Journal of Constructional Steel Research 66 (2010) 10001010

Contents lists available at ScienceDirect

Journal of Constructional Steel Research

journal homepage: www.elsevier.com/locate/jcsr

High strength steel tension splices with one or two bolts

Primo Moe , Darko Beg
Faculty of Civil and Geodetic Engineering, University of Ljubljana, Jamova 2, 1000 Ljubljana, Slovenia

article

info

abstract
The paper presents experimental research of 38 tension splices with bolts in double shear. Steel grade S690 was used. The objective of the research was to determine ductility and resistance of such connections. The experimental resistance is compared to Eurocode provisions and a new bearing resistance formula is developed. The reliability assessment is given according to EN 1990, Annex D. 2010 Elsevier Ltd. All rights reserved.

Article history: Received 23 December 2009 Accepted 15 March 2010 Keywords: High strength steel Bolted connections Ductility Resistance Bolt bearing

1. Introduction In constructional practice across Europe, structural high strength steels (HSSs) are considered as steels with specified minimum yield strength equal to or higher than 420 MPa. Such steels can reach a yield stress up to 1300 MPa. For successful application of higher strength steels, design and fabrication standards need to be accepted by the authorities responsible for safety and by industry. Eurocode 3, standard for the design of steel structures, is divided into several parts. Part EN 1993-1-1 [1] includes steel grades up to S460. Additional rules for steel grades up to S700 are presented in EN 1993-1-12 [2] and were the lattermost addition to Eurocode 3. HSSs are as a structural material competitive, mostly due to lower overall economy and structural efficiency. Nevertheless, HSSs undoubtedly have lower ductility than mild steels in terms of standard measures of ductility, such as ultimate-to-yield strength ratio, uniform strain and elongation at fracture. The values for high strength steels S690 are typically much lower than those for mild steels: ultimate-to-yield strength ratio fu /fy = 1.05, maximum uniform strain u = 0.05 and elongation after fracture Ac = 15%. Another measure of ductility is reduction of area at fracture Z , which is obtained by comparing the cross-sectional area after fracture. For high strength steel this measure indicates that the fractures can be characterized as ductile with large inelastic deformations. Local ductility in terms of plastic deformations is especially important in tension splices with bolts in shear, where contacts between bolts and plates are in general not established simultaneously for all bolts. For that reason, the local ductility of the connection has to be sufficient in order to ensure such

Corresponding author. Tel.: +386 1 4768625; fax: +386 1 4768629. E-mail address: primoz.moze@fgg.uni-lj.si (P. Moe).

bolthole elongation, that provides the transfer of loading through all bolts. Otherwise, the stress concentration would cause rupture of the steel plate or shear fracture of the bolt. Such inelastic behaviour is hidden in most of the nominally elastic resistance checks of steel structures, which is why sufficient local ductility has to be ensured. Tension splices are often referred to as shear connections. This term is adopted in what follows. In previous researches on bolted shear connections made of higher steel grades, one- and two-bolt connections were studied. Kim and Yura [3] investigated shear connections with one or two bolts placed parallel to the loading. Among others they used steel with nominal yield strength fy = 430 N/mm2 with ratio fu /fy = 1.13. The experimental resistance was compared to bearing resistance according to American AISC standard and to Eurocode standard; the latter was found to be conservative. Their work was continued by Aalberg and Larsen [4,5]. They duplicated the tests by Kim and Yura, using steel grades S690 and S1100. The value of ultimate tensile-to-yield ratio was equal to fu /fy = 1.05 for both steel grades. The local ductility of connections was not decreased due to the low fu /fy ratio. Puthli and Fleisher [6] focused on the shear connections made of steel grade S460 (fu /fy = 1.23) with two bolts placed perpendicular to loading. They compared experimental resistances to resistance according to EN 1993-18 [7]. The focus was set on minimum end and edge distances and the result was the suggestion for the reduction of minimum distances and a modified bearing resistance formula. More recent research was published by Rex and Easterling [8]. The research on the behaviour of a bolt bearing on a single plate was part of a larger investigation of the behaviour of partially restrained steel and composite connections. A 6.5 mm thick plate of different high strength steel grades (ultimate strength from 665 to 752 MPa) was tested on bolt bearing. Due to small plate thickness and large end distance e1 , several curling failures were observed.

0143-974X/$ see front matter 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.jcsr.2010.03.009

P. Moe, D. Beg / Journal of Constructional Steel Research 66 (2010) 10001010

1001

Statistical evaluation and reliability assessment of proposed design resistances were not carried out by any of the authors. Moe, Beg and Lopati [9] previously studied net cross-section failures of members in tension made of steel S690. The study included 43 tests. The result was design net cross-section resistance of symmetrically and unsymmetrically connected members in tension, which was determined by statistical evaluation. All failures showed local ductility by necking of net cross-section or by bolt hole elongation. Even a small reduction in gross section (e.g. 10%) did not result in its yielding and consequently to large plastic deformations of the whole member; thus global ductility was limited. According to these results the proposed design net cross-section resistance in EN 1993-1-12 was changed to be equal to that in EN 1993-1-1: Ft ,Rd = 0.9Anet fu

Fig. 1. Definition of end e1 , edge e2 distance and pitches p1 , p2 .

M2

(1)

In EN 1993-1-8 the design bearing resistance per bolt is defined as: Ft ,Rd = k1 b fu dt

M2

(2)
Fig. 2. Specimen type B1.

where fu is nominal ultimate tensile strength of the plate, d is bolt diameter, t is plate thickness and partial factor M2 with the recommended value of 1.25. Parameters b and k1 are defined as follows and take into account mainly geometrical parameters:

in the direction of load transfer b = min d ; d = d =

e1 3d0 p1 3d0 fub fu

;1

(3) (4) (5)

for end bolts

1 4

for inner bolts

perpendicular to the direction of load transfer

k1 = min 2.8 e2 d0 p2 d0

1.7; 1.4

p2 d0

1.7; 2.5
(6)

for edge bolts k1 = min 1.4

1.7; 2.5

for inner bolts.

