Vous êtes sur la page 1sur 21

Paper Ref: S2603_P0580 3rd International Conference on Integrity, Reliability and Failure, Porto/Portugal, 20-24 July 2009

Albano C. Sousa1*, Rui C. Barros2

Graduate student, FEUP - Faculdade Engenharia Universidade Porto, Dept of Civil Engineering Porto 4200-465, Portugal, Email: ec04001@fe.up.pt Assoc. Prof. of Civil Engineering, FEUP - Faculdade Engenharia Universidade Porto Dept of Civil Engineering, Porto 4200-465, Portugal, Email: rcb@fe.up.pt

SYNOPSIS A comparison is made between European (EC2 and EC3) and American (ACI 318 and AISC 360) codes of practice with regards to the effective length K-factors, in assessing critical loads of slender columns. Discrepancies between the codes are apparent since the evaluation of an individual column end restraints is quite different, except between the American codes. What this means, however, is that the derivation of the expressions used in the calculation of K-factors are fundamentally different. This paper presents a method of comparison between each code which is based in finding a common link in the assessment of end restraints and finds that the effective length which is obtained is in essence the same, apart from some numerical errors or round-off approximations. EFFECTIVE LENGTH IN CODES OF PRACTICE THEORETICAL BACKGROUND The concept of effective length is a useful tool in individual stability checks of columns in multi-storey frames. It essentially is a mean of comparison between the critical load of a member subject to any type of end restraints and its corresponding theoretical Euler load. PE =

2 EI


Pcr =

2 EI

(K l )

2 EI


Kf =

PE Pcr


where PE is the Euler load, l the system length of the member, l 0 the effective length and K f the well known K-factor. In order to derive the appropriate K-factor for a particular column and its end restraints, as done by Wood [1] and used in the derivation of the EC3 [2] charts, one must first understand the concept of stability functions. For that, consider the infinitesimal segment of a statically undetermined beam-column in Figure 1.

Figure 1 Infinitesimal Segment of a Statically Undetermined Beam-Column [3]

The following equilibrium equations (4) and (5) can be obtained by interpretation of Figure 1.

(V + dV ) V + p dx = 0


dx ( M + dM ) M + V dx + P dw ( p dx ) = 0 2


Dividing these equations by dx and considering that dx 0 , the following equations are obtained:
V '= p

(6) (7)

M '+ P w' = V

Differentiating (7) and substituting in (6), and further considering that in this system of axis M = EI w' ' , one gets:

(EI w' ')' '+(P w')' = p


As one can see this equation is a forth order differential equation and, provided that P > 0 (compression) and that EI and P are constant along the beam-column, the general solution derived (for example) in Bazant and Cedolin [3] assumes the form:
w( x ) = A sin jx + B cos jx + Cx + D + w p (x )


where A,B,C and D are arbitrary constants, w p (x ) is a particular solution corresponding to the transverse distributed loads p ( x ) and j = P EI .

The Figure 2 represents the loading of a HingedFixed beam-column, which is materialized by a compressive force P and an imposed rotation of a . The purpose is to ascertain the value of the end moment M a that has to be applied in order to produce that rotation and M b the corresponding reaction moment.

Figure 2 Loading of a HingedFixed Beam-Column

Applying the following boundary conditions to equation (9)

x=0 x=l

w=0 w = a w=0 w' = 0

(10.) (11)

with w p ( x ) = 0 since p ( x) = 0 , yields:

B+D=0 Aj + C = a A sin jl + B cos jl + Cl + D = 0 Aj cos jl Bj sin jl + C = 0



= jl =

P P l = EI PE


and solving the set of equations (12), the value of the constants A,B,C and D can be obtained.

Since M a = M (0) = EIw' ' (0) = EI j 2 B M a = K a (14)

in which K is the stiffness coefficient equal to s EI l , and where (Bazant and Cedolin [3])

(sin cos ) 2 2 cos sin

reaction moment





M b = M (l ) = EIw' ' (l ) = EI j 2 ( A sin + B cos ) . Therefore c can be defined as the relation between each end moment, which in practical terms becomes a carry-over factor for the beam-column (analogous to the one of the moment-distribution Cross Method), as:







Mb sin = M a sin cos


The understanding of the derivation of the carry-over factor is crucial to the understanding of the derivation of the K-factors. Equations (15) and (16) represent stability functions s and c which, in effect, modulate the stiffness of the beam-column as a function of its axial force (Barros and Oliveira [4]). For P = 0 , leads to the known result s = 4 and c = 1 / 2 for uniform inertia frame member. EUROPEAN CODES OF PRACTICE EUROCODES 2 AND 3 Consider now Figure 3, similar to beam-column of Figure 2 but instead of the fixed end (on the bottom) there is now a set of beams, with a rotational stiffness ( K b ) bottom at the bottom intersection of beams.

