EFFECT OF END CONNECTION RESTRAINTS ON THE STABILITY OF STEEL BEAMS IN BENDING

Advanced Steel Construction Vol. 4, No. 3, pp. 43-59 (008) 43 EFFECT OF END CONNECTION RESTRAINTS ON THE STABILITY OF STEEL BEAS IN BENDING S. Amara,*, D.E. Kerdal and J.P. Jaspart 3 Department of Civil Engineering, Civil Engineering Laboratory, Laghouat University, Algeria Professor,Department of Civil Engineering, LSC Laboratory, U.S.T.Oran, Algeria 3 Professor, ArGEnCo Department, Liège University, Chemin des Chevreuils, Liège, Belgium * (Corresponding author: tel:+3 776 3 05 5; fax:+3 9 93 6 98; email : slhamara@yahoo.com) Received: June 008; Accepted: 5 August 008 ABSTRACT: The influence of the end restraint conditions on the lateral-torsional bucling of beams is investigated in detail using finite element method. The paper focuses on the limitation of Eurocode 3 regarding the lateral bending and torsional restraint coefficients and θ of the end supports. Theoretical expressions of the coefficients and θ taing into account the minor axis flexural restraint at the support and the end torsional restraint respectively are presented. The introduction of ne coefficients and θ representing the actual support conditions in the expression of the elastic itical moment is suggested. A comparison beteen the elastic itical moments for various beam oss-sections, lengths and various end restraints, obtained from the finite-element method, and those derived from EC3 ENV method, in hich the proposed coefficients and θ are introduced, confirms the reliability of these coefficients that model the end support conditions. Keyords: Elastic lateral-torsional bucling; End restraints; Elastic itical moment; Out-of plane bending; Uniform torsion; Warping. INTRODUCTION A slender beam under the action of bending loads in the plane of maximum flexural rigidity can bucle by combined tist and lateral bending of the oss-section, unless it has continuous lateral support. This phenomenon, hich as first investigated theoretically and experimentally during the nineteenth century, is non as lateral bucling. Slender beams manufactured from narro rectangular sections or I-sections ith narro flanges, lac both lateral flexural rigidity and torsional rigidity, and if left unsupported, or supported intermittently only, they may bucle under bending stresses considerably loer than the yield or proof stress of the material. The lo torsional rigidity is an important factor, so thin-alled open section beams such as channels or eds are also susceptible to this form of instability. Box beams on the other hand, hich are torsionally stiff and have similar flexural properties about the to principal axes of inertia, are very resistant to lateral bucling. The elastic bucling stress is also influenced by the conditions of support at the ends of the beam, and by the type and position of the applied loads that cause bending. In thin-alled open sections the point of application of the load ith respect to the shear centre is important, and for all types of sections initial imperfections can influence the behaviour, particularly of members of intermediate length. Research developments on lateral torsional bucling of steel members have been accompanied by the realiation of updated design codes and standards. odern steel codes for structures, such as AISC LRFD [, ], BS 5950- [3] and EC3 [4, 5], provide, on the basis of the limit state concept, design procedures to compute the lateral-torsional bucling resistance of beams. Primary, these procedures generally require the determination of the elastic itical bucling moment. Initial imperfections, residual stresses and inelastic bucling are taen into account through the use of bucling curves [6].

44 Effect of End Connection Restraints on the Stability of Steel Beams in Bending The elastic itical moment is directly dependent on the folloing factors [7]: material properties such as the modulus of elasticity and shear modulus; geometric properties of the oss-section such as the torsion constant, arping constant, and moment of inertia about the minor axis; properties of the beam such as length, and lateral bending and arping conditions at supports; and finally loading, since lateral-torsional bucling is greatly dependent on moment diagram and loading position ith respect to the section shear centre. The bending moment diagram is taen into account by means of the equivalent uniform moment factor C. The elastic itical moment of a simply supported beam ith uniform moment is multiplied by the equivalent uniform moment factor C to obtain the elastic itical moment for any bending moment diagram. Lateral bending and torsional restraints provided by beam end supports have a significant effect on the lateral torsional bucling of beams. The degrees of lateral bending and torsional restraints developed by the supports depend on the type of the connection used. any simple connections met in practice have only partial lateral bending restraint and are generally assumed to provide full torsional restraint. Hoever, some connections such as long fin-plate connections provide both partial lateral bending and torsional restraints. Therefore, the beams connected ith fin-plates are prone to undergo some tisting about the longitudinal axis at the supports, in addition to lateral bending. EC3 ENV taes into account the effect of the lateral bending restraint of the end support in the evaluation of the elastic itical moment by means of a coefficient. Hoever, it is assumed that full torsional restraint is provided by the connection. For any particular end connection, the influence of the degree of lateral bending and torsional restraints can be expressed in terms of reduction of the lateral torsional bending moment. The scope of this or is to study the effects of lateral and torsional end restraints on the lateral torsional bucling moment of the beam. An analytical model has been developed in order to evaluate the lateral bending and the torsional restraints coefficients and θ. The model also allos to evaluate the percentage of reduction of the lateral torsional bending moment against 0 for full restraint. On the basis of the value of the percentage reduction of for a particular connection, its classification as simple, partial restraint or full restraint can be made. Therefore, it is possible using the analytical model developed to determine the required lateral bending and torsional restraints of the connection to ensure full lateral restraint. Theoretical expressions of the coefficients and θ, taing into account the support minor axis bending and the torsional restraints respectively are proposed. A variety of connections ith different end restraints are investigated using the finite element softare LTBeam [8] in order to determine their influence on the lateral torsional bucling itical moment. It should be noted that this research or is of theoretical nature. It is done for three simple load cases. A uniform distributed load, acting on the beam in the vertical direction at the shear centre, a constant moment along the beam and a concentrated load at mid-span acting at the shear centre. A comparison beteen the elastic itical moments for various beam lengths and various end restraints, obtained from LTBeam, and those derived from the EC3 ENV formula in hich the coefficients and θ computed from the proposed formulae are introduced, confirms the reliability of these coefficients that model end support conditions. Reference is made to EC3 ENV [5] and not to EC3 EN [4] as the formulae for have been removed from [4] in the so-called conversion period

