Você está na página 1de 7

Articles Bernt Johansson Milan Veljkovic*

DOI: 10.1002/stco.200910031

Review of plate buckling rules in EN 1993-1-5


Dedicated to Emeritus Professor Dr.-Ing. Dr. h.c. Gerhard Sedlacek on the occasion of his 70th birthday
EN 1993-1-5 provides harmonized European design rules for plated structures. It is a step forward in the harmonization process, but like all the Eurocodes it includes many options for national choices, in terms of parameters and also methods, which leads to quite different results. Actually, there are four different methods for dealing with buckling problems. One purpose of this paper is to illustrate differences by way of examples and also to suggest topics for studies intended to improve the design rules and bring them closer together. The revision of EN 1993-1-5 is due to take place around 2014 and there is ample time for performing research projects that can form the basis for the new version. Examples of such topics are the recalibration of buckling curves, reviewing the interaction checks and the verification format of the reduced stress method.
but abbreviations are used where there is no risk of misunderstanding.

2 Effective width method


The effective width method for describing the effects of plate buckling due to direct stress is presented in section 4 of EN 1993-1-5. However, to cover other buckling phenomena and detailing it also relies on sections 5 to 9. This is a set of rules that covers common structures such as I-girders and box girders. The effective width method describes the effects of plate buckling due to direct stress by reducing the actual area to an effective area for each plate that is subject to local buckling and in the case of longitudinally stiffened plates also due global plate buckling. The resistance of the cross-section is given by a linear stress distribution with the maximum stress equal to the yield strength. It is a basic assumption of the effective width method that the yield strength can be reached in a corner of the crosssection or at a supported edge. This assumption is mostly valid and has been proven in hundreds of tests, but it may require some restrictions on web slenderness in order to avoid flange-induced buckling and the breakdown of corners between plates forming a small angle. One objection to the effective width method may be that a plate in compression may show a sharp drop in resistance after the maximum load has been passed. This may be correct for an isolated plate, but if the plate is connected to other plates that have not reached their yield strength, the strains will be controlled by these other plates. This means that the strains will be elastic and small and the drop in the resistance of the plate that has buckled will be small until the other plates reach their yield strength.

1 Introduction
EN 1993-1-5 [1] provides design rules for plated structures with the main focus on girders for which plate buckling influences the behaviour, so-called class 4 sections. However, the code also provides some design rules applicable to all types of cross-section, e.g. rules for shear lag and patch loading. It represents a big step forward in the harmonization of European design standards but due to different traditions there are still several openings for national choices. It was developed between 2000 and 2004 by a project team chaired by the first author and in which Professor Gerhard Sedlacek was a prominent member. He and his team in Aachen made substantial contributions to the standard, including providing Dr. Christian Mller, RWTH, as technical secretary. Following normal CEN procedures it would soon be time to revise EN 1993-1-5, but that has been postponed until around 2014, which means that there are several years available for finding improvements and the purpose of this paper is to indicate issues that need further studies. Several of the proposals are quite substantial tasks suitable for PhD projects and hopefully someone will be attracted to them. Since EN 1993-1-5 was finalized, new findings have become available through research projects and some will be presented here. This paper is aimed at researchers or experienced engineers and basic concepts are assumed to be familiar. The content may also be of interest for writing National Annexes. Notation follows that of EN 1993-1-5
Received 31 July 2009, accepted 4 September 2009 * Corresponding author: Milan.Veljkovic@ltu.se

2.1 Buckling of plates loaded in uniform compression


For the basic case, a plate with hinged edges under uniform compression, the effective width is given by the Winter formula:

p 0.22 beff == b 2 p

(1)

This formula results in a safe estimate of the resistance for plates with small residual stresses but may overestimate the resistance for plates with high residual stresses. This was dis-

228

Ernst & Sohn Verlag fr Architektur und technische Wissenschaften GmbH & Co. KG, Berlin Steel Construction 2 (2009), No. 4