(7)

In the previous definitions fub is nominal ultimate tensile strength of the bolt and d0 is a bolt hole diameter. End distance e1 , edge distance e2 and pitches p1 and p2 are defined in Fig. 1. Maximum and minimum end and edge distances and pitches are also limited in EN 1993-1-8. Maximum dimensions are related to the exposure of steel to corrosion and local buckling of plate in compression. Minimum dimensions set lower limit for the resistance and are indirectly related also to execution tolerances. Minimum distances are: end distance e1 1.2d0 , edge distance e2 1.2d0 , pitch p1 2.2d0 and pitch p2 2.4d0 . Standard EN 1993-1-12 gives no additions to any of these rules. American standard ANS/AISC 360-05 [10] defines bearing resistance separately in serviceability limit state (SLS) and in ultimate limit state (ULS). In SLS the bearing strength is defined by hole elongation equal to 6.35 mm (1/4 in.) and in ULS an upper bound anticipates hole elongation at maximum bearing strength. As opposed to that other hand, Eurocode does not prescribe the limit for hole elongation. In Background documentation to Eurocode 3 [11] it is stated that it is necessary to give a limit for mean bearing stress to control deformation. This limit stress was originally set to the average bearing stress of 3fy and not to a specified deformation.

For group of bolts the following statement is given in 3.7(1), EN 1993-1-8: The design resistance of a group of fasteners may be taken as the sum of the design bearing resistances Fb,Rd of the individual fasteners provided that the design shear resistance Fv,Rd of each individual fastener is greater than or equal to the design bearing resistance Fb,Rd . Otherwise the design resistance of a group of fasteners should be taken as the number of fasteners multiplied by the smallest design resistance of any of the individual fasteners. As stated in [12]: This statement is meant to persuade the designer to choose a balanced bolt pattern and to avoid having a relatively small end distance in combination with a relatively large pitch. A wrong design may lead to premature failure of the end bolts before the inner bolts reach their capacities. The capacity of the group of bolts will be overestimated in such cases. In order to correctly predict a balanced bolt pattern and consequently the desired ductility and failure, the bearing resistance formula should accurately describe the phenomena and should also be supported by experimental results. In this paper a total of 38 tests on connections with bolts in double shear are discussed. Quenched and tempered steel S690 QL was used. The aim of this investigation was to study the effect of connection geometry on strength and ductility and to revaluate bearing resistance according to EN 1993-1-8. 2. Experimental program The specimens were fabricated from a single steel plate, grade S690 QL (nominal material strengths fy = 690 N/mm2 , fu = 770 N/mm2 ). The nominal plate thickness was t = 10 mm. The actual material strengths of the plate were determined by standard tensile tests according to EN 10002-1 [13]. The test pieces were prepared according to EN ISO 377 [14]. The average obtained yield and tensile strengths were fy,act = 847 N/mm2 and fu,act = 885 N/mm2 , respectively. The average elongation after fracture was A = 14.4% and the average percentage reduction of area Z = 58.4%. Tension splices with one or two bolts in double shear were studied. The B1 test series included 25 single bolt shear connections (Fig. 2). Bolts M27 an M22 were used in drilled standard size holes d0 = 30 mm and 24 mm (see Table 1). Some of the holes were positioned eccentrically to loading axis. The B2 test series included 13

1002 Table 1 Specimens geometry. Specimen name Bolt

P. Moe, D. Beg / Journal of Constructional Steel Research 66 (2010) 10001010

Nominal dimensions b (mm) t (mm) 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 d0 (mm) 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 24 24 24 24 24 24 24 24 24 24 24 24 24 24 24 24 e2,min /d0 e1,min /d0 p2 /d0 1.00 1.20 1.20 1.20 1.20 1.35 1.35 1.18 1.50 1.50 1.50 1.50 1.50 1.50 1.33 1.50 1.50 2.00 2.00 2.00 2.00 2.00 1.67 1.42 1.17 1.00 1.20 1.20 1.20 1.20 1.50 1.50 1.50 1.50 1.50 1.50 1.50 1.35 3.00 1.20 1.50 2.00 3.00 2.50 3.00 3.35 1.00 1.20 1.50 2.00 2.50 3.00 3.00 1.50 1.50 1.50 2.00 2.50 3.00 3.50 4.17 4.17 4.17 3.00 1.20 2.00 3.00 3.00 1.50 3.00 1.00 1.20 1.50 2.00 3.00 2.00 Anet (mm2 ) 300 420 420 420 420 510 510 410 600 600 600 600 600 600 500 600 600 900 900 900 900 900 560 440 320 480 672 672 672 744 816 888 960 960 960 960 960 744

Actual dimensions b (mm) 61.0 73.2 71.2 71.7 72.0 81.0 80.5 81.1 90.0 92.0 90.0 90.0 90.3 90.0 90.3 87.4 89.2 118.0 118.2 119.4 122.1 118.8 78.7 79.1 79.1 97.2 115.9 115.9 116.0 124.1 130.3 137.0 144.0 144.1 144.0 143.7 144.4 121.0 t (mm) 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10.2 10.15 10.15 10.15 10.15 10.15 10.15 10.15 10.15 10.15 10.15 10.15 10.15 10.15 10.15 d0 (mm) 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 24 24 24 24 24 24 24 24 24 24 24 24 24 24 24 24 e2,min /d0 e1,min /d0 p2 /d0 0.95 1.20 1.15 1.19 1.16 1.33 1.32 1.17 1.47 1.53 1.48 1.47 1.46 1.46 1.30 1.42 1.48 1.90 1.93 1.96 2.02 1.95 1.61 1.44 1.18 0.96 1.20 1.20 1.20 1.21 1.50 1.48 1.49 1.50 1.50 1.44 1.47 1.33 3.01 1.21 1.50 2.01 2.98 2.52 3.02 3.34 1.00 1.21 1.51 2.01 2.49 3.00 3.00 1.50 1.50 1.53 2.06 2.56 3.06 3.54 4.22 4.21 4.21 2.97 1.20 2.01 2.99 2.96 1.50 3.03 1.03 1.21 1.53 2.05 3.04 2.01 Anet (mm2 ) 310 432 412 417 420 510 505 402 600 620 600 600 603 600 480 574 592 880 882 894 921 888 547 459 333 499 689 689 690 772 835 903 974 975 975 971 978 741

