effect of weld geometric profile on fatigue life of cruciform welds made by lasergmaw processes

a Department of Mechanical Engineering, University of Maine, Orono, ME 04469, USAb Applied Thermal Sciences Inc., PO Box C, 1861 Main Street, Sanford, ME 04073, USA Received 20 September

a Department of Mechanical Engineering, University of Maine, Orono, ME 04469, USA

b Applied Thermal Sciences Inc., PO Box C, 1861 Main Street, Sanford, ME 04073, USA

Received 20 September 2005; received in revised form 25 June 2006; accepted 10 July 2006

Abstract

The effect of weld geometric profile on fatigue life of laser-welded HSLA-65 steel is evaluated.Presented are results of cruciform-shaped fatigue specimens with varying weld profiles loadedcyclically in axial tension–compression Specimens with a nearly circular-weld profile were created at

133 cm/min, as part of this effort, with a hybrid laser gas-metal-arc welding GMAW (L/GMAW)process The ability of the laser-welding process to produce desirable weld profiles resulted in fatiguelife superior to that of conventional welds Comparison of finite-element analyses, used to estimatestress-concentration factors, to the hot spot and mesh insensitive approaches for convergent caseswith smooth weld transitions is presented in relation to the experimental results When a geometry-based stress concentration factor is used, the fatigue tests show much less variability and can belumped into one master curve

r2006 Elsevier Ltd All rights reserved

Keywords: Fatigue; Laser welds; S–N Curves; Cruciform; Full-penetration welds; Hybrid welds; concentration factors

Corresponding author Tel.: +1 207 581 2131; fax: +1 207 581 2379.

E-mail address: vince_caccese@umit.maine.edu (V Caccese).

joints are regions of stress concentration where fatigue cracks are likely to initiate.Geometry is one of the primary factors that control the fatigue life Accordingly,procedures that improve weld geometric profile by reducing stress concentrations will have

a beneficial impact on fatigue life Most fatigue-life improvement methods implemented todate are post-weld operations Kirkhope et al.[1,2]discusses methods of improving fatiguelife in welded steel structures by operations such as grinding, peening, water-jet erodingand remelting They stated that use of special welding techniques applied as part of thewelding process in lieu of post-weld operations are attractive because the associated costsare lower and the quality control is simpler Demonstrated in this paper is the use of acombined laser and gas-metal-arc welding (GMAW) weld procedure that results in asubstantially improved geometric profile of a longitudinal fillet weld The improved weldprofile results in lower stress concentrations without the need of post-weld operations.Laser welding is a relatively new technique that has potential to achieve excellent fatigueresistance, especially when used in combination with other more traditional weldingmethods such as GMAW Good control over weld profile is demonstrated when a laserand GMAW processes (L/GMAW) are used together Laser welding is a high-energydensity process that can be used on a wide variety of metals and alloys The automotiveindustry has used laser welding in production since the 1980s Recently, the ship-buildingindustry has looked toward laser welding to provide fabricated components in shipproduction Original laser welding for ship structures utilized CO2lasers with up to 25 KWpower Current manufacturing systems are looking toward use of state-of-the-artytterbium fiber lasers with power rating up to 10 KW Also, much hope is placed inlaser techniques to economically weld other structural components such as sandwichpanels The work presented in this paper is part of an ongoing effort to quantify the fatiguelife of laser-fabricated shapes for use in naval vessels

Some of the advantages that can be achieved through laser welding are ease of processautomation, high welding speed, high productivity, increased process reliability, lowdistortion of the finished part and no requirement for filler metal With current laser-welding techniques it is possible, as described by Duhamel[3], to achieve full-penetrationwelds in one pass on materials up to 1-in thick, depending on laser power and weld speed,with no filler and preparation as simple as precision cutting of the edges In addition,distortion of the finished component is significantly less than distortions measured inconventionally welded or hot-rolled shapes Even though filler material is not required inall cases to achieve a sound full-penetration weldment, lack of filler may cause undue stressconcentrations due to the geometry of the joint, especially if a sharp radius or reentrantcorner exists These stress concentrations can substantially reduce fatigue life of a high-quality full-penetration weld, solely due to the geometry of the weld profile Thecombination of laser welding with other processes such as GMAW, which is used to addfiller material, can dramatically improve the weld geometric profile Accordingly, theimproved weld geometry results in lower stress concentrations and hence improved fatiguelife