B. Johansson/M. Veljkovic Review of plate buckling rules in EN 1993-1-5

Fig. 1. Results from stub column tests on square welded based on [2]

cussed in the EN 1993-1-5 project team and the conclusion was that it was mainly a problem of small test specimens in which the relation between weld area and plate area was large compared to that of full-size structures. Since then, more tests have been added to the database and the tendency is clear: specimens with large welds have a resistance lower than (1), see Fig. 1 from [2]. The tests, except those from the literature, were carried out on square boxes of highstrength steel (fy = 4201100 MPa) and had welds in the corners equal to the plate thickness. Some previous studies have shown a higher relative resistance for high-strength steel but this study shows that there is no such difference, which is a disappointment. The proposal in [2] is to use the formula:

1 + 2 p

1.0

(2)

section 7.1 of EN 1993-1-5 reflect this fact. The interaction formula uses the plastic bending resistance as a reference for all cross-section classes and the influence of the shear force has been reduced compared to the original model of Basler [3]. The interaction formula is actually the same as the one in EN 1993-1-1 for plastic resistance except that the shear buckling resistance is used instead of the plastic shear resistance. Further, there is a rule that the interaction need not be checked closer than hw/2 to an internal support (actually only valid for webs without longitudinal stiffeners). This latter rule is based on a judgement that the shear buckling is not critical for a transverse load that is applied close to a support. This is likely to be correct for shear force only, but here there is also a bending moment with a quite steep gradient. That

where = 0.5[1 + p(p p0) + p] with p = 0.5 and p0 = 0.6. The expression has been proposed by Mller [11] and been introduced in Annex B of EN 1993-1-5. In [2] it is applied to plates with large welds with a shorter plateau and larger imperfection factor. The curve is shown in Figure 1 and it has been calibrated to make a safe estimate of the test result. The question of the influence of the weld size is worth a closer look, which, even if not common with large welds, may be relevant and should therefore be covered. The positive effect would be that such rules would discourage overdesign of the welds. It is also worth mentioning that there are no indications that the rules for webs in bending overestimate the effective width.

2.2 Interaction between shear and bending


There are several studies that show that interaction between shear and bending is very weak. The design rules in
Fig. 2. Interaction between moment and shear force, girders with longitudinal stiffeners [4]

Steel Construction 2 (2009), No. 4

229

B. Johansson/M. Veljkovic Review of plate buckling rules in EN 1993-1-5

means that the bending moment has dropped off significantly in the design section compared to the maximum value. The consequence is that the interaction formula rarely governs. It would be worth a study to check this interaction in order to simplify the rules or even to omit the check completely for certain cases. Fig. 2 shows test results for girders with longitudinal stiffeners [4] evaluated with maximum moment and shear and normalized with the respective resistances based on Swedish design rules which are very similar to EN 1993. It is an indication that it is sufficient to check the moment and shear resistances without interaction. Further, the Swedish design code BSK [5] neglects the interaction between shear and bending for class 1 cross-sections. This is consistent with test results and can be justified by utilizing the strain hardening of the material.

In brief, the reduced stress method uses load amplifiers ult, which is the yield strength divided by the equivalent stress according to von Mises yield criterion, and cr, which is a factor for the actual stresses, to reach the critical level according to elastic plate buckling theory. These are used to define a slenderness parameter:

ult cr

(3)

This slenderness parameter is used for defining reduction factors for buckling x and z according to section 4.5.4(1) of EN 1993-1-5. The formula gives an interpolation between plate-like behaviour and column-like behaviour. The verification of the resistance (without partial factors) is carried out with the following expression in which x and z are the applied stresses:

