elastic solution,

2) it does not appear to involve Poisson’s ratio,

3) it uses the maximum principal stress at the centre of the plate, and,

4) it does not consider corner uplift.

Copyright © Ramsay Maunder Associates Limited (2004 – 2016). All Rights Reserved.
Whilst Pounder uses Poisson’s ratio for the displacement expression it does not appear in his
expressions for the moments. Pounder provides equations for the maximum principal stress at the
centre of the plate based on a simplified equation (12) that he attributes to Navier (1785-1836), and
on his own approximate equation (13). These two equations may be re-expressed in the form of the
ultimate load parameter as follows.

Ultimate load parameter attributed, by Pounder, to Navier:

.1 +  0 Navier Approximation I (11)


 =
0.920 ∙

If, in Navier’s series solution, Poisson’s ratio is set to zero, then the first term in the series (m=1, n=1)
is of the same form as equation 11 but the value of the factor changes from 0.920 to 0.986.
Although it is not clear why Pounder modified this factor, it makes Pounder’s approximate Navier
solution about 10% more conservative.

An additional approximation of the maximum principal moment at the plate centre may be obtained
by truncating the series after the first term (as for equation 11) but this time using the correct value
of Poisson’s ratio.

;
B!CD; E Navier Approximation II (12)
 = F.GF∙D
Modify

The Ultimate load parameter from Pounder is given in equation 13.

1900.1 + % 0H Pounder Approximation (13)


 =
.1425 + 3116 % + 2191 I0

As noted, Pounder’s equation is for the elastic, rather than the plastic, solution, and so the
suggestion, in the SDM, that these are Ultimate Load Capacities is potentially misleading. It is
interesting, also, to see that neither of equations 11 or 13 involves Poisson’s ratio which is known to
be a factor in the elastic moment solution.

Pounder’s results use maximum principal moment as the governing moment for failure even though
the plate material is ductile mild steel. Nowadays it would be recognise that the governing moment
for such materials should be the von Mises stress, but although von Mises (1883-1953) had
published his paper on plasticity in 1913, [5], i.e., earlier than Pounder’s first edition, it appears that
the ideas took longer than this short intervening period to disseminate amongst practising
engineers. It could also be that Pounder was considering brittle cast materials where the maximum
principal moment criterion would be a more appropriate governing criterion. However the
experiments that he references in his booklet clearly state that ductile mild steel was used.

It is curious, also, that the SDM states that ‘the values are obtained using Pounder’s formula allowing
corners to lift’. Corner uplift, which may occur with uniaxial supports, is a non-linear phenomenon
that would not easily be treated in the era in which Pounder was working, and there is no evidence

Copyright © Ramsay Maunder Associates Limited (2004 – 2016). All Rights Reserved.
in his booklet of this non-linearity having being considered. Indeed, in his introduction, Pounder
expresses clearly what he means by the simply supported condition:

“Flat plates may be divided into two general groups, viz., those which are simply supported at their
perimeters – of which the bolted plate or cover is the nearest approximation in practice – and those
having their edges rigidly fixed or built-in, i.e., encastré.”

It is also the case that, through St Venant (1797-1886), the stresses at the centre of the plate would
not be significantly influenced by small perturbations to the boundary conditions as would be
required to allow corner uplift.

Pounder states that his approximate Navier equation is accurate for large aspect ratios but less so,
and indeed non-conservative, as the plate tends to the square shape and this is indeed seen in figure
6. Pounder realises that this needs correction and proposes another approximate solution based on
his ‘strip’ method. His result is conservative for all plate aspect ratios considered but, for the square
plate, under-predicts the exact moment by some 50%.

Figure 6: Ultimate load parameter based on elastic moments at plate centre

If, however, Pounder had adopted the correct reading of Navier with an appropriate value of
Poisson’s ratio then the first terms of the series would have given a more consistent, whilst still
conservative, approximation to the truth.

Copyright © Ramsay Maunder Associates Limited (2004 – 2016). All Rights Reserved.
Roark’s Formulas for Stress and Strain – Plastic Solutions
Unlike the SDM’s elastic solutions, those provided in Roark are indeed true limit analysis solutions
and, as has been seen in NBC02, are significantly nearer to the safe limit solution than the overly
conservative elastic solutions of the SDM. Roark’s solution, [7], for the NBC02 plate, is still some
30% lower than that achieved by EFE and it is educational to understand why this is. The plastic
collapse loads are presented in Roark in terms of a parameter beta which is a function of the plate
aspect ratio as shown in the reproduced extract of figure 7.

