7/28/2019 SRS-135

1/12

" i, ,"==-:;-c - : -C, ' ' '_- , : - , - - . , . ~ - , . , . " . " ' ' ' ~ "

..,., ,- ... .-.', '... :.- - . -- ...... ..- '- ..

BEHAVI R F WELDED BUILT . UPBEAMS UNDER REPEATED LOADS

-... .....By

J. E. StallmeyerW. H. Munse

andB. J. Goodal

Reprinted from the Welding JournalJanuary 1957

UNIVERSITY OF ILLINOISURBANA, ILLINOIS

7/28/2019 SRS-135

2/12

EHAVI EL E UIL-UEAM UN ER REPEA E L A

BY J. E. STALLMEYER, W. H. MUNSE AN D B. J . GOODAL

REPRI NTED FROM

T E EL IN J URNALJANUARY 1957

7/28/2019 SRS-135

3/12

BEU

AVI R F WELDE UILT-UP BEAR REPEATE L A

Fatigue tests on built-up beams indicate that thepresence of a splice materially reduces the fatigue strengthof a welded beam. A difference in fatigue strengthis also revealed for the different splice configurationsBY J . E. 5 TAL L ME Y E R, W. H. M UN 5 E AND B. J . GOO D A LSUMMARY. Flexural fatigue tests havebeen conducted on small all-welded beamsfaln;icated from A373 steel . All beamswere fabricated by manual arc weldingusing E7016 electrodes an d a back-stepping welding ~ r o c e d u r e . S e v ~ r a l f i ~ l dsplice configuratIOns were used il l the i l l -vestigation, and studies made of the diffcrence in their modes of failure as well asthe difference in their fatigue strengtb15.In addition, the program included fatiguetests on control specimens to determinethe fatigue strength of the A373 steel asreceived from the mill and also with buttwelded joints, either parallel or perpendicular to the direction of stress.

Table l-Chemical Composition of Steel Plates

The fatigue tests on the beams indicatethat the presence of a splice materially reduces th e fatigue strength of a weldedbeam. A difference in fatigue strength isalso revealed for th e different splice configurations.IntroductionObject and Scope of InvestigationThe increased use in recent years of allwelded girders along with the variety ofprocedures used by different fabricatorsand designers for making splices andattaching stiffeners has resulted in theexistence of a large variety of weldedbridge details. With the limited numberof tests available and because of relatively limited experience, it is no tpossible to specify which of the splicingprocedures is best for .maximum resistance to repeated loads. In order toevaluate some of the more commonsplicing procedures, this investigationon all-welded beams was undertaken.The primary purpose of the initialphase of the investigation was to studythe effect of typical butt-welded fieldsplices on the flexural fatigue strengthof all-welded built-up beams and girders.Four splice types, derived from twobasic splice configurations either withJ. E. Stallmeyer is Assistant P r o f ~ s s . o r , an.d W.H. Munseis Research Professor, ClVIl Engmeering, University of Illinois, Urbana, Ill.; B. J.Goodall is Engineer, Inland Steel Co., ~ a s tChical!o, Ind., and formerly Research AssoClate,University of Illinois, Urbana. Ill.Presented a t 1956 AWS National Fall Meetingin Cleveland, Ohio, October 8t h to 12th.

Platethickness, Chemical content, %Steel in . C Mn P S Si CuStructural 1/ 4 0.22 0.32 0.012 0.035Structural 3 /4 0.28 0.44 0.016 0.029Structural 1 0.28 0.44 0.016 0.029A373 3 /16 0.23 0.63 0.022 0.031 0.030 0.17A373 1 0.21 0.60 0.030 0.030 0.053 0.20

Table 2-Physical Properties of Steel Plate (8-ln. Gage Length Tensile Coupons)YieldThickness, strength,Steel in. psi

