Вы находитесь на странице: 1из 33

Journal of Constructional Steel Research 57 (2001) 279311 www.elsevier.

com/locate/jcsr

New design rules for plated structures in Eurocode 3


Bernt Johansson a, Rene Maquoi b, Gerhard Sedlacek
a b

c,*

Division of Steel Structures, Lulea University of Technology, SE-971 87 Lulea, Sweden MSM, Department of Civil Engineering, University of Liege, B-4000 Liege, Belgium c Institute of Steel Construction, RWTH Aachen, D-52074 Aachen, Germany

Received 18 June 1999; received in revised form 6 July 2000; accepted 31 August 2000

Abstract This paper gives an overview of Eurocode 3 Part 1.5 Design of Steel Structures. Supplementary rules for planar plated structures without transverse loading have been developed together with the Eurocode 3-2 Steel bridges. It covers stiffened and unstiffened plates in common steel bridges and similar structures. This paper presents the background and justication of some of the design rules with focus on the ultimate limit states. The design rules for buckling of stiffened plates loaded by direct stress are presented and explained. For shear resistance and patch loading the new rules are briey derived and compared with the rules in Eurocode 3-1-1. Finally, the statistical calibration of the rules to tests is described. 2001 Elsevier Science Ltd. All rights reserved.
Keywords: Steel structures; Plated structures; Design; Plate buckling; Stiffened plates; Shear buckling; Patch loading

1. Introduction New design rules for plated structures have been developed by CEN/TC250/SC3 (project team PT11). The result of the work is the ENV-version of Eurocode 3 Part 1.5 (EC3-1-5) [1]. It has been drafted in close co-operation with the project team PT2 preparing the steel bridge code and it contains rules for stiffened or unstiffened plated structures. These rules are not specic for bridges, which is the reason for
* Corresponding author. Tel.: +49-241-80-5177; fax: +49-241-888-8140. E-mail address: stb@stb.rwth-aachen.de (G. Sedlacek).
0143-974X/01/$ - see front matter 2001 Elsevier Science Ltd. All rights reserved. PII: S 0 1 4 3 - 9 7 4 X ( 0 0 ) 0 0 0 2 0 - 1

280

B. Johansson et al. / Journal of Constructional Steel Research 57 (2001) 279311

making them a part of EC3-1, which contains general rules. The table of contents shown in Fig. 1 gives an overview of EC3-1-5. In addition there is an informative annex containing formulae for elastic buckling coefcients, which has been included for the convenience of the designer. Such coefcients may alternatively be found in handbooks or by computer calculations. All verications are presented in Section 2 of EC3-1-5. For the ultimate limit states (ULS) the requirements are the same as in EC3-1-1. For the serviceability limit states (SLS) no requirements are given, only methods for nding stresses, etc. The requirements depend on the particular application; for instance, requirements for bridges are found in EC3-2 [2]. The focus of this paper is Section 4 of EC3-1-5, which contains methods for nding the resistance to plate buckling in ULS. The objective of the paper is to present the scientic background to the rules. First the mechanical models behind the rules are explained and references to source documents are given. All such models include simplications, which had to be justied by calibration of the rules against test results. Several models for each failure mode have been checked with calibrations according to Annex Z of EC3-1-1 [3] and the ones chosen to be included in EC3-1-5 are those giving the lowest scatter and the most uniform safety. EC3-1-5 explicitly permits the use of different steel grades in anges and webs

Fig. 1.

Table of contents of Eurocode 3 Part 1.5.

B. Johansson et al. / Journal of Constructional Steel Research 57 (2001) 279311

281

in so-called hybrid girders. No detail rules are given for the design of such girders but in all design rules, subscripts (f for ange and w for web) indicate the relevant yield strength. Although the rules may look unfamiliar to many engineers they are in fact only a new combination of rules from different European countries. For the time being they represent a set of useful rules for common plated structures. However, the rules are not complete in the sense that any type of plated structure is covered. There are also details that may be improved with existing knowledge but the time and funds available for the work have not allowed this. One such item is the formula for the effective area of unstiffened plates. The single formula from EC3-1-1 has been retained although it has not been harmonised with the slenderness limit between cross section classes 3 and 4. A set of formulae for different boundary conditions and states of residual stresses should be developed but this has to wait until the ENversion is prepared.

2. Design of stiffened plates for direct stress 2.1. General Plates resisting predominantly direct stresses are used as anges and webs of plateand box-girders. The distinction between ange and web is sometimes questionable. The denition used here is that a ange is subject to a distribution of direct stresses that is not very far from being uniform (no account being taken in this respect of shear-lag effects). A web is subject to a distribution of direct stresses with a signicant gradient and most often a change from tension to compression. For very wide plates used as webs or anges, it is sometimes more economical to stiffen a relatively thin plate than to increase the plate thickness in order to avoid any stiffening. A plate is normally rst stiffened transversally, i.e. by stiffeners transverse to the direction of longitudinal stresses, and, when necessary, by additional longitudinal stiffeners. When the distribution of compressive stresses is quasi uniform, the longitudinal stiffeners are equally spaced. If not, the stiffeners are located in an optimum manner in order to combine efciency and economy. The transverse stiffeners are usually parts of transverse bracings of the crosssection of the structure and for this purpose they are normally stiff in bending. There is some advantage for them being designed to full this requirement. Such transverse stiffeners are denoted rigid when they constitute nodal lines for plate buckling under the action of compressive stresses. That is a rst principle of the design rules of EC3-1-5. Accordingly, the amount of efforts devoted to check plate buckling is substantially reduced and facilitated. Possible instability is restricted to: buckling of the whole panel for unstiffened panels; buckling of unstiffened subpanels or buckling of the whole panel for longitudinally stiffened panels.