Figure 7: Ultimate load parameter from Roark

In order to discover why the ultimate load parameter presented in Roark is, for NBC02, some 30%
lower than that achieved with EFE, the primary reference has been retrieved and examined, [6]. The
authors, Save & Massonnet, approached the solution for this problem using both lower-bound and
upper-bound techniques and then took the average (mean) of the two solutions to provide the
ultimate load parameters that were later published in Roark. The upper-bound limit solutions are
developed using the yield line technique with the square yield criterion whereas the lower-bound
solutions are taken from the earlier work of Shull & Hu, [9], where the Tresca (1814-1885) yield
criterion is adopted.

With regard to the use of the Tresca yield criterion, it should be noted that limit analysis requires
solution of a mathematical programme and until fairly recently this would have been a linear
programme where the constraints (the yield criterion in this case) would need to be linear. Tresca is
clearly a safe linearisation of the von Mises ellipse as it sits inside (inscribes) von Mises ellipse – see

Copyright © Ramsay Maunder Associates Limited (2004 – 2016). All Rights Reserved.
figure 11. Nowadays, more general forms of constraint may be efficiently handled using second-
order cone programming (SOCP) and so this restriction has been removed. It is useful to record here
that the work of Lode, [10], demonstrated, experimentally, that the von Mises yield criterion is a
more accurate predictor of yield than Tresca for ductile metallic members.

The upper and lower-bound ultimate load parameters from Save & Massonnet’s work are shown,
together with the average values used in Roark, in figure 8 – note that no result for the lower-bound
solution was provided for the aspect ratio 0.6. Two additional curves have been added from the
elastic solutions in the previous section of this article. These are both based on the exact von Mises
moment at the centre of the plate with the first being unscaled and the second being scaled with a
factor of 3/2 – see equation 7.

Figure 8: Ultimate load parameter from Roark

There are a number of interesting points that are revealed from the plot shown in figure 8. The idea
that the engineer can make an assessment of collapse based on an (accurate) elastic analysis and
then scaling up the load by the factor of 3/2 (equation 7) is sound, robust and always conservative
for a uniform thickness plate irrespective of the support and loading conditions. And so one should
not expect to see the lower-bound plastic solution lying, as it does, below the scaled elastic curve.
Admittedly it is using a different yield criterion (Tresca rather than von Mises) but, even so, it is
indicative of the fact that the lower-bound solution of Shull & Hu is rather coarse. Further evidence

Copyright © Ramsay Maunder Associates Limited (2004 – 2016). All Rights Reserved.
for this statement is that, for small aspect ratios, the unscaled elastic curve is above that for the
lower-bound plastic solution. A second point to consider is that, in general, the idea of taking the
mean of upper and lower-bound solutions as the true solution is flawed unless one can guarantee
that the the two approximations are equi-spaced about the true solution. In this case, with the very
crude lower-bound solution, it will be demonstrated, see figure 14, that the mean solution remains
conservative.

It has already been recognised, in the second article of this series, [11], that using the yield line
approach with the square yield criterion can produce significantly different results than would be
achieved with the von Mises criterion; it is not possible, a priori, to determine whether these will be
greater than or less than the value achieved with von Mises. The upper-bound results of Save and
Massonnet were generated using the square yield criterion, and it is of interest to confirm their
methodology and then to see how the yield line pattern compares with the utilisation at collapse.

Figure 9: Yield line solutions to NBC02 using the square yield criterion

Yield line solutions for the NBC02 plate are presented in figure 9. The first solution assumes a simple
failure mechanism commonly used for this configuration of plate, and a geometric optimisation
process is carried out, by varying J, to find the lowest ultimate load parameter (recall that this
approach is an upper-bound approach). In this solution the blue lines represent lines where the
yield moment is reached in sagging. The second solution is that produced by a modern limit analysis
tool (LimitState:SLAB) but here the blue and red lines represent rotations with line thickness being

Copyright © Ramsay Maunder Associates Limited (2004 – 2016). All Rights Reserved.
proportional to the angle of rotation. For this plate configuration the two solutions are broadly
similar in terms of the collapse load (SLAB producing a slightly lower, and therefore more accurate,
value than the assumed pattern) and collapse mechanism. It is noted, however, that the collapse
load is almost 7% below the one produced by EFE (which is a lower-bound) using the von Mises yield
criterion (231kPa). The reader interested in discovering more about the yield line technique and
more modern computational approaches to the collapse of plates (RC slabs in this case) might like to
read [12].

