Anda di halaman 1dari 11

Journal of Constructional Steel Research 64 (2008) 12641274

Contents lists available at ScienceDirect

Journal of Constructional Steel Research

journal homepage:

Analysis of the behaviour of stainless steel bolted connections

A. Bouchar , J. Averseng, A. Abidelah
LaMI, Civil Engineering, Blaise Pascal University, rue des Meuniers, BP 206, 63174 Aubire cedex, France



a b s t r a c t
This study is focused on two types of bolted connections that are common in steel structures. They concern cover plate connections and T-stubs, where the bolts are loaded in shear or in tension. The Eurocode 3 requirements for stainless steel connection design are essentially the same as for carbon steel. The study considers the case of austenitic stainless steel for which the conventional elastic limit is relatively low compared to the ultimate strength. In bearing, criteria on deformation limits have to be considered for cover plate connections. In T-stubs, strain hardening of stainless steel exhibits a continuous increase of the applied load and can influence the failure mode. A finite element model is developed and validated for the two types of connections. A more extensive parametric study should be carried out to develop a better understanding of the behaviour of stainless steel connections. 2008 Elsevier Ltd. All rights reserved.

Article history: Received 27 November 2007 Accepted 1 July 2008 Keywords: Bearing Finite element model Non-linear analysis Stainless steel Steel connection T-stub

1. Introduction The use of stainless steel for structural applications in civil engineering is relatively limited. In construction, stainless steel is commonly used for secondary elements. Structural design is covered by EN 1993-1-4 [1]. Stainless steel is used in construction for multiple reasons. Its high ductility is advantageous with respect to energy dissipation in the case of cyclic loading and enables loads to be redistributed before failure. It has excellent resistance to corrosion, good aesthetic appeal, good ductility, good resistance in fire [26] and strong strain hardening characteristics. Its use in train structures shows that it offers, in addition to reduced life cycle costs, significant deformation capacity, which enables energy dissipation during a shock. Several studies on the mechanical behaviour of stainless steel structural elements are available [710]. As for carbon steel, there is a wide variety of grades of stainless steel due to the variation of chemical composition and thermal treatment. They can be classified into five main groups according to their metallurgical structure [11,12]. The austenitic type is the most widely used in construction in annealed or cold worked states. The duplex type is also used in some applications, and its use is growing, mainly for bridges applications. The stainless steel stressstrain curve is different from that of carbon steel. Indeed, carbon steel exhibits linear elastic behaviour followed by plastic deformation before strain-hardening, while stainless steel has a curve without a well-marked elastic limit and it is non-linear even for low levels of load. A conventional

Corresponding author. Tel.: +33 473407532; fax: +33 473407494. E-mail address: (A. Bouchar).

limit at 0.2% of the plastic strain is usually used. Stainless steel additionally exhibits a non-symmetrical behaviour in tension and compression [11] and has a general tendency to displaying anisotropic behaviour. However, the aspect which has potentially most influence for structural applications is the non-linearity. Furthermore, as for certain metal alloys, stainless steels are subject to significant creep at ambient temperatures, unlike carbon steel. For this reason, stainless steels should only be used in a controlled way if a high level of stress is to be maintained during a long period. In some cases, it is recommended to restrict the stress due to the long-term actions to a portion of the conventional elastic limit [11]. According to EN 1993-1-4, it is necessary to demonstrate the efficiency of preloaded bolts from test results and to take account of the effect of creep and shrinkage when necessary. Stainless steel structural connections may utilize welding, bolting or other mechanical fasteners. Basic design is not very different from that of carbon steel connections. However, the higher ductility offered by stainless steel should be beneficial if the mechanical characteristics are well used. In fact, the conventional elastic limit is used for stainless steel and its value is relatively low compared to the ultimate limit. The over-strength developed in the connection can affect the failure mode, mainly for the connections combining bolt tension and plate bending (such as a T-stub). Furthermore, several general design aspects have to be considered, in particular to minimize the risk of corrosion. Plate-to-plate bolted connections are easy to use. Their load transfer is made by plates in bearing and bolts in shear. The mechanical behaviour of this type of connection mobilizes the bolt shear, the bolt-hole bearing and the net-section deformation. As part of an ECSC investigation, a study on stainless steel bolted cover plate connections was carried out at the Civil Engineering Laboratory (LGC), Blaise Pascal University [1315]. Plates in several

0143-974X/$ see front matter 2008 Elsevier Ltd. All rights reserved. doi:10.1016/j.jcsr.2008.07.009