282

B. Johansson et al. / Journal of Constructional Steel Research 57 (2001) 279311

In both cases the length a of the panel is equal to the distance between the transverse stiffeners dening this panel, see Fig. 2. Those edges are so-called loaded edges. The width B of the panel is the distance between its boundaries to adjacent panels or possibly between one such boundary and a free edge. The longitudinal edges of a panel are denoted unloaded edges. A subpanel is an unstiffened plate having the length a of the panel to which it belongs and a width b. The wording loaded/unloaded refers to loading by direct stresses. That is a second principle of EC3-1-5, the scope of which is plates subject to uniaxial direct stresses only. The case of plates subject to general biaxial loading is not included at the present time. However, there are rules for patch loading, including possible interaction with bending. For longitudinally stiffened panels two extreme cases concerning the stiffening are identied in EC3-1-5. 2.1.1. The case of equally spaced multiple stiffeners in the compression zone When the behaviour of the longitudinally stiffened panel as a whole is considered the number of longitudinal stiffeners located in the compression zone is sufciently large to justify the smearing of their exural stiffness across the panel width. In addition, the behaviour of the subpanels has to be checked independently. For design purposes, the multiple stiffener approach is generally accepted when the number of stiffeners is at least three. 2.1.2. The case of a few unequally spaced longitudinal stiffeners in the compression zone Smearing of the stiffness would be too rough an approach in this case and is therefore not recommended. Instead, some account should be taken of the discrete location of the longitudinal stiffener(s). This method enables the designer to analyse special situations where the widths of the subpanels are very different because of a steep stress gradient across the panel width. It will be used especially when designing so-called web plate elements. That is in relation to a third principle, or better, with

Fig. 2.

Components of a stiffened plate.

B. Johansson et al. / Journal of Constructional Steel Research 57 (2001) 279311

283

the scope of the design rules of EC3-1-5. The latter are devoted to normal structures, thus excluding implicitly girders of very high depth where the stiffening of the web would need a large number of unequally spaced longitudinal stiffeners in the compression zone. 2.2. Unstiffened panels or subpanels This section refers to a subpanel according to Fig. 3. It is equally applicable to an unstiffened panel for which b should be replaced by B. For simplicity, the characteristic resistance is described and no partial safety factors appear. The resistance of an unstiffened subpanel of width b and thickness t is conventionally given by the squash load of the effective cross-sectional area (bt)eff: Nu (bt)efffy (1) where (bt)eff is the effective cross-sectional area of the unstiffened (sub)panel and fy is the material yield stress. The possible effect of plate buckling is clearly introduced as a penalty on the gross cross-sectional area bt rather than on the magnitude of the stress at the ultimate limit state. The effective cross-sectional area (bt)eff of a subpanel is a part r( 1) of the gross cross-sectional area (bt): (bt)eff r(bt) (2) where r, termed the effectiveness of the cross-section, is computed using the wellknown Winter formula used in EC3-1-1: r 1/lp 0.22/l2 1 p (3)

This formula accounts for favourable effects resulting from post-buckling plate behaviour, on the one hand, and for detrimental effects of unavoidable structural and geometrical imperfections, on the other. The effectiveness r depends on a single parameter, the relative plate slenderness lp, which is dened as:

Fig. 3. Unstiffened subpanel/panel.

284

B. Johansson et al. / Journal of Constructional Steel Research 57 (2001) 279311

lp

(fy/scr)

(4)

Any elastic critical stress scr is commonly written as: scr kssE (5)

where ks is the buckling coefcient and sE the so-called reference Eulerian stress: sE p2Et2/12(1 v2)b2 189800(t/b)2 (in N/mm2) More explicitly, the relative plate slenderness lp writes: lp (b/t)/(28,4e ks) and it involves: the width-to-thickness ratio (b/t), that is to plate buckling what column slenderness is to column buckling, i.e. the governing parameter; the yield stress factor e=(235/fy), which indicates that the relative plate slenderness of a given plate increases with the material yield stress (fy, to be expressed in N/mm2); the buckling coefcient ks, which amounts to 4 for a simply supported long plate subject to uniform compression but depends on the aspect ratio a=a/b of the subpanel or a/B of the (unstiffened) panel. According to the Winter formula, the penalty r applicable to the gross crosssection is seen to start (r 1) when the relative plate slenderness lp exceeds 0.673. Unfortunately, this limit does not coincide with the limit between section classes 3 and 4, which creates a discontinuity in the design rules. This inconsistency is expected to be remedied in the nal version of Eurocode 3. The validity of this formula has been extended to any type of boundary and longitudinal loading conditions by introducing the relevant buckling coefcient ks, in Eq. (7). In practice, except when a plate edge is free, the conservative assumption of simply supported edges is usually made. That is due to the difculty in assessing the magnitude of edge restraints. However, the designer is free to take the edge restraint into account if the value is justied. In addition, for a plate of constant thickness, any non-linear distribution of longitudinal direct stresses across the plate can be characterised by the stress ratio y, ratio of the extreme edge direct stresses. Finally, should the plate possibly be stiffened, then the properties of the stiffening would also affect the plate critical buckling stress. As a result, the buckling coefcient for the most general case is a function: ks ks (boundary conditions, a, y, stiffening properties) (8) (7) (6)

B. Johansson et al. / Journal of Constructional Steel Research 57 (2001) 279311

285

2.3. Longitudinally stiffened panels In this section, only the behaviour of the stiffened panel as a whole is of concern. An individual (unstiffened) subpanel is treated in Section 2.2. As mentioned in Section 2.1, two distinct design approaches may be contemplated, the application of which is mainly governed by the number of longitudinal stiffeners in the compression zone of the stiffened panel. 2.3.1. Effectiveness of the stiffened panel The fact that a plate panel is stiffened in the longitudinal direction makes its behaviour more like that of a column type structural element than that of a plate type element in the range of small slenderness. While plate-like behaviour exhibits a signicant post-buckling resistance, which may largely exceed the elastic critical plate-buckling load, the elastic critical column buckling load is an upper bound of the resistance. The behaviour of any stiffened panel lies somewhere between these two limits. For design purposes, the effectiveness of the stiffened panel is obtained by a simple interpolation between the value of the effectiveness rp for the plate-like behaviour, on the one hand, and the value cc for the column-like behaviour, on the other. This interpolation is empirically based but its suitability was supported by calibration: rc (rp cc)x(2 x) cc (9)