The yield line solution shown above provides a collapse load that is virtually identical to that
presented in Save and Massonnet’s paper (see figure 9) demonstrating that they probably
approached the solution in the same manner, using an assumed collapse mechanism, which was
then geometrically optimised to produce the lowest upper-bound collapse load.

In order to ensure that the results from EFE, for this plate configuration, are sufficiently accurate a
mesh convergence study was first conducted for the square plate and the results are presented in
figure 10 which shows the mesh, contour plots of utilisation and the ultimate load parameter.

4.96

Figure 10: Mesh convergence for EFE with the square plate

The convergence observed in figure 10 is rapid and approaches from below the theoretical solution,
as it is a lower-bound solution. Based on these results the fifth mesh is deemed to provide sufficient
accuracy for the purposes of this study, and from the basic mesh is obtained using four levels of
uniform mesh refinement.

The ultimate load parameter for the EFE result using the basic mesh (1x1x4 elements) is virtually
identical to that of Save and Massonnet even though a different yield criterion was used. It is
therefore interesting to investigate the solution produced by EFE in more detail and to see why this

Copyright © Ramsay Maunder Associates Limited (2004 – 2016). All Rights Reserved.
might be. In ensuring that the yield criterion is satisfied, EFE constrains the moments to lie within
the yield surface at a set of sample points in each element. The scheme uses seven sample points
comprising points at the vertices, at the mid-sides and at the centroid of an element. The moments
at these points are compared with the yield surface in figure 11. This figure presents the moments
in the form of normalised principal moments where the normalisation involves dividing the principal
moment by the moment capacity for the plate.

von Mises Ellipse

Tresca Polygon

Figure 11: Sample point moments and yield surfaces

The figure shows the von Mises ellipse together with the (inscribed) Tresca polygon. The sample
point moments for two meshes have been added to the diagram and, for the basic mesh (mesh 1),
the sample point moments have been numbered to correspond with the sample point numbering
shown on the image of mesh 1. Contour plots of principal moments are also shown for the two
meshes, and may be compared with the utilisations presented in figure 10. For the basic mesh, all
sample point moments lie on or within the Tresca polygon except those at the corners of the plate.
As such, it is not the case that EFE is producing the same result as offered by S&M despite the
closeness of the predicted ultimate load parameters. It is interesting to observe that, for the refined
mesh (mesh 6), all sample point moments lie virtually on the von Mises ellipse (mostly in the sag-sag
quadrant – as expected) as indicated by the near uniform unit utilisation for this mesh shown in
figure 10.

Whilst limit analysis, as a mathematical programming (optimisation) problem, leads to the collapse
solution in a single step, the same solution should be obtained, albeit in a significantly less efficient
manner, via an incremental finite element analysis with a non-linear (elastic/perfectly-plastic)

Copyright © Ramsay Maunder Associates Limited (2004 – 2016). All Rights Reserved.
material model. Although the verification of EFE, using such incremental finite element methods,
has been discussed in detail in the first of this series of articles [1], it is instructive to see how
plasticity develops as the load is increased. For this purpose the NBC02 plate configuration will be
used and the contours of utilisation (von Mises stress divided by the yield stress) for a quarter model
of the NBC02 plate are shown with increasing load in figure 12.

103kPa (fully elastic) 118

  = 10.08

170 231(collapse)
Figure 12: Utilisation contours (elastic to collapse)

The results of figure 12 were created using a solid representation of the plate with (kinematic)
boundary conditions to simulate the plate problem. It is seen that plasticity develops at the centre
of the plate first and then at the corners and that with increasing load these regions grow towards
each other until they meet and lead to collapse.

Whilst it is possible, as shown above, to generate accurate plastic collapse loads with incremental
finite element analysis, when compared to the mathematical programming approach used in limit
analysis, the process is comparatively inefficient. The incremental approach is also fraught with
many difficulties for the inexperienced engineering analyst with the result achieved being very
dependent on a plethora of settings in the finite element system, the default values of which may
not be suitable for this sort of analysis. In this context, it was telling that of the readers who
responded to NBC02, only two managed to get close to the limit analysis solution and in order to do
this they used solid rather than plate representations of the problem.

Limit analysis techniques using mathematical programming provide a much more efficient approach
to predicting the collapse of plates and if a lower bound approach is adopted, is an approach that is
safe even in inexperienced hands. This last point certainly cannot be said for conventional finite
element analysis where significant experience is required on the part of the engineer if finite
element malpractice is to be avoided.