S. Amara, D.E. Kerdal and J.P. Jaspart 45 Given the type of loading, Eurocode 3 [5] provides the values of the equivalent uniform moment factor C only for limited end restraints conditions for = and = 0,5. In the cases of the lateral bending coefficients being different from,00 and 0,5 the corresponding values of C have been obtained by linear interpolation.. LATERAL TORSIONAL BUCKLING AND ELASTIC CRITICAL OENT Under ineasing loading (see Figure ), the beam first bends strictly in the plane of loading and is the deflection in that plane. Once the moment reaches a certain magnitude, called elastic itical moment, the beam may deflect suddenly out of the plane of bending. This instability phenomenon is non as lateral torsional bucling. Lateral torsional bucling is said to occur by bifurcation of equilibrium. The beam simultaneously exhibits lateral displacements v in the y direction (bending about the minor axis of the oss-section) and tist rotation θ about its longitudinal axis x. It is clear that lateral-torsional bucling is resisted by a combination of lateral bending resistance EI d v / dx and torsional resistances GI t d θ / dx and EI d 3 θ / dx 3. Thus, a member is especially prone to lateral torsional bucling hen it has lo lateral flexural stiffness EI and its torsional stiffness GI t and arping stiffness EI / L are lo compared to its stiffness in the plane of loading. With the nomenclature used in Eurocode 3 [4], here (x-x) is the axis along the member, (y-y) is the major axis of oss-section and (-) is the minor axis of the oss-section, the governing differential equation for the lateral torsional bucling is [7]: d φ d φ EI GI + = 0 () 4 4 dx t dx yφ EI EI y ith dx = ; dv q dx = ; d Vy dx = ; dv y dx = 0 () d y V here q is the distributed load acting on the beam, V y and V are the shear forces, y and are the bending moments, and φ is the torsion deformation. In order to be able to impose appropriate boundary conditions at supports, the internal shear forces and the bending moment components in Eqs. () and () are referred to the axis in the undeformed configuration. Exact solutions for Eq. are obtained for a doubly symmetrical beam ith simply supported conditions, free arping and subjected to a uniform moment diagram. The elastic itical moment for this basic case is: EI I LGIt L I π EI π = + (3) The elastic itical moment obtained for the basic situation by formula (3), is multiplied by the equivalent uniform moment factor C hich taes into account the actual bending moment diagram. Thus, the value of may be computed by the expression:

46 Effect of End Connection Restraints on the Stability of Steel Beams in Bending π EI I ( L ) GIt = C ( ) + (4) ( L ) I π EI here the lateral bending coefficient and the arping coefficient are introduced in order to tae into account support conditions other than simply supported. These coefficients are equal to,00 for free lateral bending and free arping and 0,5 for prevented lateral bending and arping.. Eurocode 3 (ENV) Approach and Its Limitations The assessment of the stability behaviour of steel beams based on simplified calculations, as desibed in the standards of most countries, is not alays a realistic evaluation. The assumption that member end connections behave as either pinned or completely rigid is a highly simplified approach because experimental investigations sho that true joint behaviour has characteristics beteen these to simplified extremes. In order to simplify the calculation of structural elements, beams or girders and columns are treated in isolation in most steel design codes, and the effect of the connections is estimated using simple factors hich tae account of the arping and bending restraint conditions. The design proposals for the bucling of beams assume that the end supports should completely prevent end tisting. If the supports have only limited elastic torsional restraint stiffness, the beam ill bucle at a loer load than that estimated from the idealised case (Figure ). Figure. Bucling of a Simply Supported I-beam One of the most commonly employed general formulae to estimate elastic itical moment is the so-called 3-factor formula, hich as included in the ENV version of EC3 [5]. The recently completed EN version of this design standard [4] provides no information concerning the determination of. In theory, this formula should be applicable to beams subjected to major axis bending having doubly or singly symmetrical oss-sections and arbitrary support and loading conditions. Hoever, some particular cases are currently not covered, most notably the case of beams ith partial end restraint against minor axis bending, and partial end restraint against tisting, hich are in practice partially free to deflect laterally and tist at the supports. The moment gradient along the beam is considered by the use of the coefficient C hich is also affected by the lateral bending conditions at end supports. Eq. 4 gives the elastic itical moment as a function of C, and. It can be seen from the expression Eq. 4, that the effect of the end tisting at the supports of the beam hich is supposed to be introduced by a coefficient θ has not been taen into account. This means that the beam is assumed to be completely prevented

S. Amara, D.E. Kerdal and J.P. Jaspart 47 from tisting about the longitudinal axis at the end supports. Hoever, many real situations met in practice are not in compliance ith these standard conditions. For these practical situations (such as those ith simple connections ith long fin-plates hich possess partial torsional restraint), the beam is prone to undergo some tisting about the longitudinal axis at the end supports causing a deduction in elastic itical moment. Therefore the effect of the partial end torsional restraint should be considered by introducing a coefficient θ in the expression of. This paper attempts to fill in to of the insufficiencies identified in the above expression of used by the Eurocode 3, by (i) proposing a formula for computing the value of the lateral bending restraint coefficient hich depends on both the connection end restraint and the beam flexural stiffness about the minor axis and (ii) proposing a formula for computing the value of the torsional restraint coefficient θ of the end support (Figure ) hich depends on both the torsional stiffness of the support K Θ and of the torsional rigidity of the beam GIt / L. In the EC3 ENV [5], it is suggested to tae made. Therefore, in this paper, the arping coefficient =,00 unless special provision for arping fixity is is set to be equal to,00. Figure. Torsional End Restraint. Theoretical Bacground on Torsional Restraints at Supports Flint [9] presented an analysis for a member ith end connections providing only limited torsional stiffness. For a beam under single point load or to symmetrical point loads, Flint derived the folloing relationship beteen the itical load and the support torsional restraint stiffness. 4 m= RT (5) 3 In Eq. 5 m is the ratio beteen bucling loads for beams ith finite and infinite support torsional stiffness. R T is the ratio of the torsional stiffness of the beam to its supports and is given by GJ / L hereθ is the rotation of a support. T / θ A theoretical study as carried out by Schmidt [0] to determine the effect of elastic end torsional restraint on the itical load of a beam. It shos that the beam is incapable of supporting any load if the end supports offer no resistance to end tisting. It further shos that the itical load ineases

48 Effect of End Connection Restraints on the Stability of Steel Beams in Bending little as the end torsional restraint stiffness parameter e ineases beyond 0. The parameter e is defined as e= / RT. According to Flint s equation (5), Bose [] found that the itical load for e = 0 ill be 93% of the load for rigid torsional end support. Bennetts et al [] and Grundy et al [3] have attempted to investigate the value of the torsional stiffness K of an end restraint and in particular, the torsional stiffness K s of the connection component. They investigated the behaviour of fin-plates and noted that the torsional stiffness varied almost continuously ith the applied moment. Bennetts et al stated that the end torsional stiffness is a function of the beam eb thicness, the depth of the beam and the depth of the connection component. They felt that the interaction beteen these components is quite complex and attempts to produce a theoretical model for the overall behaviour ould probably prove unsuccessful. They also found that the experimental results ere sensitive to the method of testing and results differed in subsequent loading cycles. Hoever, it is orth noting that the test rig used produced only a torsional moment in the fin-plate connection. This method of testing removed any possibility of beneficial restraining effects afforded by the supported beam upon the connection. 3. PARAETRICAL STUDY ON BEA ELASTIC STABILITY As seen before, the elastic itical moment is a function of C, and. It should be noted that the design proposals for the bucling of beams EC3 does not consider the possibility of being different from,00 or 0,5 hich represent the to extreme cases, free lateral bending and prevented lateral bending. In practice, the coefficient may tae any value beteen,00 and 0,5 depending on the degree of lateral restraint provided by the end supports of the beam. It should also be mentioned that EC3 ENV does not tae into account the possibility of the end restraint being partially prevented from tisting about the longitudinal axis of the beam. The elastic itical moment expression considers the beam as completely prevented from tisting at the end supports. If the supports have only limited elastic torsional restraint stiffness, as it is the case of some practical connections such as fin-plates, the beam ill bucle at a loer load than that estimated from the idealised case. Therefore, the aim of this or is to derive analytical expressions for the lateral bending restraint at the support and the torsional end restraint θ to be introduced in the expression of the elastic itical moment. 3. Proposed Expressions of Lateral and Torsional Restraint Coefficients at Supports The behaviour of beams is dependent on their end support conditions and possibly on their intermediate supports. These conditions depend not only on major axis bending (primary bending) but also on minor axis bending, uniform torsion and arping torsion. The latter three types of support conditions influence deeply the LTB resistance. In the existing codes, the support conditions are accounted for by means of so-called effective length factors and end arping factor. Each of these to factors varies from 0,5 for full fixity to,00 for no fixity at all, and taes the value of about 0,7 for one end fixed and one end free. In the EC3 ENV [5], it is suggested to tae =,0 unless special provision for arping fixity is made. The values of the C factor involved in the analytical expression of the elastic itical LTB moment is significantly influenced by the values; hoever, in EC3 ENV, only or 3 values of are suggested. In order to be