A. Bouchar et al. / Journal of Constructional Steel Research 64 (2008) 12641274


types of stainless steel (austenitic, ferritic and duplex) connected by three bolt diameters in different arrangements were examined. Only the results of the austenitic steel (304 L) are analysed in this paper because it is, so far, the mostly used in construction and it has a high ductility. The experimental results are analysed to assess the resistance formulae for the cover plate bolted connections and to obtain an estimate of the bearing deformation [16]. They show that attention has to be given to the resistance of the connection compared to that of the connected element. Thus, for bolted connections, a design for ULS (Ultimate Limit State) loads may not be sufficient to avoid unacceptable plastic deformations at SLS (Serviceability Limit State). Available design codes are mainly focused on structural carbon steel. For stainless steel, the formulae and design approaches are modified to take into account the non-linear behaviour of the material. Preloaded bolts are rarely used because of the low friction coefficient between the plates and the difficulty obtaining a suitable value of pretension. However, an experimental study [17] showed that stainless steel fasteners can perform well in resisting slip. Furthermore, in addition to the creep effect, galling can be a problem for stainless steel bolts. Another common type of connection is considered in this study. It concerns a T-stub which is a common component that can represent the tension zone in a beam-to-column bolted connection. Its load transfer is made by the flange in bending and bolts in tension. In this type of component, the stiffness ratio between the flange and the bolts induces prying effects that amplify the forces in the bolts. The experimental results available are for carbon steel T-stubs. They are used for the validation of the numerical model. According to EN 1993-1-4, connection design should be done using the provisions given in EN 1993-1-8 [18] except where modified or superseded by special provisions such as for plate bearing, shear resistance of bolts or the resistance of welds. Thus, implicitly, the provisions for carbon steel can be used to design a stainless steel connection under bending such as an endplate. In this type of connection, plastic analysis is used particularly to characterize the tension zone represented by the equivalent T-stub component. Due to the lack of experimental results for stainless steel, it was decided to use a numerical model to compare stainless steel and carbon steel T-stubs. This comparison concerns two geometrical configurations of T-stubs for which the experimental results for carbon steel are available. Thus, the experimental results are used to validate the numerical model that is subsequently used to analyse the stainless steel T-stubs. For stainless steel, it is stated that a plastic global analysis should not be used unless there is sufficient experimental evidence to ensure that the assumptions made in the calculations are representative of the actual behaviour of the structure. In particular there should be evidence that the connections are capable of resisting the increase in the internal moments and forces due to strain hardening. Thus, the moment resisting connections have to be analysed further to compensate for the lack of data and to evaluate the over-strength, which can be developed in stainless steel members and connections due to the evolution of strain hardening which can be different between the connection and the members. In the connection, this parameter can be mainly dependent on the failure mode of its basic components. In the T-stub, the development of strain hardening in the material is different in the flange in bending and the bolts in tension. This necessitates the development of an extensive study to understand the main parameters that are influencing the mechanical behaviour of the T-stub and its failure modes. For example, it is necessary to define the yield limit and the elastic modulus to be used in the calculation of the strength and the

stiffness of stainless steel T-stubs while continuing to guarantee a sufficient reliability. A finite element model is developed and validated on the basis of the experimental results for the two types of connections considered (cover plate and T-stub). The aim of this model is to obtain an efficient tool to be used in an extensive parametric study relating to stainless steel structural applications. 2. Design approach for stainless steel connections Design methods for stainless steel are often based on those for carbon steel with allowance made for the high ductility and deformability exhibited by stainless steel. At present, the use of stainless steel for structural applications is covered by European standard EN 1993-1-4 [1]. The part dedicated to connections concerns mainly cover plate types. This could be explained by the fact that the main applications of stainless steel are using thin elements. However, for beam-to-column connections, an approach based on the component method can be used. The few documents dealing with stainless steel connections concern mainly either welded or bolted cover plate connections of thin elements [19]. The other types of connections are based directly on the formulations established for carbon steel. All types of bolted connections used in carbon steel structures can be used with stainless steel. The design of carbon steel connections is usually evaluated at the ULS, while the deformation criteria at SLS are considered to be implicitly satisfied. This is justified by the linear behaviour of carbon steel up to its elastic limit and by the low ratio between ultimate and yield strengths, generally comprised between 1.1 and 1.5. In the case of stainless steel, the stressstrain curve is non-linear and the ratio between the ultimate strength and the yield limit may exceed 2. So, a design to ULS may not guarantee that no excessive deformations will occur at the SLS, since the ratio between the ULS and SLS loads is usually between 1.35 and 1.5. In ENV 1993-1-4 [20], a limitation was used in the net section resistance to limit its elongation at SLS. Another possible verification should concern the deformation of the hole in bearing, which is not easy to predict. EN 1993-1-4 considers that the ULS verification is sufficient if a reduced ultimate limit is used in bearing. The design formulae of connections concern mainly the bolts in tension and shear and the components of the cover plate connection (net section, gross section and bearing). For the moment resisting connections, the provisions of EN 1993-18 are applicable to the various components. In this paper, a preliminary study is presented concerning the T-stub, which is a key component in the tension zone of bolted connections. 2.1. Bolts in shear and tension (ULS) The shear strength of carbon steel bolts, for one shear plane, depends on the position of the shear plane with regard to the threading and the ductility of bolts. EN 1993-1-4 considers that stainless steel bolts behave like carbon steel ductile bolts (class: 4.6, 5.6 or 8.8). Thus, the coefficient used for shear is equal to 0.6 independently of the shear plane position (in the threaded part of the shank or not). The use of only one coefficient is realistic regarding the ductile character of stainless steel bolts and the experimental results which showed that all the values for three diameters are higher than 0.6 [16]. The ratio between the shear resistance in tension to that of direct tension gives mean values varying from 0.69 to 0.62 for three diameters from 12 mm to 20 mm. For shear in compression, the dispersion of the results is lower than that for shear in tension for the three diameters, and the mean resistances are higher. In practice, the shear test in tension, which gives lower mean values of resistance, is more conservative


A. Bouchar et al. / Journal of Constructional Steel Research 64 (2008) 12641274

Fig. 1. Moment resisting connection and basic cover plates with one or two shear planes.

