You are on page 1of 18

ELSEVIER

Int. J. Fatigue Vol. 18, No. 3, pp. 153-170, 1996 Copyright 1996 Elsevier Science Limited Printed in Great Britain. All rights reserved 0142 1123/96/$15.00

0142-1123(95)00117-4

Review of fatigue strength assessment of nonwelded and welded structures based on local parameters
D. Radaj Daimler-Benz AG, D-70546 Stuttgart, Germany (Received 20th September 1995)
The basic types of the approaches of fatigue strength and service life assessment based on local stress and strain parameters are at first described for nonwelded structures: fatigue phenomena, influence parameters, objectives of the assessment, global and local approaches and historical development. The nominal stress approach, structural stress approach, notch root approach and fracture mechanics approach are reviewed and set into comparison. The peculiarities of welded structures are emphasized: inhomogeneous material, welding residual stresses and geometrical characteristics. The variants of the local approaches mentioned above which are applied to welded structures are specified together with relevant conclusions.
(Keywords: fatigue-strength; welded structures; local approaches)

INTRODUCTION Local approaches assessing or proving the fatigue strength are procedures which aim at the design, dimensioning and optimization of structural components on the basis of local stress and strain parameters l~. The local stress and strain parameters cause local damage which may finally result in failure of the whole component. Damage due to fatigue, the core of the fatigue strength problem, is primarily crack initiation and propagation. The variants of the local approaches differ in respect of the local stress and strain parameters chosen, inclusive of their limit values. They can be assigned to the structural stress and strain approach, to the notch stress and strain approach and to the fracture mechanics approach. These local approaches supplement or substitute the older nominal stress approach which is of a global nature. The object of the following review is the variants of the local approaches proposed for nonwelded and welded structures. The review has been initiated by the progress in methods achieved in recent years while industrial application is lagging behind. Another reason for the review is the discussions on local approaches to be included in design codes. A well-known older review has been presented by Iida 7. KOttgen et al. s published a paper on the same issue at a later time. Progress within the procedures and wishes for application by industrial users induced the author to prepare a completely new review which was presented in the German language on the occasion of a conference on joining techniques in light-weight design 9. The following paper is an unabbreviated English version of this review.

FUNDAMENTALS OF ASSESSMENT PROCEDURES

Micro- and macrophenomena of material fatigue


The local approaches to be reviewed aim at covering the dominating parameters of an extremely complex physical reality in order to make it controllable by the engineer. This reality comprises primarily microstructural phenomena (moving dislocations, microcrack initiation on slip bands, further crack growth by local slip mechanisms at the crack tip), but can be approximately described by a macroscopic elastic-plastic stress and strain analysis according to continuum mechanics which refers to cyclic crack initiation, stable cyclic crack propagation and unstable final fracture, Figure 1.

Influence parameters for cyclic crack initiation


The cyclic initiation of the technical incipient crack is primarily determined by the amplitudes of the cyclic stress and strain components at the notch root, the volume of the highly stressed material being also important. The influence parameters controlling the limit values of the stress and strain parameters are summarized in Table 1, which refers to the local approach insofar as local stresses and strains are introduced instead of external loading. The number of the influence parameters is large, but can be handled within the procedure of strength assessment. A problem arises from the lacking possibility to decouple the effects of the influence parameters in the case of engineering tasks.

153

154

D.

Radaj

I Dislocation H Crack H Microcrack H Macrocrack H Final ] movement nucleation propagation propagation fracture Crackinitiationphysical ~--I Crack propagationstable =-I C.p.unstable_ j Crackinitiationtechnical =--I Crackpropagationtechnical -,~Figure 1 Micro and macro phenomena of material fatigue Table 1 Influenceparameters controlling cyclic crack initiation
Material Type Alloy Grain size Microstructure Environment Temperature Corrosion Component Shape (notch effect) Size (technological, stress-mechanicaland statistical size effect) Surface Roughness Hardness Residual stress Loading course Amplitude spectrum Amplitude sequence Mean value sequence Rest periods

Loading type Mean stress (including residual stress) Multiaxiality (including phase angle)

Influence parameters for cyclic crack propagation The cyclic crack propagation is primarily determined by the amplitudes of the cyclic stress intensity factor or of the cyclic J-integral at the crack tip. The influence parameters controlling the limit values of the stress and strain parameters at the crack tip are summarized in Table2, which once more refers to the local approach by introducing the local parameters. Considering crack initiation in comparison, some influence parameters have been dropped and others added. The influence of surface is omitted, the influence of crack shape and crack size added. Contents and objectives of strength assessments Strength assessments comprise the explicit or implicit comparison of load, stress and strain parameters including their limit values, which cause a definite damage, a definite deformation, an incipient crack or total fracture. The loads comprise forces and moments, the stresses and strains are of the nominal, structural and notch type, stress intensity factors and the J-integral are derived therefrom. The appertaining limit values are designated as strength values. The stresses and strains follow from the forces and moments according to the continuum theories, i.e. elasticity, plasticity and viscosity theory. The strength values are determined from loading tests using simple or component-like specimens. Considering the fatigue

strength, the WOhler test, the service life test and the crack propagation test are most important. The assessment of fatigue strength is also performed in such a way that the specimen or component life is determined at a given load value instead of the strength at a preset number of cycles. The latter procedure is in agreement with the prevailing kind of fatigue test evaluation. Global and local approaches Strength assessments are termed 'global approaches' if they proceed directly from the acting forces and moments or from the nominal stresses derived therefrom under the assumption of a constant or linearized stress distribution (therefore 'nominal stress approach'). The global approaches originally use limit values of load or nominal stress which are related to global phenomena such as fully plastic yielding or final fracture of the specimen. Strength assessments are termed 'local approaches' if they proceed from local stress and strain parameters. The local processes of damaging by fatigue of material are considered, cyclic crack initiation, cyclic crack propagation and final fracture. Crack initiation is connected with the 'notch root approach' which is based on the stresses and strains at the notch root derived by continuum mechanics. Crack propagation and final fracture, on the other hand, are described by the 'frac-

Table 2 Influenceparameters controling cyclic crack propagation


Material Type Alloy Grain size Microstructure Environment Temperature Corrosion Component Shape Size (technological, stress-mechanicaland statistical size effect) Crack Shape Size

Loading type Crack opening mode Mean stress intensity Multiaxiality (including plate thickness)

Loading course Amplitude spectrum Amplitude sequence Mean value sequence Rest periods

Fatigue strength assessment of nonwelded and welded structures ture mechanics approach' which proceeds from an existing incipient crack. The strength assessment according to the complete local approach therefore consists of the notch root approach and the fracture mechanics approach. An approach between the global and local versions is the structural stress approach which emphasizes the stress concentrations caused by the macrogeometry, while the actual notch effect is definitely suppressed. Different variants of the local approach can be distinguished according to the local stress and strain parameters chosen and the type of failure criteria introduced. The most important basic variants of the global and local approach are shown in Figure 2, each variant characterized by the typical load, stress or strain parameters and the appertaining strength diagram. The local quantities result from the global quantities proceeding from the left hand side to the right hand side of the figure by increasingly taking local conditions into account. The following strength diagrams are presented: load S-N curve, nominal stress S - N curves for standardized notch cases, structural stress S-N curve, notch strain S - N curve, Kitagawa diagram (cyclic limit stress over the length of short cracks) and crack propagation rate over the cyclic stress intensity factor of longer cracks. Historical development towards the local approach The global approach which even today is basic designates the beginning of research and development aiming at fatigue resistant design. The local approaches evolved from the global approaches, insofar as the local consideration corresponding to the local fatigue phenomena supplements, deepens and extends the global approach, at first without claiming independent usage. Considering the history, the local approach received the following essential development impulses: 1. evaluation of fracture surface--fatigue fracture surface, final fracture surface, clam shell markings, notch effect (1850); 2. measuring techniques for local stresses and strains (photoelasticity since 1930, inductive strain gauge with short measuring length since 1955, small-scale resistance strain gauges since 1965); 3. application of the notch stress theory (first edition of Neuber's book in 1937, Neuber's macrostructural support formula published in 1960); 4. application of fracture mechanics (crack propagation formula published by Paris in 1963);

155

5. application of the finite element method in structural design based on appropriate computer technique (since about 1970); 6. application of short crack fracture mechanics (since about 1990). At first, researchers from the USA were leading in the application-related development of the local approach (e.g. application of the Neuber formula and the Paris equation). The starting point was the lowcycle fatigue strength at elevated temperatures. At about the same time, the high-cycle fatigue strength was mainly considered by researchers in Germany, especially the determination of the fatigue notch factor of notched specimens. Later on, the main efforts related to the local approach shifted to Germany, especially to research laboratories in Darmstadt, the research tradition of which originates from Thum, Gassner and Svenson. The extensive literature dealing with the local approach has been compiled in Ref. 2. TYPES OF ASSESSMENT PROCEDURES Nominal stress approach The nominal stress approach for assessing the fatigue strength and service life proceeds from the nominal stress amplitudes in the structural component and compares them with the nominal stress S-N curve, Figure 3. The gradient and scatter range of the S-N curve can be approximately defined based on the normalized S - N curve scheme if specific test data are not available. The nominal stress S-N curve comprises the influence of material, shape (inclusive of notch effect and size effect) and surface (inclusive of residual stresses). The acting forces and moments can also directly be used instead of the nominal stresses, an obvious choice in cases where nominal stresses cannot be meaningfully defined. The service life results from the nominal stress S - N curve and the nominal stress spectrum according to a simple hypothesis of damage accumulation, mostly according to the relative Miner rule. The nominal stress spectrum follows from the load spectrum taking the supporting cross-section and the type of loading into account. The service life calculation is generally performed in respect of final fracture, but it can also be made in respect of crack initiation. The effects of loading sequence and overloading remain unconsidered as far as a load spectrum or load matrix is taken as the basis. These effects can principally be taken into account, if the load-time

