Capacity of Columns with Splice Imperfections

E. p. POPOV AND R. M. STEPHEN

During the erection of a structure, the milled or cut surfaces column conipression tests, careful initial measurements of
at column splices may not make perfect full and even con- the specimens were made and the residual stresses deter-
tact. AISC has pointed out some of the reasons for these mined.
column splice imperfections:^ Because very little data are available on the behavior of
spliced columns subjected to tensile forces simulating sit-
" 1. The most advanced milling techniques utilized vs^ith
uations which may develop in an earthquake, all of the
appropriate shop layout to theoretical center line
spliced specimens were tested to failure in tension after first
will realistically result in some measurable deviation
having been subjected to large compressive loads.
from perfection.
2. The accuracy of substructure levels and elevations
is limited and is beyond the control of the fabrica- GENERAL DESCRIPTION
tor.
The tests were conducted using W14X176 shapes of A36
3. Erection techniques involve measurable movement
material, due to their common use in buildings. Five
from theoretical dimensions to compensate for
square-ended specimens were fabricated and tested; four
permissible individual member tolerances in
of these were approximately 14 ft-8 in. long and one was
achieving overall compliance to plumb, level, and
10 ft long. Four of the five specimens were spliced at mid-
line.
height to obtain the most critical condition for buckling,
4. Erection techniques in field welded structures may with various sizes of gap introduced about the two major
deliberately cause the columns to be tipped slightly axes. It is important to note that these wedge shaped gaps
out of plumb just prior to welding beams to columns were achieved by first cutting the unspliced specimen in half
in order to compensate for weld shrinkage." and then cutting one of the faces to be rejoined at a slightly
Recognizing these possibilities, AISC^ accepts a maxi- oblique angle with the longitudinal axis of the member
mum gap of yi6-in. without shims. For larger gaps, the use before welding the two members together. This eliminated
of non-tapered mild steel shims to pack out the gap is re- any large discontinuity at the splice and maintained the
quired. In either case partial penetration welds or bolted overall initial camber and sweep requirements.
connections across both flanges are used to join the members The wedge shaped gaps ranged from a full contact
together. The finished spliced members must conform to condition on one side of the splice to the desired gap width
specified tolerances.^ on the other. Two of these intentionally introduced gaps
Except for work by Hayes, who investigated bolted splice were made to occur about the weak axis and two about the
connections using small column sections,^ data on the be- strong axis. Specimens No. 1 (weak axis splice; see Fig. 1)
havior of full-size columns having imperfect contact at the and No. 3 (strong axis splice; see Fig. 2) had maximum gap
splices are not available. Therefore, it was decided to con- widths of yi6-in. and no shim was used to pack out the
duct a modest investigation to provide some needed infor- splice. A short length of yi6-in. diameter wire was used to
mation on welded splices."^ Since it is known that the initial maintain the desired yi6-in. gap during the welding process.
imperfections and residual stresses are important factors Specimens No. 2 (weak axis splice; see Fig. 1) and No. 4
influencing the behavior of columns,^'^ in addition to the (strong axis splice; see Fig. 2) had maximum gap widths
of y4-in., and yi6-in. fiat shims of A36 steel were used to
E. P. Popov is Professor of Civil Engineering, University of California, pack out the splice. The procedure used for installing shims
Berkeley. differed from standard field practice, as it was not feasible
R. M. Stephen is Principal Development Engineer, University of
to drive in the shims between the unrestrained halves of the-
California, Berkeley.
specimen in the fabricating shop. Shim material was placed
16