The interpolation is governed by a factor x that measures the vicinity of the elastic critical plate buckling stress scr,p to the elastic critical column buckling stress scr,c according to: x (scr,p/scr,c) 1 0 x 1 (10)

It is understandable that scr,p should not be smaller than scr,c. However, owing to approximations and simplications included in the procedures described in Sections 2.3.2 and 2.3.3, this requirement is not necessarily fullled. In order to prevent the parameter x from being negative, 0 must be adopted as a lower bound. On the other hand, the effectiveness rc must increase from cc and approach rp when x increases. Therefore, x has 1 as an upper bound. How to assess the values of scr,p, scr,c, rp and cc is discussed in Sections 2.3.2 and 2.3.3. Because expression (10) does not reect a monotonic decrease when x increases, it can be suggested, in a revised version of EC3-1-5, to simplify the process without a signicant penalty on the results by adopting simply: rc cc but not smaller than rp, where rp, is the value of rp, computed for a very long stiffened plate, i.e. a plate where the buckling coefcient no longer depends on the plate aspect ratio.

286

B. Johansson et al. / Journal of Constructional Steel Research 57 (2001) 279311

2.3.2. Plate-like behaviour Two methods are specied; each is especially applicable to specic types of stiffened panels. 2.3.2.1. Multiple longitudinal stiffenersConcept of equivalent orthotropic plate The basic idea is to smear the exural stiffness of the longitudinal stiffeners across the plate width. Conceptually, that would lead to the substitution of the actual discretely stiffened plate by an orthotropic plate, referred to as the equivalent orthotropic plate in the following. Usually, multiple stiffeners are equally spaced or not far from being such. Then the properties of the equivalent orthotropic plate may be assumed uniformly distributed across the width. The buckling coefcient for the stiffened panel, designated as ks,p, may be obtained by any means: computer analysis, appropriate charts [4,5] or simply by the following approximate expressions: ks,p 2((1+a2)2+g) if a (1 g)0.25 a2(y+1)(1+d) 4(1+ 1+g) ks,p where: g=Ix/Ip d=Asl/Ap relative exural stiffness, i.e. ratio of the second moment of area Ix, of the actual stiffened panel to the second moment of area Ip(=Bt3/12(1 v2)) of the plate for longitudinal bending relative cross-sectional area, i.e. ratio of the cross-sectional area Asl of the longitudinal stiffeners without any contribution of the plate to the cross-sectional area Ap (=Bt) of the plate aspect ratio edge stress ratio, s1 and s2 being respectively the larger and the smaller edge stresses, see Fig. 4 (compression is taken as positive). (y+1)(1+d) if a (1 g)0.25 (11a)

(11b)

a=a/B y=s2/s1

Fig. 4.

Denition of the stress ratio y.

B. Johansson et al. / Journal of Constructional Steel Research 57 (2001) 279311

287

The above expressions imply that, on the one hand, the wall elements of the stiffeners do not exhibit local buckling and that, on the other hand, no instability of any stiffener in its whole (stiffener tripping) occurs before the stiffened panel reaches its ultimate strength. In that case, the stiffeners are said to be fully effective; that is, a fourth principle of EC3-1-5. The full effectiveness of any stiffener can be achieved by complying with deemed-to-satisfy requirements. Plate buckling initiated by tripping of open stiffeners is likely to result in a sudden and so-called catastrophic type of collapse, which should be prohibited. However, closed stiffeners, e.g. trapezoidal boxes, may very well have class 4 sections because the local buckling of the walls of the stiffener will not trigger a collapse. In the case of such closed class 4 stiffeners the local buckling should be considered in the same way as for subpanels of the plate. Possibly a plate can be tted with notably unequally spaced multiple stiffeners. Then the assumption that the distribution of the exural properties of the equivalent orthotropic plate is varying linearly across the panel width may look more appropriate than a uniform distribution. Then use shall be made of computer simulations or charts [5]. The elastic critical plate-like buckling stress scr,p is: scr,p ks,psE (12)

where sE is given by Eq. (6) with B instead of b. When dening the relative plate slenderness lp,o of the equivalent orthotropic plate it should be taken into account that the critical stress is referred to the gross crosssectional area A and yield load to the effective cross-section Aeff. Hence, the relative plate slenderness lp, becomes: lp,o (Aefffy/Ascr,p) (bAfy/scr,p) (13)

where bA=Aeff/A. In EC3-1-5, bA is calculated only for the compressed part of the plate, which leads to a smaller value than if the whole plate had been considered in case the stresses change sign. The equivalent orthotropic plate is characterised by an effectiveness ro for the plate-like behaviour: ro 1/lp,o 0.22/l2 1 p,o (14)

The symbols lp,o and ro are used instead of the EC3-1-5 symbols lp and r for the sake of clarity. 2.3.2.2. One or two stiffeners in the compression zoneconcept of equivalent column on an elastic foundation The following procedure is especially dedicated to situations where both the number and the location of the longitudinal stiffeners result from a notably non-uniform distribution of direct stresses, as in a web element. A special procedure is suggested which accounts for the discrete nature of the stiffening in a simple way. The elastic critical plate buckling stress scr,p is no longer based on the concept of an equivalent orthotropic plate but on one of an equivalent column

288

B. Johansson et al. / Journal of Constructional Steel Research 57 (2001) 279311

Fig. 5.

Physical model of a compression strut on an elastic foundation.

supported by an elastic foundation, see Fig. 5. The elastic critical column buckling stress of this equivalent column is used as an approximation of scr,p. The properties of the equivalent column, including its elastic foundation, must be determined so that both the number and location of the stiffeners, on the one hand, and the behaviour of the plate sheet in the direction transverse to the stiffeners, on the other, can be satisfactorily accounted for. 2.3.2.3. Case of one stiffener When there is only one longitudinal stiffener in the compression zone, the location of the equivalent column is that of the stiffener. For sake of simplicity, possible stiffeners in the tension zone are fully disregarded. Accordingly, the single stiffener divides the width B of the panel into two subpanels of width b1 and b2, respectively, and the elastic foundation is represented by the plate, see Fig. 6.