Figure 3 Loading of a PinnedPartially Fixed Beam-Column

If we apply a moment M a it is expected that the other end moment will be M b = CM a , (where C is the generalized carry-over factor for any type of stability function, since the previous s and c stability functions are only for the non-sway instabilities) but since this end is not fixed anymore there is a distribution of moments at the node according to its local stiffness.


EI + ( K b )bottom s l

EI l


Therefore there is a resulting moment equal to CM a k ' C that is distributed back to the original node, and to maintain the end rotation a the applied net moment is



= M a 1 C 2k'


From (18) and attending to the relation in (14), one can derive the stiffness coefficient for this situation as being
K =s EI 1 C 2k' l


Now in this gradual derivation, assume that on the top node of Figure 3 there is another set of beams with rotational stiffness ( K b ) top . In order for the column to collapse by instability, the top node must lose all its stiffness, as expressed by equation (20) for no-sway condition. EI s 2 l 1 c EI + ( K b )bottom s l + ( K b )top = 0

EI s l


As in equation (20) s and c are variables, and since s and c are only a function of the axial load (as implicit from equations 13, 15 and 16), then the solution of this equation (20) is the critical load of the beam-column and the corresponding K-factor value can be inferred by equation (3). A noteworthy fact to be emphasized is that, as mentioned by Wood [1], the loss of stiffness at each node occurs at the same time for the critical load. However all that is needed for the construction of charts, such as those in the Eurocode 3 Annex E [2], is the nominal distribution coefficient defined by I l i = I + l i where

I l bi


I is the effective stiffness coefficient as presented in table E.1 of Annex E of the l EC3, depending on the intersecting beams end restraints.

The effect of adjacent columns is accounted for as expressed in (19), in correspondence with (21) and (17), as mentioned by Wood [1] and also in EC3 [2].

i =

I l

I I + l l study adjacent columns I + + study l adjacent columns i

I l beam , i


The approximate expressions that the Eurocode 3 proposes for K factors, with the index 1 and 2 indicating respectively the top and bottom nodes, are as follows Nonsway Sway K f = 0.5 + 0.14(1 + 2 ) + 0.055(1 + 2 )
1 0,2 (1 + 2 ) 0,12 1 2 Kf = 1 0,8 (1 + 2 ) + 0,6 1 2

(23) (24)

In Eurocode 2 [5], however, what defines the K-factor is the relative flexibility at each node

ki =

M l


where is defined as the rotation of the elements that oppose the rotation imposed by moment M [5]. Through this definition, one can infer that this relative flexibility also means,
EI l columns ki = EI l beams


where is a parameter dependent of the end restraints of the beam (equal to: 4 if the end I restraint is fixed; 3 if it is pinned, etc), analogous to in equation (21). l Eurocode 2 [5] proposes the following expressions for the K-factors:


k1 k2 K f = 0.5 1 + 0.45 + k 1 + 0.45 + k 1 2



k k k k 1+ 1 1+ 2 K f = max 1 + 10 1 2 ; k1 + k 2 1 + k 1 1 + k 2


It is noteworthy that the expressions above have no physical interpretation, but are merely a numerical approximation to the exact solution of Westerberg [6].

When one observes the American Codes of practice it is apparent that the definition for the assessment of a columns end restraints is the same: equation (29) for AISC [7] and equation (30) for ACI [8].
EI l columns Gi = EI l beams


EI l columns i = EI l beams


From (29) and (30), it is apparent that Gi and i are identical and that this definition is closely related to the EC2s k i , as given before in equation (26). AISC [7] offers the following two implicit transcendental equations for the calculation of the K-factors, also given and programmed in Fortran by Barros [9] using Newton-Raphsons iterative method, which constitute the exact solution to the problem:


G1G2 4

K f

Kf G + G2 + 1 1 2 tan Kf

tan 2K f =1 + 2K f



36 G1G2 K f = 6(G1 + G 2 )

Kf tan



Since equations (31) and (32) can be quite impractical in design office recurrent use, AISC [7] developed two alignment charts which serve the same purpose, as represented in Figure 4.