and this is the most common configuration in practice. This tendency is comparable to that of carbon steel bolts [21]. For bolts in tension, the design formula is similar to that for carbon steels. The experimental values of resistance in direct tension are higher than the guaranteed values [22]. The resistance is based on the ultimate strength of the constitutive bolt material. However, bolts with a class of resistance of 50 have a high ratio between the ultimate and elastic limits (about 2.4). This could load them beyond the elastic limit under SLS [11]. 2.2. Cover plate bolted connections The cover plate connection can be a single connection or a part of a more complex connection (Fig. 1). The design approach for cover plate connections is based on that of EN 1993-1-8 for carbon steel connections [18]. A fundamental difference concerns bearing resistances. The gross section design is similar to that of carbon steel but the elastic limit used is the conventional one. Partial safety factors to be used in design formulae are M 0 = M 1 = 1.1 and M 2 = 1.25. The proposed formulae are defined considering various associated failures modes: net section, gross section and bearing. 2.2.1. Gross cross section resistance The resistance of the gross cross section is limited by the plastic resistance. This can be a limitation for stainless steel members in tension because the nominal value of the conventional elastic limit is relatively low. The use of the real characteristics can be a realistic and efficient solution. The gross cross section resistance should be determined using Eq. (1). Npl,Rd = Afy /M 0 . (1)

and eccentricity of the load transferred to the hole, in comparison with the case of the non-loaded hole in a plate in tension. So, it is also based on the connections with only a single shear plane for which the tension load transferred to the bolt affects the tension resistance of the net section, in particular for thin elements. The American standard [19] proposes a similar expression for the coefficient kr but makes a distinction between connections with one or two shear planes. Also, it proposes a limiting criterion on stress to limit the deformation at the net section. 2.2.3. Bearing resistance The bearing resistance is given by a similar formula as used for carbon steel (Eq. (3)). Fb,Rd = k1 b fu dt /M 2 (3)

where: b = min{e1 /(3do ); p1 /(3do )0.25; fub /fu ; 1.0} and in this study k1 = 2.5. According to ENV1993-1-4, this verification of resistance is not sufficient because the bearing resistance could be governed, for stainless steel, by the need to limit the hole elongation under serviceability loads. This elongation is not easy to calculate. However, the real ultimate tensile strength fu can be replaced by a reduced one fu,red given by the combination of the elastic and the ultimate limits as follows: fu,red = 0.5fy + 0.6fu . Thus, it can be considered that the hole elongation is limited and a separate serviceability check may be avoided. In EN 1993-1-4, only the verification using the reduced ultimate limit is proposed. 2.3. T-stub connections Bolted endplate beam-to-column connections (Fig. 2) are widely used in carbon steel structures. They are mainly loaded by bending moment and shear force. They show a complex mechanical behaviour, because the load is partially transferred through contact between the connected parts (bolts, plates and I or H rolled or welded sections). The component method proposed in EN 1993-1-8 consists of decomposing a bolted beam-to-column connection into its elementary components, the most important of which is the T-stub in the tension zone. These components are characterized in terms of resistance, stiffness and deformation capacity (ductility). The properties of these components are then assembled according to suitable rules so that the resulting model exhibits the equivalent characteristics of the whole connection and it is possible to determine which the most influential components are. A T-stub can be considered as an elementary connection on its own or as a component in the tension zone of a moment resisting connection (Fig. 2). The behaviour of elementary T-stubs was

2.2.2. Net cross section resistance The net cross section resistance in a plate with holes for bolts, in tension, is given by Eq. (2). Nu,Rd = kr Anet fu /M 2 (2) where: kr = (1 + 3r (d0 /u 0.3)), and : kr 1 r = ratio of the number of bolts at the cross section to the total number of bolts in the connection; u = 2e2 , with : u p2 ; Anet = the net cross-sectional area; d0 = the nominal diameter of the bolt hole; e2 = the edge distance from the centre of the bolt hole to the adjacent edge, in the direction perpendicular to the direction of the load transfer; and p2 = the spacing between the bolt holes, in the direction perpendicular to the direction of load transfer. In the case of carbon steel, the reduction factor kr is used only for thin elements [23]. It is supposed to cover stress concentrations

A. Bouchar et al. / Journal of Constructional Steel Research 64 (2008) 12641274


Fig. 2. Beam-to-column connection (A) and definition of the T-stub (B).

Fig. 3. T-stub failure modes according to Eurocode 3.

widely investigated using experimental, analytical or numerical approaches [24]. The application of the component method to stainless steel needs some adaptation and justification due to the non-linear character of the stressstrain curve and to the manufacturing possibilities of stainless steel. Until now, I and H sections are not so common, but tubes with bolted base plates or endplates are more usual. EN 1993-1-8 proposes formulae to cover the strength and the stiffness of a T-stub considering that its flanges behave as simply superposed plates. The equivalence between the T-stub and the tension zone in the connections is realized by the concept of effective length considered for each bolt row, which is based on the plastic mechanisms. This effective length is used for the calculation of strength and stiffness of the T-stub. The equivalent T-stub associated to an effective length can be used to model the resistance and the stiffness of the column flange in bending, the endplate in bending, the column and the beam web in tension. The resistance of a T-stub is calculated using a simple plastic analysis (yield analysis) in 2D. The real 3D yielding is taken into account through the effective length which is a notional length and does not necessarily correspond to the physical length of the basic connection component that it represent. For each T-stub, three possible failure modes are distinguished according to the value of the ratio of the flange resistance in bending and the bolt in tension (Fig. 3) [18,25]: The first failure mode, associated with the complete yielding of the flange by forming plastic hinges in the web-flange junction and near the bolts row. The second failure mode, corresponding to the appearance of one plastic hinge in the flange and to the failure of the bolts in tension. The third failure mode, characterized by the failure of the bolts in tension. The resistance of the T-stub is given by the weakest of the resistance values associated with the three failure modes mentioned above. Failure mode 1 is of a ductile nature because it involves the appearance of large plastic deformation in the flange in bending

