compurers & stnccrures Vol. 61, No. 5, pp. 967-974.

Copyright 6 1996 Else&r Science Ltd
Printed in Great Britain. All rights reserved
0045.7949/96 SIS.00 + 0.00

PII: soo45-7949(96)ooo83-1


H, I, Epstein and R. Cha~~aj~gar

Department of Civil Engineering, University of Connecticut, 261 Glenbrook Road, U-37, Stem,
CT 06269-3037, U.S.A.
Parsons, Brinckerhoff, Quade and Douglas, San Francisco, CA 94107, U.S.A.
(Received 4 March 1994)

AI&r&--Tests of angles in tension have produced block shear failures for certain ~nn~tion geometries.
Finite element investigations have previously shown that the mode of faiiure can be exhibited analytically
and have also helped to design future testing. This study attempts to give credence to trends previously
predicted from actual block shear tests by comparing nondimensionalized finite element results with the
results of the full scale testing. The effect of the outstanding leg, shear length and stagger are compared
in this study. Copyright 0 1996 Elsevier Science Ltd.


Finite element studies can be an extremely useful tool
when used in conjunction with an experimental testing
program. A parametric finite element computer study
was successfully employed for preliminary studies
of block shear failure for angles in tension [l]. These
studies helped in the design of a full scale testing
program for angles which were connected by only one
leg and had bolts in two gage lines on that leg [2,3].
One of the interesting results obtained from both the
experimental and the computer study was the effect
that bolt stagger has upon the tensile failure load of
angles. A finite element study was also accomplished
[4] which helped to understand the effect of the bolt
stagger by considering gage and stagger spacing on
flat plate finite element models. This study reinforced
some of the conclusions reached from the analysis
of the experimental results for angles. The flat plate
was studied because of the difficulty in separating the
effects of stagger and shear lag.
The so-called shear lag effect occurs in connections
such as these angles because load is not always
uniformly transferred from a tension member to its
connection by all its cross-sectional elements (i.e.
the legs of an angle, flanges and web of a W section,
etc.). The shear lag effect can significantly reduce the
capacity of a tension member. The strengths of the
angles studied in the block shear experiments were
defi~tely infhrenced by this shear lag which, in reality,
is due to the effects of bending present as the results
of loading eccentricity [5]. While current design codes
such as AISCs ASD [6] and LRFD [7] do incorporate
a shear lag correction, none is included for block
shear failure nor is the stagger effect of bolt geometry
included in the codes for block shear failures.

The finite element studies reported in this presentation were accomplished upon completion of the
experimental program on angles [2,3]. It should be
emphasized here that any finite element predictions
of load are strongly tied to failure criteria, boundary
conditions and the other parameters in the modeling
of the problem. However, relative results obtained
from finite element studies, such as this study are
useful barometers of various parametric effects.
The primary reason for the experimental and all
these finite element studies was the concern that
current steel design codes do not adequately handle
connections of this type. The results of this current
study help to reinforce some of the effects of various
parameters that were found experimentally and
add credence to some of the previously suggested
code ~ommen~tions
[2,3]. Furthe~ore,
this finite
element program can be used with some confidence
to help extrapolate results beyond the restrictions
of the testing program and to design additional tests,
where required.


Nonlinear analysis

The nonlinear finite element discretization in
structural problems leads to an equation of the type
f(q) = K(q) * q - g = 0,


where q is a vector of unknown displacements, g is the
vector of applied loads and K is the stiffness matrix.
Nonlinearities usually arise from either geometry
or material considerations. Among the former are
large deflections, stress stiffening, contacting surfaces
and buckling. Creep and nonlinear elastic behavior