Fatigue strength of laser-welded joints can be markedly different than that ofconventional welds Therefore, an experimental program was undertaken to assess thefatigue resistance of laser-welded joints to be used in beam fabrication Tests were used toquantify the actual fatigue life of welds that were laser fabricated with various weldgeometric profiles, using differing process parameters Another objective of this effort is tocompare the current results to existing methods used in analyzing fatigue life The current

study is focused upon estimating the influence of the weld geometric profile on the fatiguelife based upon the stress concentration factor due to the weld geometry Ideally, tofabricate the optimal weld geometric profile, with a stress-concentration factor near unity,requires unrealistically slow speeds and unrealistically high amounts of filler metal.Accordingly, in developing an economical and practical weld profile for a line of productsuch as T-beams, tradeoffs must be made regarding desired weld geometry, operationspeed, and amount of filler metal.

2 Fatigue-life prediction in welded connections

In a marine structure, the environment may consist of load cycles in the order of millionsper year Fatigue failures typically take place at sites of high stress in either the basematerial or weldments Base material failures typically occur at openings, sharp corners or

at edges Fatigue failure in weldments is highly dependent on the structural connection andweld geometry details Unfortunately, according to Kendrick [4], weld profile data formost of the nominal stress S–N curves have not always been reported Therefore, results oftests with unknown weld profiles have been traditionally lumped together In reality,variation in fatigue life exists within a weld detail category due to weld geometry Thisaccounts for a significant variation in test results when fatigue data are lumped together.Modern welding techniques such as L/GMAW can be used to increase fatigue life byimproving the weld geometry Analysis techniques that capture this effect in the designprocess will allow fabricators to take economic advantage of the welding-techniqueimprovements It is more likely that practical implementation of advanced weldingtechniques will occur if analytical tools are used in design that capture the economic benefit

of an improved weld geometric profile

At present, there are two primary approaches used for predicting fatigue life, namely,the fracture mechanics approach and the S– N curve approach Assakkaf and Ayyub[5]described the relationship between these approaches as depicted in Fig 1 Fracturemechanics is mostly used in life prediction of a structure with an existing crack and is basedupon crack-growth data The initiation phase is assumed negligible for welded joint in thefracture mechanics approach and the life is based upon a stress-intensity factor, whichaccounts for the magnitude of stress, crack size and joint details In 1983, Maddox [6]stated that a fillet weld has small sharp defects along the weld toe from which fatiguecracks propagate This effect combines with the stress concentration so that the fatigue life

is effectively in propagating the crack

For welded joints, the S– N approach based upon fatigue test data is most frequentlyused in design The fatigue behavior of a connection is typically evaluated using constant-

N Crack Propagation Crack initiation

Total Fatigue Life

Fig 1 Relationship between the characteristic S– N curve and fracture mechanics approaches.

amplitude fatigue tests and the results are presented as the stress amplitude versus thenumber of cycles to produce failure Fatigue damage is then treated as a linear process andlife due to a varying load history is estimated using methods such as Miner’s rule The S– Nmethod will be focused upon in this paper of which there are several variations Theapproach chosen dictates whether or not the analysis considers the local effect of the weldgeometric profile.