3 Reduced stress method 3.1 General


The reduced stress method is general and it is able to deal with complex states of stress including biaxial direct stress and shear stress. For an isolated plate with longitudinal direct stress or shear stress, it results in the same resistance as the effective width method. For combined states of stress, the reduced stress method uses a criterion similar to the von Mises yield criterion. It has the merit of being general but it yields results very different from the interaction formulae discussed in section 2.2 above. The reduced stress method calculates the buckling strength expressed as a stress in each individual plate and the resistance of the cross-section is given by the plate with the lowest strength at a junction between two plates. What this means is that no redistribution of stresses is utilized, which is a very conservative approach. The practical consequence is that the slenderness of each plate has to be limited such that the buckling strength is close to the yield strength for economic reasons. This is usually achieved by providing the girder with stiffeners, which is unfortunate because longitudinal stiffeners are expensive. Actually, it was shown in [6] that for a composite I-girder bridge the depth had to be over 4 m before it became cheaper to add longitudinal stiffeners. It was designed using the effective width method, but if it had been designed with the reduced stress method, longitudinal stiffeners would have been needed at a depth of about 2.5 m. Improvements to the reduced stress method have been discussed in the commentary to EN 1993-1-5 [7] and in a recent paper from RWTH [8]. One suggestion is to allow redistribution of stresses such that different plates have different stress distributions according to their strengths. As stated in section 2 above, this would be safe. A modification to the strength criterion is suggested in [8] for the interaction between bending and shear, and the proposal shows that it is possible to bring the methods closer to each other.

x z x z 1 + x fy z fy z fy x fy

(4)

If the plate is stocky (x = z = 1), the expression coincides with the von Mises yield criterion. However, we would expect another type of interaction for slender plates. For instance, the interaction is linear for critical stresses if the buckling modes coincide. An alternative procedure for checking this interaction is found in [9], developed by Professor Torsten Hglund at KTH, Stockholm, for shell instability. The full version also covers shear and the influence of tension as well as compression. Here, the description is limited to biaxial compression. The verification is similar to (4) and is as follows:

x z x z 1 + +k x fy z fy x fy z fy
where k = 23xz

(5)

The coefficient k changes the shape of the interaction curve such that for slender plates it becomes more linear,

3.2 Example: Biaxial compression


For simplicity, this discussion will be limited to biaxial stress; shear stress is excluded. Further, it is assumed that M1 = 1.0 in order to simplify the formulae.

Fig. 3. Interaction curve for biaxial buckling according to [8]

230

Steel Construction 2 (2009), No. 4

B. Johansson/M. Veljkovic Review of plate buckling rules in EN 1993-1-5

Fig. 4. Square plate with hinged edges loaded in biaxial compression crx = fy/2

see Fig. 3 in which X = x/(xfy) and Z = z/(zfy). In this method the reduction factors x and z are calculated independently in contrast to expression (4) in which they are calculated for the combined state of stress. In order to quantify this we consider a square plate with hinged edges according to Fig. 4. Firstly, it is loaded with uniform longitudinal stresses x and its properties are such that crx = fy/2 leading to = 1.41 and x = 0.597. Now set x = 0.597fy, meaning that resistance is fully utilized, and then add z = x/2. Following section 10 it ends up in a utilization (square root of (10.5)) of 1.09 or, in other words, by adding z the plate has become overloaded by 9 %. This is a quite small effect and a FE simulation reveals that the utilization is 1.38 when z is added. The interaction formula in (5) in this case has k = 0.93 and results in a utilization of 1.31, which is fairly close to the computer simulation but on the unconservative side. Fig. 4 also shows a full comparison between (5) and the FE simulation for this plate, indicating that (5) has to be adjusted somewhat in order to be safe. This is just one example and a thorough study is needed before an improved method can be proposed. The study in [8] is a good starting point but it needs to be calibrated for several cases.

5 Design by FEM
Annex C of EN 1993-1-5 describes a method for applying FEM to the design of plated structures. In principle, it is general and can be applied to any kind of structure, but it has only been calibrated against simple structural elements. It is suggested that more systematic calibrations should be made against test results. The method will not be described in detail here but an example is presented in section 7.3 and further information can be found in [7] and [11].

6 Patch loading
The rules for patch loading are to be found in section 6 of EN 1993-1-5 and are fairly new. They use the same procedure for finding the resistance as for other buckling phe-

4 Method in Annex B
Annex B presents a method for dealing with plate buckling of non-uniform members. It is based on the reduced stress method and uses the same definition of the slenderness parameter p, see expression (3). As this is calculated for the total stress field, there may be a need for modifying the buckling curves and Annex B presents a set of curves with a new format (2). It is, however, reasonable to require that they should produce the same results as the rules for elementary cases in sections 4 and 5 when applied to such cases. This leads to the suggestion for a project to recalibrate all the buckling curves to create a single format. This would preferably be carried out based on critical stresses considering the continuity between the different plates of the cross-section. Such critical stresses can easily be calculated with the EBPlate software which can be downloaded free of charge [10]. It would then be safe to use such critical stresses in design, and if hinged conditions are used as a simplification it would be conservative.