The finite element model to be used may be influenced by the type of results desired. 1. For this study. It is a package that performs an incremental solution to the equations of motion involving large deformations and material nonlinearities. it is always wise to match the mesh refinement to the stress gradients and deformation patterns. negative stagger if less than 90” and zero if equal to 90”. Modeling for stresses near regions with holes and cutouts requires mesh refinement in order to predict peak stresses at the holes. This was decided upon after many preliminary results [8]. the material was perfectly plastic above the assumed yield of 36 ksi (corresponding to A36 steel) with other properties consistent with steel. This corresponds to connection numbers 1. The element and its coordinate system used in this study are shown in Fig. In models where complete stress patterns. they conformed to whatever rectangular dimensions were required.!Y I 2s=3” F (4 Fig. but do not as yet explicitly deal with stagger effects when dealing with block shear. The material nonlinearity is considered in this study. depending upon the size of the angle (L. in Fig. All the finite element studies used a model length of 30 in. Since one of the objectives of this study was to compare stresses at points where strain gages were used (sometimes on both sides of the angles) a solid element was adopted. 3. in Fig. The angle sizes are listed as the length of the connected leg (L. It can then be seen that the connection geometry shown in Fig. 2 is used in specifying the stagger sign. results in eccentricities which produce bending and twisting about the centroidal axis of the member. lc shows the grid for a complete model.968 H. This diameter corresponds to an edge distance (d in Fig. Furthermore. but have aspect ratios within acceptable ranges (usually approximately 3 : 1 for stresses and 1O:l for displacements). If the variation of stresses and strains are an order of magnitude apart from the variation of displacements. This eliminated the possibility of using plane stress elements. and g. the grids around the holes are shown as square. which are connected by only one leg. Previous experimentally tested geometries [2. . The connection geometries in Table 1 have a notation giving the number of holes for bolts in the outer gage line followed by the number of holes in fd=l. 9. I. The finite element model requires a defining of the geometry. All of these considerations were taken into account in establishing the finite element model. 1) as given by AISC [6. In particular. The convention adopted for the sign of the stagger is shown in Fig. 1. For simplicity. For the actual studies. 1) of 1. The finite element program used for this study was COSMOS/M. 29 or 33. This means that elements must vary in size. Chamarajanagar are forms of material nonlinearities.7] present an s2/4g correction for stagger when dealing with net tension failure. Element geometry The grid used around a hole is shown in Fig. AISC codes [6. Finite element grid around holes and some of the parameters of the investigation. the inner gage line and the sign of the stagger. These dimensions for d and s are used throughout this investigation. The loading on the angles in this study. Epstein and R.7] are used for all models and the holes were all for 3 in diameter bolts. 25. may be required. Standard gage distances (g. The connection is said to have a positive stagger if this angle is greater than W. Table 1 shows the results for 114 connections tested (three each of the 38 geometries listed). boundary supports.5 in and a usual pitch (2s in Figure 1) of 3 in. materials and loading. as shown in Fig.3] are used in this finite element study. la. and L. a model which accurately predicts displacements may not do well with stresses and vice versa. The included angle in the potential block shear path shown as “a” in Fig.). 1) followed by the length of the outstanding leg (L2) and the thickness of the legs of the angle (t in Fig. including concentrations. 17. 1 is 2/2+. 2. the nonlinearity was due to the material only and the elasto-plastic von Mises yield criterion was adopted. a three-dimensional element may be appropriate even though the structure consists of relatively flat elements. Figure lb shows how the grids around the holes were typically patched together and Fig. 2). Boundary conditions The model fixed the top half of each hole. Element type The MODSTAR module of COSMOS has a relatively comprehensive set of elements with which to perform nonlinear analyses.

Since the primary objective of this finite element study was to correlate analysis wi. 2. The experimental specimens showed little deformation of the upper half of each hole. 2. as shown in Fig. ZERO (a=90”) Another important question to be addressed was the criteria to be used in comparing the failure of models representing the various geometries studied.Studies for correlation with block shear tests 969 Table a corresponding experiment. Failure criteria POSITIVE (a>90”) NEGATIVE (a<907 Fig. The failure criteria used in this study was the longitudinal strain at this point. the incorporation oft hese fixed boundaries was deemed appropriate. geometry and results including the results of previous investigators and the performance of tested connections. Connection designations. Sign of the stagger. All the experiments had their failure initiate at the outside edge of the connected leg adjacent to the lead bolt on the outer gage. Longitudinal displacements of two .