Structural 3/1S 44,800Structural 1/, 37,100Structural 3/. 35,300Structural 1 35,000A373 3/1s 38,800A373 1 34,600

or without cope holes, were investigated while beams without splices andbeams with cope holes only were included in the program as control specimens.Preliminary tests were carried out onspecimens fabricated from a typicalstructural grade steel to study theeffect of varying the thickness of theweb and to develop sound welding andfabricating procedures. In addition tothese preliminary tests, control specimens of an A373 steel were tested todetermine the fatigue strength of plainplate and butt-weld specimens.Description of SteelsThe majority of the tests reportedherein were conducted on specimensfabricated from A373 steel ; however,the preliminary program was carriedout on specimens made from a structural grade steel. This steel was takenfrom stock which was available in thelaboratory, while the A373 steel was

3

Ultimate Elongation Reductionstrength, in 8 in., of area,ps i % %

67,800 25.5 52.750,700 32.8 66.064,600 30.3 51.361,200 34.2 63.764,800 29.6 58.067,000 28.0 48.6

ordered for use on this program. Thechemical composition and physical properties of materials used are given inTables 1 and 2. The chemical properties were determined from check analyses and the physical properties weredetermined from standard flat specimens cut from the parent plates.All of the A373 steel met both thephysical and chemical requirements ofASTM Designation A373-54T.Preparation of Beams Without Splices

All beams used in the investigationwere of uniform cross section throughouttheir length. The flange and web plateswere cut from stock with a dual-torchoxygen cutting machine. Where plateswere taken out next to a sheared edge,a strip adjacent to the sheared edgeand about 1 in. wide was discarded.Thus, all edges of plates used in the testspecimens were flame cut. Any plateswhich had severe notches. or o t h e ~ de-

7/28/2019 SRS-135

4/12

--t-It-HH-- - ---- - -+-'r--t--p= 0 TO COMPo 112,000 LB.

'" ,I'co

Fig. 1 200,OOO-lb. Wilson fatigue testing machine adapted to test flexuralspecimensfects, due to cutting or handling of the

m a t e ~ i a l , were discarded. Plates withminor defects were used, bu t the severityof such defects was lessened by grinding them to a smooth transition. Allslag an d burrs on all edges of the platesand mill scale in the region of the weldwere removed.In assembling the specimens, aflange plate was first laid on the flangeof an H-beam. (A stiff H-section,approximately as long as the specimen,was used. Such a section could resistthe clamping forces required to alignthe parts and t.o bring them into

Type A

IType 0

Fig. 2 Splice types

proper contact with each other, without undergoing appreciable deformation.) The web plate was then carefully aligned, securely clamped andtacked to the flange by full-size filletwelds spaced about 16 in. apart. TheT-section thus formed was turned over,set on the second flange plate, clampedin position and tacked. All tack weldswere cleaned and the specimen wasready for deposition of the web-flangefillet welds.Details of the principal fillet weldingsequence are given in Fig. 3. However,this sequence is slightly different than

::

Type B

I I.;-:;'".......,

Type E

4

the initial sequence used in fabricatingthe specimens in the preliminary seriesand a number of the earlier specimensof the tests reported herein. The twoprocedures were identical in all bu t twoparticulars. In the initial sequence thewelder changed location after eachdeposition of weld. This was changedin the later tests so that three electrodelengths were deposited by the backstep method before the location ofdeposition was changed. Also, a nineinch release length allowed in the initialsequence was eliminated in the latermembers.Distortions which were found insome of the first test specimens resultedfrom the cutting of the stock and fromthe welding. The use of the dual-torchcutting machine practically eliminatedthe first of these two items, but the distortion due to welding varied with thewelding sequence. The distortion inthe specimens welded in accordance withthe initial welding sequence was negligible while in the specimens welded withthe later welding sequence the distortion was noticeable. However, thelatter sequence is more nearly thatfound in use in the large fabricatingshops.Fahrication of Spliced Beams

The two parts of a spliced beam wereobtained in two different ways. In onecase the two halves were fabricatedindividually in accordance with thewelding procedure outlined for beamswithout splices. In the second case the

Type C

I CJ

Type F

7/28/2019 SRS-135

5/12

12 10

'I 'I 'I '1 I"