Fig. 5. Computer simulation of patch loading, load deformation curve (top) and stress distribution in the web (bottom) [13]

Steel Construction 2 (2009), No. 4

231

B. Johansson/M. Veljkovic Review of plate buckling rules in EN 1993-1-5

Fig. 7. I-girder used as example, steel grade S 355, compression flange supported laterally

7.1 Effective width method


The cross-section is fully effective. Verification of the bending resistance is checked at midspan:
Fig. 6. Proposed new buckling curve for patch loading and test results [13]

MEd = 3500 kNm MRd = 3540 kNm MEd/MRd = 0.99 Note that the bending resistance should be calculated with the section modulus referred to the centroid of the flange. The shear buckling resistance is verified at a crosssection hw/2 = 0.5 m from the support and the stiffeners fulfil the rigidity requirements. VEd =1260 kN VRd = 1320 kN VEd/VRd = 0.95 It should also be shown that the maximum shear does not exceed the plastic shear resistance at the support and that is also fulfilled. In this example the load is relatively high and a check of the webs resistance to transverse forces may be required. However, the rules in EN 1993-1-5 section 6 are not applicable. The interaction is likely to be weak, see section 7.3, but possibly worth examining.

nomena, which is a step forward. There is, however, a difference between the patch loading rules and other plate buckling rules, which was pointed out in [12]. The patch loading curve is much lower than other plate buckling curves. The problem is that the plastic resistance for patch loading is not well defined because the resistance increases gradually with the deformation, see Fig. 5 [13]. The upper diagram in Fig. 5 shows the load as a function of the shortening of the web from a computer simulation. The solid line in lower diagram in the figure indicates the stress distribution in the web. The blue dashed line is the stress block assumed in EN 1993-1-5, which corresponds to a fairly high plastic resistance, and the red dashed line shows the stress distribution proposed in [13], which corresponds to the red dot on the curve in the upper diagram. This corresponds to a smaller plastic resistance and a recalibration of the buckling curve led to an expression more in line with other plate buckling phenomena, see Fig. 6. The curve is according to (2) with p = 0.43 and p0 = 0.65 but without the limitation less than 1.0. Actually, the curve rises as high as 1.4 for very stocky plates and from the test results it can be seen that it is safe. The increase in resistance is caused by strain hardening and redistribution of stresses. The critical load used for defining the slenderness for patch loading is a rather rough estimate for the case that rotation of the flange is assumed to be prevented at the point of load application. This is the case for most tests and also for most applications. However, this assumption is worth some consideration in the recalibration of the buckling curves suggested in section 4.

7.2 Reduced stress method


The verification of the web at midspan requires the determination of the load multipliers for yielding ult and for elastic plate buckling of the web cr. At the top edge of the web x = 348 MPa and z = 28 MPa, which results in ult = 1.06. An assumption of the boundary conditions of the web has to be made in order to determine cr. EN 1993-1-5
Table 1. Verification of the web according to section 10 of EN 1993-1-5
1 crx crz cr (10.2) x z utilization 2 1.30 1.30 0.80 1.06 0.78 0.45 1.18 3 1.10 0.98 0.90 0.55 1.02 1.10 0.98 0.79 0.74 1.20 4

7 Example of application
I-girders are the most common type of plated structure and the girder shown in Fig. 7 will be used as an example of the application of the various design rules. The crosssection is class 3 close to the limit between 3 and 4. This example has been chosen in order to make the comparison between the methods fair. A class 4 section would be very unfavourable for the reduced stress method.