In figure 13 the utilisations from EFE for the square yield criterion and the von Mises yield criterion
are plotted. It is interesting to compare the yield line patterns of figure 10 with the utilisations at
collapse from figure 13(b). The same form of collapse mechanism can be seen with the ‘hinges’
being developed along the centre line of the plate and then moving off towards the corner.
However, whereas the yield line approach concentrates yield along lines, the actual solution in terms
of moments allows more diffuse yielding. It is seen that, as predicted by the SLAB solution, the
yielding in the line from the centre to the corner is also more diffuse than would be suggested by the
assumed yield line pattern.

Copyright © Ramsay Maunder Associates Limited (2004 – 2016). All Rights Reserved.
  = 9.38   = 10.08

(a) Square yield criterion (b) von Mises yield criterion

Figure 13: Utilisation for NBC02

The ultimate load parameter from EFE for the square yield criterion in figure 13(a), (9.38) is slightly
lower than that shown in figure 9(b) for the refined yield line model (9.39). This is consistent with
the two approaches being, respectively, lower and upper bound solutions to the same problem and
also shows the accuracy of the two solutions – the percentage difference being around 0.1%.
Comparing the results for the two yield criteria in figure 13 it is seen that with the von Mises ellipse
the plate member is able to take some 7.5% greater load than would be the case if the square yield
criterion had been used. The influence of different yield criteria has been discussed in more detail in
the second article of this series, [11], and it is seen that for this configuration of plate the yield line
technique produces a safe approximation to the true solution.

Figure 14: Ultimate load parameter from EFE

Using EFE, a tight lower-bound solution for the ultimate load parameters has been produced and
these are plotted in figure 14. The ultimate load parameter from EFE, which is a safe lower-bound

Copyright © Ramsay Maunder Associates Limited (2004 – 2016). All Rights Reserved.
value, is significantly greater than all the other methods plotted in figure 14. It is clear then that the
use of a modern limit analysis tool, such as EFE, allows the engineer to extract more capacity from
his or her plate member with a potential increase of over 40% over that offered by Roark. It has
already been shown that Pounder’s approximate elastic solution is a poor representation of the
exact elastic solution particularly for the square plate – see figure 6. If engineers are mislead by the
wording of the SDM into believing that the values therein presented are real ultimate load
capacities, then the potential under-prediction of the actual collapse load is some 175%.

Serviceability Considerations
The main part of this article deals with the accurate prediction of the maximum elastic moment,
which might be used in an allowable stress design, and for the plate strength based on plastic
collapse which might be considered as the ULS in a limit state design. As plates, depending on the
configuration, are generally rather flexible members, the engineer needs also to consider the SLS
condition of maximum displacement. For a particular plate configuration one or other of the ULS
and SLS conditions will govern the design, and it is therefore of interest to establish which condition
governs for plates of different configurations.

Based on an elastic design of the sort intimated in the SDM one might consider the ULS condition to
be taken as the load to cause first yield (von Mises moment at the centre of the plate) and the SLS
condition as the deflection at the plate centre reaching the span, , divided by 100. Strictly
speaking, in limit state parlance, a first yield condition might also fall into the category of SLS since it
does not cause gross failure of the plate. If the ULS and SLS loads are defined, respectively, as Pu
and Ps, then the load ratio Pu/Ps may be used as an indicator as to which of the two conditions
govern the design of a particular plate; if the ratio is less than unity then ULS governs whereas for
ratios greater than unity SLS governs. The variation of load ratio with aspect ratio is shown in figure
15 for the configuration therein defined.

Figure 15: Governing conditions

This plot is for the particular set of values for the thickness, the yield stress and the dimension 
defined in the figure. For different values of these parameters the curve will move up or down the
diagram. Since the variation seen is essentially quadratic, then, observing the correct units for the
various parameters, one may adopt the following equation for a general plate configuration and if

Copyright © Ramsay Maunder Associates Limited (2004 – 2016). All Rights Reserved.
the SLS condition needs to be modified then a factor needs to be applied to Ps, likewise if the ULS
condition needs modification then a factor will need to be applied to Pu.

L 
10 [ ]  [ A] (14)
= .−1.94 + 4.22α − 0.240 ∙ ∙ ∙
M [ ] 1 275

There are practical cases where ULS governs the design of a plate particularly, for the sort of plates
considered in the SDM, when the plate thickness is large and when the aspect ratio is small. This
trend is seen in the SDM tables where some of the 10mm and 12.5mm thick plates are highlighted
with an asterisk to indicate this.