S. Amara, D.E. Kerdal and J.P. Jaspart 49 able to evaluate satisfactorily the values of by means of linear interpolation., it is probably acceptable to calculate the C factor Different expressions for computing lateral bending restraint coefficient and torsional restraint θ of the end support have been considered, and the folloing ones Eq. 6 and Eq. 7 are finally selected, even though further research might provide a more exact formulation. EI / L + 0,5K v' = (6) EI / L + 0,5K v' θ GIt / L = + 5 (7) K here: the unbraced length, t Θ I is the minor axis moment of inertia of the beam, E is the modulus of elasticity, L is K v' is the lateral flexural restraint of the support, G is the shear modulus, I is the torsional constant and K Θ is the torsional restraint of the support. It can be seen from Eq. 6 that for K v' varying from ero to infinity, the value of ranges from,00 (no fixity) to 0,5 (full fixity). Eq. 7 shos that if no torsional restraint is provided by the support ( K Θ = 0 ), then the torsional restraint coefficient θ is infinity, and for full torsional restraint of the support ( K Θ =infinity), the torsional restraint coefficient θ is,00. For doubly symmetrical oss-section and for end moment loading or transverse loads applied at the shear centre, the elastic itical moment to be considered as the itical value of the maximum moment in the beam may be assessed by the ne proposed formula: π EI I ( L) GI θ θ t = C ( ) + (8) ( θ L) I π EI here is the lateral bending restraint coefficient, ranging from 0,5 (for full fixity) to,00 (for free lateral bending at the support), hich can be evaluated from Eq. 6, θ is the torsional restraint coefficient hich varies from,00 (for support prevented from tisting about longitudinal axis) to infinity (for free tisting of the support about longitudinal axis). Eq. 8 may be ritten in its simplified form as: π EI I ( L) GI t = C ( ) + (9) θ ( L) I π EI It can be seen that Eq. 9 is the same as the one given by EC3 ENV [5], only for a beam fully prevented from tist rotation at the supports ( θ = ). It can also be seen from Eq. 9 that for a support connection providing no torsional restraint ( θ = infinity), the elastic itical moment tends toards ero and therefore the beam ill be in the state of instability.

50 Effect of End Connection Restraints on the Stability of Steel Beams in Bending Finally Eq. 9 can be expressed as: = (0) θ 0 It can be seen from Eq. 0, that the itical moment moment can be obtained by multiplying the itical 0 obtained for a beam ith full torsional restraint at supports ( θ = ) by θ. 4. ANALYTICAL EVALUATION OF THE ELASTIC STABILITY 4. Influence of Lateral Bending Restraint In this section the variation of the lateral bending restraint coefficient is examined for various sies of IPE profiles (IPE300, IPE360, IPE and IPE500) of varying lengths (L=4m, 6m, 0m, and m), subjected to three load cases, equal end moments, uniformly distributed vertical loads applied to the shear centre and a concentrated vertical load at mid-span applied to the shear centre. In order to perform a comparative study, numerical analysis as conducted using the LTBEA softare hich as developed by CTIC [8] ithin the frameor of a European project and based on the finite element method. In this analysis the end supports of the beam are assumed to be fixed for out of plane deflection, v = 0 and tist rotation, θ = 0 ( θ = ), but not restrained against arping, ( = ). The lateral flexural restraint of the support is modelled by a spring of lateral restraint value K v'. It is taen as ero (for no lateral restraint) or infinity (for full lateral restraint). To evaluate using the FE, it is necessary to determine the value of the lateral bending restraint of the connection K v'. Therefore, for any lateral bending restraint coefficient, the corresponding value of K can be evaluated using Eq. 6. v' Figure 3 shos the numerical and analytical results of the variation of elastic itical moment against variation of lateral bending restraint coefficient for equal end moments. It can be seen from Figure 3 that the results obtained from EC3 ENV formula for varying from 0,5 to,00 are in very good agreement ith those computed from FE, except for the case of full lateral restraint. For =0,5, the values of obtained from EC3 ENV formula are loer than those derived from FE. The reason is that the EC3 ENV formula assumes that coefficient C does not vary ith end support conditions for equal end moments. Hoever, using finite differences approach, iguel A. Serna [7] shos that for the case of equal end moments, the values of C for beams ith prevented lateral bending at supports ( =0.5) are higher than those for simply supported beams, hich confirms the difference found in this study beteen the numerical and analytical results of for the case of restrained end supports against lateral bending ( =0,5). Folloing the difference beteen the FE results and those obtained analytically by EC3 ENV formula, for =0,5 it is recommended that the value of C be taen as,05 hich is the value suggested in [4].