Fig. 6.

Notional cross-sectional area of equivalent column.

B. Johansson et al. / Journal of Constructional Steel Research 57 (2001) 279311

289

The gross cross-section of the elastically founded equivalent column is used for the determination of the section properties (cross-sectional area A, second moment of area Isl about an axis through the centroid and parallel to the plate sheet). It is composed of the gross cross-sectional area Asl of the stiffener and a notional crosssectional area of the plate sheet, that is determined as follows from both subpanels adjacent to the stiffener, see Fig. 6: half the width of the subpanel when fully in compression; one third of the width of the sole compressed part of the subpanel when stresses change from compression to tension. The effective cross-sectional area of the equivalent column is used for the computation of bA. This consists of the effective parts of the plate adjacent to the stiffener and if the stiffener is partially effective only, due account shall be taken of an effective cross-sectional area of the stiffener. This may be the case if a closed stiffener is used. In the absence of any elastic foundation, the buckling length of the equivalent column would be equal to the distance a between the transverse stiffeners. It is noted that the latter are designed so as to be rigid, on the one hand, and that simple supports have been conservatively assumed for the stiffener, on the other hand. In addition, the variation, over the length a, of the compressive force in the stiffener is disregarded in the following. Owing to the plate effect, the buckling length ac of the equivalent column will be smaller than the distance a. In accordance with the physical model, it is found to be: ac 4,33 Islb2b2 1 2 t3B
0.25

(15)

The elastic critical column buckling stress that is taken as an appraisal of scr,p is given as scr,p scr,p 1.05E Islt3B Ab1b2 if a ac (16a) (16b)

p2EIsl Et3Ba2 if a ac 2 2 Aa 4p (1v2)Ab2b2 1 2

2.3.2.4. Case of two stiffeners When there are two stiffeners in the compression zone, the procedure described above is applied three times, see Fig. 7. In a rst step, each of these stiffeners is considered assuming that the other one acts as a rigid support. The value of scr,p is given by Eq. (16) with b1=b and 1 b2=b and B=B is taken as the sum of b and b. To account for possible simul2 1 2 taneous column buckling of both stiffeners, a second step is required. That is done

290

B. Johansson et al. / Journal of Constructional Steel Research 57 (2001) 279311

Fig. 7.

Procedure for two stiffeners in the compression zone.

by means of an intuitive conservative trick. In that step, both stiffeners are lumped into a single one having the following properties, see Fig. 7: the cross-sectional area A is the sum of those computed earlier for the individual stiffeners; the second moment of area Isl is the sum of those computed earlier for the individual stiffeners; the lumped stiffener is located at the position of the resultant of the forces in the individual stiffeners. The whole procedure provides the designer with three values of scr,p, of which the lowest one should be selected. 2.3.3. Column-like behaviour The elastic critical column-like buckling stress scr,c is dened as the Euler stress for out-of-plane buckling of an equivalent column represented by the part of the stiffened plate that is in compression. scr,c p2EIx,c/Aca2 (17)

where Ixc is the second moment of area for longitudinal bending and Ac the gross area of the equivalent column. This buckling stress appears as a characteristic of the compression part of the stiffened panel, assuming that this part is released from any support along its longitudinal edges, is subject to uniform compression and has a buckling length equal to the length of the stiffened panel. That clearly appears as a set of conservative assumptions. When there is a signicant gradient of the direct stresses along the length of the

B. Johansson et al. / Journal of Constructional Steel Research 57 (2001) 279311

291

compressed part of the stiffened panel, undue conservatism can be avoided by reducing appropriately the buckling length, which then becomes smaller than the length a of the stiffened panel. The relative column slenderness lp,c then writes: lp,c (bAfy/scr,c) (18)

The effectiveness cc for the column-like behaviour is given by the reduction factor for column buckling given as: cc 1 f+ (f2l2 ) p,c (19)

with f=0.5[1+ae(lp,c 0.2)+l2 ]. p,c Because of the non-symmetry about the buckling axis due to one-sided longitudinal stiffeners, on the one hand, and the nature of built-up section (the stiffeners being welded onto the plate), on the other hand, due allowance is made for a geometric imperfection larger than 1/1000 of the buckling length (the latter is the one covered implicitly by the regular European column buckling curves). The initial out-ofstraightness accounted for is 1/500 of the buckling length, which is done by increasing the imperfection coefcient ae to [6]: ae a0 [0.09/(i/e)] where: i (Ix,c/Ac) (20)

and e is the largest of the distances from the neutral axis of the stiffened panel to the centre of the plate or the centroid of the one-sided longitudinal stiffeners (alternatively of either set of stiffeners when both-sided stiffeners), see Fig. 8[7]. Because of the better stability of closed section stiffeners, distinction is made between types of stiffener sections according to: Curve b (a0=0.34) for hollow section stiffeners

Fig. 8.

Distances e1 and e2.

292

B. Johansson et al. / Journal of Constructional Steel Research 57 (2001) 279311

Curve c (a0=0.49) for open section stiffeners.

2.3.4. Effective cross-sectional area of the stiffened panel The effective cross-sectional area of the stiffened panel is composed of: The gross cross-sectional area At,o of the part of the equivalent orthotropic plate located in the tension zone: At,o (Asl)t ( bt)t (21)

where (Asl)t is the gross cross-sectional area of all the stiffeners located in the tension zone, and (Sbt)t is the gross cross-sectional area of all the subpanels that are fully in tension. The effective cross-sectional area Aeff,c,o rcAc,o (22)

of the part of the equivalent orthotropic plate located in the compression zone, where Ac,o accounts for possible plate buckling of the subpanels: Ac,o (Asl)eff,c ( bt)eff,c

where (Asl)eff,c is the effective cross-sectional area of all the stiffeners located in the compression zone, and (Sbt)eff,c is the effective cross-sectional area of all the subpanels that are fully or partially in compression. For very wide anges there is a further reduction of the effective area with respect to shear lag according to EC3-1-5.