B101 B102 B103 B104 B105 B106 B107 B108a B109 B110 B111 B112 B113 B114 B115a B116 B117 B118 B119 B120 B121 B122 B123 B124a B125a B201 B202 B203 B204 B205 B206 B207 B208 B209 B210 B211 B212 B213
a

M27 M27 M27 M27 M27 M27 M27 M27 M27 M27 M27 M27 M27 M27 M27 M27 M27 M27 M27 M27 M27 M27 M22 M22 M22 M22 M22 M22 M22 M22 M22 M22 M22 M22 M22 M22 M22 M22

60.0 72.0 72.0 72.0 72.0 81.0 81.0 81.0 90.0 90.0 90.0 90.0 90.0 90.0 90.0 90.0 90.0 120.0 120.0 120.0 120.0 120.0 80.0 80.0 80.0 96.0 115.2 115.2 115.2 122.4 129.6 136.8 144.0 144.0 144.0 144.0 144.0 122.4

2.00 2.40 2.40 2.40 2.70 2.40 2.70 3.00 3.00 3.00 3.00 3.00 2.40

2.03 2.36 2.37 2.36 2.70 2.38 2.70 3.00 3.00 2.98 3.00 3.01 2.39

Eccentric hole.

Fig. 3. Specimen type B2.

two-bolt shear connections, with bolts positioned perpendicular to loading direction (Fig. 3). Bolts M22 in standard size holes d0 = 24 mm were used. Bolts were designed to withstand resistance of the steel plate, thus grades 10.9 and 12.9 were selected to avoid bolt shear rupture. Geometry of specimens was selected so that for each of the selected edge distance e2 the end distance e1 was varied. The distance between bolts was also varied for B2 specimens. The actual and the nominal geometry of specimens are listed in Table 1. The tests were carried out on the 1000 kN testing machine at a prescribed displacement rate (about 1.5 mm/min). Special forks were made to connect the specimens B1 and B2 (see Figs. 2 and 3). These forks were exposed only to elastic deformation. A relative displacement between the specimen and the fork was measured by two inductive displacement transducers (IDT), one on each side of the specimen. An extensometer was also mounted between the

Fig. 4. Test setupIDT on the both sides, extensometers in the middle.

fork and the bolt to control deformation of fork and bending of the bolt (see Fig. 4). Bolts were snug tightened to ensure that the load was transferred primarily by bearing and not by friction. All tests had been carried out until significant drop in resistance occurred. 3. Test results and observations Several failure modes were observed. Plate shear failure due to bolt bearing occurred when end distance e1 was small enough compared to edge distance e2 (Fig. 5(a), (b)). This type of failure

P. Moe, D. Beg / Journal of Constructional Steel Research 66 (2010) 10001010

1003

(a) B109-shear failure.

(b) B202-shear failure. Fig. 5. Shear and splitting failures.

(c) B112-shear failure.

Fig. 6. Forcedisplacement curves for two groups of specimens with the same width.

Fig. 7. Maximum principal (left) and shear (right) stresses at hole elongation of B112. Darker colour presents higher stress.

only occurred if the specimens were wide enough to avoid net section failure. Bolt bearing pushed out the steel in front of the bolt until the fracture started. The failures were very ductile and the load displacement curves were characterized by a long yield plateau (exp. as curve B109 in Fig. 6). The fractures were instantaneous after the development of excessive local plastic deformations of the specimen. If end distance e1 was increased (at a constant edge distance e2 ), specimens failed as shown in Fig. 5(c). Numerical simulation revealed that maximal principal stresses shown in Fig. 7 formed an arc of high tensile stresses. The edges of the plate were subjected to the highest tensile stresses, which resulted in the appearance of a fracture on the edge

perpendicular to loading direction. It progressed to the area with the highest shear stress (see Fig. 7) in a curved pattern as shown in Fig. 5(c). These kinds of failures, referred to as splitting failures, were characterized by higher resistance and larger displacement at maximum resistance than plate shear failures (see Table 2). Splitting failures were also observed and reported by Kim and Yura [3]. At even larger end distances e1 (narrow specimens) net cross-section failure prevailed (see Fig. 8(a), (b)). The fracture formed after necking of the net area and after large bolthole elongation. The net cross-section resistance was also the maximum resistance possible for specimens of equal widths. At certain end distance e1 a mixed failure was observed, where plate failure due

1004

P. Moe, D. Beg / Journal of Constructional Steel Research 66 (2010) 10001010

(a) B114-net cross-section failure.

(b) B204-net cross-section failure.

(c) B121-mixed failure.

Fig. 8. Net cross-section failure and mixed failure.