in accordance with the fact that steel used in construction usually has a high ductility. For stainless steel, the conventional yield limit may be used but the effect of the ratio between the resistances of the plate in bending and the bolt in tension is to be considered regarding the different evolutions of strain hardening. There is no explicit formula for stainless steel T-stubs in EN 1993-1-8, which uses the rigid plastic model to calculate the resistance. However, it is stated in EN 1993-1-4 that provisions given in EN 1993-1-8 should be applied for stainless steel, except where modified or superseded by the special provisions of EN 19931-4. These provisions concern mainly the bearing strength and the bolts in shear. In this study, the formulae of carbon steel are used to evaluate their applicability to two stainless steel T-stub specimens. This study has to be extended to a wide variety of T-stub configurations using the finite element model calibrated, for instance, on the experimental results of carbon steel T-stubs. The stiffness of a T-stub is defined by considering a continuous beam on four supports; two of these supports are elastic supports representing the bolts bearing loads considered as applied in their axis [25]. The analytical model of EN 1993-1-8 does not directly consider the interaction between the T-stub and the bolts. The load in the bolt is taken as equal to the applied load, while it depends actually on the ratio between the flange and the bolt stiffness due to prying effects [18,24]. Prying forces are implicitly taken into account when determining the design tension resistance and stiffness of the T-stub. 3. Behaviour of cover plate connections Tests on cover plate connections were conducted at LGC laboratory and the results were compared to the predictions of EN 1993-1-4 [16]. The geometrical configurations of the connections tested were chosen so as to cover all failure modes foreseen for cover plate connections. They concern twelve different configurations for three diameters (12, 16 and 20 mm) and four geometrical arrangements of bolts (A2L, A2T, A3 and A4) (Fig. 4). In the comparison between EN 1993-1-4 formulae and experimental results, the ultimate strengths (bolts and plates) and the nominal dimensions given by the manufacturer were used. The thickness of the two lateral plates and central plate are 5 and 10 mm, respectively, for all the tests.


A. Bouchar et al. / Journal of Constructional Steel Research 64 (2008) 12641274

Table 1 Failure load (kN) using actual material characteristics (cover plate connection) Connection Test results Fu-test A2T12 A2T16 A2T20 179.4 341.9 444.5 Mode BS Be/BS Be Calculation, EC3 formulas, = 1, fu and fy : test BS 174.6 359.8 501.0 Be (fu ) 187.6 236.2 306.9 Be (fu_red ) 157.8 198.8 258.2 NS 359.0 428.5 555.8 GS 251.6 307.5 391.3 Fu-EC3 157.8 198.8 258.2 Mode Be Be Be 1.14 1.72 1.72 Fu-test/ Fu-EC3

Fig. 4. Cover plate connections with different bolts configurations.

An example of the global force-displacement curve is given in Fig. 5 for the connection A2T20. This connection exhibits a large ductility and it has a failure mode in bearing which is more complicated than analytical approaches can describe. The loaddisplacement results for the bearing failure mode are summarized in Table 1. All the other results (12 tests) are detailed in [16]. They show a significant difference between the ultimate test loads and the design values calculated according to EN 19931-4 rules with a partial safety factor taken equal to 1. The bolt shear resistance (BS) is dominant only for the 12 mm bolt. The bearing resistance is calculated using the measured value of the ultimate limit of the material. If the reduced value is used, the resistance values with failure in bearing (Be) have to be reduced by a coefficient equal to 0.84. Most of the failure modes predicted by calculation according to EN 1993-1-4 formulae are associated with the gross section mainly when the nominal characteristics are used. This limit is reached before the net section resistance because the ratio between the areas of the net and gross cross sections is close to 0.7 (0.670.69), while the ratio between the steel elastic limit and its ultimate limit is less than 0.5 [16]. The tests for which failure occurred in the net cross section show that the real values of the connection resistance are always

higher than those predicted by calculations. It was found that the ratio between the experimental and calculated resistances varied between 1.05 and 1.16 [13]. This fact is probably due to notch effects and it is confirmed by other experimental and numerical studies, using finite element models to analyse the elastic-plastic behaviour of a single plate with a hole, subjected to a tension load (bi-directional state of stress and Von Mises criterion). Test specimens were able to support high deformations, due to the high ductility of austenitic steel. Net section failure is predicted by EN 1993-1-4 using the kr factor, which is higher than 1 in all cases except for the configuration with 3 bolts in the section with one bolt. In fact, this section is not critical even if the coefficient kr is equal to 0.88. In our opinion, this factor is to be used for thin walled elements [23] when the connection presents one shear plane. In EN 1993-1-8 [18], the kr factor is not used. However, for single lap connections with only one bolt row, it is stated that the bolts should be provided with washers under both the head and the nut and the bearing resistance for each bolt should be limited to the value given by Eq. (4). For stainless steel no indication is given on the value of the ultimate tensile strength to be used (fu or fu,red ). Fb,Rd = 1.5fu dt /M 2 . (4)

This reduction of bearing strength is due to the fact that bolts tend to be in tension with a local bending of the plate. For double shear connections, no reduction is considered except to take into account the failure mode for a group of bolts where the bearing mode has to be compared to the shear failure of the bolts. Thus, the resistance of a group of fasteners may be determined as the sum of the bearing resistances Fb,Rd of the individual fasteners provided that the design shear resistance Fv,Rd of each individual fastener is greater or equal to the design bearing resistance Fb,Rd . Otherwise, the resistance of a group of fasteners should be determined by using the smallest resistance of the individual fasteners multiplied by the number of fasteners. This condition seems to be used to guarantee a minimum ductility and redistribution of load among the fasteners. In bearing, the failure mode is more influenced by the end distance mode (e1 /3d0 = 0.510.54) than by that of the higher spacing (p1 /3d0 0.25 = 0.770.82). An estimate is done for the elongation for each component of the connection (bearing, shear bolt, net section and gross section),

Fig. 5. Loaddisplacement curve and deformed shape of the connection A2T20.