Figure 4 AISC Alignment Charts for No-sway (a) and for Sway (b) conditions

ACI [8] also offers alignment charts which, through a quick comparison, are found to be identical to the ones that AISC [7] uses. It also presents some expressions equations (33) through (37) which allow the designer a quick assessment of K-factors. As one examines these equations it is apparent that they directly contradicts the own alignment charts. An example of this is to consider, for no-sway condition, a column with both ends fixed and compare the two results: with the alignment chart one obtains the theoretical value of 0.5 but using expressions 33 and 34 one finds that the K-factor is 0.7. Since these expressions are not closely correlated to the exact formulation of the charts the subsequent analysis will not be made. Instead, the following examination will be based on the conclusion that the ACI charts are identical to the ones presented in AISC and, therefore, a thorough investigation to the AISCs expressions will also render the same conclusions as the ones with ACIs.

(33) K f = 0.7 + 0.05 ( 1 + 2 ) 1.0 (34) K f = 0.85 + 0.05 min 1.0 (use the minimum of the two expressions) m < 2 2 m 20 m 1 + m 20 K f = 0.9 1 + m Kf = K f = 2.0 + 0.3 (35) (36) (37)

Restrained at both ends Sway Hinged at one end

where m is the average of i values at both ends of the column and the value at the restrained end.

The immediate application of 29 should be carefully considered because it is based on a few simplified assumptions. In AISC [7] little reference is made to this fact except for the following statements: for sidesway inhibited frames these adjustments for different beam end conditions may be made: (i) If the far end of a girder is fixed, multiply (EI/L)g of the member by 2.0; (ii) If the far end of the girder is pinned, multiply (EI/L)g of the member by 1.5 [7]; and for sidesway uninhibited frames (i) If the far end of a girder is fixed, multiply (EI/L)g of the member by 2/3; (ii) If the far end of the girder is pinned, multiply (EI/L)g of the member by 0.5 [7]. Also for sidesway uninhibited frames, it allows the use of a modified beam length given by equation (38).
Mf l ' beam = lbeam 2 M n


where M f is the moment at the far end, and M n the moment at the near end. To clarify this, consider the beam in Figure 5.

Figure 5 Beam Pinned at Both Ends with Applied End Moments

Applying the well know matrix equilibrium equation (39), the first one leads to equation (40).
M n EI 4 2 n M = 2 4 l f f beam Mn


Defining b as

f = 4 1 + 0 .5 n

EI l beam


f b = 4 1 + 0.5 n


and b0 as a reference value, one can infer that equation (29) can be generally defined by equation (42) as introduced by Hellesland and Bjorhovde [10].

EI l columns = Gi = b EI b l beams 0

EI l columns EI m l beams


The parameter m quantifies the effect of the beam end restraints on the overall stiffness of the node, and the difference between the codes of practice, to some extent, depend on the reference value. For example, in EC3 the reference value is set to 4 and in EC2 that value is equal to 1. What can be concluded from AISC statements (Hellesland and Bjorhovde [10]) is that the reference value for which the transcendental equations (31) and (32) were derived, depends itself on the type of structure being analyzed ( b0 = 2 for no-sway ; b0 = 6 for sway). This is proven by the derivation of those presented expressions (31) and (32), as were originally developed by McGuire [11]. To better understand this, consider the following structures adapted from [11].



Figure 6 Buckling modes for a one-storey frame: No-sway (a) and Sway (b)

The corresponding b value for Figure 6 (a), since the rotations at both ends are equal and opposite, is equal to 2; the same as the reference value b0 for no-sway frames. Similarly, for the sway frame of Figure 6 (b), since the rotations at both ends are equal both in magnitude and direction, the b value is 6; the same as the reference value b0 for sway frames. As such, one may conclude that the derivations for the K-factors were made this way so that the design of these two basic and recurrent types of frames could be done without having to consider EI is used in the analyses). every time the beams end restraints (i.e. only l beams To derive equation (31) consider the following sketch in Figure 7 of a no-sway frame buckling, where lc is the column length and lb the adjacent beam length. The governing equation of the deformed shape, which can be obtained from equation (9) by attributing the appropriate boundary conditions and assigning P as the critical load, is:

x y = A sin K l f c



Figure 7 Multistory Frame Buckling No-Sway (adapted from [11] )