Table 2 Experimental and theoretical resistances. Specimen name B101 B102 B103 B104 B105 B106 B107 B108b B109 B110 B111 B112 B113 B114 B115b B116 B117 B118 B119 B120 B121 B122 B123 B124b B125b B201 B202 B203 B204 B205 B206 B207 B208 B209 B210 B211 B212 B213 Failure modea N SP SP N N N N N SH SH (SP) SP SP N N N SP SP SH SP SP SP (N) N N N N N SH N N N SH (SP) N SH SH SH (SP) SP (N) N N DU (mm) 2.4 5.1 6.1 3.5 3.5 5.8 5.6 4.0 5.2 5.8 6.4 8.9 8.5 9.1 6.2 5.8 6.6 9.8 12.0 19.5 24.8 24.3 15.9 10.0 5.6 2.4 5.8 4.3 3.9 6.2 6.7 10.4 3.9 4.9 5.6 10.2 12.6 5.8 Fmax (kN) 262 273 342 360 355 445 440 370 228 286 363 483 516 510 435 371 362 392 530 629 763 788 483 400 322 457 471 643 638 689 596 789 398 491 603 776 851 678 F (d0 /6) (kN) DNR 273 341 347 348 443 439 367 228 285 361 468 490 485 433 371 360 381 484 479 510 501 372 361 315 427 463 643 638 666 590 719 398 489 598 718 779 663 k1 0.96 1.65 1.51 1.64 1.55 2.01 1.99 1.58 2.41 2.50 2.44 2.43 2.40 2.38 1.94 2.29 2.44 2.50 2.50 2.50 2.50 2.50 2.50 2.33 1.61 1.00 1.61 1.61 1.60 1.68 1.63 2.09 2.48 2.49 2.48 2.34 2.42 1.64

b
1.00 0.40 0.50 0.67 0.99 0.84 1.00 1.00 0.33 0.40 0.50 0.67 0.83 1.00 1.00 0.50 0.50 0.51 0.69 0.85 1.00 1.00 1.00 1.00 1.00 0.99 0.40 0.67 1.00 0.99 0.50 1.00 0.34 0.40 0.51 0.68 1.00 0.67

Fb,EC (kN) 229 159 180 263 367 404 475 377 192 242 293 388 475 568 464 273 292 305 410 510 597 597 487 462 319 389 265 436 659 656 494 965 334 399 502 632 956 533

k1 1.90 1.31 1.70 1.90 1.90 1.90 1.90 1.90 0.89 1.03 1.32 1.77 1.90 1.90 1.90 1.37 1.32 1.05 1.39 1.70 1.90 1.90 1.90 1.90 1.90 1.90 1.31 1.90 1.90 1.90 1.45 1.90 0.89 1.05 1.33 1.82 1.90 1.90

k2 0.61 0.83 0.78 0.82 0.79 0.94 0.94 0.80 1.07 1.13 1.08 1.08 1.07 1.06 0.92 1.03 1.08 1.46 1.49 1.52 1.57 1.51 1.20 1.04 0.82 0.64 0.82 0.82 0.82 0.90 0.96 1.02 1.10 1.10 1.10 1.07 1.09 0.88

Fb,new (kN) 275 259 317 374 360 429 424 365 227 278 342 455 484 482 418 338 341 365 493 614 712 685 443 394 306 241 214 308 309 338 275 385 193 228 287 385 409 332

Fb
1.10 1.14 1.43 1.50 1.49 1.86 1.84 1.55 0.96 1.20 1.52 2.02 2.16 2.14 1.82 1.55 1.52 1.64 2.22 2.63 3.20 3.30 2.48 2.01 1.63 2.31 2.38 3.25 3.23 3.49 3.02 3.99 2.01 2.48 3.05 3.93 4.31 3.43

Fb,EC
0.96 0.66 0.75 1.10 1.54 1.69 1.99 1.58 0.80 1.01 1.23 1.62 1.99 2.38 1.94 1.14 1.22 1.28 1.72 2.13 2.50 2.50 2.50 2.33 1.61 0.99 0.65 1.08 1.60 1.66 0.82 2.09 0.85 1.01 1.26 1.60 2.42 1.10

Fb,neq
1.15 1.08 1.33 1.57 1.51 1.79 1.78 1.53 0.95 1.16 1.43 1.91 2.03 2.02 1.75 1.41 1.43 1.53 2.06 2.57 2.98 2.87 2.28 1.98 1.55 0.79 0.99 1.13 1.14 1.28 1.31 1.52 0.93 1.10 1.39 1.61 1.64 1.25

Fmax : maximum resistance. DU : displacement at maximum resistance Fmax . DNR: Displacement Not Reached. a Nnet section failure, SPsplitting failure, SHshear failure. b Eccentric hole.

to bolt bearing occurred simultaneously with net cross-section failure (see Fig. 8(c)). Excessive plastic deformations of net section and plate in bearing were typical for this kind of failure. All

types of failures were characterized by severe plastic deformations, especially in front of the bolt where steel literally flowed into a hole between the bolt and the bolthole in the adjacent plates. The

P. Moe, D. Beg / Journal of Constructional Steel Research 66 (2010) 10001010

1005

Fig. 9. Experimental results in relation to Eurocode bearing resistance function.

specimens tended to deform in thickness direction. This caused pressure on adjacent plates and therefore introduced tension into the bolts and obstructed the out of plane deformations. Fig. 6 illustrates load displacement curves for three groups of specimens. The first group includes B109 to B114, the second group includes B118 to B122 and the third group includes specimens B208 to B212. The widths of specimens in the first and second groups were 90 mm (e2 = 1.5d0 ) and 120 mm (e2 = 2d0 ), respectively. End distance e1 was the varying parameter. B113, B114 (Fig. 8), B122 and B212 showed net cross-section failure after hole elongation. The fractures were characterized as ductile failures (necking of net cross-section, reduction of thickness). B109, B118, B208 and B209 failed in plate shear, where the displacements at failure were approximately equal to clear end distance (e1 0.5d0 ). All the remaining specimens in Fig. 6 failed in splitting failure mode. The load displacement curves of all the remaining specimens were similar to those shown in Fig. 6. 4. Comparison of experimental results with Eurocode bearing resistance Eurocode design bearing resistance per bolt (2) was evaluated on the basis of 167 tests of plates in bearing [11,15]. Only one test result with more than one bolt was included. 137 tests results were related to steel grade S235 and 30 results to higher grades with StE690 as the highest one (equivalent to S690). The resistance model studied in the Background documentation to EC 3 [11] did not account for the reduction of resistance for edge distances between 1.2d0 e2 1.5d0 , nor for the reduction due to smaller pitch p2 3d0 . This reduction is considered in the k1 factor (Eq. (6)). However, there were a few test results for edge distances e2 1.5d0 and all these results were on the safe side, even without the reduction of bearing capacity. Moreover, there were no test results where fub /fu would dominate in Eq. (3). In order to compare our experimental results to Eurocode, the bearing resistance formula (2) is used without partial factor M2 and with actual material and geometric parameters (all other parameters are defined with formula (2)): Fb,EC = k1 b fu dt . (8)

Numerical presentation of experimental results and Eurocode bearing resistance is shown in Table 2. Normalized bearing resistance per bolt is defined as:

Fb,EC = Fb =

Fb,EC fu dt

= k1 b in terms of Eurocode formula

(9) (10)

Fu fu dt

in terms of experimental.