Cross i Structural] I EI.r ~tch I EI.-pl.~otch I Short crack I t crack Lng section d sconUn./ effe " I effec ICyclic I Cyclic H Cyclic I~.~ structuralIH I Cy :lic load nominal no' ~h stress I I stress str ss
f

H Cyt H Cyclic H not ;lie J-integral ~h stri in


i

Cyclic stress intensity


i

&K

Figure 2 Global and local approaches for describing the fatigue strength

156

D. Radaj

Material Geometry Surface

Load

[ Geometry Loading

Component test

S-N curve

Nominal stress

Measuring
~,/~V u n t i n g C

.A.,A,/~
V'vV"

Component test I Iv VVWV'v" I


Figure

culJ
--1 Rel. Miner rule I

"

Service life (fracture)

3 Nominal stress approach for assessing the fatigue strength and service life of nonwelded structural components, according to Kloos~

function is available or if the damage contributions are taken into account cycle by cycle in the right sequence.

t
f

AF

Structural stress approach The structural stress approach for assessing the fatigue strength and service life proceeds from the structural stress amplitudes in the component and compares them with a structural stress S-N curve. The approach is sufficiently developed in this form for welded structures only. A more incomplete version of the approach is the well-known finite element analysis of welded or nonwelded structures. It aims at optimizing especially macrogeometrical influence parameters. The actual notches are not taken into account with the effect that quantitative statements in respect of fatigue strength and service life are possible for the unnotched areas only of the structure. Qualitative valuations, relative considerations and trend statements are the prevailing results of finite element analysis. The structural stress approach for welded joints gives an indication of how quantitative statements may be achieved for notched areas also, without resorting to the notch root approach which requires higher expenditure. Notch root approach The notch root approach for assessing the fatigue strength and service life up to crack initiation proceeds from the elastic-plastic strain amplitudes at the notch root and compares them with the strain S-N curve of the material in the unnotched comparison specimen. The idea behind this approach is that the mechanical behaviour of the material at the notch root in respect of local deformation, local damage and crack initiation is similar to the behaviour of a miniaturized, axially loaded, unnotched or mildly notched specimen in respect of global deformation, global damage and complete fracture, Figure 4. The comparison specimen is either imagined as positioned at the notch root, or really cut out from this area. It should have the same microstructure, the same surface condition (inclusive of residual stresses) and about the same volume as the material at the notch root. At least, well founded corrections of the test results with the specimen should be possible for the purpose of comparison. The strength assessment consists of determining the stresses and

Unnotched comparison specimen

10 mm

Figure

4 Comparison specimen for simulating the cyclic stressstrain and crack initiation behaviour at the notch root

strains at the notch root in the elastic-plastic condition and comparing them with the strain S-N curve of the material in the miniaturized specimen up to complete fracture or in a larger specimen up to the technical crack initiation (crack depth about 0.5 mm, crack width at the surface about 2 mm), Figure 5. The stresses and strains at the n o t c h r o o t of the structural component are calculated proceeding from the cyclic stress-strain curve and the macrostructural support formula according to Neuber ~ (amended formulae according to Seeger and Beste12). Considering the case of a sharp notch, the microstructural support effect has additionally to be taken into account according to one of the well-known hypotheses. The mean values and the principal directions of the stresses and strains are introduced besides their amplitudes. The notch root strain can also be measured instead of being calculated. The comparison specimen is required in this case to determine the stresses which are connected with the elastic-plastic strains. Finally, a sequence of hysteresis loops in the stress-strain diagram (i.e. the

Fatigue strength assessment of nonwelded and welded structures


Material data Geometry Loading

157

function

Load-time

Iv VV'VV'" Notch la~ stress =


Strain S-N curve [ Cyclic stressI _ I strall'n curve

[Stress-strain I pathl

I Damage parameter =

accumulation I hypothesis I I -IServicslife I, (crack initiation)

Figure 5 Notch root approach for assessing the fatigue strength and service life of nonwelded structural components; according to Kloos~

stress-strain path) results on the basis of the loadtime function. The strain S-N curves of the comparison specimen on the other hand which are dependent on the mean stress can be represented by a single damage parameter S - N curve which comprises the effect of mean stress. The damage contributions from the stress-strain path are determined cycle by cycle, added up and compared with the damage parameter S - N curve. The damage parameter S - N curve may also be used with a definite factor in order to fit the results from component fatigue tests. The Miner rule in its relative form may be used proceeding from the strain S-N curve if the strain amplitudes have been evaluated in a spectrum or matrix. Uniaxial conditions are the basis of Figure 5 in order to simplify the representation. Biaxial conditions (on the surface) and triaxial conditions (under the surface) on the other hand prevail at the notch root of structural components. Additionally, the principal direction of the stresses and strains at the notch root changes permanently in the case of complex loading sequences acting on the component including mean load variation and mutual phase shift. The numerical procedure becomes extremely complicated in this case both in respect of theory and application. The yielding behaviour including hardening and softening under multiaxial, proportional or nonproportional stressing and straining (comprising combined torsional and axial loading also) has to be considered in this case. The appertaining damaging and crack initiation behaviour must be taken into account. Research has clarified only part of the problem up to now. The notch root approach, which requires high expenditure in its general form, is simplified for application. This can be done in different ways. The most simple version of the approach refers to the fatigue strength for infinite life, i.e. to a constant amplitude loading with fatigue fractures avoided. The assumption is well founded in this case that no appreciable plastic deformation occurs at the notch root, i.e. the notch effect of the structural component can be described as linearelastic and set against the endurance limit of the

material, Figure 6. The elastic stress concentration factor is dependent on the shape, the dimensions and the loading of the structural component. The fatigue notch factor is derived therefrom, introducing additionally a microstructural support hypothesis in the case of sharp notches (mostly defined for zero mean stress). The endurance limit of the material is taken from an amplitude-mean diagram (according to Haigh) taking into account roughness, hardness and residual stresses in the surface layer. The von Mises strength hypothesis is introduced in the case of ductile materials. The fatigue strength for infinite life of the structural component can thus be determined. The notch root approach in this simple form has been successfully applied at first for engine components with an expected infinite life such as crankshafts, connecting rods or gearwheels (Lang'3). The simplified notch root approach referring to the fatigue strength for infinite life can be extended into the high-cycle fatigue strength range referring to finite life without major problems. But it will be sufficient in general, to consider the normalized S-N curve of the nominal stress approach as valid which can be based on the fatigue strength for infinite life of the component. The elastic-plastic notch root approach is indispensable on the other hand in the low-cycle fatigue range and for solving problems of service fatigue strength.

Fracture mechanics approach The fatigue strength and service life of the structural component with an existing incipient crack is determined according to the fracture mechanics approach which supplements the notch root approach in general, Figure 7. The crack propagation rate of a technical incipient crack, i.e. of an incipient crack which is not too small, can be analysed based on the cyclic stress intensity factor AK according to the simple formula proposed by Paris and Erdogan 14. Cyclic crack propagation occurs as soon as the threshold value of the stress intensity factor is exceeded and ends as soon as the critical stress intensity or the load-carrying capacity of the remaining cross-section is reached with larger

158

D. Radaj
Geometry
Loading Effective notch stress Fatigue notch factor _ &IV

Surface
- Roughness - Hardness - Residual stress

Load spectrum

_ I Component - I endurance limit &ME, ASE


Strength hypothesis (von Mises)

Normalised S-N curve

Service life curve

endurance limit
m

Material

Notch support effect

tl

St~ent

N
Notch stress approach

Nominal stress approach

Figure6 Notchroot approachfor assessingthe endurancelimit of nonweldedstructuralcomponents


Geometry Loading
i---1.
AM

Crack

Load sequence

I"q
&K,&J
Cyclic material data

Al%n,&Jeff

length Endurance
Crack

Crack propag. service life

S-N curve and

Static material data

J N
~ O u , ay

[ --<~ A'o]

dal m ]1 dN I,Ko U K I c

[i/

&K,&J

Figure7 Fracturemechanicsapproachfor assessingthe fatiguestrengthand servicelife of nonweldedstructuralcomponents


crack length. The cyclic J-integral is used for describing crack propagation in the case of higher stresses and strains with accompanying elastic-plastic deformation at the crack tip. Crack closure has to be taken into account, especially in the case of compressive stresses. This can be achieved in a simplified manner by cutting the compressive part of the stress range. The crack length over a number of cycles follows from integrating the crack propagation equation cycle by cycle up to the condition of final fracture. S-N curves and service life curves for structural components containing a crack are determined in this way. The crack initiation life according to the notch root approach and the crack propagation life according to the fracture mechanics approach are added up to give the total life. The argument is occasionally met that structural components contain small incipient cracks or cracklike flaws from the beginning which are caused by production processes or preloading, so that the component life can be determined based simply on crack propagation. An extremely small, more or less fictitious initial crack length is introduced into the life calculation for that purpose. The argument that fatigue strength is mainly crack propagation strength is correct in principle, but the simplified calculation of crack propagation according to fracture mechanics is disputable if the cracks are very small. The following objections can be raised: Any final result can be achieved by manipulating the fictitious incipient crack length. The behaviour of very short cracks cannot be described on the basis of the usual AK-approach. The correction factor of geometry cannot simply be determined for cracks at the notch root. The crack shape changes during crack propagation, thereby complicating the analysis. Smaller cracks grow together forming larger cracks, a process which is not easily accessible to calculation.