Then. the 30in RESULTS OF THE FINITE ELEMENT STUDIES Global Cartesian Coordinate SySteI77 Z Fig. As will be discussed in the results section to follow. there is also a change in the sign of the stagger. While every strain level produced different numerical results. were found was by interpolation of the strains found for the inputted loads. Epstein and R. What is important in this study is the close relationship in the trends shown for the finite element studies compared to the experimental results. the appropriate length for the finite element model. were used in calculating the strain. these loads were first divided by ultimate strength in an effort to account for varying material properties. To eliminate the effect of this parameter. each of these groups is investigated separately. The way in which the numerical values. nodes. In some of the cases shown. 1 in on either side of the failure initiation point (see Fig. it is necessary to identify those connections where only this variable changes. Consistently. Table 2 shows the comparison of experimental results with those from the current finite . The nonlinear analysis used this load as a first step and then the load was increased in 10% increments of this initial yield load. Load application Another important question to be addressed was how the load should be applied and a related issue. but this makes no difference in the code treatment [6. The shear length To see the effect that shear length has on failure. some of the experimental strain gage results indicated that the length of the specimen had little effect. lb.3. one uniform over the connected leg only. the nondimensional parametric variations produced similar trends regardless of the failure strain level adopted [S]. having the ultimate strength F. Only the results corresponding to strains five times the yield are reported in this presentation. This has been thoroughly discussed in previous work [2. The fifth column in Table 1 shows the results of the experimental investigation of the 38 connection geometries of this investigation. The program was halted when the calculated strain exceeded the failure criteria. as shown in Fig. The connections tested had either a six inch connected leg (connections numbers l-24) or a 5 in connected leg (25-38). This required only an elastic analysis. and the second uniform load over the entire cross section. finite element results corresponding to loading only the connected leg were much closer to agreeing with those previously obtained experimentally [S]. corresponding to a given failure strain criteria.5] and was shown to be the result of the eccentricity of the loading with reference to the centroid of the angle. The loads input into the analysis were determined by first finding the load corresponding to yield. In order to study the effects of varying geometric parameters. This produces bending in the angle in addition to the tension. this agrees with previous analytical as well as experimental results. The results of this process are shown in the sixth column of Table 1. No entry is made for the last two connections since there are no others to compare them with. Fig. and explains the shape of the majority of the graphs in Figs 4-6. The seventh column shows the results of the finite element studies nondimensionalized in the same manner. The outstanding leg Figures 4 and 5 show the comparisons of experimental and finite element results for the eight different connection geometries investigated for angles with 6 in connected legs. Element and its coordinate system. there is not much difference. 3. Preliminary studies [S] used two different loadings. I. the failure loads were nondimensionalized by setting the equal leg angle geometries (6 x 6 or 5 x 5) to one and comparing the other connections to the corresponding equal leg angle with the same connection geometry. Chamarajanagar to help with the computational model length was adopted. Strains varying from twice the yield through five times the yield were investigated in the preliminary studies. Face Number for efficiency. Consequently. It is not the intent here to discuss the perceived reasons for the trends in the effect of the length of the outstanding legs. Each of these 38 loads represents the average of three specimens made from the same material.H. When the scale of the vertical axes in these figures is noted. lb). except that there was no material variation. Figure 6 presents the same comparison for the four different connection geometries having 5 in connected legs. to first investigate the effects of the outstanding leg. Id. indicated in the table.7]. As far as the length of the model is concerned.