Fig. 9 presents normalized bearing resistances versus normalized end distance e1 /d0 . Test results (presented as marks) are grouped according to similar normalized edge distance e2 /d0 . In thin lines the normalized resistance according to Eurocode (Eq. (9)) for different normalized edge distances e2 /d0 is presented. The lower and the upper limits of Eurocode bearing resistance are defined by minimum distances and by maximum values of k1 and b factors, respectively. Considering formula (8), the resistance linearly increases by increasing the end distance e1 until it is equal to 3d0 . From that point on, the resistance is constant. This limitation implicitly includes net cross-section resistance check and limits the hole elongation. Test results in Fig. 9 show similarity to Eurocode formula only in terms of the shape of the curves. Bearing failures of the plate with hole elongation are typical for linearly increasing part and net cross-section failures for the plateau of the curve. Failures of the specimens type B2 indicated that the pitch p2 did not have any significant effect on the bearing resistance. Therefore, the results of specimens B2 with two bolts could also be plotted in Fig. 9, but only half of the experimental resistance of B2 specimens is considered. The transition between linear and constant part of the curve (transition from bearing to net cross-section failure) is not correctly accounted for, in Eurocodes bearing resistance formula (8). It seems that the transition is not only a function of normalized end distance e1 /d0 . Our experimental results show that this transition occurred between 1.41.5e1 /e2 . This is confirmed in Fig. 10, where experimental results are plotted in relation to endto-edge distance ratio e1 /e2 . Since the Eurocode bearing resistance (2) is set to limit the hole elongation and prevent tearing out of the bolt, it is reasonable to show the experimental resistances at a certain hole elongation.

1006

P. Moe, D. Beg / Journal of Constructional Steel Research 66 (2010) 10001010

Fig. 10. Experimental results (max. resistances) in relation to the proposed bearing resistance function.

1000 Net cross-section failure Splitting failure Shear failure 700 . .

Experimental resistance at d0 /6

400

100 100

400

700

1000

Eurocode bearing resistance

Fig. 11. Experimental results at hole elongation d0 /6 and Eurocode bearing resistance (in kN). Fig. 12. Experimental re vs. theoretical resistance Fb,EC for single bolt connections.

The selected hole elongation is d0 /6. For M27 bolts this elongation equals to 5 mm and for M22 bolts it equals to 4 mm. At this hole elongation the connection is at the beginning of the yield plateau or starts losing its initial stiffness (see Fig. 6). In the case of pure net cross-section failure all deformations concentrate in the net cross-section (for S690: Ac = 15%), resulting in hole elongation approximately equal to d0 /6 = 0.167d0 . In the case of shear failure the bearing stress results in hole elongation. The resistances at the displacement equalling d0 /6 are presented in Table 2 and their relation to Eurocode bearing formula is illustrated in Fig. 11. The resistance at d0 /6 is very similar to the maximum resistance for shear and splitting failures, while for net cross-section failures this resistance was reached in post-critical region (descending part of the curve). Only for B101 the ductility of joint was not sufficient to reach the desired deformation due to pure net cross-section failure (disp. at fracture Dfr = 4.9 mm 0.163d0 ). It can clearly = be seen that Eurocode resistance model does not correlate to the actual behaviour of the bearing failure of HSS plates (Fig. 11). Shear

failures are in general on the safe side, while net cross-section failures are in general on the danger side. In Figs. 12 and 13 test results (maximum resistances) of single or double bolted shear connections versus Eurocode resistance (8) are gathered. Besides our experimental results, the results of different researchers are also shown [36,8]. Except for some of B1 results with net cross-section failures where a separate net cross-section check is relevant, formula (8) is conservative, especially for Puthlis and Fleischers results. The conservatism is the consequence of the reduction of the resistance for small edge distances e2 1.5d0 . It was shown that the bearing resistance according to Eurocode is poorly controlled by hole elongation and therefore very high scatter of results especially for two-bolt connections (V = 0.434 see Fig. 13) could indicate improper resistance function. 5. Modified approach towards bearing resistance State-of-the-art design dictates the prediction of correct failure mechanism. This can be done if the resistance of each component

P. Moe, D. Beg / Journal of Constructional Steel Research 66 (2010) 10001010

1007

. .

Fig. 14. Failure modes: (a) net cross-section (b) bearing.

the least square method), where the design resistance could be formed by partial factor M2 = 1.25. In addition, the coefficients should also have some physical meaning, if possible. Considering all these demands, coefficients kij were evaluated through nonlinear optimization by Microsoft Excel Solver. Function (11) is plotted in thin lines in Fig. 10. The slope and the position of the curves are defined by both coefficient k1 and k2 , while k1 defines the transition between shear and net cross-section failure which occurs at (see Eq. (12)):
Fig. 13. Experimental re vs. theoretical resistance Fb,EC for two-bolt connections.