A. Bouchar et al. / Journal of Constructional Steel Research 64 (2008) 12641274 Table 2 Mechanical characteristics of the austenitic steel plates Plate (mm) fy (0.2%) (MPa) Test 5 10 Mean 271 288 279.5 Nom. 200 200 fu (MPa) Test 577 581 579 Nom. 520 520 fu /fy Test 2.13 2.02 2.1 Nom. 2.60 2.6 A (%) Test 63 62 62.5


Nom. 45 45

based on the global displacement measured against a known length. This method allows an analytical estimation of the contribution of each component, in order to isolate the one corresponding to the bearing load [13,16]. The calculated bearing deformations at SLS remain relatively small (lower than 1.6 mm) with a bearing resistance based on the ultimate limit fu . It can be easily shown that, if the mean stress at the net section is about 1.5 times greater than the elastic limit, the net section elongation is still lower than 1 mm, for the three hole diameters used in the tests. The estimated bearing deformation at the hole is not very high and the net section elongation is even lower [16]. Actually, the length concerned with the net section elongation is small (the hole diameter as a maximum) and even if a plastic deformation occurs, its effect on the hole elongation always remains small except at failure, when necking takes place at the net section. Experimentally, in the case of failures in bearing, which presented the most complex behaviour, ultimate loads were at least 1.4 times greater than their calculated values, but failure occurred with large elongation of the holes. The main difficulty lies in making a realistic calculation of this elongation, which results not only from the bearing effects of the bolt but also from the elongation of the net section. Thus, the goal of the finite element model developed is to distinguish the different parts of deformations due to bearing or net section. 4. Numerical model A numerical model of a cover plate connection is developed to quantify the contribution of each component in order to show the critical aspects for the design. The model is characterized by the elastic plastic behaviour of the materials and geometrical nonlinearity due to contact and large displacements. As a first step, the application concerns the A2T type, which is influenced by a bearing failure mode for the three diameters. In the case of a T-stub, experimental results are only available for carbon steel and they are used to validate the numerical model. This validated model is used to analyse the influence of stainless steel on the behaviour of two types of T-stubs. For the material behaviour, the engineering stressstrain curve defined from the coupon tensile tests is used. Work is in progress to compare different types of stressstrain curves and mainly those considering the real values. Plastic yielding is governed by the Von Mises criterion with isotropic strain hardening. The results obtained from the numerical model are compared to the experimental results to check and to calibrate the approach before continuing with parametric studies. The 3D models are developed using the finite elements software Cast3m [26]. The numerical results fit well with the available experimental results. 4.1. Material characteristics For the cover plate connections, the materials characteristics of stainless steel are supplied by the manufacturers (Table 2). The tested steel (304L) of European designation (X2CrNi19-11 or 1.4306) is in the S220 resistance class, according to ENV 19931-4. It has a high ratio between the ultimate and elastic limits. A one parameter RambergOsgoods relation (Eq. (5)) is often

Table 3 Mechanical characteristics of the stainless steel bolts (MPa) Bolt M12 M16 M20 fy -nom 600 fy -test 692 748 686 fu -nom 800 fu -test 863 955 852

used by EN-1993-1-4 to represent the non-linear behaviour of stainless steel. It could be improved for the part above the elastic limit [27]. However, if the n parameter is well chosen the relation can describe accurately the stressstrain relation.


+ 0.002



This relation has three independent constants which define the shape of the stressstrain curve. Where 0.2 = elastic limit and n = a coefficient which gives the degree of the material strain hardening. The n value can be obtained experimentally or in EN 1993-1-4 [1,11]. In this study, this parameter is taken equal to 7.6 to obtain a stressstrain curve that fits well with the coupon test realized in the laboratory [16]. The materials characteristics for stainless steel used in the numerical model and for the analytical calculations are given in Table 2. The curves considered for the bolts are based on the experimental results [14,16] and their values are summarized in Table 3. The carbon steel characteristics from [24,28] are used for the validation of the T-stub model and as a basis for comparison. The stressstrain curves used in the numerical model for the plates and the bolts are summarized in Fig. 6. The stressstrain curves called T-carbon1 and T-carbon2 (Fig. 6), used for carbon steel to validate the numerical models, describe the short and the long T-stubs, respectively. The elastic and ultimate limits for T-carbon1 and T-carbon2 are (431/596) and (270/398) MPa, respectively. The stainless steel curve was determined from a coupon test. The curves for stainless steel bolts are used in both the cover plate connections and T-stub models. 4.2. Cover plate model Due to symmetry, the connection is decomposed into two main components C0 and C1, describing the behaviour of the cover plate and the main plate, respectively (Fig. 7). Each component represents half of the main or the cover plate with a hole and a bolt made of a perfect circular shape. Another component, common to the two basic components, is C2 representing the behaviour of the thread in the zone of contact between bolt and hole. This model is developed using the finite element software Cast3m [26]. It takes into account different sources of non-linearity such as the elastic plastic behaviour, the contact and the large displacements. The loaddisplacement curve of the connection is subsequently developed using the elementary force-displacement relationships for each component (Fig. 7). It is observed that component C2 plays a dominant role in the overall behaviour of the connection [29]. Each of the components C0 and C1 is modelled as a plate with a perfect circular bolt in contact with a hole. Their behaviour involves three sources of deformation: gross section, net section


A. Bouchar et al. / Journal of Constructional Steel Research 64 (2008) 12641274

Fig. 6. Stressstrain curves for carbon and austenitic steels (plates and bolts).