ENGINEERING JOURNAL / AMERICAN INSTITUTE OF STEEL CONSTRUCTION

 1/16 DIA. WIRE
TO MAINTAIN
GAP

Fig. 1. Details of Specimens No. 1 and No. 2

in stacks at discrete locations in the joint to maintain the described above, producing a column approximately 10 ft
required gap. Again, a short length of yi6-in. diameter wire long.
was used to maintain this desired %-in. gap during the The splices in each specimen were maintained by having
welding process. Specimen No. 5 was left unspliced and 7i6-in. partial penetration groove welds on the outer face
treated as the control member. In the process of fabricating of both flanges. The weld size was the minimum permitted
Specimen No. 3, two 2 ft-4 in. pieces were removed from for the thickness of material, according to the AWS Code
the center of the member, to be tested as a stub column and in effect at the time the specimens were fabricated. No
for residual stress measurements. The remaining two welding was done on the inside face of the flanges or on the
5 ft-0y4 in. long segments were then spliced together as web at the splice. The average measured weld sizes, de-

O MILL

^ SAME AS FOR SPECIMEN I & 2

3-1/16 NON-TAPERED
ASTM A - 3 6 S T E E L SHIM

1/16 DIA. WIRE TO 1/6 DIA. WIRE TO

' M A I N T A I N GAP MAINTAIN GAP
LJ A
'JL

4 ^1/16"

-V-
SPECIMEN 5 - S P L I C E DETAILS SPECIMEN 4 - S P L I C E DETAILS

o i_L
Fig. 2. Details of Specimens No. 3- and No. 4
17
FIRST QUARTER / 1977
 termined after the column tension tests were completed, are
noted in Table 1. Gap weld details are shown in Fig. 3.
Once the specimens were spliced, the ends of each were
1/4" milled at 90° to the longitudinal axis. In addition. Speci-
mens No. 1 through No. 4 were fitted with end fixtures to
NOTE: SEE PG. 13
FOR MEASURED VALUES
accommodate pins for carrying out the tension tests. After
OF WELD DEPTH, Tw
the compression test, similar plates were added to Specimen
No. 5.
Careful measurements were taken of each specimen to
determine the section dimensions. It was found that mea-
Fig. 3. Flange weld detail of splice surements compared very closely with the dimensions given
by AISC.^ Measurements also were taken to determine the
initial camber and sweep of the five specimens. These re-
sults indicate that the maximum camber and sweep did not
exceed dzYg in.
STRESS
COMPRESSION
TESTING PROCEDURE
Residual Stress Measurements—The residual stress
distribution was determined by the sectioning method from
the 2 ft-4 in. length of column cut from Specimen No. 3.^
The residual stress pattern is shown in Fig. 4.
The general residual stress distributed does not follow
the typical, approximately parabolic shape given by Tall.^
This is partly because the W14X176 is a heavy shape, much
heavier than any considered by Tall. However, the most
important information to be gained from Fig. 4 is that the
shape was apparently straightened. The dip in compressive
residual stresses towards the flange tip is typical for
straightened shapes. The values of E used to convert all
strain readings to stress was taken as 30 x 10^ ksi.

Coupon Tests—Following the residual stress determi-

nation, nine y2-in. thick slices of column material were
chosen to determine the average material properties over
the cross section. All specimens were approximately 12 in.
STRESS
COMPRESSION
in length. Six of these were from various flange locations,
while the remaining three were taken from the web. The
specimens were originally 12 in. x 1.2 in. x 0.5 in. strips and
were machined into tension test coupons in accordance with
Fig. 4. Results of residual stress measurements ASTM A370. The gage length varied from 2 in. to 4 in.

Table 1. S u m m a r y of Test Results

Maximum Center
Maximum Deflection (in.) Ultimate Average Weld Size (in.)
Specimen Compressive Tensile
No. Splice Type Length (ft) Force (kips) Weak Axis Strong Axis Force (kips) North Side South Side