3. Design of plates for shear 3.1. General The resistance of slender plates to shear is based on the rotated stress eld theory as proposed by Hoglund [8]. It is a tension eld theory that is capable of predicting the resistance of short as well as long panels and it replaces the two methods in EC3-1-1. At a certain slenderness the plate reaches its yield resistance but this does not necessarily mean the maximum resistance. Strain hardening and the contribution from the anges makes it possible to utilise higher resistance without excessive deformations. In EC3-1-5 the maximum strength in shear is put to 0.7fy for steels of grade S355 and lower. For higher grades the strain hardening is less pronounced and there are no test results available. Hence, a more conservative strength, 0.6fy, is proposed. In EC3-1-1 there are special rules for rolled beams for which a shear area

B. Johansson et al. / Journal of Constructional Steel Research 57 (2001) 279311

293

larger than web area is dened. That is another way of taking the increased resistance into account and it can not be combined with the above mentioned increased strength which refers to the geometrical web area. The partial safety factors gM0 and gM1 have been suggested to have different values in EC3-2: 1.0 and 1.1, respectively. The reason for this is a study of the statistical distribution of yield strength and geometrical properties of beams, which justies gM0=1.0. This result is quite new and there has been no time for re-calibrating the resistance functions for various instability modes in order to use the same partial safety factor. This would be the rational solution but in the meantime a temporary solution has been introduced for the shear resistance. This is simply that the plateau is shifted in relation to the ratio between the partial safety factors. 3.2. Rotated stress eld theory for plain web The rotated stress eld theory was rst developed for girders with slender webs with stiffeners at the supports only and for girders with transverse stiffeners but without horizontal stiffeners [8]. It was later widened to include such stiffeners [9]. First, consider a girder with a slender web and widely spaced transverse stiffeners. The state of stress in the web caused by a shear force must be such that no vertical stresses appear at the edges. The state of pure shear that may exist for low loads rotates as shown in Fig. 9, for which the conditions of equilibrium are s1 t/tan(j) (23)

Fig. 9. State of stress in a slender girder web after buckling. Vertical stiffeners are supposed to be widely spaced and no vertical stresses act on the anges.

294

B. Johansson et al. / Journal of Constructional Steel Research 57 (2001) 279311

s2

t tan(j)

(24)

An observation from tests is that the compressive stress remains close to the critical shear stress and this is used as an assumption in the theory s2 tcr p2E t2 w kt 12(1v2)hw2 (25)

The ultimate strength of a web is assumed to be reached when it yields according to the von Mises yield criterion s2 s2 s1s2 fyw 1 1 From Eqs. (23)(26) the shear resistance can now be solved to tu fv in which fv fy/ 3 lw fv tcr
3

(26)

lw

1 1 4l4 2 3l2 w w

(27)

Eq. (27) is shown in Fig. 10 together with some test results for a girder with widely spaced vertical stiffeners. It is clear that the solid dots representing tests of girders with rigid end-posts t very well with the prediction, while the tests with non-rigid end-posts do not. The reason for this is the resulting tension in the web, which has to be anchored at the girder ends. Assuming that the state of stress as given by Eqs. (23) and (24) is uniform over the depth leads to the following expression for the tensile force Nt hwtwfv 1 l2 w lw tu fv
2

(28)

This force is larger than the actual force because the state of stress close to the anges will be more like pure shear. The force has to be resisted by the end-post if the full strength should be developed. If the end-post consists of a single plate the resistance to shear will be less than predicted by Eq. (27). The resistance actually used in the design according to EC3-1-5 is reduced slightly compared with Eq. (27) in order to allow for scatter in the test results and also for

B. Johansson et al. / Journal of Constructional Steel Research 57 (2001) 279311

295

Fig. 10. Shear resistance according to rotated stress eld theory together with test results for girders with widely spaced vertical stiffeners.

the systematic deviation for girders with non-rigid end-posts. This has been done by curve tting using simpler expressions than in Eq. (27). The design resistance is given by VRcd cvfywdhwtw/ 3 where cv=cw+cf. cw is found in Table 1 and cf will be discussed later. 3.3. Contribution of stiffeners The inuence of stiffeners is accounted for by their increase of the critical stress. Transverse stiffeners are assumed to be rigid, that is, they form nodal lines in the buckling pattern, and requirements for stiffness and strength are given in EC3-1-5. Longitudinal stiffeners may be exible, that is, they deform under buckling. It is clear from test results that the effect of longitudinal stiffeners will be overestimated
Table 1 Contribution from the web to shear resistance cw according to EC3-1-5 lw lw 0.83h 0.83h lw 1.08 1.08 lw h=1.20gM1/gM0 for S235, S275 and h=1.05gM1/gM0 for S420 and S460 Rigid end-post h 0.83/lw 1.37/(0.7+lw) S355 Non rigid end-post h 0.83/lw 0.83/lw

(29)

296

B. Johansson et al. / Journal of Constructional Steel Research 57 (2001) 279311

if the theoretical critical stress is used for calculating the slenderness parameter lw. There is less post-critical strength in a web with exible stiffeners than in a plain web. This is dealt with by reducing the second moment of area of the longitudinal stiffeners to one third of the actual value when calculating the critical stress. This reduction has been considered in the following approximate formulae for the buckling coefcient from Annex A3 of EC3-1-5 kt 5.34 4(hw/a)2 ktsl kt 4 5.34(a/hw)2 ktsl ktsl 9 hw a
2 3/4 1/3

(30) (31) 2.1 Isl tw hw (32)

Isl 2 twhw

but not less than

In (32) Isl denotes the sum of the second moments of area of all longitudinal stiffeners. In addition to the check for buckling of the whole stiffened panel there is a check for buckling of the largest subpanel, assuming that the stiffeners are rigid. A comparison between the resistance to shear according to EC3-1-1 and EC3-15 is shown in Fig. 11. This comparison assumes that the anges do not contribute. The resistance according to EC3-1-5, shown by solid curves, is the same in both diagrams because the inuence of the panel length is reected only by its inuence on the slenderness parameter lw. This is also true for the simple post critical resistance according to EC3-1-1, which is quite close to the resistance for a non-rigid end-post according to EC3-1-5. This is because EC3-1-1 does not have any requirements other than that there should be a stiffener at the end of the girder. The draw-

Fig. 11. Comparison between resistance to shear without contribution from anges according to EC31-1 and EC3-1-5 assuming gM1=gM0.