232

Steel Construction 2 (2009), No. 4

B. Johansson/M. Veljkovic Review of plate buckling rules in EN 1993-1-5

does not state explicitly that hinged conditions should be used but the design buckling resistances for longitudinal stress and shear have been evaluated from tests on girders with flanges but assuming hinged conditions for the web for defining its slenderness. If those design rules are used together with critical stresses considering the restraint from the flanges, the result will be unconservative. We therefore assume hinged conditions for the web. The results are shown in column 2 of Table 1. The critical stresses have been calculated using kx = 23.9 and kz = 2. The latter figure is an approximation for a column buckling from the stress distribution in the transverse direction. The load amplifier for the combined stresses has been calculated using expression (10.6) in EN 1993-1-5 and the slenderness parameter according to expression (10.2). For the longitudinal stresses it is obviously a case of plate-like behaviour and expression (4.2) of EN 1993-1-5 has been used for x. For the transverse stresses it is a clear case of column-like behaviour (no change in the critical stress if the stiffeners at the support are removed) and expression (6.49) of EN 1993-1-1 with = 0.49 has been used for z. Finally, the utilization has been calculated as the square root of expression (10.5) of EN 1993-1-5. From column 2 in Table 1 it can be seen that this method produces a load that is 18 % higher than the resistance, which is not correct. Using EBPlate with the actual longitudinal and transverse stresses acting simultaneously gives cr = 1.10 instead of 0.80 according to expression (10.6). This difference is probably caused by the big difference between the buckling deformations for longitudinal stress with buckling over a length of 0.5 m and the buckling deformation for the transverse stress with an almost cylindrical buckle along the whole girder. The results are shown in column 3 of Table 1 and the resistance is more reasonable, only 3 % lower than in section 7.1. The method in Annex B is quite similar to the reduced stress method and the difference is the buckling curves. Using the curves for welded girders gives the results shown in column 4 of Table 1. Note that z is substantially higher than in column 3 because Annex B does not mention column-like behaviour. The great surprise is that now the plate becomes weaker. The load is now 20 % higher than the resistance. This is another example that shows the need to revise the verification format of section 10, but this is not the only problem that has to be fixed in order to get the resistance right, assuming that section 7.3 gives us the right answer. Assuming that column-like behaviour is relevant, there is a problem with the resistance given by curve c according to Table 6.2 of EN 1993-1-1. This cannot be correct for a plate, which has very small residual stresses. The plateau should be longer and the imperfection factor should be smaller. An indication in this direction can be found in Table 9.1 of EN 1994-2, which specifies maximum distances between connectors of composite plates. The limit for class 3 corresponds to = 0.44. Furthermore, the stress gradient in the transverse direction is quite steep and the governing cross-section is not at the edge where z is maximum. A remedy may be to apply the idea in section 6.3.4.2(7) of EN 1993-2 and to perform the buckling verification at a cross-section 0.25hw from that with maximum stress.

7.3 Design by FEM


The procedure according to Annex C of EN 1993-1-5 is applied. The software Abacus is used and the model is built up with type S4R shell elements four-node shell elements for general application and very suitable for thin plates. Thin plates have a thickness less than 1/15 of a characteristic length on the surface of the shell, such as the distance between supports or the wavelength of a significant eigenmode [4]. The material behaviour is bilinear type c) as shown in Fig. C.2 of EN 1993-1-5. Imperfections are introduced as eigenmodes scaled to a maximum am-

Fig. 8. Second buckling mode, which produced the lowest resistance

Fig. 9. Deformation at failure

Fig. 10. Load deflection curves with initial deformations from the first five eigenmodes

Steel Construction 2 (2009), No. 4

233

B. Johansson/M. Veljkovic Review of plate buckling rules in EN 1993-1-5

plitude of b/200 = 5 mm. The first eigenmode is an almost cylindrical buckle corresponding to buckling caused by the transverse stresses. The following eigenmodes were more wavy or mixed. The five first eigenmodes were used as imperfections in separate runs with results according to Fig. 10. The results differ only slightly, 4 % between the lowest and highest resistance. The second eigenmode, shown in Fig. 8, produced the lowest resistance, 299 kN/m, corresponding to a utilization of 0.94. As long as the eigenmodes lead to the same failure mode it is not so important as to which form is used. The deformation at failure is shown in Fig. 9 and the failure mode is web buckling followed by flange buckling. This estimate of the resistance is likely to be the best and taking it as a reference the effective width method results in 95 % , the best application of section 10 a figure of 92 % and Annex B 78 %. This is all on the safe side but may be a bit too widely spread for a simple case.