2.1 Use of S– N curves

The characteristic S– N curve approach uses fatigue test data and assumes that fatiguedamage accumulation is a linear phenomenon Three different approaches often used inS– N type fatigue design of metal structures will be discussed in this paper, namely, (1)nominal stress, (2) hot spot stress, and (3) the notch stress

Using an S– N approach, the expression for fatigue life of a welded joint can be cast into

a general form as follows:

where N is the number of cycles to failure, S is the appropriate stress level for the analysisapproach being used, and A and m are material parameters This equation can belinearized by taking the logarithm of each side of Eq (1) resulting in the expression

2.1.1 Nominal-stress approach

The nominal-stress approach uses fatigue data derived from experimental testing of astructural detail, which are used to generate an S– N curve unique to this particular detail.The nominal-stress approach does not include the stress concentration due to weldgeometric profile, since it is assumed that the connection specific S– N curve alreadycharacterizes this effect The stress, S, in Eqs (1) and (2) is then equal to the nominalstress, Snom, which is the far-field stress due to the forces and moments at the potential site

of cracking In that regard, neither the local geometry of the weld toe or the local materialproperties are taken into account in the analysis Most design codes use differentclassifications when implementing the nominal-stress approach for different structuraldetails A different S– N curve, characterized by m and A, is provided for eachclassification

Munse et al [7]categorized numerous weld and attachment details typical in steel-shipconstruction They provided fatigue parameters including uncertainties for over 50 weldedconnection details The cruciform connection studied under this current effort is listed inthe Munse report as structural detail 14 and the fatigue parameters compiled for theseconnections encompass data that span years of testing with reported fatigue parameters of

m ¼ 7:35 and logðAÞ ¼ 23:2 for stress, Snom, in MPa Mansour et al [8] reports anabbreviated joint classification for BS 5400 and DNV where a load-carrying full-penetration fillet weld without undercutting at the corners dressed out by local grinding isplaced in category F Design parameters associated with category F are m ¼ 3 andlogðAÞ ¼ 11:8 with stress, S , in MPa

2.1.2 Hot spot-stress approach

In a hot spot approach, the hot spot stress, Shs, is determined at the location where thefatigue stress is the highest This is typically at the toe of the weld where fatigue cracking islikely to initiate Computational difficulty may arise because the stress at the transitionpoint of the joint is usually a singularity To overcome this effect, the hot spot stress at theweld toe is estimated using results in the vicinity of the weld and not at the singular point.Various extrapolation standards are used and some of the uncertainties of the effect ofweld geometry are removed The hot spot stress is derived from a detailed analysis of theconnection and will include global effects and to some extent the influences of the localgeometry With this approach, each material requires a single S– N curve for fatigue-lifeassessment However, a detailed finite-element (FE) analysis is necessary The hot spotstress, Shscan be related to Snomusing a stress-concentration factor for the gross geometry,

Kgas

Shsis then used in Eq (1), along with a baseline S– N curve to predict the fatigue life Theresulting hot spot stress may differ depending upon the FE program, element type, elementmesh and method used for dealing with the singularity

Several methods have been prescribed for determination of the hot spot stress Fricke[9], Niemi and Marquis[10]recommend using results at 0.4 and 1.4 t from the weld toe toextrapolate the stress at the hot spot for certain types of weldments Extrapolation at 0.5and 1.5 t has also been recommended as described by Kendrick [4] Other recommenda-tions include using a fine mesh to predict the stress distribution, noting that the stress at thehot spot is a singularity (unless the fillet is radiused) The hot spot stress is thenextrapolated at a preset distance from elements in the vicinity of the singularity

Error can also be introduced in the hot spot-stress calculation if the weld profiles have ahigh degree of variability or if the FE model does not accurately represent the as-weldedjoint geometry Also, the extrapolation technique used to compute the hot spot stress willsignificantly influence the results A standard method that is consistently applied isrequired for analysis In a test program, the weld profile needs to be accurately recorded sothat a proper assessment can be made

Procedures for experimental determination of stress-concentration factors, similar to theapproach used in hot spot analyses, have been demonstrated by Niemi and Marquis[10]and Dong[11], among others These techniques extrapolate the response recorded by two

or more strain gages to the hot spots Strains are converted to stress and extrapolationtechniques similar to those used in hot spot analyses are employed