e1 e2

k12 k11

1.9 1.3

is known. The first step is to identify the maximum bearing resistance of the plate. Hence, the modified definition of bearing resistance may be given. Herein it is proposed that the bearing resistance is the maximum strength of the plate in bearing in view of connection geometry. This definition is in contrast to the definitions in Eurocode and American standard, where the bearing resistance is (in)directly defined by the hole elongation. The new approach may lead to identifying the weak components in HSS shear connections and to the correct estimation of the failure type. In general, bearing resistance of the single bolt shear connection with infinite end and edge distances is limited by bolt shear failure. Nevertheless, the hole elongation could be easily limited by a prescribed reduction factor. A new formula for HSS that would anticipate the bearing resistance of the plate in ultimate limit state is suggested. The formula is supported by the results of our tests and the results of other tests on similar bolted shear connections found in literature. 5.1. Single bolt connections Similarly as in Eurocode, the new bearing resistance formula is defined by the average bearing stress. In Eurocode the bearing stress is in the ultimate state limited to b k1 fu , where b and k1 are linear functions of normalized end e1 /d0 and edge e2 /d0 distances, respectively (for a single bolt connection). The idea from Fig. 10 is presented in a new formula. The ultimate bearing stress is presented as k1 k2 fu , where the ki coefficients are linear functions of end to edge distance ratio e1 /e2 and normalized edge distance e2 /d0 , respectively. Written in mathematical language, it follows: Fb,new = k1 k2 d t fu k1 = min k11 k2 = k21 e2 d0 e1 e2 (11)

= 1.46.

(14)

The transition can also be simply theoretically derived. The main assumption is that shear resistance and net cross-section have equal resistance (Fig. 14). Such failure was observed at specimen B121 (Fig. 8(c)). It follows: Fnet = Fshear (15) d0 2 d0 e2 t fu 3 (16)

(2e2 d0 )tfu = 2 e1
e1 e2

=
3

31 2 (17)

d0 1.7 e2 e2 e1 1.46 = 1.35 = d0 e2 1.43 e2 = 1.2.

d0

(18)

; k12

= min 1.3
e2 d0

e1 e2

; 1.9 ,

(12) (13)

+ k22 = 0.9

.
4

In case of single bolt connection with eccentric bolthole e2,min should be considered. Coefficients kij are determined by the following procedure. According to EN 1990, Annex D [16], estimator for the coefficient of variation of the error term V is calculated. Term V is the measure for the scatter of points in the re rt diagram (scatter diagram), where re is a vector of test results (measured resistances) and rt is a vector of theoretical resistances calculated according to expression (11). Coefficients kij are evaluated by minimizing V . The preferred solution is the one with correction coefficient b > 1 (b is the slope of regression line obtained by

Theoretically the transition between shear failure and net crosssection failure occurs when end-to-edge distance ratio is between 1.4 and 1.7. This matches the experimental results in Fig. 10, where the transition line separates two tests, one of them failing in bearing and the other in net cross-section (exp. B103, B104; B112, B113. . . ). The new formula also controls the bearing resistance of plates with unusual geometry. For narrow plates with large end distance e1 (see Fig. 15(a)), where net cross-section is critical, the formula has an upper limit, which is comparable to net crosssection resistance (dashed lines in Fig. 15(a)). The new formula also gives a limit resistance for very wide plates with small end distance e1 , where shear failure is critical (see Fig. 15(b)). As can be seen in Fig. 10, the new bearing resistance function (11) matches the experimental results very well. The function becomes slightly conservative only for large e1 /e2 ratios with larger edge distance (for instance e2 > 1.9d0 ). For such cases bolt shear is usually relevant. The scatter of results for the new bearing resistance function Fb,new (V = 0.061 in Fig. 16) is considerably lower than for the Eurocode formula (8) shown in Fig. 12. To verify the new strength function and to determine the appropriate partial factor M the statistical analysis was performed. The guidance for such analysis is given in EN 1990, Annex D and is on a practical case applied in [9]. The partial factor M is determined in two steps. In the first step M is defined as a ratio between the characteristic and the design value. In structural design nominal values are used. Thus, in the second step the nominal values have to be defined. The nominal value of the material strength may be set equal

1008

P. Moe, D. Beg / Journal of Constructional Steel Research 66 (2010) 10001010

. .

. . .

. . . . . . . . .

. . . . . .

. .

(a) the effect of normalized end distance e1 /d0 .

(b) the effect of normalized edge distance e2 /d0 .

Fig. 15. The effect of distances on bearing resistance of the proposed formula (horizontal lines in (a) present net cross-section resistance for different edge distances).

. . . .

tolerances that may be larger than assumed in (23) to (24), especially for small end and edge distances. On the other hand, the Eurocode bearing resistance formula (8) requires partial factor equal to M ,req = 1.396 (Fig. 12). If the results where net crosssection failure is relevant were removed from the analysis (33 results are included), the required partial factor for EC formula (8) is M ,req = 1.098 (see Table 3). But it has to be considered that both parameters b = 1.277 and V = 0.113 indicate that the strength function is not the most appropriate one. 5.2. Connections with two or more bolts positioned perpendicular to loading The main assumption is that the elongation of all bolt holes positioned perpendicular to loading is equal. Therefore, the bearing forces on all bolts are equal. Thus, bolts positioned perpendicular to loading should be considered as a group of bolts. By analogy with single bolt connection, coefficients k1 should differentiate between failure modes. Accordingly, the number of bolts m, positioned perpendicular to loading, is included in k1 and the edge distance e2 is replaced by half of the effective connection width beff /2. The latter modification is implicated also in the coefficient k2 . Eqs. (12)(13) are rewritten in more general form: k1 = min 1.3m k2 = 0.9 e1 beff /2

Fig. 16. Experimental re vs. new bearing resistance Fb,new for single bolt connections.

