Sei sulla pagina 1di 8

Engineering Failure Analysis 14 (2007) 434441 www.elsevier.

com/locate/engfailanal

Use of safety factors for the design of steel structures according to the Eurocodes
Gerhard Sedlacek *, Oliver Kraus
Institute of Steel Construction, RWTH Aachen, Mies-van-der-Rohe Str. 1, D-52074 Aachen, Germany Received 30 August 2005; accepted 30 August 2005 Available online 17 August 2006

Abstract This paper informs about the safety concept of the Eurocodes as dened in EN 1990/1/ and how it is implemented in the EN 1993/3/ for the design of steel structures. 2006 Elsevier Ltd. All rights reserved.
Keywords: Structural reliability; Limit state design; Partial safety factors; Combination factors

1. General The structural Eurocodes comprise the following groups: EN 1991 in which load assumptions are dened. EN 19921996 and EN 1999 which dene the material dependent regulations. EN 19971998 which cover geotechnical and earthquake design. The detailed titles of these documents in the form of Eurocodes 19 are given in Table 1. EN 1990 [1] covers Basis of Structural Design in which the safety concept of the Eurocodes is explained. 2. Limit state design According to the rules in the Eurocodes the design of structures and structural members has to be veried for dierent limit states. For the verication that a structure can bear the relevant loadings, called the Ultimate Limit State (ULS), it has to be checked whether the resistance capacity is equal or greater than the sum of the relevant action eects. For the verication of the serviceability, called the Serviceability Limit State (SLS), it has to be checked whether the functioning of the structure is guaranteed e.g. that a certain threshold value of deection will not be exceeded under given loading conditions.
*

Corresponding author. Tel.: +49 241 80 25177; fax: +49 241 80 22140. E-mail address: stb@stb.rwth-aachen.de (G. Sedlacek).

1350-6307/$ - see front matter 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.engfailanal.2005.08.002

G. Sedlacek, O. Kraus / Engineering Failure Analysis 14 (2007) 434441 Table 1 Parts of the structural eurocodes EN EN EN EN EN EN EN EN EN EN 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 Eurocode Eurocode Eurocode Eurocode Eurocode Eurocode Eurocode Eurocode Eurocode Eurocode 0 1 2 3 4 5 6 7 8 9 Basis of Structural Design Actions on Structures Concrete Structures Steel Structures Composite Structures Timber Structures Masonry Structures Geotechnical Design Earthquake Resistance Aluminium Structures

435

For the ULS design in EN 1990 the following fundamental equation is given: X X Rk P Ed cG;j Gk;j \ "cP P \ "cQ;1 Qk;1 \ " cQ;i w0;i Qk;i cM jP1 iP1 where Rk is the characteristic value of the resistance; Ed is the design value of the action eects; Gk is the characteristic value of permanent actions; P is the characteristic value of prestressing; Qk is the characteristic value of time variant actions; cM is a material dependent partial factor on the resistance side; cG is a partial factor for permanent actions; cP is a partial factor for actions due to prestressing; cQ is a partial factor for time variant actions; w0 is a combination factor for fundamental (or characteristic) load combinations; + implies to be P combined with; implies the combined eect of. Three dierent equations are provided by EN 1990 for the SLS design with X X Ed Gk;j \ "P \ "w1;1 Qk;1 \ " w2;i Qk;i
jP1 i>1