2.1.3 Notch-stress approach

The notch-stress approach uses S– N curves based upon smooth material specimenswithout notches A stress-concentration factor is then determined to account for variousimperfections According to Kendrick[4], this method can be used to predict the effect of

an imperfect weld profile on fatigue life It will include an additional stress-concentrationfactor for the actual weld geometry, Kw as well as factors for increased stress due tomisalignment and angular mismatch Applied fatigue stress, Sn, can then be written interms of an aggregate stress-concentration factor, K, and the nominal stress as

K is the product of the individual stress concentration factors given as

Kg is the stress-concentration factor due to the gross geometry, Kw is the concentration factor due to the weld geometry, Kteis the additional stress-concentrationfactor due to eccentricity tolerance (used for plate connections only), Ktais the additionalstress-concentration factor due to angular mismatch (used for plate connections only), Kn

stress-is the additional stress-concentration factor for un-symmetrical stiffeners on laterallyloaded panels applied when nominal stress is derived from simple beam analysis

For an ideal case with no eccentricity or angular mismatch, the notch stress, Sn, isrelated to the nominal and hotspot stresses as follows:

2.1.4 Mesh-insensitive approach

Dong [12]recently suggested a method for determination of the hot spot stress that isinsensitive to the FE mesh A FE analysis is performed and the resulting nodal forcesacross the thickness of the plate in the area in question are used to compute a mesh-insensitive structural stress, Smi, which can be used in a fatigue analysis This methodincludes effects of both the gross connection geometry and to a lesser extent the local weldprofile The mesh-insensitive stress can be related to the nominal stress by

In this approach, an equilibrium-equivalent stress state and a self-equilibrating stress stateare used to compute the mesh-insensitive stress Nodal forces are used instead of theresulting stresses at or near the singularity (hot spot) location, since the stresses are highlymesh sensitive This results in stress-level predictions with little sensitivity to the fineness ofthe FE mesh Therefore, this procedure may be useful in the analysis of ship structureswhere coarse FE meshes are used, especially at the preliminary design stage

3 Fatigue testing of laser-welded cruciforms

An experimental fatigue study was undertaken to further investigate the effect of localweld profile The weld geometry of cruciform specimens was intentionally varied,measured and categorized Numerous fatigue tests were performed to determine theinfluence of geometry on fatigue life Laser welding proved to be an invaluable technique

to carry out this effort due to the ability to develop a full-penetration weld Whensupplemented with a GMAW process, a smooth, nearly circular geometric profile wasrealized The fatigue tests summarized in this paper are a subset of a larger database beingcompiled for the qualification of laser-welded HSLA-65 steel for use in US Navy vessels asdocumented by Kihl [13] and Berube et al [14] Results specifically demonstrating theeffect of weld geometry on fatigue life were selected The fatigue testing was performed atthe University of Maine[14]using a 50 metric ton (110 kips) MTSTM810 universal testingmachine with a TestStarTMdigital controller, as shown in Fig 2a The 355.6 mm (14 in)long, 95.25 mm 3
3

4in:
wide test specimens (Fig 2b) were cruciform shaped andfabricated from 12.7 mm ð0:56 in:Þ thick HSLA-65 steel plating The gage length used for

the testing was 177.8 mm (7.0 in) with an 88.9 mm (3.5 in) grip length The testing wasperformed in load control, at a rate of 2.22 MN/s (500 kip/s), and the specimens wereloaded axially, under completely reversed sinusoidal loading, at stress levels of 103.4,206.8, and 310.3 MPa (15, 30, and 45 ksi) The controller automatically terminated the testwhen the extension of the specimen had doubled compared with that recorded at thebeginning of the test The doubling of the extension was typically indicative of a significantcrack in the specimen.