I. I. If/ I:1 ,I:1.I 7 9BOTTOM FLANGE

12 8 4 ...13"r- 2 6 10

J .1 T r II I I I I I I I I r I I T I T TJ j I I I I I I I I I f T T I I 1I I 7 I :5 I I I 5 I 9I T

TOP FLANGEWelding:, \A" Fillet Weld. dia, E7016 Electrode-175 amosNote:

Inside arrows indicate the deposition' of individual electrodesOutside arrows and the numbers indicate the welding sequence

fig. 3 Principal fillet welding sequence

spliced beams were fabricated from partsobtained from specimens which hadalready been tested. This was carriedout in the following way.All specimens were fabricated 2 ft6 in ...longer than the span on which theywere testp-d and provided a I ft 3 in,overhang at each end. After failurehad occurred, the fracture and a portion

FLANGE

of the beam on each side of the fracturewere removed. This left two pieceswhich were of sufficient length to provide an additional spliced specimen.The overhangs had been subjected tono load and, therefore, when turned endfor end and rewelded, these two piecesprovided an additional spliced beam.

The ends of all flange plates were

ICOPE HOLE CUTTING SEQUENCE

fLANGE "'l

. ~::: on... ...i:> ..... Q.

~ ' : = - . I - - - - - + - - - I ' J \- -

!FLANGE

o

. ~ ~-It

...o:::...Q.8

I

104

DIRECTION OF WELD PASSES

113

I1

WEB FL.ANGE

bevelled with the oxygen cutting machine fora double-V butt joint havingan included angle of 60 deg. Since theweb materi al was only 3 /16 in. thick,the bevels for the single-V web jointswere made with a portable disc grinder.Two procedures were followed incutting cope holes. In the initial welding sequence, the holes were cut to fullsize (3/4-in. radius) before the specimenwas assembled (see Fig. 4). In theprincipal welding sequence a semielliptical hole [with 2-in. major axis(flange direction) and I-in. minor axis]was cut before the specimen was assembled. After specimen was assembledand the web splice completed, the semielliptical hole was made into a semicircular hole with a I-in. radius (seeFig. 5). This removed the ends of thewelds in the web splice.Designation of Beams

The numbering system used for thetest specimens has been chosen so thatthe various fabrication and test conditions can be identified from thespecimen number. The identificationnumber of the initial or preliminarybeams consisted of a capital letter and anumber, e.g., Specimen D-l. The corresponding specimen of the later testswas identified as Specimen D-Ia.Specimens fabricated with the principalwelding sequence have an identifying R.in the specimen number. .Test Procedure

All of the fatigue tests reported herein

COPE HOLE CUTTING SEQUENCE

II

2

II

WEBD!RECTION OF WELD PASSES

,I

- -I

I

fig. 4 Initial welding sequence for butt splices fig. 5 Principal welding sequence for butt splices

5

7/28/2019 SRS-135

6/12

were conducted in 200,000-lb Wilsonlever-type fatigue testing machines.A schematic diagram of the machine,adapted to test flexural members, isshown in Fig. 1. Essentially, its operation is as follows: The force is derivedfrom a variable throw eccentric mountedon the end of a shaft driven by an electric motor. As the eccentric revolves itimparts to the overhead walking beam,which is pivoted about a bearing located in the vertical support post, apumping motion. This pumping motionof the walking beam is in turn transmitted to the specimen through theloading yoke, thereby subjecting thespecimen to a cyclic flexural stress.Installation of the specimen in themachine was generally a simple operation. Longitudinal and transverse centerlines to locate the loading and reaction blocks were marked on the specimen before it was se t into the machine.The specimen was then set on the reaction blocks, centerlines on the specimenwere matched wi th their respective lineson the blocks, the blocks were clampedin place, and the loading yoke waslowered into position to set the load.The specimens were tested on a spanof 8 ft 6 in. with two concentrated loads:12 in. apart, placed symmetricallyabout the centerline of the span. Thestress in the extreme fiber of the beamswas calculated by the usual straightline formula, 8 = Mell, using a momentof inertia based on the actual dimensions. All tests were conducted on acycle in which the stress at the criticalsection, on the extreme fiber, variedfrom a minimum of 1000 psi to a maximum of 30,000 psi. This one stresscycle was used throughout the study inorder that each of the variables included in the program might be evaluated under the same loading condition.Throughout this report the specimenwas considered to have "failed" whenthe center deflection of the beam, because of the fatigue crack, was 0.05 in.in excess of the deflection at maximumload on the uncracked section. Thiswas found to occur, in general, whenapproximately one-half of the flangearea had fractured.Test ResultsBeams Without Splices