2 = y x = x = A sin

x2 K l f c



1 = y x = x l = Asin
2 c

x2 cos K K f lc f

cos x 2 K l f c

sin K f



= ( 2 1 ) l c =

x2 A sin lc K l f c

1 cos K f

+ cos x 2 K l f c

sin K f



2 =

dy dx

x = x2

x2 A sin lc K l f c

1 cos K f

+ cos x 2 K l f c

sin K f

K f


1 =

dy dx

x = x2 lc

x 2 1 + cos + sin sin K K K f f A K f lc f = lc x2 sin cos + cos K l K K K f c f f f



Assuming a frame deflection typical of no-sway buckling modes in which the beams at level 1 and 2 all have a corresponding moment of inertia I b1 and I b 2 respectively, and similarly all the columns with I c , the bending moment throughout the beams length will be constant with absolute values of P 1 at level 1 and P 2 at level 2 (with column axial load P ). As such and since as discussed the end rotations of the beams are equal and opposite (i.e. b = 2 ), the rotation can be obtained by:

2 =

P 2 lb 2 EI b 2


Substituting P for the critical buckling load and taking 2 from equation (44), equation (49) becomes
A 2 = 2l c K f

I c lb l I c b2

x2 sin K l f c


In this case, the relative stiffness ratio at node 2 is then the already defined G2 :
G2 = I c lb lc I b2


Similarly at node 1:
AG1 1 = 2l c Kf

x2 sin K f lc

cos K f

cos x 2 K l f c

sin K f


From equations (47) and (50) the following expression is obtained:

G 2 2 + cos K K f f sin K K f f

x2 cot K l f c


Proceeding in the same manner for equations (48) and (52)

G 1 2 Kf cos + 1 cos K K f f sin K f

x2 cot K l f c

sin K K f f cos K f

cos K Kf f

G1 2 K f


The expression for no-sway that is offered in AISC [7] can be obtained from (53) and (54)

G1G2 4

K f

Kf G + G2 + 1 1 2 tan Kf

tan 2K f + =1 2K f


To derive equation (32) for sway frames, consider the following sketch in Figure 8 of a sway frame buckling.

Figure 8 Multistory Frame Buckling Sway (adapted from [11] )

Equation (43) is still applicable because the boundary conditions are still the same. However, equations (44) through (48) differ slightly and become:

2 = y x = x = A sin

x2 K l f c x2 cos K f K f lc cos K f cos x 2 K l f c

(56) sin K f

1 = y x = x l = Asin


2 =

dy dx dy dx

x = x2

x2 A cos K l K f lc f c = A K f lc

(58) + sin x 2 K l f c sin K f

1 =

x = x 2 lc

x2 cos K f lc


Similarly, for frame deflections according to sway buckling modes, where the moment of inertia of the beams is constant at each level, and all columns in the frame have the same moment of inertia, the rotations at each end of a beam are equal both in magnitude and direction. As such b = 6 and the rotation, for example, at node 2 can be expressed as:

2 =

P 2 lb 6 EI b 2


Substituting P for the critical buckling load and taking 2 from equation (56), equation (60) becomes
AG2 2 = 6l c K f sin x 2 K l f c


Proceeding in the same manner for 1

1 =

dy dx

x = x 2 lc

AG1 6l c

x2 sin K f lc

cos K f

+ cos x 2 K l f c

sin K f


Eliminating 2 from equations (58) and (61) and 1 from equations (59) and (62), one obtains two expressions for cot x 2 K f l c (similar to (53) and (54)). Equalizing them

(eliminating the common term) and simplifying the result, the final expression that is used in AISC [7] is obtained:
36 G1G2 K f = 6(G1 + G 2 )






In the present form, immediate comparison of the K-factors expressions is impossible because of the different nature of the distribution coefficients (EC3), the relative flexibility ratio (EC2) and relative stiffness ratios (AISC). Although in essence they are the same thing, one shall differentiate between them for the sake of better understanding. However, consider equation (63) as a transformation of equation (22) from a nominal distribution coefficient to an effective distribution coefficient
l I

i =

I + l columns i

I l beam,i

4 E 4 E


I + l columns i

I l beam,i

(63) 4 EI l columns = = 4 EI 4 EI + l beam,i l columns i

4 EI l columns 4 EI EI + l beam,i l columns i

If one divides ki per i one obtains


EI l columns EI l beam,i ki i = = i 4 EI l columns 4 EI EI + l beam,i l columns i