1 Weak Axis 14.703 1,650 3.25 0.12 532 0.50 0.48

Min 1/16 in. Gap

2 Weak Axis 14.693 1,655 3.00 0.07 800 0.54 0.63

Max. 1/4 in. Gap

3 Strong Axis 10.047 1,960 0.47 0.05 1,070 0.46 0.43

Min. 1/16 in. Gap

4 Strong Axis 14.708 1,760 3.50 0.86 200 0.61 0.29

Max. 1/4 in. Gap

5 Unspliced 14.646 1,675 4.70 0.50 1,800 — —

18
ENGINEERING JOURNAL / AMERICAN INSTITUTE OF STEEL CONSTRUCTION
 The yield point varied over the section as expected, and
is consistent with results obtained by Tall.^ The web ex-
hibited the highest average yield stress of 36.5 ksi, while the
flange material had an average of 33.9 ksi. The extreme
edge of the flange had a yield stress of 38.0 ksi. The ultimate
tensile stress values averaged 65.5 ksi with extremes of 63.5
ksi and 67.0 ksi. The mill test report indicated a yield point
of 39.7 ksi and a tensile strength of 68.4 ksi.

Compression Tests of Columns—The column com-

pression tests as well as the tension tests were performed
in a 4,000,000-lb Southwark-Emery Universal testing
machine. In the column compression tests the longitudinal
axis of each specimen was aligned with the axis of the
testing machine. Aluminum plates approximately Vg-in.
thick, along with steel shims, were used at the top and
bottom bearing surfaces, where it was determined that
insufficient contact between column and testing machine
head existed. The combination of bearing plates and steel
shims resulted in uniform strain readings (within ±5%)
over the mid-height cross section under loads of 400 to 800
kips. The swivel head of the testing machine was locked in
a horizontal alignment to prevent its rotation and this, in
addition to the square ends of the columns, produced es-
sentially fixed end conditions for the specimens.
The instrumentation for each specimen consisted of the
following: (1) Four SR-4 type strain gages, two on the
Fig. 5. Typical compression test arrangement
outside of each flange, each set 4 in. from the edge and 15
in. away from the splice at mid-height. In the unspliced
column these were placed at mid-height. (2) Four linear
potentiometers, two on each flange directly over the splice,
set 1 in. from the edge with 10-in. gage lengths. (3) Seven TEST RESULTS
linear potentiometers; three were placed at the quarter- Principal Results—The load deflection characteristics in
points of the column specimens to measure lateral motion compression for Specimens 1, 2, 3, 4, and 5 are shown in
in the weak axis direction. Three others were placed at Figs. 6 through 10, respectively, and a summary of the
quarter points to measure lateral motion in the strong axis principal test results is given in Table 1. A typical buckled
direction. The seventh potentiometer was used to monitor configuration of a column in an advanced stage of testing
the axial shortening. The test arrangement is shown in Fig. is shown in Fig. 11. Local buckling of the type shown in
5. In both the elastic and inelastic range the loading was Fig. 12 occurred in all spliced specimens. All local buckles
applied in increments. In the elastic range the loads were were confined to 14 in. on either side of the splice.
applied in 200 kip increments up to 1000 kip and thereafter The results of the tension tests are shown in Figs. 13
this loading was reduced to a minimum of 25 kip incre- through 17 as hysteresis loops. The initial compressive
ments. When the loading was in the inelastic range, the load loading curves have been replotted, and the tensile load-
was allowed to stabilize before any strain or deformation deflection characteristics added to the diagram. Since the
readings were taken. Testing was continued until either boundary conditions for the two types of tests are different,
deformations were sufficient to completely define buckling these results must be carefully interpreted.
or until the load reached a maximum point and began to It should be noted that besides the overall buckled shape,
decrease. the specimens also had significant local buckles in the
flanges at the splices. Measurements were taken after the
welds had failed, to determine their actual widths. Pertinent
Tension Tests of Columns—After each specimen was data for these tension tests, including fracture load and
tested in compression, and thus had sustained considerable average weld thickness, are included in Table 1.
lateral deformation, tension was applied until failure took The unspliced specimen. No. 5, also was tested in ten-
place. The pin ended fittings allowed the rotations of the sion, and reached an axial force of 1800 kips. An abrupt
longitudinal axis as the axial loads were applied. Mid- failure took place at this load, causing fracture at the pin-
height lateral displacements and axial deformations were plates which had been added after the compression test in
the only instrumentation used during the tension tests. order to salvage an otherwise excellent specimen.
19