Four beams without splices weretested in this study. In addition tofurnishing the fatigue strength of plain,welded built-up beams, the purpose ofthis series was to establish a datum tofacilitate the comparison of data gathered in the program. The results of thetests are presented in Table 3.In view of the magnitude of scattergenerally obtained in fatigue investigations, the results of the first three specimens are somewhat unusual. Suchconsistency is rarely encountered infatigue results.

7/28/2019 SRS-135

7/12

Table 4-Description of Initial Welding SequenceStepNo.1A

PassNo. RemarksJig web and flange andclamp2A Deposit fillet weldsStart butt splice welds with web in vertical12345678910

1112131415161718

12345678910

11121314

positionFlange weld (see Fig. 4)

Turn beam over and backchip passes No. 1 andNo.2

Turn beam over and complete flange splicesSet beam with web inhorizontal positionTurn beam over and backchip pass No. 13

NOTE: All welds were made with reversed polarity, in the flat position andwith AWS-ASTM E7016 electrodes.Several factors could have contributedto the lower fatigue life of SpecimenA-4R. First, this was the only specimen prepared in accordance with theprincipal welding sequence. However,it seems doubtfur thai" the change inwelding sequence could ve affectedthe fatigue life to this extent. Second,an examination of the fracture surfacerevealed that a very minor metallic in-clusion existed at the root of one of thefillet welds-no inclusions were observed in any of the other three speci-mens. Closer examination of the fracture surface revealed, however, that thefracture initiated at the root of the weldon the opposite side, where no inclusionwas observed. Consequently,'theredoes no t appear'to be any justifiable explanation for the short life of the mem-

Photographs of typical fracture loca-tions and fracture surfaces are shown inFig. 6. In Specimens"A-2, A-3 andA-4R the failure initiated at the root ofthe fillet weld in the region' of a changeof electrode, bu t in the case of SpecimenA-I, failure initiated at the edge of theflange. The .fracture surface of thislatter specimen showed no unusual defects or stress raisers at the point ofcrack initiation.The fractures of Specimens A-I andA-4R occurred in the region of purebending, between the load points, 2in. and 4 in. from the centerline,'respec+,ively. However, Specimens A-2 andA-3 failed in a zone of combined shearand flexure, 9 in. from the centerline.

In one case, Specimen A-3, there

a major lamination in the plate used forthe tension flange; the material appears to be made up of two half-inchthick plates. The fatigue life was,however, unaffected. I t appears that,even though the defect is large, it haslittle or no effect on the life of the specimen if it is oriented in a directionparallel to the applied stress.Beams with Cope Holes Only

Generally, the quality of double-Vbutt joints in the flanges of beams andgirders is comparable to that found insimilar joints in plain plates. However,special precautions are often requiredto make it so. One area in particular,because of its inaccessibility, is inducive to unsound and imperfect welds;the area at the junction of the web andthe flange. A method long used in commercial practice to make the area moreaccessible has been to cut out a semicircular hole (cope hole) in the web ofthe beam immediately above the flangesplice. By providing the cope hole it ispossible, by bringing the electrodethrough the hole, to start the weld beadin a region where the ends of the weld areaccessible for proper cleaning beforethe adjacent pass is deposited. T h e r ~can be little doubt that this procedureresults in more sound welds. However,there is some question concerning theseverity of the stress concentrationcreated by the cope hole. The purposeof this series of cope hole tests was toevaluate the effect of the cope hole.