For instance.06 E :. 2). #’ 3’ ’ 0.CONNECTIONS 1.86 * Expdnwlt -* Plnlt~ Elom~nt 0. connections 3 and 1 both have a 6 in connected leg and a shear length on the block shear path of .___ ---__-*-__ E b ii * 0.86 I 4 - s 6 7 OUTSTANDINQ LEG 3 4 6 8 7 OUTSTANDING LEG Fig.9 - 0 0 0.06 --__ k ’ ii D 0. but of different sign (as defined in Fig.16 : 0. but varying length of the outstanding leg.98 : T l 0 4 OUTSTANDINQ LEO I : R LXpWlmWlt-* Clnll*Element 0. 2). Several sets of connections in the experimental study differed only by either having stagger where another did not or both having stagger.86 3 4 S 8 0 6 7 OUTSTANDING LEO 212 0 CONNECTIONS 213 .CONNECTIONS 212 + CONNECTIONS r\. with the long-standing correction s2/4g (both these parameters were previously shown in Fig. : 1.9 .9 - 0.96 ! T l \“‘I-.96 T 1 0. study.16 : 1.1 1. element Stagger The current codes account for stagger of holes in a tension failure. .. 1. 4._ .1 I lj R 1.9 0 0.971 Studies for correlation with block shear tests 2/2 .16 p-. R E 8’ ij ’ ---___ 1. 0._ ----___ A.06 y+-.86 ’ 3 7 1.1 I 1.96 ! T l 0. The conclusions of the previous studies are reinforced by these results which show that shear length is being appropriately accounted for in the code equations. Comparison of four of the connection geometries for a 6 in connected leg.

[9]) think it reasonable to include this factor.06 E k t * *.5 in. Connection 3.7% increase and the current . Epstein and R. -+ 1.66 i T l 0 0. When included. The same ratio is true for 9/l 1 and 17/19.1 t R 0. has no stagger in the block shear path whereas 1 does. Comparison of the remaining four connection geometries for a 6 in connected leg. I.0 4 0. the code equations give that connection 1 is 5.1 1. Chamarajanagar 213 312 + CONNECTIONS 0 CONNECTIONS 1.1 I : R 1.972 H. connection 1 should be stronger if the s2/4g addition to the tensile length is appropriate. Ref.66 Exporlmont 4 a -+ 6 Phlh Elwn0nt 6 0. Authors of current texts and reference materials which address the problem (e. As such.161 1.66' a OUTSTANDINQ LEG Exporhmnt PlnlC thmmt I 4 I 6 6 7 OUTSTANDING LEO 4/4 313 0 CONNECTIONS 0 CONNECTIONS lei6 r----- 1.g. ’ “L_____ --__ ---___ D 0. \. as previously described.2% stronger than 3.0 t 0.66' 7 a 4 6 6 7 OUTSTANDINQ LEG OUTSTANDING LEO Fig. 4. The experiments gave an average of a 1.66 t 0. 5. once the varying material strengths are factored out.16 : ’- . however. It should be noted that the codes do not mention block shear paths with stagger.

9 -+ 4 6 3 6 Fhlb thmmt 6 6 LEQ 3/Z + CONNECTIONS 1.86 i2 0.1 : I !YJ R 1.1 : i : R : R 1.1 i 1. The code equations give a 4.0% increase whereas the experiment and finite element results both show a decrease.06 E E l5 ’ k :: ’ 6 0.06 1. finite element study a 3.1s 1. There were seven sets of connections. Three sets differed only in that stagger went from zero to negative.86 3 OUTSTANDING LEG : I ExperlmwU !z T 0. The experiment and the finite element studies both produced an increase.CONNECTIONS ij -* 0.06 ’ D D 0.9 0.96 0. 6.1 I R A T 0. but varying length of the outstanding leg.913 Studies for correlation with block shear tests Z/2 212 .16 : 1.2% decrease(two of the three connections produced a decrease). which differed only in the sign of the stagger. as shown in the table.96 0. Eleven of the .CONNECTIONS CONNECTIONS + 1. These results are shown in Table 3.16 1. Comparison of the four connection geometries for a 5 in connected leg.15 1.9 I- 1 0 0 0.86 1 2 3 4 5 6 OUTSTANDINQ LEG 0.96 3 1 1. Codes do not recognize the sign of the stagger as being signi~cant.M ! T 1 0 4 OUTSTANDIN Z/3 .86 ’ 2 I I 9 4 6 6 OUTSTANDING LEG Fig.06 : R E E k 1 1. indicated in Table 3.