References
[1] Eurocode 3 Design of steel structures. Part 1-5 Plated structures, EN 1993-1-5:2006. [2] Clarin, M.: Plate Buckling Resistance. Patch Loading of Longitudinally Stiffened Webs and Local Buckling. Doctoral thesis 2007:31, Lulea University of Technology, 2007, ISRNLTU-DT-0731SE. [3] Basler, K.: Strength of plate girders under combined bending and shear, ASCE Journal, St 7, October 1961. [4] Hglund, T.: Strength of Steel and Aluminium Plate Girders Shear Buckling and Overall Web Buckling of Plane and Trapezoidal webs. Comparisons with Tests. Royal Institute of Technology, Dept of Structural Engineering, technical report 1995:4, Steel structures. [5] BSK Swedish design rules for steel structures. Boverket 1999. ISBN 91-7147-527-3. [6] COMBRI+ Final report of RFCS project Valorisation of Knowledge for Competitive Steel and Composite Bridges COMBRI+. [7] Johansson, B., Maquoi, R., Sedlacek, G., Mller, C., Beg, D.: Commentary and worked examples to EN 1993-1-5-Plated structural elements, EUR 22898 EN 2007, October 2007. [8] Naumes, J., Feldmann, M., Sedlacek, G.: Gemeinsame Grundlagen von Methode 1 (wirksame Breiten) und Methode 2 (Beulspannungsbegrenzung) beim Plattenbeulnachweis nach Eurocode 3 Teil 1-5. Stahlbau 78 (2009), No. 3, pp. 139147. [9] Samuelsson, L., Eggwertz, S. (Eds.): Shell Stability Handbook. London: Elsevier Applied Science, 1992. [10] EBPlate http://www.cticm.eu/spip.php?article45 [11] Johansson, B., Veljkovic, M.: Steel Plated Structures. Progress in Structural Engineering and Materials, Vol. 3, No. 1, JanuaryMarch 2001, pp. 1327. [12] Mller, C.: Zum Nachweis ebener Tragwerke aus Stahl gegen seitliches Ausweichen. Aachen: Shaker-Verlag, 2003. [13] Gozzi, J.: Patch loading resistance of plated girders Ultimate and serviceability limit state. Doctoral Thesis 2007:30, Lulea University of Technology, 2007, ISRN: LTU-DT-07/30-SE. [14] Documentation for ABAQUS v. 6.8 software, Dassault Systmes, 2008. Keywords: plate buckling; effective width; reduced stress method; patch loading; FEM

8 Conclusions
The present edition of EN 1993-1-5 specifies four methods for the design of plated structures and they are in some ways inconsistent. The rules are safe or even conservative for normal girders but in odd cases they may be unconservative. The following conclusions in the form of suggestions for studies are drawn from this review: The buckling curves should be recalibrated against test results and computer simulations to create a single mathematical format. This should preferably be carried out using critical stresses considering the continuity between the plates because they can be easily calculated with tools like EBPlate. The interaction between moment and shear in section 7 of EN 1993-1-5 should be reviewed. The question is: For which cases, if ever, is it needed? The reduced stress method in section 10 of EN 1993-1-5 should in the first place be so improved such that it allows redistribution of stresses between plates. Further, Annex B should be merged with section 10. Finally, the verification format should be reviewed because the present one only works for stocky plates. The patch loading rules should be changed in line with the suggestion in [12] and possibly also extended to cover continuous loads. The possible interaction with flange-induced buckling should be investigated.

Authors:
Prof. Dr. Bernt Johansson and Prof. Dr. Milan Veljkovic, Lulea University , of Technology, F-Huset, 97187-Lulea Sweden

234

Steel Construction 2 (2009), No. 4

Você também pode gostar