B. Johansson et al. / Journal of Constructional Steel Research 57 (2001) 279311

297

back of the tension eld method (it overestimates the resistance for short panels and underestimates it for long panels) has been eliminated by the method for a rigid endpost in EC3-1-5, which gives fair estimates of the resistance for any length of the panel, including girders with no intermediate transverse stiffeners. Another advantage of the new rules is that they are simpler to use. 3.4. Contribution from anges The intermediate vertical stiffeners prevent the anges from moving towards each other. This effect is taken into account by adding a tension eld that can be supported by the anges acting as beams supported by the stiffeners according to Fig. 12. This is a much smaller tension eld than that of EC3-1-1 because the rotated stress eld already catches the post buckling resistance of the web alone. After some simplications the contribution from the anges can be expressed as cf bft2fyf 3 f ctwhwfyw 0.25 1 MSd Mf.Rd
2

(33)

1.6bft2fyf f a th2 fyw w

(34)

An example of the resistance to shear including the effect of the anges is shown in Fig. 13. The resistance according to EC3-1-5 is compared with the one according to EC3-1-1. The simple post critical method does not take the contribution of the anges into account. The tension eld method does take the effect into account but in such a way that the effect disappears when the slenderness is lower than 0.8. This does not reect the real behaviour of a girder.

Fig. 12.

Tension eld supported by the anges.

298

B. Johansson et al. / Journal of Constructional Steel Research 57 (2001) 279311

Fig. 13. Resistance to shear for a girder with bf=25tf, tf=3tw and fyw=fyf=355 MPa, MSd=0 and assuming gM1=gM0.

4. Design for patch loading 4.1. General The rules for the resistance of a web to patch loading are new in the Eurocode context and have been developed by Lagerqvist [10] and Lagerqvist and Johansson [11]. The new rules use the same format as other buckling rules. The three verications in EC3-1-1 for crushing, crippling and buckling have been merged into one verication. The new rules also cover a wider range of load applications and steel grades. The rules have been checked for steel grades up to S690 and there is no longer any need for the special formula for S460 in Annex D of EC3-1-1. 4.2. Model for patch loading resistance The design rules in EC3-1-5 cover three different cases of patch loading. Because of space limitation only the most common case is dealt with here, see Fig. 14. The design procedure includes the following parameters: a yield resistance Fy a slenderness parameter l=Fy/Fcr where Fcr is the elastic buckling force

B. Johansson et al. / Journal of Constructional Steel Research 57 (2001) 279311

299

Fig. 14. Patch loading. Denition of parameters.

a resistance function c=c(l) which reduces the yield resistance for l larger than a certain limiting value. The characteristic resistance is written as FR Fyc(l) and the parameters are written as Fy fywtwly Fcr kF pE t 12(1v2)hw 0.47 1 l
2 3 w

(35)

(36) (37) (38)

c(l) 0.06

The expression in Eq. (38) was originally proposed [10] but during the drafting of EC3-1-5 it was simplied to c(l) 0.5 1 l (39)

The mechanical model according to Fig. 15 is used for the yield resistance. The

Fig. 15.

Mechanical model for the yield resistance for patch loading.

300

B. Johansson et al. / Journal of Constructional Steel Research 57 (2001) 279311

mechanical model has four plastic hinges in the ange and the plastic moment resistance for the inner plastic hinges, Mi, is calculated under the assumption that the ange alone contributes to the resistance. For the outer plastic hinges, Mo, it is assumed that a part of the web contributes to the resistance. This assumption is based on the observations from the tests that the length of the deformed part of the web increased when the web slenderness increased. With a simplied expression for Mo, the effective loaded length ly, for the model in Fig. 15 is given by ly ss 2tf(1 where m1 fyfbf fywtw hw tf
2

m1+m2)

(40)

(41)

m2 0.02

(42)

The buckling coefcient kF in Eq. (37) was determined on the basis of the results from an FE analysis. The FE analysis included the inuence from the stiffness of the anges as well as the length of the applied load and expressions for kF, where the inuence of these parameters are included can be found in Ref. 10. These expressions were simplied in EC3-1-5 to kF 6 2 hw a
2

(43) (44)

ss kF 2 6 6 hw

It is also necessary to consider the interaction with bending moment. This inuence is accounted for by Eq. (45), from which it follows that when the ratio Ms /MR 0.5 the bending moment has no inuence on the patch load resistance. Fs Ms 0.8 1.4 FR MR (45)

The resistance to patch loading according to EC3-1-5 is compared with that of EC31-1 in Fig. 16, which shows the quotient between the two resistances as a function of ange thickness over web thickness. The left diagram for zero loaded length shows a fairly large difference between the two design methods. The right diagram for a loaded length of 0.2 times the web depth, a more realistic case, shows a large difference only for a stocky web.

B. Johansson et al. / Journal of Constructional Steel Research 57 (2001) 279311

301

Fig. 16. Comparison between patch loading resistance according to EC3-1-5 and EC3-1-1 for a girder with bf/tf=25 and a large distance between vertical stiffeners.