S. Amara, D.E. Kerdal and J.P. Jaspart 5 [N-m] 360 30 80 40 00 60 0 80 40 0 (a) Results for IPE 300 _EC3 ENV _FE 4 m 6 m 0 m m [N-m] 700 500 300 00 00 (b) Results for IPE 360 _EC3 ENV _FE 4 m 6 m 0 m m 000 000 [N-m] 900 800 700 500 _EC3 ENV _FE 4 m [N-m] 800 00 000 800 _EC3 ENV _FE L=4m 300 00 00 6 m 0 m m 00 L=6m L=0m L=m (c) Results for IPE (d) Results for IPE 500 Figure 3. Numerical and Analytical Elastic Critical oments Versus Lateral Bending Coefficient for Equal End oments For the case of uniformly distributed vertical loads, the numerical and analytical results of the variation of elastic itical moment as a function of lateral bending restraint coefficient are given in Figure 4.

5 Effect of End Connection Restraints on the Stability of Steel Beams in Bending 360 _EC3 ENV _FE 700 _EC3 ENV _FE [N-m] 30 80 40 00 [N-m] 500 60 0 80 40 0 (a) Results for IPE 300 4 m 6 m 0 m m 300 00 00 (b) Results for IPE 360 4 m 6 m 0 m m 800 _EC3 ENV _FE 650 500 _EC3 ENV _FE [N-m] 700 500 4m [N-m] 350 00 050 900 750 4m 300 00 00 (c) Results for IPE 6m 0m m 450 300 50 (d) Results for IPE500 6m 0m m Figure 4. Numerical and Analytical Elastic Critical oments Versus Lateral Bending Coefficient for Uniformly Distributed Loads It can be seen that for all IPE profiles and beam lengths considered in this study, there is quite good agreement beteen the values given by the EC3 ENV formula and the numerical results obtained using the finite elements approach. Finally, Figure 5 shos the results of the variation of elastic itical moment against lateral bending restraint coefficient for the case of concentrated load acting at the mid span. As can be seen from Figure 5, the graphs of the elastic itical moments obtained analytically for various IPE profiles and lengths are very close to those computed numerically by the finite element method. Therefore the proposed ne formula (6) provides very good approximation of for all cases considered in this study.

S. Amara, D.E. Kerdal and J.P. Jaspart 53 [N-m] 350 300 50 00 50 00 50 0 _EC3 ENV _FE (a) Results for IPE300 L=4m L=6m L=0m L=m [N-m] 700 500 300 00 00 0 _EC3 ENV _FE (b) Results for IPE360 L=4m L=6m L=0m L=m 900 _EC3 ENV _FE 800 _EC3 ENV _FE 800 [N-m] 700 500 L=4m [N-m] 00 000 L=4m 300 00 00 0 (c) Results for IPE L=6m L=0m L=m 800 00 0 (d) Results for IPE 500 L=6m L=0m L=m Figure 5. Numerical and Analytical Elastic Critical oments Versus Lateral Bending Coefficient for a Concentrated Load at the id Span 4. Influence of End Torsional Restraint In this study, the effect of the variation of support torsional restraint θ on the elastic itical moment is investigated numerically using finite element approach and analytically ith the application of the EC3 ENV formulation. As mentioned previously, the elastic itical moment for a beam determined according to EC3 ENV formula, assumes that the end supports of the beam are fully prevented from tisting, thus full torsional restraint is provided. Hoever, some types of supports met in practice such as fin-plates, provide only partial torsional restraint. Therefore it is recommended that the effect of torsional restraint provided by end supports of the beam should be taen into account in the EC3 ENV formula by means of a coefficient θ. The values of θ may be obtained using the proposed formula (7). The lateral-torsional bucling of four IPE profiles (IPE 300, IPE 360, IPE and IPE 500) ith three different lengths (L=6m, L=0m and L=m) have been studied for to load cases, uniformly distributed vertical loads applied to the shear centre and a concentrated vertical load at mid-span