Comparison of the approaches


The nominal stress approach is considered as robust as far as it is statistically founded. It is considered as superior to the local approach in this respect 15, but this is only half the truth. The robustness is bound to the condition that the structural component and the test specimen correspond in respect of all essential influence parameters. Considering the multitude of parameters which have an influence on the fatigue strength, such a correspondence is more exceptional than general. Any deviation between the structural component and test specimen must be assessed without statistical proof. The local approach is one possibility to do this in a meaningful manner. The notch case classification based on the local approach, for example, is wellknown to the experts. Therefore, the local approach is

Fatigue strength assessment of nonwelded and welded structures


indispensable as a supplement to the nominal stress approach if structural component and test specimen differ in respect of individual parameters. The local approach is also useful in cases where nominal stresses cannot be meaningfully defined. Additionally, it is noteworthy that the engineer needs mainly relative statements. The question is put forward very often what fatigue strength or service life can be expected with a design variant in relation to a wellproven former design. Only the local approach offers essential aids for answering such questions. Different views exist between experts concerning how detailed the local consideration must be within the assessment procedure, based on structural stresses only or on notch stresses also, or even comprising crack propagation. No general answer is possible. The circumstances of the considered special case determine the decision between the approaches. The following statement is generally valid: the structural stress analysis is always required because notch stresses and stress intensity factors are based on structural stresses. Therefore, the structural stress approach has the widest field of application. The step from the structural stress approach to the notch stress approach and possibly further to the fracture mechanics approach is justified if the scatter range of the local notch geometry caused by the manufacturing process is small or if the scatter range can be passed over by a worst case consideration. The disagreement of the experts is easy to understand in other cases, e.g. in respect of nonmachined welded joints. The supporters of the structural stress approach then hold against the notch stress approach that the scattering of the notch geometry cannot be accurately evaluated and that it is not well suited for the notch stress analysis, but any approach can be adjusted to the engineering demand in a restricted field of application.

159

tures. This statement holds true at least as far as the attempt is made to take all details into account which are of relevance in respect of continuum mechanics. The pecularities can be subdivided into inhomogeneous material, welding residual stresses and geometrical characteristics. They often remain unconsidered in the local approaches. In general, the material characteristic values of the base material are used, the effect of residual stresses is only roughly taken into account and the worst case of the geometrical notch parameters is considered.

lnhomogenous material
Welded joints are characterized by inhomogeneous material. The filler material added to the base material is of similar type in general, but especially alloyed in order to achieve a high quality of manufacture, e.g. in respect of the transfer of droplets, the weld pool shape and the suppression of hot cracking. The filler material mixes with the base material in the weld pool while individual alloying elements may be burnt or evaporated and other elements may intrude from the ambient atmosphere or materials. Micropores may occur if the evaporation is impeded, microseparations are fostered if the material is susceptible to hot cracking. The microinclusions may be changed in respect of type and number. Such irregularities may especially occur in the area of the toe groove and weld root. The heat affected zone adjacent to the weld pool shows different microstructures according to the thermal cycles experienced. These are characterized by different grain sizes and hardness values resulting in different yield limits, crack initiation strengths and crack propagation resistances. Only part of the considered problems is removed in cases of welding without filler material. In addition to the inhomogeneous material within the welded joint, defects or imperfections typical of welding may occur, e.g. cracks, pores, cavities, lack of fusion, overlap, inadequate penetration and burn through holes.

Partial tasks of the fatigue strength assessment


The preceding and subsequent reviews focus on the assessment procedures, their basis and their varying structure. In doing this, no guidance is offered for actually performing the assessment. The basic structure of the procedure is the same for all the approaches. The assessment procedure is subdivided into the following eight partial tasks according to HaibachS.6,16: 1. 2. 3. 4. 5. 6. 7. 8. definition of the task and the procedure; identification of the fatigue-critical cross-sections; determination of the characteristic service loads; determination of the characteristic stresses; determination of the critical stresses; derivation of the appropriate safety range; compilation and evaluation of the assessment; documentation of the assessment.

Welding residual stresses


Welding is generally performed by melting the surfaces of the parts to be joined and the filler material by a concentrated heat source. The subsequent rapid cooling produces welding residual stresses via thermal strains and microstructural transformation. These stresses may reach the yield limit in the weld area and decrease steeply in its neighbourhood. They have a generally low level globally (constraint stresses), but produce local stress concentrations at notches. The welding residual stresses are reduced by cyclic loading or changed in a favourable manner if the ductility of material is adequately high and the cyclic loading sufficiently severe, i.e. with amplitudes higher than the endurance limit. A more detailed analysis is necessary if this is not the case. The endurance limit and the high cycle fatigue strength are changed by the effect of residual stresses. This is the case not only in respect of unfavourable tensile residual stresses produced by welding, but also in respect of favourable compressive residual stresses produced by post weld treatment.

The assessment procedure according to the local approach applied to special nonwelded structural components is demonstrated together with special instructions by Lang 13, Sonsino 17 and Seeger and Zacher/8. PECULARITIES OF WELDED STRUCTURES

Complication of the approaches


Welded structures show several peculiarities which further complicate the local approaches which are rather complex, even in the case of nonwelded struc-

Geometrical characteristics
The assessment of the fatigue strength and service life based on local stress and strain parameters encounters special difficulties because of the partly scattering

160

D. Radaj
point is close to the maximum notch stress because the stress peak from the external loading is slim and steep in relation to the endurance limit curve in the case of the sharp notches typical of nonmachined welds. The conclusion from this consideration is that the maximum notch stress should be compared with the endurance limit of the base material, modified according to the local hardness and residual stress when performing the fatigue strength assessment for welded joints. The proposed desirable modification of the fatigue strength assessment for welded joints has not been considered in more detail up to now. The situation becomes much more complex if the medium- and low-cycle fatigue strength has to be assessed based on the local approach. This holds true both for welded joints and for surface hardened specimens. The inhomogeneity of the material changes both the cyclic deformation behaviour and the cyclic strength behaviour. The layering of material zones with highly different yield limit which is caused by the thermal cycles can introduce high stress concentrations, even in unnotched specimens, as soon as the yield limit is exceeded. 2 The problem of mismatching is addressed in the case of base and filler material with different yield limits under large deformations of the weld. Research is going on in respect of static loading.2~ The behaviour of welded joints in the mediumand low-cycle fatigue range based on the local approach is only insufficiently clarified by research. STRUCTURAL STRESS AND STRAIN APPROACHES

and partly undefined geometrical parameters of welded joints. Measuring these geometrical parameters and procesing them within a notch-mechanical analysis is expensive. The radius of curvature at the weld toe or weld root and the slope of the weld contour near the weld toe (alternatively the amount of weld stiffening) are highly variable and scattering. The same holds true for the amount of burn-in or for the nugget diameter. These geometrical notch data of welded joints depend on the type and parameters of the welding process, on the welded materials, on the plate thickness, and on the margin of tolerance when positioning the structural components to be joined producing some misalignment and gaps. They can be precisely determined, at least to some extent. The weld toe geometry may be determined using external casting techniques. The weld root or nugget edge may be made visible after transverse cutting and polishing.