for frequent load combinations, and X X Ed Gk;j \ "P \ " w2;i Qk;i
jP1 iP1

for quasi-permanent load combinations. where w1 is a combination factor for frequent load combinations; w2 is a combination factor for quasi-permanent load combinations. From the equations above it can be seen that the safety elements the partial factors and the combination factors are varied in dependence of the relevant limit state, whereas the characteristic values, introduced into the verication, are the same for all limit states. Using characteristic values in connection with safety elements enables to achieve dierentiated safety levels for several limit states. It has to be mentioned that in the Eurocodes further design equations are given for dierent design situations, as e.g. for accidential actions these cases are not considered in the frame of this report. 3. Characteristic values of resistances and of actions Values given in the Eurocodes as well as values for resistances, given by producers, should correspond (at least approximately) to certain fractiles. Values on the resistance side should correspond to the 5%-fractile, characteristic values of permanent actions are dened as the 50%-fractile (mean value) and characteristic values of variable actions (e.g. climatic actions) should be in the range of the 98%-fractile of the distribution of the annual extremes, which means an average return period of 50 years (see Figs. 1 and 2). Determining characteristic values on the action side and on the resistance side is one of the most important research topics concerning the Eurocodes. Extensive investigations of measured data for developing realistic load assumptions (e.g. to obtain a European snow load map [2,4]) and a signicant number of tests on structural members particularly in the eld of steel elements for EN 1993 [3] were carried out and still have to be carried out.

436

G. Sedlacek, O. Kraus / Engineering Failure Analysis 14 (2007) 434441

Fig. 1. Characteristic values corresponding (approximately) to certain fractiles.

Fig. 2. Sketch of the limit state function R E = 0.

4. Safety elements dened in the Eurocodes 4.1. The target reliability In order to obtain structures for which resistance and serviceability are in a target range of reliability, the characteristic values have to be adjusted by safety elements. For the simple case of normally distributed variables E on the action side and R on the resistance side with the limit state function RE 0 the safety margin is in turn normally distributed with the mean value l and the standard deviation r: q l lR lE r r2 r2 R E With these values the reliability index b can be determined corresponding to a probability of failure pf: l with pf /b b r where u is the cumulative standard normal distribution. To answer the question which value on the action side and which value on the resistance side lead to the desired reliability level the weighting factors aE (for the action eect) and aR (for the resistance) have to be introduced: rE rR aR aE r r Then the so called design values Ed and Rd can be calculated:

G. Sedlacek, O. Kraus / Engineering Failure Analysis 14 (2007) 434441 Table 2 Target reliabilities recommended in EN 1990 Limit state ULS SLS Target b 3.8 1.5

437

pf = /(b) 7.2 105 6.8 102

Ed lE aE b rE R d l R a R b rR It has to be pointed out that this formulation of the reliability and of the design values is only exact in case of normally distributed variables and in case of a linear limit state function. Alternative procedures or simplifying methods, which are not topic of this paper, have to be applied if one of these conditions is not met. In EN 1990 the target reliabilities shown in Table 2 are recommended for the design life for ULS and SLS. 4.2. The calibration of safety elements Once the statistical parameters of normally distributed variables E and R are known the calibration of partial factors cF (for the action side) and cR (for the resistance side) can be carried out by comparison of the design values with the characteristic values: cF lE aE b rE 1 aE b V E ; l E k rE 1kVE cR lR k rR 1kVR lR aR b rR 1 aR b V R

where VE is the coecient of variation of the variable on the side of the action eect; VR is the coecient of variation of the variable on the resistance side; k is a factor to obtain a target fractile of the Normal Distribution; (k = 1.65 for 5%-fractile, k = 0.0 for 50%-fractile, k = 2.05 for 98%-fractile). From the equations above, containing the weighting factors a, it can be concluded that aiming at a certain level of reliability both for cF and cR one has to take into account the action side and the resistance side as well. In the exemplary case of tests on structural members in order to derive the statistical parameters of their resistance this would cause a problem. Partial factors for the resistances could not be calibrated to the test results due to the fact that the parameters of the relevant actions are not known (in most cases it would be even unknown which kind of actions could be important for the design, as the tested member could be used for a great variety of structures and purposes). To solve this problem in the Eurocode an approach is applied to calibrate the partial factors. The weighting factors are xed to aE = 0.7 and aR = +0.8, provided that 0.16 < aE/aR < 7.6. This procedure, xing the contribution of the action side and of the resistance side to an estimated value, is called the semi-probabilistic safety concept [5,6]. Partial factors for the ULS are then calibrated by the following equations: cF 1 0:7 3:8 V E ; 1kVE cR 1kVR 1 0:8 3:8 V R