54 Effect of End Connection Restraints on the Stability of Steel Beams in Bending applied to the shear centre. In this analysis the end supports of the beam are assumed to be fixed for out of plane deflection ( v = 0 ) and for lateral bending rotation ( = ) but not restrained against arping ( = ). The torsional restraint of the support is modelled by a spring of torsional restraint value K Θ. If no torsional restraint is provided by the support ( K Θ =0) then =0 and for full θ torsional restraint of the support ( K Θ = infinity) then =. For partial torsional restraints of the θ end supports, hich correspond to the values of θ varying from 0 to,00, the corresponding values of the spring torsional restraints K Θ can be calculated from Eq. 7. Figure 6 shos the numerical and analytical results of the variation of / 0 against variation of for the case of uniformly distributed vertical loads. According to Eq. 0, the graphs of θ / versus for all IPE oss-sections and lengths obtained from the analytical θ 0 expression through the application of the ne proposed formula are represented by a single straight line. It can be seen from Figure 6 that the numerical results of / 0 against obtained θ through LTBEA softare in hich the values of the spring torsional restraints computed from the proposed analytical formula (7), are represented by curved lines. K Θ are / 0,0 0,9 0,8 0,7 0,6 0,5 0,4 0,3 0, 0, EC3 ENV-Formula FE _L = 6m FE _L = 0m FE _L = m 0,0 0,0 0, 0, 0,3 0,4 / 0,0 0,9 0,8 0,7 0,6 0,5 0,4 0,3 0, 0, EC3 ENV-Formula FE _L = 6m FE _L = 0m FE _L = m 0,0 0,0 0, 0, 0,3 0,4 / θ / θ (a) Results for IPE 300 (b) Results for IPE 360

S. Amara, D.E. Kerdal and J.P. Jaspart 55 / 0,0 0,9 0,8 0,7 0,6 0,5 0,4 0,3 0, 0, EC3 ENV-Formula FE _L = 6m FE _L = 0m FE _L = m 0,0 0,0 0, 0, 0,3 0,4 / 0 0,0 0,0 0, 0, 0,3 0,4, / θ / θ (c) Results for IPE (d) Results for IPE 500,,0 0,9 0,8 0,7 0,6 0,5 0,4 0,3 0, 0, EC3 ENV-Formula FE _L = 6m FE _L = 0m FE _L = m Figure 6. Results from FE and Eq. 0 for uniformly distributed loads Figure 7 shos the numerical and analytical results of the variation of / 0 against variation of for the case of a concentrated vertical load at the mid-span acting at the shear centre. Again it θ also shos that the numerical results of / 0 against obtained through LTBEA θ softare in hich the values of the spring torsional restraints Θ are computed from the proposed analytical formula (7) are represented by curved lines. / 0,0 0,9 0,8 0,7 0,6 0,5 0,4 0,3 0, 0, EC3 ENV-Formula FE _L = 6m FE _L = 0m FE _L = m 0,0 0,0 0, 0, 0,3 0,4 / 0,0 0,9 0,8 0,7 0,6 0,5 0,4 0,3 0, 0, EC3 ENV-Formula FE _L = 6m FE _L = 0m FE _L = m 0,0 0,0 0, 0, 0,3 0,4 / θ / θ (a) Results for IPE 300 (b) Results for IPE 360