Fig. 9. Loaddisplacement curves for A2T connections (tests and FEM results). Fig. 7. Components of A2T connections Global model. Table 4 Dimensions of the specimens of cover plate connections (mm) Connection A2T12 A2T16 A2T20 Bolts M12 M16 M20 e1 22.5 27.5 35 d1 55 65 80 e2 22.5 27.5 35 p2 45 55 70 b 90 110 140

Fig. 8. Deformed shapes of C0, C1 and C2 components.

and bearing. In addition, the contact between the bolt and the hole is modelled at the local scale of the thread, considering a unit quarter slice of ISO thread profile (Fig. 8) generalised to the whole length of the bolt. The model is compared to the experimental results considering the global loaddisplacement curves on A2T connections using the three bolt diameters [14] (Fig. 9). The three bolt diameters considered correspond to the configurations tested for which the bearing failure was dominant. The dimensions of the connections are defined in Table 4. All the bolt holes have a nominal clearance of 2 mm. It can be shown that the numerical model closely represents the experimental loaddisplacement curves for A2T20 and A2T12

connections. In the case of A2T16, the experimental curve is stiffer than the numerical one. This can be explained by the fact that the 16 mm bolts are not threaded for a distance of at least 10 mm from the head, while the model considers that the whole bolt is threaded as is the case with the 12 and 20 mm bolts. The model is then used to analyse the influence of the material characteristics of plates with three different diameters. The material characteristics considered for stainless steel bolts and plates are those given by tests (Fig. 6). The same material characteristics are used for the 12 mm and 20 mm bolts because of their similarity (Table 2). Furthermore, two equivalent carbon steel plates are used to obtain a basis of comparison with stainless steel. They concern carbon steel with yield limit close to that of stainless steel with strain hardening (called carbon2 in Fig. 6) or with elastic perfectly plastic curve (EPP). The loaddisplacement curves of the three connections using three different material laws (stainless steel, carbon 2, elastic-perfectly plastic) are presented in Fig. 10. Except with the diameter 16 mm, the different models closely replicate the experimental initial stiffness. The difference between the models appears with the development of plasticity. The loaddisplacement curves of the model with elastic perfectly plastic (EPP) law, which represents the design approach, show the lowest resistance. Furthermore, the results of the model with EPP law are very close to those of the model with carbon steel and strain hardening due to the plastic plateau in its stressstrain curve. Stainless steel shows the highest values of resistance due to its continuous strain hardening. Thus, the use of an EPP law in the design of stainless steel connection is conservative in

A. Bouchar et al. / Journal of Constructional Steel Research 64 (2008) 12641274


Fig. 10. Loaddisplacement curves of A2T connections considering different stressstrain relationships.

Fig. 11. Influence of end distance e1 on the behaviour of C1 component and plastic deformations (A2T20).

comparison with carbon steel. The maximum displacement given by the numerical model, in comparison with the experimental values, is relatively limited due to numerical instabilities. Strain hardening capacity of stainless steel which can be represented by fu /fy ratio has a great effect on the global deformation of the connection. This effect can be evaluated by comparison to the EPP material, in order to define a simplified approach for the analysis of stainless steel connections taking into account part of the real capacity of the material. These preliminary results show that stainless steel exhibits a higher stiffness and plastic resistance compared to the other models due to its strain hardening. Further results are needed for other connection configurations where different failure modes are combined with bearing. For example, the comparison of the behaviour of each bolt is to be analysed depending on its position inside a typical connection (internal or end position) for one or two shear planes. Another parameter analysed with regard to the bearing behaviour is the effect of the end distance e1 on the load

displacement curve of the A2T20 connection. The preliminary results (Fig. 11) show that the model has more sensitivity to the reduction of the end distance. This aspect has to be developed further to define the shear deformation on the end distance and the real local compression under the bolt. It seems necessary to consider this aspect using the ultimate and the reduced limit of the material in comparison with carbon steel design. The numerical model can help to define optimal dimensions of the end distance and different bearing resistances depending on the position of each bolt hole in the connected plate. 4.3. T-stub model The T-stub, a main component in bolted connections under bending, is analyzed at first for carbon steel as a calibration of the finite elements modelling approach. Then, the T-stub is considered with stainless steel for two geometrical configurations. The aim is to make a comparison between stainless and carbon


A. Bouchar et al. / Journal of Constructional Steel Research 64 (2008) 12641274

Table 5 Characteristics of the T-stubs used in the modelling T-stub fy (MPa) Bolts M12-8.8, fy /fu = 893/974 M16-8.8, fy /fu = 957/1052 Dimensions (mm) P Tshort TLong 431 270 40 90 tf 10.7 10.7

90 100

L 80 210

b 150 150

e 30 25

e1 20 60

m 29.35 34.45

Table 6 Resistances of the T-stubs (analytical and numerical values) (kN) T-stub Fp (FEM) CS Tshort TLong 125.5 230 SS 153.5 276 EPP 125 230 Fu_EC3 ( = 1, fy and fu = real values) Mode1 86.9 195 Mode 2 168.4 308.7 Mode 3 291 599.7 F-min 86.9 195

Fig. 12. T-stubs considered for the numerical model (two rows of bolts).