5. Calibration of design rules versus test results 5.1. General The new design rules provided in EC3-1-5 were calibrated versus test results by a statistical evaluation according to Annex Z of EC3-1-1, which uses the following denitions and assumptions. It is assumed that both the action effects S and the resistance R of a structure are subject to statistical normal distributions, which are characterised by mean values m and standard deviations s, see Fig. 17. To guarantee that the distribution of the action effects S and the resistance R have a sufcient safety distance a safety index b is dened in EC1-1 as follows: b mRmS s2 +s2 R S 3.8 (46)

where mS is the mean value of the action effect, mR is the mean value of the resistance, sS is the standard deviation of the action effect and sR is the standard deviation of the resistance. The safety requirement for a structure is dened by the criterion [Rd] [Sd] 0 where [Rd] and [Sd] are design values. To dene the design values in Eq. (47), Eq. (46) may be expressed by (47)

302

B. Johansson et al. / Journal of Constructional Steel Research 57 (2001) 279311

Fig. 17. Statistical distribution of the action effects S and the resistances R.

mR

sR s +s
2 R

2 S

bsR

mS

sS s2 +s2 R S

bsS

(48)

With the notations aR sR s2 +s2 R S sS s2 +s2 R S

aS

it is possible to express the design values as Rd mR aRbsR Sd mS aSbsS (49) (50)

With the approximations aR=0.8 and aS=0.7 the design values of the action effects and of the resistances can be described independently from each other and a more

B. Johansson et al. / Journal of Constructional Steel Research 57 (2001) 279311

303

detailed investigation of the design value of the resistance can be carried out using the statistical procedure given in Annex Z of EC3-1-1. In a rst step of this procedure a resistance function rt=gR(x), the so called design model for the resistance, has to be established. This is an arithmetic description of the inuence of all relevant parameters x on the resistance r which is investigated by tests. By comparing the strength values from the resistance function rt using measured input data with test results re, see Fig. 18, the mean value correction factor b for the resistance function rt and the standard deviation Sd for the deviation term d can be determined. This gives the following formula describing the eld R brtd (51)

In most cases the probabilistic density distribution of the deviation term d cannot be described by a single normal distribution as is assumed in Fig. 17. It may be represented by a non-normal distribution, which may be interpreted as a composition of two or more normal distributions. Therefore, the density distribution for the resistance is checked by plotting the measured probability distribution on Gaussian paper. If the plot shows a straight line, the actual distribution corresponds to a unimodal normal function as assumed and the statistical data (b and Sd) are determined with the standard formulae provided in Annex Z of EC3-1-1. For the case that the plot shows a curved line the relevant normal distribution at

Fig. 18.

Plot of re rt values, mean value correction b and standard deviation Sd of the deviation term d.

304

B. Johansson et al. / Journal of Constructional Steel Research 57 (2001) 279311

the design point is determined by a tangent to the lower tail of the measured distribution, see Fig. 19. The statistical data b and Sd of the relevant normal distribution are then determined from the tangent approach to the actual distribution. In general, the test population is not representative of the total population of struc tures and therefore is only used to determine the mean value deviation b and the scatter value Sd of the design model. To consider scatter effects of parameters not sufciently represented by the test population the standard deviation of the resistance has to be increased. To this end, in addition to the standard deviation Sd, the following variation coefcients are taken into account for the yield strength and geometrical values: nfy = 0.07 for strength fy nt = 0.05 for thickness t nb = 0.005 for width b nh = 0.005 for depth h These variation coefcients are combined with the standard deviation Sd according to Eq. (52)

Fig. 19. Plot of rei/rti-values on Gaussian paper and denition of the relevant normal distribution at the design point.

B. Johansson et al. / Journal of Constructional Steel Research 57 (2001) 279311

305

sR

(ni)2+S 2 d

(52)

Using a log normal distribution for R the characteristic value Rk of the resistance function may be represented by the 5% fractile value and can be obtained from Eq. (53) Rk bmR exp( 1.64sR 0.5s2 ) R Also, the design value Rd of the resistance function may be dened by Rd bmR exp(aRbsR 0.5s2 ) R (54) where aRb = 0.83.8 = 3.04. The gM-value of the resistance function is obtained from the ratio of the characteristic value to the design value gM Rk Rd (55) (53)

In most cases instead of a 5% fractile value Rk a value Rnom with nominal values for the input parameters is used as characteristic value. To consider Rnom instead of Rk a modied partial safety factor gM* is used from: g M kgM k may be expressed by: (57) (56)

where k=Rnom/Rk. For the resistance functions for plate buckling

exp(2.0sfy0.5s2 ) 0.867 fy k b exp(1.64sR0.5s2 ) b exp(1.64sR0.5s2 ) R R

The procedure explained above is used in the following to determine the g values M for the resistance functions for plate buckling due to compression stresses, shear buckling and buckling due to patch loading. Where gM is not in compliance with the standard value gM=1.10 used for stability checks, the function Rnom is subsequently modied to reach the standard value gM. 5.2. Calibration of the design rules for shear buckling The design rules for shear buckling were checked versus test results according to the procedure given in Section 5.1. For the statistical evaluation the test results were obtained from a data bank given by Hoglund [9] which contains 166 test results for the following types of steel plate girders: girders with stiffeners at support only; girders with transverse intermediate stiffeners;

306

B. Johansson et al. / Journal of Constructional Steel Research 57 (2001) 279311

girders with longitudinal and transverse stiffeners. Only 150 of 166 tests could be used for the statistical evaluation because some of these specimens did not fail by shear buckling. Since the procedure of the design model for shear buckling depends on the arrangement of stiffeners, the available test results were subdivided into subsets, see Fig. 20. For these individual subsets the statistical evaluations were carried out and the statistical results are presented in Fig. 21. The gure shows that for all subsets g -values lower than 1.10 were determined so that a gM-value of 1.10 can be applied M in the design model g . M The sensitivity of the design model to the variation of the yield strength fyw was checked by plotting the ratio Vei/Vti versus the yield strength of the web fyw (Fig. 22). Owing to the small variation of the mean values of Vei/Vti the conclusion can be drawn that the inuence of fyw is adequately considered in the design model. 5.3. Calibration of the design rules for patch loading The statistical evaluation for the design model of patch loading was carried out with test results, which were obtained from a data bank given by Lagerquist [10]. The data bank contains test results for welded and rolled I-girders, which were loaded by patch loading, end patch loading or opposite patch loading, see Fig. 23.

Fig. 20.

Subsets of test results.

B. Johansson et al. / Journal of Constructional Steel Research 57 (2001) 279311

307

Fig. 21. g -values for the design model of shear buckling. M

Fig. 22.

Sensitivity plot for fyw.

308

B. Johansson et al. / Journal of Constructional Steel Research 57 (2001) 279311

Fig. 23.

Different cases of patch loading.