On the basis of the tests of thisseries, presented in Table 3, the average life of plain beams with cope holesonly is found to be about 29.2% of theaverage life of plain beams and in allcases, the crack initiatE).. ( at the top ofthe cope hole at the

7/28/2019 SRS-135

8/12

Fig. 6 Typical fractures of beams without splices

Fig. 7 Typical fractures of spliced beams8

7/28/2019 SRS-135

9/12

Table 6-Results of Fatigue Tests on Plain Plate, Transverse and LongitudinalButt-Welded Specimens, . . - - - - -Fa t i gue strength*'----..

SpecimenF-1F-2F-3

F-1TF-2TF-3T

F-1LF-2LF-3L

Stress, psi36,00036,00036,000

25,00025,00025,000

27,00025,00027,000

Cycles f07' A373 steel, A 7 steel,failure, 10 3 fMoo.oDO h. 00. 000Plain plate specimens1046.0 32,100964.3 31,6001751.6 35,200

Avg 33,000 34,600Transverse butt-welded joints859.7 22,400978.8 22,800882.1 22,400Avg 22,500 23,800+Longitudinal butt-welded specimens1623.1 26,2003723.3 2 5 , 0 0 0 ~3656.0 2 7 , O O O ~

Avg 26, 1 0 0 ~ 26,300* Results of tests oIl A7 steel taken from bibliography Reference 3.

k

0.18

0.13

0.13

Splice Type C, which is essentially ashop splice because of the manner inwhich it is fabricated, has been included in this paper to show the advantage in strength of a ('partial"splice over a "complete" splice. Thesplices in the flanges were made beforethe specimens were. assembled. Thereinforcement on the inside of theflange was ground flush with the plateso that it would no t interfere with the

web and the fillet weld at the junctionof the web and the flange.Splice Types A and D

The basic configuration of these twosplice types was a complete splice(joints in both flanges and web) madein the same cross section. The specimens were identical with the exceptionof the cope holes provided in Type A.Fatigue lives determined for fourspecimens of each type are given in

Table 3 along with the average of thetest results. Based on the average lifeof the first three beams without splices,the life of the beams of splice Type Ais only 15% as great while the life ofType D was 24% as great. In only onecase was there any overlapping of thetest results of the Types A an d Dsplices. In all other instances theType D splices exhibited the greaterresistance to repeated loads.Metallurgical examinations of thefractures revealed that Specimens A-1aand A-3a failed at the edge of the copehole in the base material of the flange.This type of failure is undoubtedlycaused by the stress concentrationeffect of the cope hole. SpecimenA-2a, which had the lowest fatigue life,had poor root bonding of the butt weldin the flange and started to fracture atthis point of poor bonding. The fracturein Specimen A-4a initiated at the weldmetal-base metal interface in the flange.The results of the fatigue tests are:however, fairly consistent and wouldindicate that the presence of the copehole is practically as harmful under repeated loadings as the presence of asevere stress concentration within theweld itself.A metallurgical examination of theType D splices revealed that all of thefailures initiated at the weld metalbase metal interface in the flange. ,I nSpecimen B-2a there were two failures.On one side of the butt weld in the flangethe specimen failed completely throughthe flange, while on the other side there

Table 7-Spedmen Description an d Summary of Test Results of the Preliminary Series,..... I ," 6 H J.... f-'r-- 5 \ )+1/4"I2 " I , 3 ,- ~

I _. ..8 _II I .... .,11 f 0." I80' - 0" I 11_ 0"

Maxstress in Cyclesext. for LocationtSplice * fiber, Stress in failure, ofSpecimen d, in. b, in . tf , in. tw, in . tr/tw type 1000 ps i weld, ps i 10 3 fractureP-1 111/2 5 3/4 3 1:0 None 30.0 2440 830.1 1P-2 111/2 5 3/4 1/ 4 ,3.1 . None 30.0 2340 1037.3 1