EI l columns EI l beam,i i = EI 4 l columns 4 EI EI + l beam,i l columns i


EI 4 EI 4 EI + l columns l beam,i 1 i l columns + = 4 EI EI 4 l beam,i l beam,i i i


From equation (64) and also attending to equation (26):

1 k i = + k i i 4 ki =
4 (1 i )


Equation (65) gives a relation between distribution coefficients and relative flexibility ratio, used in both EC3 [2] and EC2 [5] respectively. In Figures 9 and 10 one can see the differences between the K-factors given by EC3 (dashed lines) and EC2 (continuous lines) for no-sway and sway frames, respectively.

Figure 9 Comparison of K-factors given in EC2 and EC3 for No-Sway frame members


Figure 10 Comparison of K-factors given in EC2 and EC3 for Sway frame members

Analyzing the data, one can assess the maximum difference between the two codes of practice, which for no-sway frame members is about 3% and for sway frame members about 15%. Since both procedures are a result of equations adjusted to transcendental expressions (23, 24, 27 and 28) one can state that the observed differences are a result of numerical errors of round-off approximations; EC3 being overall the most conservative. Proceeding in the same way according to equations (64) and (65), one can derive the relation between AISCs nodal stiffness ratios and the corresponding counterparts in EC2 and EC3. Table 1 summarizes all of these relations. Table 1 Relations of nodal stiffness ratios between EC3, EC2 and AISC EC3 EC3
ki =
Gi =



i =

4k i 1 + 4k i

i =

Gi b Gi + 0 Gi b0


4 (1 i )
b0 i 4 (1 i )

Gi = b0 k i

ki =


Since the AISC expressions for K-Factors are implicit, a practical alternative for comparing different results is to solve the equation as a function of one of its end restraints ratios for a given K-factor, as presented in equations (66) and (67), derived from equations (31) and (32) respectively.



tan 2K f Kf G 2 1 1 1 + 2 tan 2K f Kf G2 = 2 Kf G1 + 1 2 K f tan Kf


6G1 Kf tan


+ 36


G2 = G1 K f

Kf 6 tan Kf


Proceeding similarly for EC3 one obtains:

No-sway 2 =

b b 2 4ac 2a


a = 0.055 b = 0.14 + 0.111

c = 0.5 + 0.141 + 0.0551 K f



Sway 2 =

K f (0.8 0.61 ) 0.121 0.2


K f (1 0.81 ) + 0.21 1


As for EC2, deriving from expression (27) yields:


k 2 = 0.45

4K f 1 k1 1+ 0.45 + k1 4K f 2 k1 1+ 0.45 + k1



Since equation (28) defines the K-factor as a maximum of two expressions, one has to apply the same procedure to both and, when plotting the comparison, judge which one translates into the biggest K-factor. Expression A: Expression B: Kf 1 k1 1+ 1 + k1 k2 = Kf 2 k 1+ 1 1 + k1 0.1( K f 2 1) k1



k2 =

k1 0.1( K f 2 1)


With the expressions that have been presented equations (66) through (73) and using the relations of Table 1, one can obtain the following results shown in Figures 11 through 13. An important aspect of such figures is that the conclusions as to the difference in K-factors (as those made for Figures 9 and 10) are not immediately possible, because what is displayed is in fact the difference in nominal distribution coefficients i . The variation in a i is not representative of the variation in a K-factor for instance, in no-sway 1 = 0 2 = 0.599 K f = 1.3 and 1 = 0 2 = 0.273 K f = 1.1 therefore to a certain difference of 120% in i corresponds a certain variation of 18.2% in the K-factor. What can be inferred, however, is the relative significance of each curve, i.e. whether one is more conservative than the other, and under what conditions those assertions are valid. As such, the constructed charts serve as a qualitative basis for analysis.

Figure 11 Comparison between Codes of Practice for No-Sway frame members


Figure 12 Comparison between Codes of Practice for Sway frame members, with K f { 1.3; 2.0; 3.0}

Figure 13 Comparison between Codes of Practice for Sway frame members, with

K f {7.0 ; 8.0 ;10.0 ;15.0}

From the figures presented, one can validate the conclusion that the expressions given by each code of practice are very closely related and are not equal because EC3 and EC2 are numerical approximations of the exact formulation of the problem, which correspond to AISCs curves.