It should be mentioned here that alternative equations are given in EN 1990 for variables which are not distributed according to the Normal Distribution but to a dierent statistical distribution, as e.g. according to the Gumbel- or Log-Normal Distribution. In the partial factors not only is the variation of the variable considered taken into account but also the model uncertainty, i.e. that the coecients of variation VE and VR in the formulas above represent values that may include several scattering parameters [7]. By introducing the model uncertainty into the determination of partial factors the scatter in predicting an action eect on a certain structure and in deriving the resistance from a design model is taken into account (see Section 4.5). For instance, the model uncertainty for predicting the eect due to the self weight of a steel beam is much lower than for predicting the eect due to uctuating wind loading. As already mentioned in Section 2 safety elements besides partial factors include combination factors to be applied on variable loads. Reducing variable actions with a combination factor takes into consideration that

438

G. Sedlacek, O. Kraus / Engineering Failure Analysis 14 (2007) 434441

the simultaneous occurrence of two or more extreme values of uctuating values corresponds to a probability which is much lower than the target failure probability pf. For the case of the eects E1 and E2 of two variable actions Q1 and Q2 with the limit state function R E1 E2 = 0 the following resulting standard deviation r is obtained: q q r r2 r2 r2 r2 r2 R R E E1 E2 Consequently two dierent weighting factors on the action side have to be determined: rE1 rE1 rE aE1 p p a aE E1 rE r2 r2 r2 r2 r2 R R E E1 E2 rE2 rE2 rE aE2 p p a aE E2 2 2 2 2 rE rR rE1 rE2 rR r2 E where: rE q r2 r2 E1 E2

The weighting factors aE and aR are called global weighting factors while the factors a and a are called E1 E2 local weighting factors. Knowing these values the determination of combination factors is possible as follows: wE1 wE2 lE1 a aE b rE1 E1 lE1 aE b rE1 l a aE b rE2 E2 E2 lE2 aE b rE2

Once again in the semi-probabilistic concept a simplication is applied: the local weighting factor is xed to 1.0 for the dominating variable action and to 0.4 for the accompanying variable action enabling the determination of the combination factor for a certain load without statistical evaluation of the other variable loads of the load combination considered. On the assumption that the action E1 is the accompanying action we get the following denition of the combination factor: wE1 lE1 0:4 0:7 3:8 rE1 1 0:28 3:8 V E1 lE1 0:7 3:8 rE1 1 0:7 3:8 V E1

Note that in most cases this approach leads to safe sided results, particularly in case of variable actions with negative correlation, e.g. for some regions in Europe for the combination of the climatic actions wind and snow. 4.3. Safety elements on the action side dened in EN 1990 The safety elements recommended in EN 1990 are listed in Tables 2 and 3 for the dierent limit states at ULS, comprising the criteria: EQU ! loss of equilibrium rigid body instability, STR ! structural resistance, GEO ! sizing of foundations and at SLS, comprising the criteria that a threshold value is not exceeded for deformation, vibration and damage.

G. Sedlacek, O. Kraus / Engineering Failure Analysis 14 (2007) 434441 Table 3 Safety elements for ULS and SLS recommended in EN 1990

439

440

G. Sedlacek, O. Kraus / Engineering Failure Analysis 14 (2007) 434441

It has to be pointed out that all safety elements in the Eurocodes are recommended values that can also be adjusted in the National Annexes by the member states. 4.4. Safety elements for steel members dened in EN 1993 In EN 1993 Steel Structures [3] a distinction is made between the possible failure modes at the ULS: mode 0! failure with excessive deformation by yielding before failure (verication of resistance of cross sections related to the yield strength), mode 1! stability failure induced by imperfections and yielding (verication of resistance of members related to the yield strength) and mode 2! fracture failure after yielding (verication of net sections for tension related to the ultimate strength).
Table 4 The procedure Design Assisted by Testing for multiplicative safety elements (left column) and for additive safety elements (right column) taking into account the model uncertainty Multiplicative form Additive form

1. Tests on structural members are carried out and the resulting experimental resistances rei are compared with the theoretical resistances rti according to the design model considered

2. The deviation of the theoretical values from the experimental values are considered yielding the correction terms bi. bi = rei rti bi rei rti 3. The mean value of the correction terms is calculated  1 Pn bi b
n i1