American institute of Steel Construction. Adidam. M. American institute of Steel Construction. I. 1153-I 156 (1993). Load and Resistance Factor Design. Gulia. 29. Which connection to choose Connection 1 (2/2+) and connection 2 (2/2-) differ only in the arrangement of the holes on the connected stagger for block shear tension failure in angles. On the other hand.h 5. IL (1988). Epstein and N.. 1”1LEl I-d #3 CONCLUSIONS The finite element results presented herein are very encouraging.*. H. 1st edn. University of Connecticut. IL (1986). IL (1989). The conclusions of the effect of various parameters.S. Analysis of block shear experiments for structural steel angles in tension. Comput. Camput. India. Connection 2 has a 6 in shear path compared to a 4. #15 6” #23 5. Chamarajanagar. the higher the block shear load. once material strengths are considered. Confidence is therefore obtained in using #19 such studies to extrapolate beyond the range of variables used in an experimental program. Chamarajanagar Table 2. (1990). -_. AISC..S. 39.0% (averaL leg. I. #7 4. H. AISC. J.9% 5 +30. 6. H. Srruct. Manual of Steel Construction. CT (1990). CT (1990). Bangalore.- l_-. H. #33 #35 Average of all 11 I +23. . Comparison of connections differing only in stagger . 8. Analysis and Design. reported in previous experimental work. 6” #ll Connections - 1% increase w&. As a designer. ICSTAD Proc. 48.5 in path for 1. Tbacker. Eiements for Teaching Load and Resistance Factor Design. 758-763.5%! SUMMARY 1.04btr15. S. The effect of eccentric tension of the block shear failure of angles. M.. According to the results herein and the experimental studies. S. The experiments gave all 13 connections producing higher loads for the increased shear length connections. N. Strut?.8% (aver0 . thesis. An experimental study of block shear failure of angles in tension. if faced with the choice of which connection to choose. There are 13 sets of connections having the same number of bolts and the same size connected leg. the increase in shear length outweighs the decrease in the stagger. once the number of bolts are selected for a short connection where block shear is a distinct possibility. 3. there would appear to be con- flicting parametric effects. 9. Allowable Stress Design. Even though there is still much room for improvement in the finite element modeling and failure criteria. 75-84 (1992). S.. Manuul of Steel Construction. #2 6’ AND #lO #18 REFERENCES I. R. I. the relative values obtained appear to he good trend indicators.0% +22. Vol. Finite element studies of bolt stagger effects in tension members. the designer should choose the pattern maximi~ng shear length. support the observations of the previous experimental and finite element studies [l-3] that the sign of the stagger appears to be an important consideration in any correction factor used to predict the effect of stagger. C&ago. The effect of bolt 2. thesis. Comoarison of comections differing onfg in shear length 96 increase Ixp*rlnlsn FS”. Finite element studies for the correlation of stresses and failures of block shear tests. The finite element results [8] reinforce this conclusion. Chicago.974 H. Epstein and F. Therefore. I. 2. 9th edn. AISC Engng J. 571-576 (1991). the longer the shear path. 7--i 6’ +12. t ___ . University of Connecticut.. pp.24. 7. As it turns out. Yura. Epstein and B. Table 3. both experimentally and from the finite element results. Epstein and R. in: Adounces in Structural Testing. H. have indeed been supported by these studies. connection 2 having negative stagger compared to the positive stagger of 1 should produce a weaker connection. Adidam. which differ only by stagger. I. Epstein. Chicago. AISC. i 13 sets of connections.