If the allowable stress is factored down from the yield stress then this will have the effect of
reducing Pu and therefore moving the graph downwards. The same effect will occur if an overtly
conservative approximation of the true elastic stress is used, e.g., from Pounder’s equation. The
effect then is to give a false impression of where the transition between ULS and SLS governance
occurs and because the line is moving down then it will be that ULS is seen to govern when in reality
it should be SLS.

Practical Conclusions
Having uncovered a range of issues with currently published results for the collapse analysis of steel
plates and, hopefully, provided sound explanation as to why these have occurred, it is useful now to
provide some guidance to the practising engineer.

It has been shown in this article, which studies one particular, and common, configuration of plate,
viz., the simply supported rectangular plate with a uniformly distributed load, that published data
investigated in this article is overtly conservative. Whilst engineers need to demonstrate a
conservative attitude to their work, such overt conservatism, as exhibited here, is extremely
wasteful of both human and material resources and unnecessary in the light of the availability of
modern software tools like EFE. If, for example, the SDM’s ‘ultimate load capacities’ are taken at
face value then, for example, the collapse load for the NBC02 plate would have been under-
predicted by some 125%! Alternatively, the plate thickness would have needed to be increased from
8mm to 10mm or by some 25% in order for the plate to pass assessment based on this assumed
capacity. Equally, if a plate was being reassessed for a new duty, then unnecessary human effort
would be required to work out how the plate could be strengthened and, similarly, unnecessary
additional material, perhaps in terms of stiffening beams, would have needed to be added for the
design to be compliant with what turns out to be erroneous load capacities.

Through the application of modern and safe limit analysis software tools more appropriate plastic
collapse loads can be established. Indeed, when it comes to the prediction of first yield, which might
concern designers who base their assessment on elastic principles and an allowable stress design
approach, then the elastic solution used in the SDM can also be updated with the theoretically exact
solution.

Copyright © Ramsay Maunder Associates Limited (2004 – 2016). All Rights Reserved.
One of the aims of this article has been to attempt to clear the fog of misunderstanding and the use
of outdated methods prevalent in currently published results, so that the practising engineer might
be able to clearly see the results that he/she should use.

It might be a bold statement but the following results are both sound and conservative. The elastic
solutions are based upon the proper theoretical solution for the plate configuration being
considered and the limit solutions come from a lower-bound approach which is guaranteed to be
safe. Clearly it should be born in mind that if the overtly conservative results that a designer is
already using are masking other uncertainties, then the conversion from archaic to modern strength
solutions might cause a problem. However, if the designer is aware of and is controlling these
uncertainties then the results presented here should offer a means of obtaining more strength from
his/her steel plate.

1.00 0.90 0.80 0.70 0.60 0.50 0.40 0.30 0.20



 3.48 3.48 3.49 3.51 3.59 3.78 4.22 5.18 7.53

 6.24 6.27 6.37 6.56 6.90 7.45 8.39 10.08 13.67
Table 3: Updated ultimate load parameters

As an engineer one is aware that, for a uniform thickness plate, the load to cause plastic collapse is
equal to or greater than 3/2 times the load to cause first yield – see equation 7. The actual factor for
this plate configuration is greater than this and lies in the range 1.79 ≤   ⁄  ≤ 1.99 depending on
the particular plate aspect ratio.

Design Example
The owner of a building has asked you, as a consultant structural engineer, to assist with the design
of a floor plate to cover a 2m by 0.6m rectangular cut-out in a reinforced concrete floor. The plate
will be simply supported and subjected to a uniformly distributed load of 110kPa. The owner is
prepared to accept a maximum in service deflection of 0.6m/100 = 6mm and wants the plate to be
protected from collapse if the load is accidentally doubled. The owner has also suggested, because
of the potential uncertainty/variation in the working load, that this does not lead to a stress that
exceeds 0.5 of the yield stress.

The design of the plate may be governed by maximum displacement exceeding the allowable figure,
(SLS considerations), or by the maximum stress exceeding 0.9 times the yield stress, (ULS
considerations). To determine which of these conditions govern the design of this plate we will use
figure 15. For an aspect ratio of 0.3 this gives a load ratio of 0.84. This load ratio needs to be
factored by 0.5, for the factor on the yield stress, and 2 for the size of the plate and leads to a load
ratio of 0.84. As this is less than unity the design is governed by ULS considerations.