to the characteristic value. On the basis of actual material strengths, characteristic material strengths need to be obtained. This can only be done if an upper bound (conservative assumption) of the coefficient of variation of material strength Vf is available from prior knowledge. The nominal values for the geometrical variables may be adopted as mean values. To include uncertainty of steel grade and fabrication of elements, the standard deviation is increased by the coefficients of variation VXi which are also known from prior knowledge. To determine variations of independent variables VXi , the data were gathered from [11]. The following variations were used: Vd0 = 0.005 Vd = 0.005 Vt = 0.05 Vb = 0.005 Ve1 = 0.005 Ve2 = 0.005 variation coefficient for bolt hole diameter; variation coefficient for bolt diameter; variation coefficient for plate thickness; variation coefficient for width; variation coefficient for end distance; variation coefficient for edge distance; (19) (20) (21) (22) (23) (24)

; 1.9 ,

(26) (27)

beff /(2m) d0

.
4

The effective connection width considers also the pitches p2 . beff = 2e2 + (m 1)(p2 d0 ). (28) The resistance of a group of fasteners Fb,new derived directly from the assumption of equal bearing forces per bolt: Fb,new = mFb,new = m k1 k2 d t fu . (29)

Vfu = Vf = 0.007

variation coefficient for tensile strength. (25)

The required partial factor for the new bearing resistance (11) for single bolt connection is M ,req = 1.165. The scatter for this function is very low with V = 0.061 and with correction factor b = 1.049 (see Table 3, Fig. 16), and it describes the phenomena very well. Therefore, the design resistance can be formed with partial factor M2 having some extra safety for parameters which were not included in our analysis, such as the effects of fabrication

The case of block tearing, where boltholes are put very close together and far away from the edge of the plate (e2 p2 ), is considered in coefficients (26)(27). The connection geometry that leads to block tearing results in hyperbolically decreasing k1 factor and linearly increasing k2 factor, thus the product k1 k2 tends to stabilize, similarly as in Fig. 15(b). The experimental results support this statement. Unfortunately there are no experimental results available for the case where boltholes are positioned very close to the edge while the pitch p2 is very large (e2 p2 ) . In the case of modified approach, the bearing resistance function for two bolts (29) gives less scattered results than the Eurocode bearing resistance function (8). The combination of

P. Moe, D. Beg / Journal of Constructional Steel Research 66 (2010) 10001010 Table 3 Results of statistical analyses. No. bolts Failure types No. results Eurocode bearing formula b One All Bearing check critical All Bearing check critical 44 33 38 34 1.131 1.277 1.515 1.810 V 0.152 0.113 0.434 0.372 New bearing formula

1009

M ,req
1.396 1.098 2.691 1.829

b 1.049

V 0.061

M ,req
1.165

Two

1.213

0.089

1.044

. . . .

more than two bolts parallel to the loading axis. Since there are only a few experimental results for the discussed case, the new resistance model will not be presented. Extensive experimental testing of shear connections with three and four bolts in double shear is currently performed by the authors. 6. Conclusions The design bearing resistance per bolt according to EN 19931-8 was developed for steel grade S235 and was studied on single bolt connections. It was defined as strength at a certain bolthole elongation limited by mean bearing stress. Therefore, failure resistances like net cross-section failure and block tearing are not directly connected to bearing resistance and have to be checked separately. The paper presents the experimental results of the connections with one or two bolts (positioned perpendicular to loading) in double shear. Steel grade S690 was used for the fabrication of 38 connections. The test setup and the results are described in detail. Three major failures were observed: shear failure of the plate, splitting failure (failure due to transverse tensile stress) and net cross-section failure. The comments on stress distribution are supported by finite element analysis. The high strength steel plates in bolt bearing showed ductile behaviour in terms of large hole elongations with plate squashing. The bearing resistance per bolt was compared to the Eurocode formula, which for high strength steel in general gives conservative results, mostly due to too conservative bearing resistance model. The upper limit of bearing stress, which is set to limit the hole elongations, corresponds also to a limit for net cross-section resistance that presents the maximum resistance of a plate in bolt bearing. A more accurate calculation of bearing forces is needed in order to determine the ultimate resistance of the connections with multiple bolts. Since Eurocode bearing resistance does not give an adequate estimation of ultimate bearing force, the problem is how to determine the maximum bearing resistance of a group of bolts. It seems that local ductility of high strength steels is sufficient to eliminate stress peaks, thus, the main objective becomes the identification of the weakest component in the connection and the prediction of its resistance. There are two possible methods to fulfil these requirements. The first one is the identification of all failure modes and the derivation of resistance checks for each possible mode. Shear resistance of the bolt, block tearing and net cross-section resistance, which are already presented in Eurocode, are possible failure modes. The splitting failure resistance is the only missing check that has to be included in the abovementioned checks. The disadvantage of this method is that the procedure for the correct distribution of forces between bolts is not given. The second method is a derivation of a formula that would give bearing force on each bolt at the maximum possible resistance of the steel plate in bearing. Thus, the formula would correctly predict the distribution of bearing forces on bolts. This paper presents the second method. A derivation of a new bearing resistance formula of an individual fastener with a modified definition of bearing resistance is presented, where its definition also has a physical meaning. It is a function of end to edge distance ratio e1 /e2 , edge distance e2 and the number of bolts. The improvement of the

Fig. 17. Experimental re vs. new bearing resistance Fb,new for two-bolt connections.

small scatter and large correction coefficient (V = 0.075, b = 1.096 Fig. 17) results in low value of the required partial factor, equalling M ,req = 1.044. Our experimental results B2 and Puthli and Fleishers results [6] show very good agreement with a new bearing function. It is important to emphasise that different types of failures are covered in the presented experimental results. Besides shear, bearing and net cross-section failures in our experimental results, Puthli and Fleisher also experienced block tearing. The net cross-section and block tearing checks are also relevant in this case and need to be preformed additionally, although the new bearing formula covers also the mentioned checks. The design resistance can be formed with partial factor M2 . Although Eurocode resistance (8) gives results very much on the safe side for two-bolt connections, the statistical analysis gives a very large required partial factor M ,req = 2.69 because of large scatter of the results with V = 0.434 (Fig. 13). If test results for which net cross-section resistance or block tearing check is critical are removed from the analysis, the required partial factor decreases to M ,req = 1.83 while large scatter remains V = 0.372 (see Table 3). The improper strength function is again to blame. Nevertheless, EC bearing resistance is safe, although large partial factors are required. The problem is that the strength function itself gives too safe results for certain geometries. The deviation of safe results entails large scatter and consequently large partial factor. The answer would be to eliminate too safe results from the statistical analysis, but the problem is that there is no rule to tell whether the result is too safe. 5.3. Connections with two or more bolts in the direction of loading The distribution of load between bolts in a shear connection with more than one bolt in the direction of loading is directly affected by the geometry of the connection (end, edge distances and pitch p1 ) and the rigidity of the connected parts or elements. As presented throughout of this paper, experiments show that HSS should have enough local ductility to redistribute load between