FIRST QUARTER / 1977

 UPPER CURVE (mm)
0 UPPER CURVE(mm) 25
.50 .75 1.00 LOWER CURVE (mm)

7 6 0 KIP = MAX. LOAD; DEFLECTION = 3.5 "

.005 .01 .015

LATERAL M I D - H E I G H T DEFLECTION, S(IN.) LATERAL MID-HEIGHT DEFLECTION, S(IN.)

Fig. 6. Load-deflection for Specimen No. 1 Fig. 9. Load-deflection for Specimen No. 4

UPPER CURVE(mm) 50
I 1 _. . I

0 UPPER CURVE (mm) 25

0 LOWER CURVE (mm) .25 .50 .75

2000 V 1655 KIP MAX LOAD
AT SIN. D E F L E C T I O N . ^ ^
-8000KN
^ ^ . - - S T R O N G AXIS ^^WEAK AXIS /STRONG

^ r1 r TK" ^WEAK

<
o
(
P "p—STRONG
WEAK AXIS
AXIS
LP
z
- 4 4000 1 1
CO 1000

o
V •
H-S

I
i"""' LATERAL MID-HEIGHT DEFLECTION. S(IN.)

RESIDUAL _ ^ 1
DEFLECTI0N^\^ 1 Fig. 10. Load-deflection for Specimen No. 5

LATERAL MID-HEIGHT DEFLECTION, S(IN,)

Fig. 7. Load-deflection for Specimen No. 2

(mm) 0 2.5 5.0 7.5 10.0 12.5

1 1 1
^ ^ S T R O N G AXIS

-8000KN ___^
I 9 6 0 KIP AT Al"-^'^^
^-WEAK AXIS DEFLECTION

\ L
f —p-^W-rt—

-4000KN \ Xs U = | ^ : : ^ I / I 6 I N . GAP

Ly(,_JL
1 •
RESIDUAL ^-
DEFLECTION

1 1 1 1 1 1
.10 .20 .40

LATERAL MID-HEIGHT DEFLECTION. S(IN.)

Fig. 8. Load-deflection for Specimen No. 3 Fig. 11. Typical column buckle configuration
20
ENGINEERING JOURNAL / AMERICAN INSTITUTE OF STEEL CONSTRUCTION
 P AXIAL LOAD TENSION

LOAD AT FRACTURE = 1070 KIP

-2000K

P AXIAL LOAD COMPRESSION

Fig. 15. Load-deflection hysteresis for Specimen No. 3

Fig. 12. Typical local buckle P AXIAL LOAD TENSION

LOAD AT FRACTURE = 2 0 0 KIPv

i P AXIAL LOAD TENSION

1000 K
LOAD AT FRACTURE = 5 3 2 KIP)

500 K

8000KN
^ I 7 6 0 M A X LOAD - 2000 K
150 l25 DEFLECTION 3.5 IN
S{mm)
LATERAL DISPLACEMENT I
IN WEAK AXIS DIRECTION P AXIAL LOAD COMPRESSION

Fig. 16. Load-deflection hysteresis for Specimen No. 4

8 0 0 0 KN
^ 1650 KIPS AT 3 25 IN. DEFLECTION 2000 K

* P AXIAL LOAD COMPRESSION

Fig. 13. Load-deflection hysteresis for Specimen No. 1

J P AXIAL LOAD TENSION

LOAD AT FRACTURE = I 8 0 0 K - - 2000 K

,
|P AXIAL LOAD TENSION FAILURE AT
INADEQUATE PIN
8000KN I

PLATE. I500K
LOAD AT FRACTURE = 8 0 0 Ks

- 1000 K
_> 4 0 0 0 KN lOOOK

- 500 K
500 K

^(IN.) 3 1 ( ^ 2 1 l| 0 ll |2
S(mm) i'75 '50 l25 0 '25 '50
B(mnn)200 I5( 50 100 150
1 LATERAL DISPLACEMENT 8 ,
11 IN WEAK AXIS DIRECTION LATERAL DISPLACEMENT S
500K IN. WEAK AXIS DIRECTION