Consulting the SDM the thickness required to support the service load is 12.5mm. However, if we
make use of the updated data offered in this article then, from figure 8, the value of the ultimate
load parameter (elastic) is 5.2. Using equation (4) this gives a required thickness of 9.6mm and the
nearest (thicker) thickness to this is 10mm.

Turning to the requirement that the plate needs to support 2x110=220kPa before plastic collapse we
can look at Roark – see figure 14. This gives a ultimate load parameter (plastic) of 7.75. Using

Copyright © Ramsay Maunder Associates Limited (2004 – 2016). All Rights Reserved.
equation (6) with a 10mm thickness this leads to a collapse pressure of 177kPa which is less than the
duty requires. Increasing the thickness to 12.5mm gives us 278kPa which is more than is required.
However, if we had used the more accurate EFE results then the plastic collapse load for the 10mm
thickness plate (231kPa) would have been deemed sufficient.

The outcome of this design exercise is that the conventional route would have required a plate
2.5mm thicker than the updated route. The saving in mass of steel for this small plate is about
23.4kg or about 25% of the total mass.

This is a small saving in cost for the contractor who is carrying out the modification to the building.
However, if one extrapolates this over the total amount of plate sold in the UK over a single year
then it might be seen as more significant. Unfortunately, Tata Steel was not willing to provide
information that would have enabled this figure to be quantified. Either way, the over-specification
of structural members, through the use of outdated approximations, is no longer necessary, and the
practise should be considered counter to the ideals of sustainability and the drive to reduce UK plc’s
carbon footprint.

Closure
This article has considered, in some detail, the particular and common case of a simply supported
rectangular plate with a uniformly distributed load, and has presented an updated set of results,
using modern software tools, for use by the practicing engineer. In the next article a similar
treatment will be given to another common plate configuration, viz., the fixed rectangular plate with
a uniformly distributed load.

The updating of outdated methods is important to human progress and the authors are pleased to
report that McGraw Hill Education have taken on board the need for an update to their Roark
publication and have asked the authors to prepare a corrected section for the next edition of their
book. The SCI’s acknowledge that the data they offer is rather conservative but point out that the
SDM presents a ‘traditional’ method and that it is ‘appropriate for initial design’.

Acknowledgements
The authors would like to thank:

Professor Matthew Gilbert of Sheffield University for running the NBC02 plate configuration through
his software LimitState SLAB, http://www.limitstate.com/ - see figure 8(b).

Matt Watkins of ESRD Inc. for running the incremental plastic analysis through his company’s
software StressCheck, http://www.esrd.com/ - see figure 11.

References
1) Article Number 1
2) NAFEMS Benchmark Challenge Number 2, ‘Assessment of a Simply Supported Plate with
Uniformly Distributed Load’, NAFEMS Benchmark Magazine, October 2015.
3) Steel Construction Institute, ‘Steel Designers’ Manual’, 6th Edition, 2003.
4) C. Pounder, ‘The Design of Flat Plates’, The Association of Engineering and Shipbuilding
Draughtsmen, Session 1949-1950.

Copyright © Ramsay Maunder Associates Limited (2004 – 2016). All Rights Reserved.
5) von Mises, R., ‘Mechanik der Festen Korper im plastisch deformablen Zustand’, Göttin.
Nachr. Math. Phys., vol. 1, pp. 582–592, 1913.
6) Article Number 3
7) Warren C. Young & Richard G. Budynas, ‘Roark’s Formulas for Stress and Strain’, Seventh
Edition, McGraw Hill, 2002.
8) Save, M. A., and C. E. Massonnet, ‘Plastic Analysis and Design of Plates, Shells and Disks’,
North-Holland, 1972.
9) Shull, H. E., Hu, L. W. ‘Load-Carrying Capacities of Simply Supported Rectangular Plates’,
Trans. ASME J Appl. Mech., 30: 617, December 1963.
10) W. Lode, ‘Versuche uber den Einfluss der mittleren Hauptspannung auf das Fliessen der
Metalle, Eisen, Kupfer und Nickel’, Z. Phys., 36, 913, 1926.
11) Article Number 2
12) A.C.A. Ramsay & E.A.W. Maunder, ‘Yield Line Analysis of Reinforced Concrete Slabs; is the
10% Rule Safe?’, NAFEMS Benchmark Magazine, January 2015.

Copyright © Ramsay Maunder Associates Limited (2004 – 2016). All Rights Reserved.

You might also like