Figure 11 charts the comparison for no-sway frame members between EC3, EC2 and AISC for three values of K-factors: 0.55, 0.7 and 0.85. The main conclusion from the analysis of this chart is that EC3 is overall more conservative than EC2 same conclusion already reached from Figure 9 and EC2 goes as far as in fact being non-conservative. But since the differences between EC2 and EC3 are at the most 3% for these cases, this is of negligible significance. In Figures 12 and 13 one can see the comparison for sway frame members between EC3, EC2 and AISC. Perhaps this is the most interesting case to analyze because of the two expressions given by EC2. As one can observe there is a discontinuity in the EC2-A curve that gives way to EC2-B, each with different curvatures. As the K-factor increases, EC2-A loses its weight in the expression entirely; see Figure 10 for K f = 3.0 , in which EC2-A was fully plotted on purpose. EC3 for small values of K-factor (< 3) is conservative. For values higher than 3 it becomes non conservative (see Figure 12). This fact in itself is not very pejorative, because as seen in Figure 10 for high K-factors EC3 and EC2 practically coincide; since both are upper and lower boundaries, the difference between EC3 and the exact formulation is negligible. In EC2, however, problems arise precisely were the difference seen in Figure 10 is most critical (yielding errors between 10-15%). This area can be observed in Figure 12 for K f = 3.0 , where it is clear that the curve EC2-B diverges from the exact solution as i increases. This aspect is critical because it is on the unsafe side, even though this translates, for example, in a difference between 2.91 and 3.33 in K-factor values.

A comparison was made between European (EC2 and EC3) and American (ACI 318 and AISC 360) codes of practice with regards to the effective length K-factors, in assessing critical loads of slender columns in no-sway and sway frames. A method of comparison between each code was presented, which is based in finding a common link in the assessment of end restraints. For column members in same circumstance the effective length that is obtained is in essence the same, apart from some numerical errors of round-off approximations. In conclusion, the best numerical adjustments for the K-factor values are the EC3 expressions, yielding minimal errors and being overall the most conservative. When it is not conservative the errors are found to be negligible.

1. 2. 3.


Wood, R.H., Effective Lengths of Columns in Multi-Storey Buildings - Part 1. Structural Engineer, 1974. 52(7): p. 235-244. European Committee for Standardization, Eurocode 3: Design of steel structures Part 1-1 General rules and rules for buildings. 2005, CEN: Brussels. Bazant, Z.P. and L. Cedolin, Stability of Structures: elastic, inelastic, fracture and damage theories. Oxford Engineering Sciences Series. 1991, New York: Oxford University Press. XXIV, 984 p.-XXIV, 984 p. Barros, R.C. and S. Oliveira, Anlise Elstica de 2 Ordem de Prticos Planos em Calculadora Programvel, Proceedings 3 Congresso Luso Moambicano de Engenharia, Maputo, Moambique, 19-21 Agosto 2003. Volume I, Tema C: Materiais e Estruturas, pp. 435-446.


5. 6. 7. 8. 9.

10. 11.

European Committee for Standardization, Eurocode 2: Design of concrete structures Part 1-1 General rules and rules for buildings. 2004, CEN: Brussels. Westerberg, B., Second Order Effects in Slender Concrete Structures, in Background to the rules in EC2. 2004, KTH Civil and Architectural Engineering: Stockholm. American Institute of Steel Construction, Specification for Structural Steel Buildings. 2005, AISC. American Concrete Institute, Building Code Requirements for Structural Concrete and Commentary 2005, ACI. Barros, R.C., Dimensionamento de Colunas em Ao, Relatrio Tcnico Interno e tambm de apoio tcnico-cientfico disciplina Instabilidade de Estruturas, 13 edio do Curso de Mestrado em Estruturas de Engenharia Civil na F.E.U.P., Ano lectivo 1997/98, F.E.U.P., Abril 1998. Hellesland, J. and R. Bjorhovde, Restraint demand factors and effective lengths of braced columns. Journal of Structural Engineering, 1996. 122(10): p. 1216-1224. McGuire, W., Steel Structures. 1968, Englewood Cliffs, N.J.: Prentice-Hall, Inc.