[1 4 0 0 0 KN

i-
- 1000 K •lOOOK

f •I500K

1655 KIP AT 3.0IN DEFLECTION - 2000 K

^ 8 0 0 0 KN •2000K

' P AXIAL LOAD COMPRESSION

t P AXIAL LOAD COMPRESSION

Fig. 14. Load-deflection hysteresis for Specimen No. 2 Fig. 17. Load-deflection hysteresis for Specimen No. 5
21
FIRST QUARTER / 1977
 the load may increase, but with corresponding large lateral
30-. LEGEND deformations. It is difficult to define a similar point for
1.0 ^ ^ 1 9 6 6 CRC STRENGTH CURVE o K =05
• K=0.65 Specimen No. 3.
X AT BUCKLING
..I»----12,. Several factors contributed to the high values of column
buckling load in these tests, especially as this pertains to the
spliced columns. Of most importance, the Kl/r ratios were
^ A I S C Fo/Fy ^ \
DESIGN CURVE \ ^ small, which tended to reduce the effect of splicing.
0 50 Moreover, the specimens tested were quite straight. The
maximum initial sweep was 0.094 in. (»//2000) at center
span in Specimen No. 2, and the average initial sweep was
0.05 in. (»//3600). This is considerably less than the de-
viation from straightness allowed by the AISC Specification
n 1 ^ \ ^ \ , (//1000 = 0.18in.).
In reviewing the tension test data, the most interesting
Fig. 18. Theoretical and experimental column strengths result was the widely varying loads at fracture. Table 1
gives the fracture loads and the average weld sizes, which
were measured after fracture occurred. The spread in
fracture loads was from 200 kips for Specimen No. 4 to
Discussion of Results—Figure 18 compares the perfor- 1070 kips for Specimen No. 3. Some of this discrepancy can
mance of the five test specimens with theoretical column be attributed to the extent of overall and local buckling
strengths as predicted by the CRC Curve.^ It is important which developed at the splices during compression, as well
to note that the maximum compressive force does not nec- as the actual size of the welds and their general quality.
essarily occur at the maximum deflection.
In reviewing the data in Table 1 for the column com-
pression tests, it is seen that the maximum compressive CONCLUSIONS
forces, for Specimens 1, 2, 4, and 5, which have the same The results of these tests indicate that the lack of perfect
slenderness ratios, are essentially the same. This is clearly contact at compression splices of columns may not be im-
brought out by the data points in Fig. 18. portant, provided that the gaps are shimmed and welding
The columns were tested in a "flat-ended" condition, and is used to maintain the sections in alignment. It is noted that
during the application of the compressive load the head of the spliced columns, rather independently of the gaps in-
the testing machine was locked in position. This testing troduced at mid-height, behaved very much like the un-
procedure tends to make the column behave as something spliced column, and their column strengths were essentially
close to a fixed-end member, which may be noted from Fig. the same. It must be pointed out, however, that the Kl/r
11. However, complete fixity of large members is very ratio of the columns tested was low, being on the order of
difficult to achieve. The columns had square milled ends, 30. A slenderness ratio in the range 25-30, however, is not
but required shimming to achieve uniform bearing against at all uncommon in actual buildings. It may be desirable
the testing machine platens. For this reason, two sets of to extend the research to include columns with higher
points corresponding to the experimentally determined slenderness ratios, to determine the effect of the splicing on
maximum load points are plotted. For the one set K is as- a wider range of columns. The significance of the splice
sumed to be 0.5, and for the other it is taken as 0.65. It is location for longer columns than those tested may have to
believed the latter value is more representative of the con- be considered.