 1 Pn bi b n i1 di bi  b  1 P di ) 0 d n rd0 q Pn 1 2 i1 di d n1

4. The error terms di are determined di  rei bi  br b


ti

5. TheP mean value of the error terms is calculated. 0 1 d0 ) 0 d


n i

6. The standard deviation of the error terms is calculated. q P 1 d S d0 n1 n d0i  2 i1

7. The standard deviation of the error term is combined with the standard deviations of further relevant variables (e.g. yield strength and geometry,. . .) r r 2 2 P ogX M P ogX M 2 0 S D S d rD rd 2 oX M;i rx;i oX M;i rx;i where g = limit state function XM = mean value of variable X where g = limit state function XM = mean value of variable X

8. The design functions rd can be derived. In the multiplicative form a Log-Normal Distribution is used, and in the additive form a Normal Distribution is used 0 0 2 rd gR X M  eaR bS D 0:5S D b rd gR X M  aR b rD b  where lR gR X M b lR gR X M  b 9. The partial factor c resp. the additive safety element DTR are determined and related to the relevant nominal values XN M c gR X N M rd DTR = gR(XN) rd

G. Sedlacek, O. Kraus / Engineering Failure Analysis 14 (2007) 434441

441

The partial factor on the resistance side for steel structures has to be chosen in dependence on the relevant failure mode: mode 0 Rk Rd0 with : cM0 1:00 ! cM0 mode 1 ! Rk Rd1 cM1 Rk Rd2 cM2 with : cM1 1:10

mode 2 ! with : cM2 1:25

Note that all conversion factors, model uncertainties (see also Section 4.5) and scatter eects are included in the design values Rd0, Rd1 and Rd2. For the verication in the SLS the general partial factor cM = 1.0 is applied. 4.5. Determination of partial factors according to annex D of EN 1990 In Annex D to EN 1990 it is dened how partial factors cM for resistances can be derived from tests on structural members taking into account the uncertainty of the design model considered. This procedure is called Design Assisted by Testing. In the current version the partial factors are limited to multiplicative elements, and it is not taken into consideration that for thermal actions as well as for resistances on thermal actions (e.g. fracture toughness of steel) the use of additive safety elements is more appropriate. Consequently an introduction not only to the current procedure of Annex D but also as to how, in the form of a supplement, the determination of additive safety elements could be achieved is given in Table 4. In the left column the current method for multiplicative safety elements is presented whereas in the right column of the table the additional method for additive safety elements is dened. 5. Conclusion Thanks to extensive research activities in terms of tests on structural components in order to get the statistical parameters of design models and in terms of the evaluation of numerous data sets of measured data of climatic actions in order to develop European maps for actions it is now in many cases possible to design structures which correspond approximately to one target reliability. However, some rules and recommended values (as e.g. combination factors) are up to now not consistent with the target reliability level, so that the research in the eld of structural Eurocodes still has to be continued to identify inconsistencies and to improve some recommendations of the Eurocodes. One very helpful instrument to derive safety elements from tests on structural members is the procedure Design Assisted by Testing dened in Annex D to EN 1990. References
[1] [2] [3] [4] EN 1990 Basis of Structural Design, CEN, 2002. EN 1991-1-3 Eurocode 1: Actions on Structures Part 1.3: General Actions Snow Loads CEN, 2005. EN 1993 Eurocode 3: Design of Steel Structures CEN, 2005. Sanpaolesi, L. Scientic support activity in the eld of structural stability of civil engineering works snow loads, Commission of the European Communities, DGIII D3 Report University of Pisa, 1999. [5] Background Documentation to Eurocode 3, Document 2.01: Background Document for Chapter 2 of EC 3, Commission of the European Communities 1989. [6] Background Documentation Eurocode 1 (ENV 1991) Part 1: Basis of Design European Convention for Constructional Steelwork 1996. [7] Brozetti J, Sedlacek G, Kraus O. Commentary to EN 1990 Basis of Structural Design Report, 2002.

Potrebbero piacerti anche