56 Effect of End Connection Restraints on the Stability of Steel Beams in Bending / 0,0 0,9 0,8 0,7 0,6 0,5 0,4 0,3 0, 0, EC3 ENV-Formula FE _L = 6m FE _L = 0m FE _L = m 0,0 0,0 0, 0, 0,3 0,4 (c) Results for IPE / 0,0 0,9 0,8 0,7 0,6 0,5 0,4 0,3 0, 0, EC3 ENV-Formula FE _L = 6m FE _L = 0m FE _L = m 0,0 0,0 0, 0, 0,3 0,4 / θ / θ (d) Results for IPE 500 Figure 7. Results from FE and Eq. 0 for a Concentrated Load at id-span It is orth noting from Figure 6 and Figure 7 that, the closer the / 0 graphs to the straight line, the more accurate the formula (7). Therefore it can be seen from the figures that, for both load cases and for all IPE oss-sections and lengths considered in this study, the graphs of / 0 obtained from FE analysis are particularly close to the ones obtained from the proposed analytical expression Eq. 0, in hich the torsional restraint coefficient θ has been introduced. In other ords, comparison of the cases performed in this analysis revealed that an acceptable small difference exists beteen analytical and numerical results (errors are ithin about 0%). Therefore, there is quite good agreement beteen the results given by the proposed analytical formula and the numerical results of the FE approaches. 5. REQUIRED CONNECTION TO ENSURE SUFFICIENT RESTRAINT LEVEL Depending on the degree of lateral bending and torsional restraints developed, a connection can be classified as simple, partial or full restraint connection. Using the analytical model developed in section 3., it is no possible to evaluate the lateral bending and torsional restraints for any beam end connection, therefore its classification can be made. If a lateral bending restraint of a connection results in a reduction of less than 0% of from 0 for full restraint, then it can be assumed as full lateral bending restraint connection. Hoever, if it results in more than 0% drop of from 0, then it is considered as partial bending restraint connection. Similarly, if a connection torsional restraint results in less than 0% drop of from 0, then full torsional restraint connection can be assumed, otherise, it is considered as partial tortional restraint connection. The interaction beteen the effects of and θ on the elastic itical moment has not been included in this study. Therefore, in order to classify a connection, it is clear that the effects of lateral bending and torsional restraints have been considered independently in the analysis. Thus, hen varying the values of, the torsional restraint has been considered as full torsional restraint ( θ =). When varying the values of θ, the lateral bending restraint has been considered as simple restraint ( =).

S. Amara, D.E. Kerdal and J.P. Jaspart 57 Finite elements analyses have been performed to verify the accuracy of Eqs. 6 and 7. For beams ith IPE profiles subjected to equal end moments, uniformly distributed loads or, a concentrated point load at the mid span acting at the shear centre, the results sho that Eq. 6 can be reasonably applied to evaluate the lateral bending restraint coefficient. The analyses also sho that for beams ith IPE profiles subjected to uniformly distributed loads or, a concentrated point load at the mid span acting at the shear centre and for =, Eq. 7 can be satisfactory applied to simulate the effect of the torsional restraint of the end support θ. It can be seen from Figure 3, that for the case of beams subjected to equal end moments, 0% drop in the value of the elastic itical moment against 0 for full restraint, is obtained for a value of =0,53. For the case of uniformly distributed loads, Figure 4, shos that 0% reduction in the value of from 0 for full restraint, corresponds to =0,56. Figure 5 shos that for a beam subjected to a concentrated point load at the mid-span and for =0,577, the corresponding value of the elastic itical moment is reduced by 0% compared to 0 for full restraint. It can also be seen that, for the same percentage reduction in, the corresponding values of are strongly dependent on the loading type. According to the results obtained from the proposed expressions of and θ given in section 3., it is recommended that if the percentage drop in the value of the elastic itical moment against for full restraint remains ithin 0%, then full restraint may be assumed. Therefore, for lateral bending restraint coefficients ranging from 0,5 to 0,53 (for equal end moments), from 0,5 to 0,56 (for uniformly distributed loads) and from 0,5 to 0,577 (for a concentrated point load at mid-span), the end lateral bending restraint may be assumed as full lateral bending restraint ( =0,5). For each load case, the minimum lateral flexural stiffness value of the end support v' required to assume full lateral restraint of the support, can be obtained from Eq. 6. Figure 6 and Figure 7 sho the values of θ for percentage drops in the value of the elastic itical moment, against 0 for full torsional restraint. When examining Figure 6. and Figure 7, it can be revealed from analytical and FE results that for =0,9, the corresponding θ value of / 0 is 0,9. Therefore, for all IPE oss-sections, beam lengths and load cases performed in this analysis, 0% reduction in the value of, results in 0% drop in the value of θ against 0 for full torsional restraint. According to results in section 3., for IPE oss-sections under uniformly distributed loads or a concentrated point load at mid span acting at the shear centre, Eq. 7 provides values of end torsional restraint coefficients θ that are in very good agreement ith the FE results.