Desirable modifications of the local approach The local approach of strength assessment for welded joints should take into account the geometrical details, the inhomogeneity of the material and the welding residual stresses at least within a rough estimate, i.e. neglecting the scattering which originates from the manufacturing process. This task is similar to the strength assessment for surface hardened specimens. A special variant of the local approach has been developed for this application, i.e. the approach with locally differing endurance limit according to Kloos et a l . 19 It is a feature of this approach that the point of crack initiation may be under the surface and is not known at first. A possible similar procedure of assessing the fatigue strength for infinite life at the weld toe notch is shown in Figure 8 by way of example. The curve of the local endurance limit over the layout of the weld toe follows from the distribution of hardness and residual stress. Crack initiation has to be expected at the point where the curve of cyclic stresses in the notch root, caused by external loading, contacts the curve of local fatigue strength for the first time with increasing load amplitudes. The crack initiation

Structural strain approach according to Haibach Haibach22 has shown in an historically early contribution that the cyclic elastic-plastic strain measured and averaged with a strain gauge of definite length (3 mm) at a definite small distance from the weld toe (2.0-2.5 mm considering the centre of the strain gauge) is well suited to characterize the high cycle fatigue strength of welded joints independent of joint type, weld shape and type of transverse loading, provided

Hardness
"~1

.:.~/WB
J~.~/WN HAZ

iHV
I I

Endurancelimit
I0~ I ...-- .d~

Residualstress

I1,,
I

,,
l

WB: Weld bead WN: Weldnotch HAZ:Heat affected


zone

Loadingl

Lstress
,
WB WN HAZ

~1 Limitcondition

WB WN HAZ

Figure 8 Additionalcomponentsof the notch root approach for assessing the endurancelimit of weldedjoints with inhomogeneousmaterial and residual stresses (approach with locally differingendurance limit)

Fatigue strength assessment of nonwelded and welded structures


that the fatigue fracture occurs at the weld toe, An averaged cyclic structural strain of Aes=0.063% (with failure probability Pf=50%) is given as a sustainable strain and an averaged cyclic structural stress of A~r~= 80 N mm-2 (with failure probability Pf = 0.1%) as a permissible stress for structural steels 5. The vertical broken line in the right-hand part of the figure characterizes the transition from the local approach to the nominal stress approach. The relation between the structural strain or stress according to Haibach and the fatigue-effective maximum notch stress according to the fatigue notch factor has recently been derived by Olivier et al. 53 assuming linear-elastic structural behaviour. The relation depends on the plate thickness. The values of AEs and Ao-~ given above have been derived for a plate thickness of 10 ram. A higher plate thickness is connected with lower sustainable strains and stresses. The opposite is true for a lower plate thickness. The structural strain approach according to Haibach has been applied to vehicle components, boilers, pressure vessels and crane structures. A broader application is impeded by the necessity to manufacture the welded structure under consideration before measurements can start. On the other hand, the procedure can obviously be transferred to a numerical structural analysis.
Figure9. Structural stress approach according to Dijkstra and Gurney

161

Rather high stresses may occur locally in framework structures consisting of hollow section bars, e.g. in the tubular steel joints of offshore drilling rigs. They are caused mainly by local bending of the tube shell and by superimposed notch effects. The strength and life of tubular joints can realistically be assessed on the basis of the local stress concentration. The structural stress approach developed by Dijkstra, 23 Gurney et al. 24 for structural joints made of circular section tubes aims at determining the local structural stresses in the tube shell at the crack initiation point and using this stress as the nominal stress in a simple welded joint specimen, e.g. the corner joint with a fillet weld, Figure 10. The local structural stress is supposed to make strength-characeristic values transferable from the specimen to the tubular joint, even in cases of slight differences of local geometry and loading type. This can be expressed
Component, Loading Residual stress Structural strain
~s (el.-pl.)
L

in such a way that the fatigue strength of the welded joint specimen in terms of nominal stress presents the limit value of the local structural stress in the tube joint. There are relations to the fracture mechanics approach insofar as the structural stress is well suited as the reference stress for crack propagation. The varying and scattering notch effect of the weld is of secondary influence within this approach. The transfer method mentioned above is bound to the following conditions: the point of crack initiation must be known in advance (named 'hot spot' because of local heating under cyclic loading). The crack must be initiated at the weld toe, i.e. not at the weld root. Transverse loading of the weld must prevail. The structural stress (or 'geometric stress') must unambiguously be separable from the notch stress. The notch effect and the welding quality in the tubular joint and in the comparable specimen must approximately be the same. The procedures for separating the structural stress from the notch stress are now considered in more detail. Following the procedure proposed by Dijkstra and Gurney et al., the axial surface stresses in the tube are measured at a small and at a larger distance from the weld toe notch and after that linearly extrapolated to the point of notch stress concentration. The exact position of the two measuring points is defined dependent on the diameter and wall thickness of the tube, i.e. dependent on the geometrical parameters which determine the stress field in the tube joint when modelled as a cylindrical shell structure (the distances from the weld toe notch are 0.2~1~ and 0.65x/~, for example, with tube diameter 2r and tube wall thickness t). The structural stress approach combined with the above separation procedure is embodied in the design rules for tube structures in offshore engineering. The approach which is well founded in the high-cycle fatigue range, i.e. in the range of predominantly elastic behaviour, becomes questionable when extended to the lower-cycle fatigue range, i.e. in the range of elasticplastic behaviour. The structural stress approach considered as a procedure which transfers the fatigue strength values of welded specimens to the local design within a structure is also applicable in cases where the structure is analysed and optimized according to the plate and shell theory, independent of whether simple engineering for-

~~ F
;I

AiStrain|[

m,I

0~ (9--

Structural strain endurance limit


=~== _ =....-, ' ' &~sE ~ 0 , 1 % R=-I

t nil,
I I

I -

' &'E z~
E~ O~ O~

Stress relieved
NE=2X 106
Mild steel

Figure 9

Structural strain approach according to Haibach for assessing the endurance limit of welded structural components

162

19. Radaj
Component, Loading Structural stress &as (el.) .~_A M &as Strain L ~gauge Component fatigue strength ~ . &ME Structural stress S-N curve

z~
E~

.&as

'~<~i" - - ' " ~ ' ~ ' / " ~ " '


N
Figure 10 Structuralstress approach according to Dijkstra and Gurney for assessing the fatigue strength of welded structural components in

offshore engineering mulae or more complex finite element methods are used. The theoretical approach suppresses the notch stress completely because of the assumption of linear stress distribution over the plate or shell thickness. According to the author3, this is a better way to define structural stresses because a single valued solution is possible. The structural stresses determined accordingly are somewhat higher than the values gained from linearly extrapolating the measured stress values at the surface. Structural stress calculations which use the measurement related procedure of surface stress extrapolation are sometimes chosen in order to apply the permissible structural stresses from the codes. The structural stress approach as described above has tentatively been applied in ship design also. The global and local geometric shape parameters of the ship structure are a major factor influencing its fatigue strength. They vary to a large extent because each ship is of a singular design. The influence of shape parameters is insufficiently taken into account by the nominal stress approach, whereas the structural stresses reflect the influence25. Another field of successful application of the structural stress approach are welded joints with small eccentricities caused by imperfect manufacture (imperfections) 26'27. The usage in respect of design-typical large eccentricities (e.g. coverplate ends on double-T-section girders, longitudinal stiffeners, cover plates in general) was less successful 28"29. The increase of surface stress at those structural components can be calculated in analogy to the model of elastically supported beam bending, just as it is the case with the tube shell problem. Based on such comparative calculations, the measuring points are recommended at a distance from the weld toe notch of 0.3-1.0t (with plate thickness t) 28"29. On the other hand, these investigations have shown that the linear extrapolation should be supplemented by a quadratic or even cubic term. This means that the stress increase in front of the considered welded joints occurs with various gradients and nonlinearities. A uniform schematic approach seems to be impossible. The structural stress approach of the considered type presupposes a pronounced notch effect at the point of crack initiation. It is to be well distinguished from a structural stress approach without a notch effect at this point. Crack initiation outside the weld notches in the notch-free area of the base metal should be aimed at by appropriate design and production measures.

Structural stress approach according to Rupp and Grubisic The structural stress approach according to Rupp and Grubisic 39"4 refers to spot welded joints in thin sheet metal structures of the type used in automotive engineering, Figure l l . The structural stresses and strains at the weld spot of the structural component (an axle suspension arm is shown in the figure) are set in relation to sustainable values derived from fatigue tests with a component-like specimen (the hatprofile box specimen and the H-shape specimen are shown in the figure). Structural stresses are chosen as fatigue relevant quantities which are determined from the resultant internal forces and moments in the jointing face of the weld spot (shear forces, transverse force and bending moments, the torsional moment is neglected) both in the structural component and in the specimen. Two types of potential fatigue fracture are considered in the assessment of fatigue strength and service life, the plate fracture occurring at weld spots of sufficiently large diameter and the nugget fracture prevailing at weld spots with unconservatively chosen diameter. The maximum radial structural stress on the inner plate surface at the weld spot edge is considered as initiating plate fractures. This stress is approximately equal to the maximum von Mises equivalent stress, which is considered as fracture-relevant in ductile materials. It is calculated from engineering formulae which are derived from a rigid core model of the weld spot. It can also be estimated on the basis of strains measured at the weld spot edge on the outer surfaces of the plate, thereby gaining information on the resultant weld spot forces and moments. The static and cyclic radial stresses from calculation or measurement are transformed to equivalent cyclic radial stresses with a zero static component based on the Haigh diagram. The transformed stresses from the different resultant weld spot forces and moments are superimposed. The maximum transformed stress is determined. Structural stress S-N curves are derived from the specimens which failed by a plate fracture. These curves differ a little with the plate thickness. The criterion for stopping