" P-6 111/2 5 3/4 3 /16 4.3 None 30.0 2340 1793.4 5,', P-7 12 5 1 3 /16 5.3 None 30.0 3055 2164.3 3P-3 111/2 5 3/4 1/ 4 A 30.0 2340 350.9 4; P-4 111/2 5 3/4 B 30.0 2340 450.7 4P-5 111/2 5 3/4 1/ 4 C 30.0 2340 563.6 2{P-2a 111/2 5 3/4 1/ 4 D 30.0 2340 1357.5 4a... P-3a 111/2 5 3/4 1/ 4 D 30.0 2340 571.9 4a'1 P-6a 111/2 5 3/4 3 /16 D 30.0 2340 767.5 4aJ P-7a 12 5 1 3 /16 D 30.0 3055 459.4 4b* See Fig. 2 for splice types.t 4-Toe of fillet weld at edge of cope hole. 4a-Through butt joint. 4b-Along edge of weld reinforcement.

9

7/28/2019 SRS-135

10/12

Fig. 8 Typical fractures of spliced beamswas only a part ial failure. Typicalfracture locations and fracture surfacesare shown in Fig. 7.A hardness survey across the interface of base metal and weld metal ofSpecimen B-2a revealed two points ofpeak hardness in the heat-affected zoneof the base metal. This area of highhardness can be interpreted to meanthat this region had low ductility compared to the oth er regions. This discontinuity in the microstructure pro bably acted as a stress raiser and led toan earlier fa ilure than would be expected if the material in the beam had amore homogeneous microstructure.Splice Types B ancl EIn this case the splice was a completesplice made at three points along thebeam length. Join ts in the flanges weremade at sections which were 4 in. oneither side of the joint in the web.The maj or variable again was the copehole. Splice Type B was made withcope holes and Type E was made without cope holes.The fatigue test results indicate thatthe Type B specimens had fatigue liveswhich were approximately 84% of thefatigue life of the Type E specimens.These results follow the pattern indicated by the previous results in thatthe splices with cope holes are consistently weaker than the splices with-

out cope holes. Based on the averaO"elife of the first three plain beams, thelife of the splice Type B was 23% andthat of Type E was 27% as great. .In splice Type B the fillet weld wasterminated in two different ways at thecope hole. On one side of the cope holethe fillet weld was carried all the wayaround the cope hole, whereas, on theother side of the cope hole, the weldwas brought just up to the cope hole.The metallurgical examination showedthat failure occurred on the side of thecope hole where the fillet weld was notcarried around. It would appear thatthe fatigue lives would be somewhatlonger if the fillet weld had been carriedaround on both sides of the cope hole.However, in at least one case there wasevidence that a small fatigue crack hadstarted on the side of the cope holewhere the fillet weld had been carriedaround. It would seem, therefore, theincrease which would be obtained bythis procedure would be relativelysmall.An examination of the fatigue failuresin the Type E specimens revealed thatD-2a and D-3a had poor root bondingof the weld in the butt weld in the flange.This type of failure has already beendiscussed for Specimen A-2a. In thecase of Specimen D-la the failure occurred at the weld metal-base metalinterface of the flange weld and was

10

similar to that reported for B-2a.Typical fracture locations and fracturesurfaces are shown in Fig. 8.Splice Type C

The Type C splice, which is' a shopsplice, is considerably stronger thanany of the other splices tested. Based '.on the average life of the :6rst three :.... ;. .plain beams, the life of this type of -- , '-splice was 44% as great . This is no t 0:::unexpected since the welds in theflange plates can be made under betterconditions and undoubtedly result in ...... ...1"higher quality welds.Metallurgical examination revealedthat all failures initiated at the weldmetal-base metal interface in the flange. .In one case, however, an additional- . "fatigue crack had started at a change ofelectrode in the fillet weld just 3 in.