1010

P. Moe, D. Beg / Journal of Constructional Steel Research 66 (2010) 10001010 [2] EN 1993-1-12. Eurocode 3: design of steel structurespart 112: additional rules for the extension of EN 1993 up to steel grades S 700. Brussels: European Committee for Standardisation; 2007. [3] Kim HJ, Yura JA. The effect of ultimate-to-yield ratio on the bearing strength of bolted connections. J Construct Steel Res 1999;49(3):25570. [4] Aalberg A, Larsen PK. Bearing strength of bolted connections in high strength steel. In: Mkelinen, Kesti et al., editors. Nordic steel construction conference 2001NSCC 2001: proceedings, NSCC. 2001. p. 85966. [5] Aalberg A, Larsen PK. The effect of steel strength and ductility on bearing failure of bolted connections. In: Lamas, Silva, editors. The third European conference on steel structures: proceedings of the 3rd European conference on steel structures. Universidad de Coimbra; 2002. p. 86978. [6] Puthli R, Fleischer O. Investigations on bolted connections for high strength steel members. J Construct Steel Res 2001;57(3):31326. [7] EN 1993-1-8. Eurocode 3: design of steel structurespart 18: design of joints. Brussels: European Committee for Standardisation; 2005. [8] Rex CO, Easterling WS. Behavior and modeling of a bolt bearing on a single plate. J Struct Eng, ASCE 2003;129(6):792800. [9] Moe P, Beg D, Lopati J. Net cross-section design resistance and local ductility of elements made of high strength steel. J Construct Steel Res 2007;63(11): 143141. [10] ANS/AISC 360-05. Specifications for structural steel buildings. Chicago: American Institute of Steel Construction, Inc.; 2005. [11] Snijder HH, Ungermann D, Stark JWB, Sedlacek G, Bijlaard FSK, HemmertHalswick A. Evaluation of test results on bolted connections in order to obtain strength functions and suitable model factorspart A: results. In: Eurocode No. 3part 1background documentation. Document 6.01. Brussels: Commission of the European Communities; 1988 [Chapter 6]. [12] Bijlaard F. Eurocode 3, a basis for further development in joint design. J Construct Steel Res 2006;62(11):10607. [13] EN 10002-1. Metallic materialstensile testingpart 1: method of test at ambient temperature. Brussels: European Committee for Standardisation; 2001. [14] EN ISO 377. Steel and steel productslocation and preparation of samples and test pieces for mechanical testing. Brussels: European Committee for Standardisation; 1997. [15] Snijder HH, Ungermann D, Stark JWB, Sedlacek G, Bijlaard FSK, HemmertHalswick A. Evaluation of test results on bolted connections in order to obtain strength functions and suitable model factorspart B: evaluations. In: Eurocode No. 3part 1background documentation. Document 6.02. Brussels: Commission of the European Communities; 1988 [Chapter 6]. [16] EN 1990. Eurocodebasis of structural design. Brussels: European Committee for Standardisation; 2004. [17] EN 1090-2. Execution of steel structures and aluminium structurespart 2: technical requirements for steel structures. Brussels: European Committee for Standardisation; 2008. [18] Snijder HH, Ungermann D, Stark JWB, Sedlacek G, Bijlaard FSK, HemmertHalswick A. Evaluation of test results on bolted connections in order to obtain strength functions and suitable model factorspart C: test data. In: Eurocode No. 3part 1background documentation. Document 6.03. Brussels: Commission of the European Communities; 1988 [Chapter 6].

formula comes mainly from the ratio e1 /e2 which successfully controls different failure types. Moreover, the proposed formula also has an upper limit to control the net cross-section resistance. If this limit was set lower, then the same formula could be used to limit the hole elongation. The formula was also statistically evaluated according to EN 1990, Annex D for two studied cases: single bolt connection and a connection with two bolts positioned perpendicular to loading. The prescribed reliability was easily achieved by the recommended value of partial factor M2 = 1.25. The proposed formula corresponds also to the geometry that is outside the minimum limits set by EN 1993-1-8. Despite that, the reduction of minimum distances is not recommended due to fabrication tolerances defined in EN 1090-2 [17]. The tolerance for an end distance e1 equalling 2 mm (Class 1 according to EN 1090-2) may have important influence on the reduction of the bearing resistance for smaller holes. Any significant reduction of pitch distances would be in many cases pointless due to the erection demands (size of a wrench key). The proposed new bearing resistance was planned to be reevaluated for mild steels, but due to missing test data for mild steel the evaluation has not been performed yet. Despite all efforts, the report [18] could not be found. Furthermore, the proposed formula does not present an additional check to the existing checks, but it could replace the current Eurocode bearing resistance check, preserving the simplicity of the current check and providing more information on the behaviour of bolt bearing. Acknowledgements Company Trimo d.d., Trebnje Slovenia, provided the material and fabrication of the specimens. The work was also partly financially supported by the Slovenian Ministry of Science and Technology within the program of young researchers. Their contribution is gratefully acknowledged. References
[1] EN 1993-1-1. Eurocode 3: design of steel structurespart 11: general rules and rules for buildings. Brussels: European Committee for Standardisation; 2005.