According to the data base and the various design models the test results were subdivided into the following subsets: Data set 1: Data set 2: Data set 3: Patch loading End patch loading Opposite patch loading

For these subsets the statistical evaluations were carried out, and a summary of the statistical results is presented in Fig. 24. The gure shows that for all three subsets a g -value of 1.10 is justied. M In Fig. 25 a sensitivity plot is given for the slenderness parameter l which shows that the scatter of the ratio Fei/Fti is only slightly inuenced by the slenderness parameter l.

Fig. 24.

g -values for the design model of patch loading. M

B. Johansson et al. / Journal of Constructional Steel Research 57 (2001) 279311

309

Fig. 25. Sensitivity plot for the slenderness l.

5.4. Calibration of the design rules for buckling of stiffened plates The calibration of the design model for plate buckling was carried out with test results for multiple longitudinally stiffened steel plate girders in compression which were obtained from a literature study. Unfortunately, not all of the collected tests could be used to check the design model, because either some relevant data were not given in the test reports or the tests were carried out with additional initial imperfections which are not considered in the design model. Finally 25 tests were applicable to calibrate the design model. In these tests the longitudinal stiffeners were designed as bulb ats, ats or angles, see Table 2. Some of these stiffeners do not full the design recommendations given in EC 3 Part 1.5, because they are not fully effective (class 4 section). In this case the design resistance of a longitudinal stiffened steel plate was determined by taking into account the local buckling of both subpanels and stiffeners. In addition, the shifting of the neutral axis of the stiffened steel plate due to local buckling was considered using the interaction formula for bending and axial compression which is provided in EC3-1-1, see Eq. (58). Ns kyNSeN 1.0 rcAcfy Wefffy where: ky 1 myNS 1.5 rcAcfy (58)

310

B. Johansson et al. / Journal of Constructional Steel Research 57 (2001) 279311

Table 2 Test results for longitudinal stiffened steel plates in compression Authors Dorman and Dwight [12] Test 3 (TPA3) No of tests 4 Stiffener types All tests with bulb ats

4 (TPA4) 7 (TPB3) 8 (TPB8) Scheer and Vayas [13] 1 (III A 50-70) 2 (III A 75-70) 3 (III A 75-100) Fukumoto [14] B-1-1 B-1-1r B-2-1 B-3-1 C-1-4 C-2-1 Lutteroth [15] All tests

All tests with bulb ats

All tests with ats

12

9 tests with at plates 3 tests with at plates and angles

my lc(2bM,y 4) 0.9 bM,y 1.1 The statistical evaluation was carried out for the different groups of tests given in Table 2. The results of the statistical evaluations are presented in Fig. 26 and show that a g -value of 1.10 can be applied in the design model. M

Fig. 26.

Statistical results for the design model of plate buckling.

B. Johansson et al. / Journal of Constructional Steel Research 57 (2001) 279311

311

6. Conclusions The rst common European pre-standard considering plate buckling has been published for trial application. In addition to widening the scope to stiffened plates it also includes some improvements of the present rules in EC3-1-1. The design rules have been veried by calibrations to tests, which include also steel grades above the present limit of S460. This will facilitate a future widening of EC3 to higher steel grades. The test application shows that this rst version of EC3-1-5 can be improved in several respects. The rst author has already received several questions and remarks indicating the need for clarication and corrections. Further comments and remarks are welcomed and they may be sent to the rst author for consideration in the ENversion of EC3-1-5.

References
[1] Eurocode 3 Design of steel structures. Part 1.5 General rules. Supplementary rules for planar plated structures without transverse loading. ENV 1993-1-5:1997. [2] Eurocode 3 Design of steel structures. Part 2 Steel Bridges. ENV 1993-2:1997. [3] Eurocode 3 Design of steel structures. Part 1.1 General rules and rules for buildings. ENV 19931-1:1992. [4] Kloppel/Scheer Beulwerte ausgesteifter Rechteckplatten, W. Ernst & Sohn (Tables V/1.1 to V/5.2 pp. 96105). [5] Kloppel/Moller Beulwerte ausgesteifter Rechteckplatten, Band II, W. Ernst & Sohn (Tables Q001 to Q005, pp. 130139). [6] Rondal J, Maquoi R. Formulations dAyrton-Perry pour le ambement des barres metalliques. Constr Met 1979;4:4153. [7] Jetteur PH, Maquoi R, Massonnet CH, Skaloud M. Calcul des ames et semelles raidies des ponts en acier. Constr Met 1983;4:1528. [8] Hoglund T. Design of thin plate I-girders in shear and bending with special reference to web buckling. Bull. 94, Division of Building Statics and Structural Engineering, Royal Institute of Technology, Stockholm, 1981. [9] Hoglund T. Strength of steel and aluminium plate girdersshear buckling and overall web buckling of plane and trapezoidal webs. Comparison with tests. Tech. Report, Dept. Structural Engineering 1995:4, Steel Structures. Royal Institute of Technology, Stockholm. [10] Lagerqvist O. Patch loading, resistance of steel girders subjected to concentrated forces. Doctoral thesis 1994:159 D, Department of Civil and Mining Engineering, Division of Steel Structures, Lulea University of Technology. [11] Lagerqvist O, Johansson B. Resistance of I-girders to concentrated loads. J Constr Steel Res 1996;39(2):87119. [12] Dorman AP, Dwight JB. Tests on stiffened compression panels and plate panels. In: Conference on steel box girder bridges. London: Institute of Civil Engineers. p. 63-75. [13] Scheer J, Vayas I. Traglastversuche mit ausgesteiften Blechfeldern unter allseitiger Navierscher Lagerung und konstanter Stauchung der Endquerschnitte. Schlubericht Nr. 6036-1, Institut fur Stahlbau, TU Braunschweig, 1982. [14] Fukumoto Y. Ultimate compressive strength of stiffened plates. In: Specialty conference on metal bridges. New York: ASCE, 1974. [15] Balaz I. Ausgesteifte Druckgurte von Kastentragerbrucken. Stahlbau 1987;5:14554.

Вам также может понравиться