Fatigue strength assessment of nonwelded and we/ded structures


Component Loading
Plate fracture Structural stress Model: Plate with S-N curve rigid core (or strain gauge) AOr max J Structural stress in plate AOrmax

163

Load spectrum &F~.,~ N

Weld spot resultant forces Formulae, FEM

- - N

Component = F-N curve and service life curve a~


--,

Specimens

,]

Nugget fracture

I J

Model: Beam cross section


Structural stress

Structural stress S-N curve

~'~\~

I in nugget
AC 1max

]
& ~ N

Figure 11 Structural stress approach according to Rupp and Grubisic for assessing the fatigue strength and service life of spot welded structural components

the fatigue test is a definite loss of global stiffness of the specimen caused by cracks at the weld spot edge penetrating the plate thickness. Finally, the maximum radial structural stresses in the component are set in relation to the limit state stresses in the specimen. A specially defined structural stress at the nugget edge is considered as initiating nugget fractures. This stress is calculated from the resultant weld spot forces on the basis of the beam bending theory. The weld nugget is axially extended to a compact cylinder, the end faces of which are loaded by the weld spot resultant forces and moments acting in the middle plane of the plates. The equivalent stress is determined from the bending normal stress and the transverse shear stress according to the maximum normal stress hypothesis, differentiating between cyclic and static stress components. The cyclic stress with a non-zero static component is transformed to an equivalent cyclic stress with a zero static component. The maximum cyclic equivalent stress at the weld spot edge is extracted. This maximum structural stress is limited by an S - N curve gained from the specimen test results related to a nugget fracture. It is a feature of the considered approach that the calculation of damage accumulation is also performed on the level of the structural stresses. This calculation is performed in respect of plate fracture and nugget fracture for all the points of the weld spot edge, evaluating the different cutting planes in each point. This combination of edge point and cutting plane is considered as critical which gives the highest damage sum. The high amount of calculation connected with this local critical plane approach is necessary in cases where different, more or less arbitrary load-time-functions are superimposed within the structural component, e.g. in automotive engineering. The structural stress approach presented above is considered as valid in the medium and high cycle fatigue range (N/> 104). A weak point of the approach is the fact that only those cyclic stresses are introduced as fatigue relevant which can be attributed to resultant weld spot forces and moments in the jointing face. Referring to spot welded components, high cyclic structural stresses are possible also without resultant forces and moments in the jointing face, i.e. self-equilibrating eigenforces in the jointed plates. The question may

also be raised why the structural stress describes the limit condition for fracture sufficiently well despite the sharp notch at the weld spot edge with its additional stress concentration. There are two answers possible which are not completely uncontradictory: on the one hand, the notch effect at the weld spot edge can be described by stress intensity factors which are proportional to the local structural stresses times the square root of plate thickness3. One the other hand, the cracks which propagate to a plate fracture occur only partially directly at the weld spot edge. They are often displaced to the outside which may be explained in such a way that the notch effect of the weld spot edge is neutralized by favourable hardening effects and by compressive residual stresses. The described structural stress approach has been verified by applying it to the transversal swinging arm of a wheel suspension and to the cross-member of an engine mount34. It is at present introduced in the design and analysis departments of the German automotive industry. NOTCH STRESS AND STRAIN APPROACHES

Notch strain approach according to Lawrence The earliest numerical investigation on the local approach applied to welded joints has been performed by Lawrence and co-workers 35~41, Figure 12. Two types of welded joints are considered, a butt welded joint with crack initiation at the weld toe and a spot welded joint with crack initiation at the nugget edge. The fatigue process is subdivided into a crack initiation phase (crack depth about 0.5 mm, crack width about 2 mm) and a crack propagation phase thereafter. The crack initiation phase is covered by the notch strain approach and the crack propagation phase by the fracture mechanics approach. Referring to the notch strain approach, the fatigue notch factor of the welded joint characterizing the endurance limit, i.e. the fatigue strength for infinite life, is determined first. It depends on the radius of curvature of the notch in addition to the stress concentration factor. This radius is not well-known at the weld toe, weld root or nugget edge. Therefore, the largest possible fatigue notch factor is determined according to the microsupport hypothesis for sharp

164

D. Radaj
Notch geometry I Microsupport Welded joint Effective notch I effect (Peterson) Loading Residual stress stress, Kt from I Kf max from

i"

F]MIK'frP a"
Iworst case notch)

Crack initiation life

~Eel/2 orA~pl/2

t from K I, KII

Welded joint S-N curve (crack initiation)

(base material)

effect (Neuber)

da/dN )l,

Crack propagation life

N
Figure 12

A~/2

I/L

Notch strain approach according to Lawrence for assessing the fatigue strength and service life of welded joints

notches proposed by Peterson (worst case notch). A notch radius p ~ a* is attributed to the toe notch, the material parameter a* characterizing the microsupport effect (a* ~ 0.5 mm for steels). The stress concentration factor or fatigue notch factor, respectively, of the weld toe is calculated using the finite element method. The relevant factor for spot welded joints is determined on the basis of the stress intensity factors at the sharp nugget edge using the Creager formulae for blunt cracks. The fatigue strength for finite life is now determined within the notch strain approach based on the fatigue notch factor. The numerical analysis proceeds from the cyclic stress-strain curve taking relaxation effects into account, from the macrosupport formula according to Neuber applied to the fatigue notch factor (substituting the stress concentration factor) and from the strain S-N curve. The material parameters characteristic of the weld unaffected base material are used in all cases. The result of the analysis is the traditional S-N curve of the welded joint with crack initiation as the damage criterion presenting permissible nominal stresses. It is proposed for approximating the life under load spectra to numerically process the elastic or alternatively plastic local strains occurring under the load spectrum against the elastic or plastic part of the strain S-N curve in analogy to the Miner rule (damage parameters where still unknown at that time). Referring to the fracture mechanics approach, the residual life, i.e. the number of cycles between crack initiation and global fracture, is calculated using the Paris formula for crack propagation. The investigations performed by Lawrence and co-workers35-~ have given some insight into the relation between crack initiation and crack propagation life. It was shown that the relation is highly material dependent besides being dependent on the total number of cycles to fracture. The higher-strength steels are characterized by a longer crack initiation period. Another result refers to the effect of residual stresses. Residual stresses which reach the yield limit substantially change the fatigue strength especially of higher-strength steels, increasing it in the case of compressive stresses and lowering it in the case of tensile stresses.

Notch strain approach according to Heuler and Seeger The procedure chosen by Heuler and Seeger42 for assessing the fatigue life of a single-bevelled groove weld with fillet welds on both sides between web and inclined flange corresponds to the notch strain approach, Figure 13. Taking the microsupport effect into account is unnecessary in the considered case, because of the large radius of notch curvature. The P-N curve with damage parameter P according to Smith, Watson and Topper constitutes the basis for gaining the S-N curve and the service life curve of the structural component evaluating the stress and strain cycles at the weld notch. The cyclic material characteristic values of the base material, the heat affected zone and the filler material were determined separately, but the differences were small under the applied welding conditions which produced an even hardness distribution. The numerical analysis of stresses and strains in welded joints within the local approach just described has been substantially improved by Clormann2 and applied to the circumferential weld of a tube joint in order to demonstrate the applicability of the procedure. Notch stress approach according to Radaj The notch stress approach according to Radaj 3'43-48 is restricted to the determination of the endurance limit of the welded joint with possible extensions into the medium-cycle fatigue range, Figure 14. It is applicable both to seam welded and spot welded joints. A salient feature of the approach is the assessment of the microsupport effect at the most sharply toe and root notches of the different welded joints according to the Neuber hypothesis. The sharp notches in the crosssectional model have to be fictitiously rounded in order to get the fatigue-relevant uniaxial notch stress resulting in the fatigue notch factor of the welded joint by reference to the nominal stress. Despite the fact that the approach has been proposed for any radii of notch curvature it has mainly been applied as a worst case consideration in which the real notch radius is introduced as zero. The fictitious notch radius then results in pf ~ 1 mm in the case of welded joints of structural steel.