Fig. 13. 3D model of the carbon steel T-stub (FEM validation).

steel for two T-stubs. These preliminary results have naturally to be completed by experimental tests on stainless steel with different configurations with and without stiffeners. The mechanical influence of stainless steel is thus analyzed considering the stiffness and the resistance. Stainless steel bolts are used with stainless steel and carbon steel T-stubs to obtain the same realistic basis of comparison. Before applying the model to stainless steel connections, it is validated on the experimental results of two carbon steel T-stubs. For the validation of the numerical model, two different and symmetrical T-stubs, with two rows of two bolts, are taken from the bibliography [24,28] (Table 5 and Fig. 12). These two T-stubs exhibit a failure mode 1, which arises by plastic hinges in the steel plate. The real material characteristics used in the model are called carbon1 for the short T-stub and carbon2 for the long T-stub with carbon steel bolts (Fig. 6 and Table 5). Three planes of symmetry characterize the T-stubs analysed which enables one eighth of the specimen to be modelled. Unilateral contact is considered under the flange and between the flange and the washer under the bolt head. The flexibility of the

bolt due to the contact between the threading and the nut can be taken into account by a fictitious length larger than the real one or by a diameter lower than the real one [25]. In this study, the bolt is considered threaded on a short part. Thus, a fictitious length is considered with the nominal diameter reduced to the tensile stress area on a short part of the bolt length, to take account of the real bolt strength. This reduction of strength is considered for the analytical calculation (Table 6) but not in the numerical model because the failure of the T-stub arises in mode 1 with a low load in comparison with modes 2 and 3. For all the cases, the washers are considered with the same dimensions as the bolt head. The results of the finite element analyses are based mainly on the global loaddisplacement curves. The comparison between the experimental and the numerical results for the two studied T-stubs show a good accuracy of the model (Fig. 13). The prying forces due to the deformability of the plate have an effect on the bolt load, which can influence the evolution of the limit between failure modes 1 and 2. This aspect has to be analysed in an extensive way for stainless steel because of the absence of a well-defined yield limit with plateau.

A. Bouchar et al. / Journal of Constructional Steel Research 64 (2008) 12641274


Fig. 14. Loaddisplacement curves for the short and long T-stubs (bolt: stainless steel, SS = stainless steel, CS = carbon steel, EPP = elastic perfectly plastic).

Fig. 15. Bolt load vs. the applied load for the short and long T-stubs (M12: short T-stub, M16 long T-stub).

The T-stub behaviour is analyzed on the basis of the load displacement curves and the evolution of the bolt load considering the prying force effect. The analysis shows that the resistance of the T-stub increases in a significant way with stainless steel, without changing the failure mode (Fig. 14), because of the effect of strain hardening. The vertical marks on each curve represent the ultimate deformation of the material (Fig. 6). It is to be considered that even with failure mode 2, the stainless steel T-stub exhibits a higher potential of ductility due not only to the flange in bending but also to the bolt in tension. This study is in its preliminary state and it has to be extended to include various geometrical configurations and experimental results. The evolution of the real load in the bolt, including prying effects, has to be known because it can influence the resistance of the T-stub and, in particular, the failure mode. The curves representing the bolt load evolution versus the load applied to the short T-stub (Ts) show that the load in the bolt is limited by the deformation of the flange. This load is lower for the Ts_EPP (elastic perfectly plastic) than for Ts_SS (real stainless steel). For the long Tstub (TL), the failure mode is mode 1 and the bolt load has a quasilinear evolution versus the applied load (Fig. 15). The plastic resistance of the connection (Fp ), given by the finite element model, is defined as the intersection between the two lines tangent to the loaddisplacement curve at the initial and the final parts of the curve. This plastic resistance is compared to the design values given by the EN 1993-1-8 formulae (Table 6). The analytical values of the T-stub resistance are calculated using EN 1993-1-8 with the real characteristics of materials and partial coefficient equal to 1. The failure mode 1 is predicted for the two T-stubs with a resistance value equal to 86.9 kN for the short T-stub and 195 kN for the long T-stub. Thus, as observed on Fig. 14, the reserve of resistance available is high, even with an elastic plastic stressstrain curve. This concern the two types of steel, although stainless steel exhibits a higher value due to strain hardening. The failure modes have to be evaluated further for stainless steel to

take advantage of its high ductility and strain hardening without reducing the available ductility by a transfer of the failure mode to bolts in tension. Considering the two T-stubs analysed in this study, the rigid plastic method proposed for carbon steel can be used for stainless steel. One of the aspects to be developed in future studies is to define the elastic limit to be used for stainless steel that will lead to the analytical method having sufficient accuracy and reliability. 5. Conclusion The experimental and numerical results showed that the failure modes of the cover plates in bearing or net cross section occur with a large deformation. For the resistance calculation of the net cross section, the use of the kr coefficient is debatable for the thick plates in structural applications. This is principally true for connections in double shear where the symmetry allows a good distribution of loads. A study based on numerical and experimental results is in progress to analyse the differences between single and double shear connections and the opportunity to make a distinction between these two types of connections. Despite interesting results for stainless steel bolts in tension and shear, the bolts shear failure mode has to be avoided, as for carbon steel, to guarantee the ductility of actual connections. The failure modes in gross cross section, bearing or net cross section are to be preferred. For these failure modes it is possible to obtain relatively large deformation before failure. This effect will have some advantages because loads will always be able to be redistributed in a suitable way amongst the bolts. The finite element model showed that it could closely represent the experimental results for cover plate connections. It showed the influence of the end distance on the resistance of a cover plate. This study has to be extended to a wide variety of geometrical and material configurations of bearing type connections. One of the