from the flange splice. A hardness sur-vey revealed that this latter failure oc-curred at a location of minimum hard-ness in the weld metal. This hardnessvalley was due to the heat treatment ofthe first weld deposit by the second elec-trode deposited. As in the case of apeak hardness, this hardness valleymay act as a stress raiser and, alongwith the geometrical notches in theweld, may cause a fatigue crack toinitiate. .Results of Control Specimen Tests

No data were available from which

7/28/2019 SRS-135

11/12

the fatigue strength of the A373 steelcould be determined and compared toresults from another structural gradeteel. For this reason several flat-plateatigue specimens were prepared andested on a zero to tension stress cycle.e specimens conform to those whichave been used in previous investigaions at the University of Illinois onthe fatigue strength of welded joints.The dimensions of the specimens

the fatigue strength9.e welding procedure, as well asthe dimensions of the specimens, wasthat which was used inrevious investigations. For a comlete description of the welding pro

ort of the previous investigation. lThree specimens of each of the threeypes to be tested were prepared: three .lain-plate specimens, three specimensith a qutt weld perpendicular to theof the applied stress, and threebutt weld parallel tohe direction of applied stress. Alled condition with the reinforcement

The results of the tests, in terms ofe presented in Table 6. In additionthe presentation of the test data ande fatigue strength corresponding to

the average fatigue strengthin another steeled in a previous paper.3 The

e test results in accordance with thee outlined in the reference paper.e two steels were compared only at

it was thoughtthe extrapolation to 100,000 cyclesto provide reliable valuesdata available.The fatigue failures in the A373 steel

had been tested previously. Ine case of the transverse butt jointe failures initiated at the edge of thethe parent plate .e fracture took a random path

the weld metal and base metalbetter than 90% of the fracturein the base metal. The fail

in a region of the outsidein electrode hadmade. This point of initiation of

e fatigue failure corresponds to theof weakness which was observedOn the basis of these tests, it may be

that the strength of A373is approximately the same as thate other steel when the tes ts are conthe same type of specimen.

There is certainly no marked differencein the fatigue strength in any of thethree specimen types which were tested.Beam Specimens Fabricated from an Ordin-ary Structural-Grade Steel

As mentioned earlier in the paper, thefirst beams tested in this investigation were fabricated from a steel whichwas available in the laboratory. Thepurpose of these tests was to developa suitable specimen into which thevariables to be studied could be incorporated and also to work out the fabricating and testing procedures.The results of the preliminary testsconducted in this part of the investigation are reported in Table 7 along withthe dimensions of each particular specimen and the type of splice, if any, whichwas included in the test specimen.Although all failures in the beams without splices initiated in the same region,the initial direction of propagationseemed to depend on the ratio of flangethickness to web thickness. In Specimen P-1, for example, the crack progressed directly into the flange. In

specimens where the ratio of flangethickness to web thickness was higher,the crack, upon formation, propagatedinto the web. Not until the crack hadprogressed about 2 in. into the webdid it start to spread into the flange. Inno case was the failure sudden; but,the rate of propagation was muchgreater through the flange than throughthe web. The fatigue life reported forSpecimen P-1 is 1.5% higher than thelife when the first visible crack wasobserved, whereas the fatigue life forP-7 is 10.8% higher than when thefirst crack was observed.It is interesting to note that in thebeams without splices, the fatigue livesincreased consistently as the ratio offlange thickness to web thickness increased. Of course, with the limitednumber of tests, no general conclusionscan be drawn bu t the results are certainly encouraging.

The failures in the spliced beams fabricated from stock steel were similarto those which were later found in thebeams fabricated from A373 steel.