Fatigue strength assessment of nonwelded and welded structures


Welded joint Loading Notch geometry Notch stress Stress-strain path Neuber, Seeger)

165

%
I_) 10mm
N

I,/

~E/2

P-N curve PSWT

F-N curve and service life curve

Cyclic material data (base and filler material) /ta/2 I ~/2

Damage parameter

77

Pswr
E

Figure 13 Notch strain approach according to Heuler and Seeger for assessing the fatigue strength and service life of welded joints

Comp., Loading
Structural stress

Weld shape

Surface
roughness Reduction factor ~'r ~ 0.9

~--~--~

iA-~-

Component = endurance limit

Residual stress effect

(1) O--

I-------I
Notch support effect (Neuber) pf ~ p + sp"

Base material endurance limit

t4

Strength hypothesis (von Mises)

g -

Or-

E-o

Mild steel

p=0, R=-I p f ~ l mm
Figure 14 Notch stress approach according to Radaj for assessing the endurance limit of welded structural components

The procedure which was demonstrated by the example of a welded joint between two channel-section bars subjected to a torsional moment comprises the following steps. At first, the structural stresses or internal forces in the butting plates at the weld notches (toe, root, nugget edge) are determined neglecting the notch effect. This can be achieved by using engineering formulae, the finite element method or strain gauges. Second, the structural stresses or internal forces are transferred as external forces to the cross-sectional model of the welded joint with fictitiously rounded notches in order to determine the fatigue notch factors by applying the notch stress theory, the boundary element method or the finite element method. Fictitious notch rounding may produce a weakening of the crosssection in the case of thin plates which is not negligible any more. No weakening occurs if the fatigue notch factors are calculated proceeding from the stress intensity factors of the real slit- or crack-like notch. The strength reduction factor often used with welded joints results as the reciprocal value of the fatigue notch factor, but the fact has to be taken into account that the fatigue notch factor is defined in relation to an unnotched polished specimen, whereas the strength reduction factor refers to an unnotched specimen in the mill-scale-plated condition. Therefore, a correction factor has to be introduced when calculating the

strength reduction factor from the fatigue notch factor. This factor is approximately 0.9 for lower-strength steels. The multiaxiality of the stress state at the notch root is taken into account by the von Mises hypothesis as far as ductile materials are considered. Proceeding in an approximative manner, the same fictitious radius of curvature is applied in tensile, compressive, bending and shear loading of the notch root. A smaller radius has to be introduced in shear loading within a more accurate analysis. A static mean stress can be taken into account on the basis of the relevant Haigh diagram, but the microsupport hypothesis is not applicable with higher mean stresses. The material characteristic values of the base material are used, modified according to the hardness in the area of crack initiation if necessary. The result of the notch stress approach described above is the endurance limit of the considered welded joint in terms of nominal stress. It is recommended to correct the result in respect of residual stresses, to cover the medium-cycle fatigue range according to the normalized S-N curve and to assess the service life on the basis of the Miner rule. This means switching from the notch stress approach to the nominal stress approach. The application examples published by the author

166

D. Radaj
crane construction, according to the standard notch cases of the nominal stress approach (a box girder with longitudinally welded edges is shown in the figure). A large number of welded joints with a T- or Y-section and with a fillet weld or a single bevel butt weld have been investigated. Fatigue fractures were initiated at the weld toe and the weld root. The test results have been evaluated on a statistical basis. The following special features are connected with the modified approach: the fictitious radius of notch curvature is fixed with 1 mm for structural steel independent of the scattering actual radius, and the question is asked what the appertaining fatigue-relevant cyclic notch stress is at the limit cycle number No = 2 106. This technical endurance limit was determined a s mO'max E 500 N mm -2 with a failure probability of 50% for stress-relieved specimens made of structural steel (the true endurance limit being somewhat lower). The scattering of the strength values is explained from the scattering of different influence parameters: real radius of curvature, slope at the weld toe, height of weld stiffening, depth of penetration and surface roughness. The procedure is designated a 'mean stress scatter approach'8. The endurance limit of the structural component is derived from the notch stresses in the welded structure compared with the notch stress endurance limit in the test specimens--both stresses determined for a fictitious notch radius of 1 mm. Once more, it is recommended to switch to the nominal stress approach at this point, in order to assess the effect of residual stresses and to determine the S-N curve and the service life of the structural component. The modified notch stress approach described above is well suited for classifying unconventional welded joints according to the standardized notch cases of the nominal stress approach. This has been demonstrated not only for the hollow section girder in crane construction, but also on the occasion of analysing the damage at the GROWIAN wind power station 55. FRACTURE MECHANICS APPROACHES Fracture mechanics approach according to Maddox and Hobbacher Frequently it is considered a feature of welded joints that microcracks resulting from manufacturing meas-

(and reviewed in Ref. 3) refer to the channel-section bar joint mentioned above, to the transverse stiffener with and without a cutout at the web and fastened by a fillet weld transverse to the tensile chord of a doubleT-section girder46, to the stress relief effect of a groove in the flat end of a pressure vessel 47 and to the circumferentially welded joint of a boiler48. The investigations mentioned above aimed at the valuation of design measures and at the classification of the considered welded joint in respect of the standardized S-N curves in the codes. The notch stress approach is especially well suited to determine the influence of the local geometrical parameters of the welded joint on its fatigue strength. These parameters comprise the radius of notch curvature, the slope angle at the weld toe, the penetration depth and the plate thickness. It was proven by the approach for the first time that fillet welded joints with a sufficiently large gap length may have a fatigue strength in terms of displacement amplitudes, which is not lower than the strength of the butt welded joints. Prescribed displacement amplitudes instead of force amplitudes is a loading case often met with structural components. The endurance limit of the fillet welded cruciform joint, a favourite subject of research activities, is sufficiently well predicted by the notch stress approach dependent on the ratios of weld thickness to plate thickness and of gap length to plate thickness 49. Several further applications (butt welds, longitudinal stiffeners, corner joints) are related to the shipbuilding area 5. It has been checked whether the notch stress approach described above correctly classifies simple welded joints in respect of the standardized notch case scheme of the nominal stress approach. The comparative investigation was initiated by the International Institute of Welding (IIW) and successfully performed by Petershagen 51. another comparison with the data from the S-N curve catalogue edited by Olivier and Ritter52 had a convincing result in respect of mean values and scattering widths TM. Notch stress approach according to Olivier et al. Olivier et al. 53"54 have modified the notch stress approach according to Radaj and have proven its reliability more generally, Figure 15. The reason for the investigation was the uncertainty of how to classify longitudinally welded hollow section girders used in

Component

Loading~ AF

Weld shape Effective notch stress /~rnalx Component endurance limit ~&FE Residual stress J ~ effect J ~ z.~

0 Effective notch -.,J-, , --4,-- endurance limit ? = . ~ ~maxE ~500 R =-1 N/mm2 pf=l mm NE=2X 106 Mild steel (St 53-3)

Figure 15 Notch stress approach according to Olivier, Krttgen and Seeger for assessing the endurance limit of welded joints

Fatigue strength assessment of nonwelded and welded structures


ures and subjected to high tensile residual stresses are typical for the notches at the weld toe and weld root, especially when these are combined with undercut. Additionally, it is assumed that crack propagation actually starts with the first loading cycle, so that the crack initiation phase is completely neglected. Applying the fracture mechanics approach according to Maddox s6 and Hobbacher57, Figure 16, the component life is predicted according to N = C/(KAS) m proceeding from the cyclic stress intensity factor AKt, with factor C representing the influence of material and geometry on the propagating crack, with K being the stress concentration factor and with m ~ 3 characterizing the curve gradient in the Paris diagram. The stress concentration factor K can be related to notch stresses (K = At) or to structural stresses (K = Ks). The procedural variant with Ks is insufficiently substantiated in respect of physics, but acceptable under an engineering point of view when applied to a sharply notched seam and spot welds (according to Hirt and Kummer58). The procedural variant with Kt presumes that the decrease of the geometry factor Y with increasing crack length is taken into account using available approximative formulae. The decrease is extremely steep in the vicinity of the notch. The result of the crack propagation analysis starting with a small semielliptical initial crack and ending with this crack penetrating the plate thickness (axis ratio mostly assumed as constant) is the S-N curve of the welded joint under a constant load amplitude (curve gradient k = m) or the service life curve under varying load amplitudes (the result is identical with the Miner rule). The fracture mechanics approach which is undispensable for the assessment of defined larger imperfections (e.g. with the aim of proving fitness for purpose) or of design inherent gap planes in welded joints (e.g. in fillet welded cruciform joints)--the approach is incorporated in some codes for that purpose59,6--should be seriously objected if used for calculating the fatigue strength and service life of unconventional welded joints or for classifying them in respect of standardized notch classes, in both cases proceeding from very small, more or less fictitious microcracks and neglecting the crack initiation phase. These are the objections: 1. Microcracks in welded joints of high quality which

167

additionally propagate with the first loading cycles cannot be proven to generally exist in reality. Merely 'slag intrusions' have been detected in carefully performed metallographic investigations on higher strength steels 61,62. A noticeable crack initiation phase can be expected with such intrusions. The high tensile residual stresses at the weld notches, on the other hand, occur only with special types of welded joints, welding conditions and material parameters. The longitudinal stiffener joint is often used for demonstrating early crack initiation. An additional objection against the fictitious microcrack considered above is that the choice of the dimensions of the microcrack is subject to arbitrariness. . Considering those rare cases in which the microcrack is detectable and immediate crack propagation can be assumed, the cyclic stress intensity factor is unsuited for describing crack propagation. The result of the analysis, if performed, is highly unconservative. It is necessary to describe short crack propagation in the plastically deformed notch area based on the cyclic J-integral. Such a procedure is laborious in respect of numerical procedures, but possible in principle 63. . Crack propagation does not take place according to the simplifying assumption that the axis ratio of the semielliptically approximated surface crack remains constant. In general, microcracks are initiated at several positions simultaneously and united to a single crack at a later stage. Additionally, the single crack may propagate more in width than in depth. These processes can only partly be simulated numerically and complicate the analysis to such an extent that it cannot be handled easily in an engineering sense.