58 Effect of End Connection Restraints on the Stability of Steel Beams in Bending It can be seen from Figure 6, Figure 7 and Eq. 7 that for 0% drop in the value of the elastic itical moment against 0 for full torsional restraint, it results in a value of =0,9, hich θ corresponds to a value of the torsional stiffness of the support K Θ =,3 GIt / L. Thus, for KΘ =,3 (ratio of the torsional stiffness of the supports to its beam), the value of the elastic GIt / L itical moment is = 0,9 0. From these results, it is recommended to assume that the torsional stiffness of the connection is acceptable and may be considered as full torsional restraint, if it results in no more than 0% drop in the value of for full restraint. Therefore, if it is proved that the ratio of the torsional stiffness of a connection K Θ to its beam GIt / L is at least equal to,3, then it is recommended to assume full restraint connection, as ill be 90% of 0 for rigid torsional end support. 6. CONCLUSIONS When dealing ith lateral torsional bucling, modern design standards require the computation of the elastic itical moment, hich mainly depends on the moment distribution along the beam and on the end supports restraints. One of the most commonly used general formulae to estimate elastic itical moments in steel beams prone to LTB is the so-called 3-factor formula, hich is included in the ENV version of Eurocode 3 [5]. This paper presents a revie of EC3 ENV approach and its limitations ith regards to lateral bending and torsional restraints of the end supports. Based on these limitations, the paper has presented ne expressions for estimating the actual degree of lateral bending restraint and torsional restraint θ of the end supports. The values of the coefficients and θ obtained from the proposed expressions, are introduced in the general formulae that estimates the elastic itical moment. The influence of the lateral bending and torsional restraint on the lateral-torsional bucling of IPE beams ith various oss-sections, different loading conditions and lengths has been investigated using analytical and FE approach. Comparison beteen these to approaches allos to sho the accuracy of the proposed and θ expressions. For beams subjected to uniformly distributed loads or a concentrated load at mid-span applied at the shear centre, the results of the variation of against, obtained from EC3 (ENV) formula for varying from 0,5 to,0 are in very good agreement ith those computed from FE. The results of variation of / 0 versus computed from FE are quite close to those obtained θ from EC3 (ENV) formula. Finally the folloing can be concluded from this study: (i) (ii) Eq. 6 and Eq. 7 can be reasonably applied to evaluate lateral bending and torsional restraints of end supports. For beams subjected to equal end moments ith =0,5, it is recommended that the value of coefficient C be taen as,05, instead of,00.

S. Amara, D.E. Kerdal and J.P. Jaspart 59 (iii) (iv) (v) Full lateral and torsional restraint of end supports may be assumed if it results in less than a 0% drop in the value of elastic itical moment against full torsional restraint moment 0. Full lateral bending restraint at supports may be assumed for beams subjected to equal end moments, if 0,53, for uniformly distributed loads if 0,56 and for a concentrated point load at mid span if 0,577. To assume full torsional restraint of end supports, it is necessary that the ratio beteen the KΘ torsional stiffness of the supports and of the beam be at least,3. GI / L t ACKNOWLEDGENT The authors gratefully acnoledge the support provided by the ArGEnCo Department at Liège University (Belgium). REFERENCES [] AISC LRFD 986. American Institute of Steel Construction (AISC). Load and resistance factor design. Chicago: AISC; 986. [] AISC LRFD 994. American Institute of Steel Construction (AISC). Load and resistance factor design. Chicago: AISC; 994. [3] BS 5950-. Structural use of steelor in buildings. Code of practice for design. Rolled and elded sections. British Standards Institution; 000. [4] EC3 005. European Committee for Standardiation. Eurocode 3: Design of steel structures, Part -: General rules and rules for buildings (EN 993--). Brussels; ay 005. [5] EC3 99. European Committee for Standardiation. Eurocode 3: Design of steel structures, Part -: General rules and rules for buildings. (ENV 993--). Brussels; 99. [6] Trahair, N.S., ultiple Design Curves for Beam Lateral Bucling, Stability and Ductility of Steel Structures, T. Usami and Y. Itoh (Eds.), Pergamon, 998. [7] Serna,.A., Lope, A., Puente, I. and Yong, D.J., Equivalent Uniform oment Factors for Lateral-torsional Bucling of Steel embers, Journal of Constructional Steel Research, Vol. 6, pp. 566-580, 006. [8] Galéa, Y., oment Critique de Déversement élastique de Poutres Fléchies, Presentation du logiciel LTBEA. Revue de la Construction étallique. Vol., CTIC 003. [9] Flint, A.R., The Influence of Restraints on the Stability of Beams, The Structural Engineer, Vol. 9, pp. 35-46, September 95. [0] Schmidt, L.C., Restraints Against Elastic Lateral Bucling, J. of Eng. ech. Div., ASCE, Vol. 9, No. E6, pp. -0, December 965. [] Bose, B., The Influence of Torsional Restraint Stiffness at Supports on the Bucling Strength of Beams, The Structural Engineer, pp. 69-74, December 98. [] Bennetts, I.D., Thomas, I.R. and Grundy, P., Torsional Stiffness of Shear Connections, Preprints, etal Structures Conference, Institution of Engineers, pp.-0, Australia, 98. [3] Grundy, P., urray, A.W. and Bennetts, I.D., Torsional Rigidity of Standard Beam-to-Column Connections, Preprints, etal Structures Conference, Institution of Engineers, Australia, pp.-64, 983. [4] ECCS Technical Committee 8- Stability, 006. European Convention for Constructional Steelor, Rules for ember Stability in EN 993--. Bacground documentation and design guidelines, No. 9, 006.