0 0 0 0 II 0 0 0 00 0 0 :v[ 0 0 00 0 0 0 0 0 0 0

4 ' - 0 "

o. PLAiN PLATE

5"'

0 0 0 0 Ii 0 0 0 00 0 0 0 0 00 0 0 0 0 0 0 0

I. 1 9 ~ "

b. LONGITUDINAL BUTT WELD

0 0 0 0 II 0 0 0 00 0 0 0 0 00 0 0 0 0 0 0 0

c. TRANSVERSE BUTT WELDFig. 9 Details of butt-welded joints

11

7/28/2019 SRS-135

12/12

Specimen P-3 and P-4 failed in thetension flange at the toe of the filletweld around the edge of the cope holeand propagated through the heataffected zone parallel to the transversebutt weld in the flange. The failuresof the other spliced beams initiated atthe junction of the web and flange inthe butt joint of the tension flange.Conclusions

The following conclusions are basedon the results of the limited experimental program discussed herein. Furthertests on similar members of otherstructural grade steels and fabricatedby other welders would probably modifysome of the conclusions somewhat.However, it is believed that these testsprovide the practicing engineers withrealistic values for the fatigue strengththat may be expected from properlyfabricated welded beams.1. In beams without splices, thefatigue fractures initiate in the filletweld at the junction of the web andthe flange in a region of change of electrode. Consequently, the use of continuous welding procedures may be expected to provide an increase in theresistance of such members to repeatedloads. With the exception of onespecimen the test results were extremely consistent.2. The average life of the plainwelded beams with cope holes only was29% of the life of the beams withoutsplices. All failures in this groupinitiated at the toe of the fillet weld

around the edge of the cope hole.Splices with cope holes failed after afewer number of cycles than the corresponding splice fabricated without copeholes. Splices with cope holes failedat the toe of the fillet weld around theedge of the cope hole. In splices without cope holes fracture generally occurred at the weld metal-base metalinterface at the butt joint in the flange.Thus, the cope hole provides a stressconcentration which is more severethan the other notches in a splicedbeam and initiates the fatigue failures,although the life of such members isnot greatly different than that of amember without cope holes.3. Splice Type C (a shop splice)had a fatigue life considerably greaterthan the other types of splices.4. The fatigue strength of beamswithout splices was found to increaseas the ratio of flange thickness to webthickness increased and the greatestfatigue resistance was obtained formembers with relatively thin webs andthick flanges. In addition, specimenswith higher values of flange-to-webthickness ratios required a greaternumber of additional cycles for failureafter a visible crack was observedthan those with low values. This is theresult of a difference in original direction of propagation of the fatiguecrack.

AcknowledgmentsThe program reported herein is partof a cooperative research project con-

ducted in the Engineering ExperimentStation of the University of Illinois,Department of Civil Engineering, andsponsored by the Department of Commerce, Bureau of Public Roads. Thework constitutes a part of the structural research program of the Department of Civil Engineering underthe general direction of N. M. Newmark, professor and head of the Department. The metallurgical investigations were conducted by Mr. C. A.Robertson and Mr. R. Falck under thedirection of Prof. W. H. Bruckner.

The Fatigue Committee of the vVeld-ing Research Council acted in an advisory capacity in the planning of thisprogram. However, the views expressed in this paper are those of theauthors and do not necessarily reflectthe views of the sponsor or the Fatigue Committee.Bibliography

1. Stallmeyer, J. E. , Nordmark, G. E.,Munse, W. H. , an d Newmark, N. M. , "FatigueStrength of Welds in Low-Alloy Structural Steels,"THE WELDING JOURNAL 35 , (6), Research Suppl.,298-s to 307-s (1956).2. Wilson, W. M., Munse, W. H., and Snyder,I. S., "Fatigue Strength of Various Types of ButtWelds Connecting Steel Plates," University ofIllinois Experiment Station Bulletin 384 (March1950).3.' Nordmark, G. E., Shoukry, Z., an d Stallmeyer, J. E., "Fatigue and Static Properties ofWelded Joints in Low Alloy Structural Steels,II," Structural Research Series 114, Universityof Illinois.4. AMERICAN WELDING SOCIETY, "StandardSpecifications for Welded Highway and RailwayBridges," 1956.5. Wilson, W. M., "Flexural Fatigue Strengthof Steel Beams," University of Illinois Experiment Station Bulletin 377 (January 1948).6. Lea, F. C., an d Whitman, J. G., "TheFailure of Girders Under Repeated Stresses,"Journal of the Institution of Civil Engineers, No .7, 301-323 (June 1938).