Fracture mechanics approach according to Radaj and Yuuki


The fracture mechanics approach as proposed initially for spot welded joints by Radaj 3'64-71 and independently by Yuuki and co-workers 72-74 is also applicable on welded joints with design inherent gap planes, such as cruciform joints, transverse stiffeners, overlap joints and cover plates. The approach has been

Welded joint Loading Residual stress

Notch effect Geometry factor Y

"i

i'-

Crack propagation
S-N c u r v e

Service life
curve

Crack propagation (Paris-Erdogan)

Crack configuration I ~ai

Figure 16 Fracture mechanicsapproach according to Maddox and Hobbacher for assessing the fatigue strength and service life of welded joints

168

D. Radaj
shear forces, the transverse force, the bending moments and the torsional moment in the jointing face, are determined both for the welded joint in the structural component (the front pillar and a quarter of the roof of a car body are shown) and in the specimen. This is achieved by engineering formulae in simple cases and by the finite element method in more complex situations (a coarse mesh is sufficient). Cyclic and static resultant force components have to be separated. The structural stresses at the weld nugget edge are derived from the resultant forces, or from finite element analysis directly (membrane and bending stresses, longitudinal and transverse shear stresses, these stresses acting in pairs with the same or opposite direction in the two plates). The stress intensity factors of mode I, II and III are determined based on the symmetrical and antisymmetrical stress pairs in the two plates applying simple formulae 3 or more laborious finite element and boundary element modelling TM. The equivalent stress intensity factor and the expected direction of cracking at the weld nugget edge follow from the above stress intensity factors. The maximum equivalent stress intensity factor (referred to the characteristic resultant force value) depends on the type of loading of the weld spot and on the geometrical parameters of the welded joint, i.e. on plate thickness and nuggest diameter among others. The ratio of plate thickness has additionally to be taken into account in cases of unequal plate thickness and also the ratio of the elastic moduli in cases of dissimilar materials. Finally, a scatter band of the sustainable maximum stress intensity factor plotted against number of cycles can be derived from the results of fatigue tests with spot welded specimens. This diagram is then used to assess the fatigue strength of spot welded structural components. Alternatively, the normalized S-N curve may be applied on the basis of the endurance limit determined from fracture mechanics, as far as this curve is available for spot welded joints. The service life curve can be calculated according to the relative Miner rule in a final step. The fracture mechanics approach according to Yuuki is the prevailing method of numerical fatigue strength assessment for spot welded joints in the Japanese automotive industry, whereas the German automotive industry prefers the structural stress approach according

applied to fillet welded overlap joints by Partanen


et al. 75

The maximum value of the equivalent cyclic stress intensity factor at the weld nugget edge or at the seam weld root is considered as the decisive parameter characterizing the fatigue strength and service life of specimens and structural components. The equivalent stress intensity factor is determined from the stress intensity factors of the gap opening modes I, II and III according to a relevant strength hypothesis. The latter can be chosen according to proposals by Irwin, Erdogan or Sih. These result mainly in a simple quadratic form correlating the mode-related factors. A threshold value of the cyclic equivalent stress intensity factor is considered as determining the fatigue strength for inifinite life of the welded joint. Crack propagation takes place if the threshold value is exceeded. The high-cycle fatigue strength up to fracture can be expressed by the initial stress intensity factor at the gap edge as far as the number of cycles between crack initiation and the stage of plate penetration by the crack remains approximately proportional to the initial stress intensity. This condition has been proven to be correct by Smith and Cooper 76 in the case of tensileshear loading of the weld spot. The approach presumes that the notch at the weld nugget edge or seam weld root is crack-like in respect of geometry and that crack initiation is controlled by the local stress intensity, which means that superimposed non-singular stress components have no influence. These assumptions are supported by reality mainly in the range of the endurance limit. The sustainable stress intensity factors for welded joints cannot be simply set equal to the base material characteristic values because the material at the weld nugget edge or seam weld root, respectively, is affected by welding, inhomogeneous in structure and influenced by residual stresses. Actually, these limit stress intensity factors are determined experimentally using welded-joint specimens. The approach presupposes that the stress intensity factors can be calculated with sufficient accuracy both for the specimen and the structural component. The essential steps of the approach when applied to spot welded joints are shown in Figure 17. At first, the resultant forces at the weld spot, especially the

Component Loading Weld spot resultant forces Formulae, FEM Structural stress Stress intensity at weld spot pairs at weld spot ,,O.b++ % + &Kt, KII, Kill, Keq
(3 + - , 'tl + -

Component fatigue strength

z8 ,i= = E-o Or-

l:q++, "lTq +AKeq/&F

S&ecimes
l i

K-N curve

t
'-:.::::::::.:::::

etc.

Figure 17 Fracture mechanics approach according to Radaj and Yuuki for assessing the fatigue strength of spot welded structural components

Fatigue strength assessment of nonwelded and welded structures


to Rupp and Grubisic. A weak point of the fracture mechanics approach is the very limited availability of sustainable stress intensity factors. But relative evaluations are possible, even with stress intensity factors alone which give an answer to questions regarding the design of specimens and components. The different spot welded specimens, for example, have been compared in respect of their endurance limit based on the above procedure 3'66. The H-shaped specimen was evaluated in this way on demand of the International Institute of Welding (IIW) 77. A novel double-cup specimen was also analysed78. CONCLUSIONS
The following conclusions are drawn:
26 20

169

21 22 23

24 25

1. The local approaches of fatigue strength assessment for welded structures are essentially supporting the design, dimensioning and optimization of those structures. 2. There are quite a number of local approaches available for assessing the fatigue strength and service life of welded structures based on structural stresses, notch stresses and fracture mechanics. 3. The state of development, the contents and the range of applicability of the approaches are rather different. This fact, and the variety of the approaches, make it difficult to standardize and verify them for industrial application and inclusion into the design codes. A new potentiality of assessing the fatigue strength and service life of welded joints arises from short crack fracture mechanics. REFERENCES
1 2 3 4 5 6 7 Neuber, H. 'Kerbspannungslehre'. Springer-Verlag, Berlin, 1937, 1958, 1985 Radaj, D. 'Ermtidungsfestigkeit~rundlagen fiir Leichtbau, Maschinen- und Stahlbau'. Springer-Verlag, Berlin, 1995 Radaj, D. 'Design and Analysis of Fatigue Resistant Welded Structures'. Abington Publishing, Cambridge, 1990 Buxbaum, O. 'Betriebsfestigkeit'. Verlag Stahleisen, Diisseldoff, 1986 Haibach, E. 'Betriebsfestigkeit--Verfahren und Daten zur Bauteilberechnung'. VDI-Verlag, DUsseldorf, 1989 Haibach, E. 'Betriebsfeste Bauteile'. Springer-Verlag, Berlin, 1992 Iida, K. "Application of hot spot strain concept to fatigue life prediction'. IIW-Doc. XIII-1103-83. International Institute of Welding, 1983 K6ttgen, V.B., Olivier, R. and Seeger, T. Schwingfestigkeitsanalyse fiar SchweiBverbindungen auf der Grundlage 6rtlicher Beanspruchungen. In: 'Expert '91' (Ed.: Radaj, D.), DVS-Berichte, Vol. 133, pp. 75-84. DVS-Verlag, DUsseldorf, 1991 Radaj, D. 'Lokale Konzepte des Betriebsfestigkeitsnachweises ftir SchweifSkonstruktionen--Grundlagen und Anwendungen'. DVM-Berichte, AK Betriebsfestigkeit, 1994, pp. 15-45 Kloos, K.H. "Kerbwirkung und Schwingfestigkeitseigenschaften'. DVM-Berichte, AK Betriebsfestigkeit, 1989, pp. 7-40. Neuber, H. Konstruktion 1968, 20, 245 Seeger, T. and Beste, A. Fortschritt-Berichte, VD1-Z. 1977, 18, 1 Lang, O. Z. Werkstofftechnik 1979, 10, 24 Paris, P.C. and Erdogan, F. Trans. ASME, J. Basic Engng 1963, 85, 528 Graf, T. and Zenner, H. Materialprufung 1994, 36, 71 Haibach, E. 'Praktische Umsetzung des Betriebsfestigkeitsdenkens'. DVM-Berichte, AK Betriebsfestigkeit, 1988, pp. 7-22 Sonsino, C.M. Konstruktion 1993, 45, 25 Seeger, T. and Zacher, P. Bauingenieur 1994, 69, 13 Kloos, K.H. Adelmann J., Bieker, G. and Oppermann T.G. VDIBerichte 1988, 661, 215