3.1 Test article fabrication

There were four series of test articles detailed for this investigation The test articles werefabricated using either a laser ‘cold-wire’ (LBW-CW) or a laser-hybrid (L/GMAW)welding process In the LBW-CW process, filler material is added by using a smallpercentage of the laser energy to melt wire fed to the weld pool The laser-hybrid weldingprocess combines the laser with a GMAW process With this hybrid welding procedure,the laser beam and GMAW arc act in the same welding zone to support each other It isbelieved that the energy from the laser beam is responsible for establishing the keyhole, as

in the laser-only process, and the GMAW system delivers filler material to the weld pool,thus creating the weld geometric profile

Weld-process parameters for the fabricated specimens are summarized inTable 1 Thefirst series (Series-A) was fabricated at the Applied Research Laboratory (ARL), of PennState University, using the laser ‘cold-wire’ process with their 14 kW CO2laser operating at

10 kW delivered power and a weld speed of 25.4 cm/min (10 in/min) This resulted in a weldwith profile as shown inFig 3 The weld is characterized by a geometry that has a smallregion that is somewhat flat in the center and a smooth radius toward the ends The nextthree series of specimens were fabricated at Applied Thermal Sciences (ATS) in Sanford,

ME The ATS system is equipped with a real-time adaptive feedback control of the weld

Hydraulic Grip

12.7 mm Cruciform Specimen

Weld, typ.

t=12.7 mm typ.

95.25 mm 355.6 mm

Fig 2 Fatigue-test specimen and test setup: (a) Specimen in test machine; (b) Test article.

process, which monitors the welding parameters including weld shape The first seriesfabricated by ATS, Series-B, used the laser cold-wire process with a 25 kW CO2 laseroperating at 14.3 kW delivered power and weld speed of 190.5 cm/min (75 in/min) Theweld profile resulted in a small flat-shaped fillet as shown inFig 4 Series-C was fabricated

Table 1

Weld process parameters

Weld series Weld process Laser-delivered power Laser weld rate Wire typea GMAW power

Wire size used for all series is 0.889 mm (0.035 in) dia.

Fig 3 Series A—FLC weld profile: (a) side view; (b) end view; (c) traced profile.

Fig 4 Series B—CR125 to CR131 weld profile: (a) side view; (b) end view; (c) traced profile.

at a reduced process rate with increased wire feed and resulted in a larger fillet of the samegeneral profile as Series-B, as shown inFig 5.Fig 6shows the resulting welds for the lastseries (Series-D), which used a laser-hybrid process These welds had a vastly improvedgeometric profile that was as near to circular as can be expected.

3.2 Test results

Fatigue test results of the four weld series are summarized in Tables 2–5 Series-A,welded at ARL, was the first test series fabricated and is used as the baseline forcomparison These data are plotted inFig 7along with the S– N curve using parametersreported by Munse et al [7]and Niemi and Marquis[10] for cruciform joints and testsperformed by Kihl[13]on conventionally welded HSLA-65 steel cruciforms In addition,the design-based curve for category F given in Mansour et al.[8]is also provided All laser-welded tests show longer fatigue life than reported by Munse, and Series A and D showfatigue life better than that reported by Kihl for conventional welds of the same material

Fig 5 Series C—CR154 weld profile: (a) side view; (b) end view of failed specimen; (c) traced profile.

Fig 6 Series D—CR187 weld profile: (a) side view; (b) end view; (c) traced profile.

Specimen conditionb

Cycles to failure Geometric

Specimen condition a

Cycles to failure Geometric

Specimen conditiona

Cycles to failure Geometric

An objective of this study was to assist the welders at ATS in developing weldingparameters that would yield results comparable to Series-A, welded by ARL at 25.0 cm/min (10 in/min) The first two sets of laser welds done at ATS, Series-B and Series-C, werepreliminary studies to determine the effect of fillet size and shape on fatigue life.These series were tested at the intermediate 206.8 MPa (30 ksi) stress range only, so that

Specimen Conditiona

Cycles to Failure

Number of Cycles to Failure

Fig 7 Summary of fatigue test results.