A. Bouchar et al. / Journal of Constructional Steel Research 64 (2008) 12641274 [9] Van Den Berg GJ. The effect of the non-linear stressstrain behaviour of stainless steels on member capacity. Journal of Constructional Steel Research 2000;54(1):13560. [10] Gardner L, Nethercot DA. Structural stainless steel design: A new approach. The Structural Engineer 2004;November:218. [11] Euro Inox. Design manual for structural stainless steel. 3rd ed. 2006. [12] Gardner L. The use of stainless steel in structures. Progress in Structural Engineering Materials 2005;7:4555. [13] Bouchar A, Baptista A. Strength and deformation of stainless steel bolted joints with reference to Eurocode 3. 3rd Eurosteel conference, vol. 2. 2002. p. 87988. [14] Ryan I. Bolted connections (WP4.2), ECSC Project - 7210-SA/327, Final Report for partner Ugine-CTICM; 2000. [15] SCI. 2000. Development of the use of stainless steel in construction, Contract 7210-SA/842, 903, 904, 327, 134, 425, Final Report to ECSC. [16] Bouchar A. Resistance and ductility of stainless steel bolted connections, COST-C12, Improvement of buildings structural quality by new technologies, final conference, 20-22/01/2005, Edition Shaur et al., p. 31121. [17] Scot JG, Trautner JJ. Behavior of stainless steel fasteners under cyclic load. Journal of Constructional Steel Research 1998;46(13):4602. full paper No. 245. [18] EN 19931-8. Eurocode 3 Design of steel structures, part 1.8, Design of joints. CEN; 2005. [19] ASCE. Specification for the design of cold-formed stainless steel structural members. ASCE Standards; 1990. 114 pages. [20] ENV 19931-4. Eurocode 3 Design of steel structures, part 1.4, General rules and supplementary rules for stainless steels. CEN; 1996. [21] Kulak GL, Fisher JW, Struik JHA. Guide to design criteria for bolted and riveted joints. 2nd ed. AISC/RCSC; 2001. 352 pages. [22] Bouchar A, Ryan I, Kaitila O, Muzeau JP. Some aspects of the analysis of bearing-type stainless steel bolted joints. In: 9th nordic steel const. conf. 2001. p. 75562. [23] EN 19931-3. Eurocode 3 Design of steel structures, part 1.3, General rules and supplementary rules for cold formed members and sheeting. CEN; 2006. [24] Alkhatab Z. Analysis of the behaviour of steel joints strengthened by backingplates, numerical approach and experimental calibration. Ph.D. thesis. Blaise Pascal University, Clermont 2 (France), 2003 [in French]. [25] Alkhatab Z, Bouchar A. Analysis of T-stub strengthened by backing-plates with regard to Eurocode 3. Journal of Constructional Steel Research 2007;63(12): 160315. [26] Pasquet P. User Manual- finite element software package Cast3m. CEA, Saclay, 1998 [in French, English information on web site]. [27] Rasmussen KJR. Full-range stressstrain curves for stainless steel alloys. Journal of Constructional Steel Research 2003;59(1):4761. [28] Bursi OS, Jaspart JP. Calibration of a finite element model for isolated bolted end-plate steel connections. Journal of Constructional Steel Research 1997; 44(3):22562. [29] Bouchar A, Averseng J, Ryan I. Analysis of bearing-type stainless steel bolted joints. In: Zingoni A, editor. Proc 3rd int conf on structural engng, mechanics and computation. Rotterdam: Millpress; 2007. p. 4567. Full paper on CD-Rom.

goals of the extensive study is to evaluate the behaviour of the bolts in bearing depending on their position in the connected plate. The numerical model of the T-stub confirms a ductile behaviour of stainless steel with a higher resistance than carbon steel due to strain hardening. It shows the applicability of the rigid plastic approach of EN 1993-1-8 to the studied T-stubs. However, this aspect also has to be analysed in a more extensive way to confirm in a precise manner the applicability to stainless steel T-stubs of the formulae dedicated originally to the carbon steel. This concerns the resistance and the stiffness of the T-stubs and it can subsequently be extended to the design of beam-to-column connections. The numerical model has to be applied to various configurations of stainless steel T-stubs to confirm the preliminary results. It will be used to analyse the behaviour of beam-to-column connections with regard to the effective length of T-stubs which can be different from that of carbon steel. With such a model, it is possible to identify the zones of high deformations in order to define the criteria which govern the real behaviour of stainless steel connections taking account of the deformation capacity of the material. In the global analysis of connections, it is necessary to develop a coherent approach for their different basic components and to continue to be inspired by the approach for carbon steel. References
[1] EN 19931-4. Eurocode 3 Design of steel structures, part 1.4, General rules and supplementary rules for stainless steels. CEN; 2006. [2] Mkelinen P, Outinen J. Mechanical properties of an austenitic stainless steel at elevated temperatures. Journal of Constructional Steel Research 1998; 46(13): paper No. 101 on CD-Rom. [3] Chen J, Young B. Stressstrain curves for stainless steel at elevated temperatures. Engineering Structures 2006;28(2):22939. [4] Gardner L, Baddoo NR. Fire testing and design of stainless steel structures. Journal of Constructional Steel Research 2006;62(6):53243. [5] Gardner L, Ng KT. Temperature development in structural stainless steel section exposed to fire. Fire Safety Journal 2006;41:185203. [6] Gardner L. Stainless steel structures in fire. In: Proceedings of the Institution of Civil Engineers. Structures and Buildings 2007;160(3):12938. [7] Burgan BA, Baddoo NR, Gilsenan KA. Structural design of stainless steel members - comparison between Eurocode 3, Part 1.4 and test results. Journal of Constructional Steel Research 2000;54(1):5173. [8] Kouhi J, Talja A, Salmi P, Ala-Outinen T. Current R&D work on the use of stainless steel in construction in Finland. Journal of Constructional Steel Research 2000;54(1):3150.