27

28

29

30

31

32

33 34

35

36

37

38

39 40

10 11 12 13 14 15 16 17 18 19

41

42 43 44 45 46 47 48 49

Clormann, U.H. 'Ortliche Beanspruchungen von Schwei/3verbindungen als Grundlage des Schwingfestigkeitsnachweises'. Ver6ffentlichungen d. Instituts f. Stahlbau u. Werkstoffmechanik, TH Darmstadt 1986 Schwalbe, K.H. and Kocak, M. (Eds.) 'Mismatching of Welds'. Mechanical Engineering Publishers, London, 1994 Haibach, E. 'Die Schwingfestigkeit von Schweigverbindungen aus der Sicht einer 6rtlichen Beanspruchungsmessung'. LBFBericht no. FB-77. Lab. f. Betriebsfestigkt, Darmstadt, 1968 Dijkstra, O.D. and de Back, J. 'Fatigue strength of welded tubular T- and X-joints'. In Proc. 12th Offshore Technology Conf., OTC Paper no. 3639, Houston, May, 1980 Gurney, T.R. 'Fatigue of Welded Structures'. Cambridge University Press, Cambridge, 1979 Petershagen, H., Fricke, W. and Massel, T. 'Application of the local approach to the fatigue strength assessment of welded structures in ships'. IIW-Doc. XIII-1409-91. International Institute of Welding, 1991 Iida, K. and Iino, N. 'Effect of angular distortion on fatigue strength of transverse butt-welds in high strength steels'. IIW Doc. XIII-827-77. International Institute of Welding, 1977 Iida, K., Ochiai, O., Kawai, T., Hirakawa, K., Fujisciwa, K, and Okazawa, T. 'Effect of misalignment in welded joint on low cycle fatigue strength of hemispherical pressure vessel made of high tensile strength steel'. IIW Doc. XIII-815-77. International Institute of Welding, 1977 Partanen, T., Tarjavuori, P. and Niemi, E. 'Definition of the hot spot stress gradient in stiffened plate structures based on the beam on elastic foundation theory'. IIW-Doc. XII-1478-92. International Institute of Welding, 1992 Niemi, E. 'On the determination of hot spot stresses in the vicinity of edge gussets'. IIW-Doc XIII-1555-94. International Institute of Welding, 1994 Rupp, A. 'Beanspruchung und Beanspruchbarkeit von PunktschweiBverbindungen unter Schwingbelastung---ein Auslegungsverfahren. Dr.-Ing.' Thesis, TU Braunschweig, 1992 Rupp, A., Grubisic, V., St6rzel, K. and Steinlulber H. 'Ermittlung von ertragbaren Schnittkr~iften ffir die betriebsfeste Bemessung von Punktschweil3verbindungen im Automobilbau'. FAT-Schriftenreihe, no. 78, Frankfurt, 1989 Rupp, A., Dilthey, U., Marek, U., Graffe, M., Graf, G. and Dittmann K.-J. 'Ermittlung ertragbarer Beanspruchungen am SchweiBpunkt auf Basis der tibertragenen SchnittgrrBen.' FATSchriftenreihe, no. 111, Frankfurt, 1994 Rupp, A., Grubisic, V. and Radaj, D. MaterialpriiJung 1990, 32, t 75 Rupp, A., Grubisic, V. and Radaj, D. 'Rechnergesttitzte Auslegung punktgeschweiBter Bauteile'. DVM-Berichte, AK Betriebsfestigkeit, 1994, pp. 159-175 Mattos, R.J. and Lawrence, F.V. Estimation of the fatigue crack initiation life in welds using low cycle fatigue concepts. SAE SP424. Society of Automative Engineering, Warrendale, PA, 1977 Lawrence, F.V., Mattos, R.J., Higashida, Y. and Btirk, J.D. 'Estimation of Iatigue initiation life of welds'. ASTM STP 648, 1978, pp. 134-158 Lawrence, F.V. and Mazumdar, P.K. 'Application of straincontrolled fatigue concepts to the prediction of weldment fatigue life'. DVM-Berichte, AK Betriebsfestigkeit, 1979, pp. 468-478 Lawrence, F.V., Burk, J.D. and Yung, J.Y. 'Influence of residual stress on the predicted life of weldments'. ASTM STP 776, 1982, pp. 33-43 Yung, J.Y. and Lawrence, F.V. Fatigue Fract. Engng Mater. Struct. 1985, 8, 223 Lawrence, F.V., Wang, P.C. and Corten, H.T. An empirical method for estimating the fatigue resistance of tensile shear spot welds. SAE TPS-830035, Society Automotive Engineering, Warrendale, PA, 1983 Wang, P.C., Corten, H.T. and Lawrence, F.V. 'A fatigue life prediction method for tensile-shear spot welds'. SAE TPS850370, Society of Automotive Engineering, Warrendale, PA, 1985 Heuler, P. and Seeger, T. Konstruktion 1983, 35, 21 Radaj, D. Konstruktion 1984, 36, 285 Radaj, D. Konstruktion 1985, 37, 53 Radaj, D. VDl-Berichte 1988, 661, 67 Radaj, D. Stahlbau 1985, 54, 243 Radaj, D. Konstruktion 1986, 38, 237 Radaj, D., Gerlach, H. and Gorsitzke, B. Konstruktion 1988, 40, 447 Petershagen, H. 'A comparison of two different approaches to

170

D. Radaj
the fatigue strength assessment of cruciform joints'. IIW Doc. XIII-1410-91. International Institute of Welding, 1991 Petershagen, H. 'Erfahrungen mit dem Kerbspannungskonzept nach Radaj'. DVM-Berichte, AK Betriebsfestigkeit, 1989, pp. 89-101 Petershagen, H. 'A comparison of different approaches to the fatigue strength assessment of welded components'. IIW-Doc. XIII-1208-86. International Institute of Welding, 1986 Olivier, R. and Ritter, W. 'WiShler-Linien-Katalog fiir SchweiBverbindungen aus Baust:ahlen'. DVS-Bericht no. 56. DVS-Verlag, Diisseldorf, 1979-1985 Olivier, R., KOttgen, V.B. and Seeger, T. 'SchweiBverbindung I.' FKM-Forschungshefte no. 143, Frankfurt, 1989 Olivier, R., Ktittgen, V.B. and Seeger, T. 'Schweil~verbindung II' (Schwingfestigkeitsnachweise). FKM-Forschungshefte no. 180. Frankfurt, 1994 Ktttgen, V.B., Olivier, R. and Seeger, T. Konstruktion 1993, 45,1 Maddox, J. Weldg. J. 1974, 53, Res. Suppl., 401 Hobbacher, A. Archiv Eisenhattenwesen 1977, 48, 109 Hirt, M.A. and Kummer, E. VDI-Berichte 1978, 313, 539 'Recommendations for the fatigue design of steel structures'. ECCS-TC6-Fatigue. Brtissel, 1985 Hobbacher, A. 'Recommendations for assessment of weld imperfections in respect to fatigue'. IIW-Doc. XIII-1266-88. International Institute of Welding, 1988 Signes, E.G., Baker, R.G., Harrison, J.D. and Btirdekin F.M. British Weldg. J. 1967, 14, 108 Graville, B.A. and Watkinson, F. 'Research programme to investigate the fatigue of high strength steels: initiation in fillet welds'. In: BWRA Report 1967, C 215/3/67 63 64 65 66 67 68 69 70 71 72 Vormwald, M. 'AnriBlebensdauervorhersage auf der Basis der Schwingbruchmechanik for kurze Risse'. Vertffentl. Inst. f. Stablbau u. Werkstoffmech. d. TH Darmstadt, no. 47, 1989 Radaj, D. ,,onstruktion 1986, 38, 41 (Part 1) Radaj, D. Konstruktion 1986, 38, 397 (Part 2) Radaj, D. Konstruktion 1987, 39, 51 (Part 3) Radaj, D. Konstruktion 1988, 40, 159 (Part 4) Radaj, D. Konstruktion 1989, 41, 255 Radaj, D., Zheng, Z.H. and MiShrmann, W. Engng Fract. Mech. 1990, 37, 933 Radaj, D. Weldg. World 1990, 28, 29 Radaj, D. Stahlbau 1990, 59, 201 Yuuki, R., NakatsOkasa, H. and Ohira, T. 'Fracture mechanics analysis of fatigue strength of spot welded joints (in Japanese).' Proc. Conf. Automobile Technology 1985, Vol. 31, Res. Rep. no. 10, pp. 71-77 Yuuki, R., Ohira, T., NakatsOkasa, H. and Yi, W. Trans, Jap. Soc. Mech. Engng 1985, 51, 1772 Yuuki, R. and Ohira, T. 'Development of the method to evaluate the fatigue life of spot-welded structures by fracture mechanics'. IIW Doc. XIII-1358-89. International Institute of Welding, 1989 Partanen, T., Lindberg, I. and Niemi, E. 'On fatigue behaviour of transverse fillet welds in lap joints'. IWW-Doc. XIII-150593. International Institute of Welding, 1993 Smith, R.A. and Cooper, J.F. 'Fatigue of Welded Constructions,' The Welding Institute, Abington, Cambridge, 1988, pp. 287-295. Radaj, D. and Giering, A. Materialpriifung 1994, 36, 275 Radaj, D. and Giering, A. Materialpriifung 1995, 37, 10

50 51 52 53 54 55 56 57 58 59 60 61 62

73 74

75

76 77 78

You might also like