ditions of these experirnents. The data obtained on the behavior of initially buckled
It can be seen from Fig. 18 that there is only approxi- members, subsequently loaded in tension, should prove
mately 7 percent difference between the lowest failure load useful in evaluating the behavior of such members in a-
and the highest. The higher strength of Specimen No. 3 was seismic design. Columns and braces in buildings, as well
due to its lower Kl/r ratio, yet on the dimensionless plot of as in offshore structures, may buckle in compression under
Fig. 18 the results are in reasonable conformity with those extreme excitation and then become subjected to tensile
for the other four specimens. Because of the closeness of the forces. Some data from experiments simulating this type
test points, it is difficult to discern any pattern or correlation of behavior on small members are available.^ This inves-
between the direction of the splice (weak or strong axis), tigation provides data on the behavior of full-size members.
the size of the gap, and the column strength. It is to be noted that for the two loading conditions of
The loads at which each specimen had sustained a compression and tension, the boundary conditions were not
0.10-in. weak axis lateral deflection was defined as the identical.
column "buckling" load, and the corresponding values are If the designer anticipates tensile forces developing in the
shown in Fig. 18 by the symbol "x." This arbitrary buck- spliced columns, one must be certain that the specified weld
ling criterion was chosen on the basis of the behavior of size is obtained. As ultrasonic inspection at partial pene-
Specimens Nos. 1, 2, 4, and 5, as seen in Figs. 6, 7, 9, and tration welds is questionable, visual inspection at first root.
10. This "buckling" load represents a point beyond which pass is recommended.
22
ENGINEERING JOURNAL / AMERICAN INSTITUTE OF STEEL CONSTRUCTION
 ACKNOWLEDGMENTS 3. Hayes, John M. Effects of Initial Eccentricities on Column Per-
formance and Capacity ASCE Structural Division Journal, Paper
W. A. Milek of AISC provided the initial encouragement No. 1440, November 1957.
and support for carrying out this investigation in a virtually 4. Popov, Egor P., Roy M. Stephen, R. Philbrick Capacity of Col-
unexplored area of research. R. Philbrick, a graduate umns with Splice Imperfections Earthquake Engineering Re-
student of the University of California, Berkeley, assisted search Center, Report. No. EERC 76-21, September 1976.
with the experiments and preparation of the initial report. 5. Bruce G. Johnston, Ed. The Column Research Council Guide
to Design Criteria for Metal Compression Members 2nd Ed.,
Dr. R. Bjorhovde of AISC offered numerous suggestions
1966.
w^hich helped to clarify the material presented. The authors 6. Beedle, Lynn S. and Lambert Tall Basic Column Strength ASCE
are most grateful to these individuals, as well as to AISC, Transactions, Vol. 127, Part II, 1962.
AISI, and NSF for providing the necessary funds to carry 1. Tebedge, N., G. Alpsten and L. Tall Residual-Stress Measure-
out the work. ments by the Sectioning Method Proceedings of the Society for
Experimental Stress Analysis, Volume XXX, Number 1.
8. Tall, Lambert and Goran A. Alpsten On the Scatter in Yield
Strength and Residual Stresses in Steel Members International
REFERENCES Association for Bridge and Structural Engineering, Symposium
on Concepts of Safety of Structures and Methods of Design, Lon-
1. Quality Criteria and Inspection Standards American Institute don, 1969.
of Steel Construction, 1971. 9. Kahn, Lawrence F. and Robert D. Hanson Inelastic Cycles of
2. Manual of Steel Construction Seventh Edition, American Institute Axially Loaded Steel Members ASCE Structural Division
of Steel Construction, 1970. Journal, Paper No. 1211, May 1976.

23
FIRST QUARTER / 1977