International Journal of Fatigue 33 (2011) 11181126

Contents lists available at ScienceDirect

International Journal of Fatigue

journal homepage: www.elsevier.com/locate/ijfatigue

Crane runways Fatigue evaluation of crane rail welds using local concepts
M. Euler, U. Kuhlmann
University Stuttgart, Institute of Structural Design, Pfaffenwaldring 7, 70569 Stuttgart, Germany

a r t i c l e

i n f o

a b s t r a c t
The introduction of the wheel loads into a crane runway with a hot rolled I-section and a block rail fastened by longitudinal llet welds forms a multiaxial fatigue problem. At the point of wheel load application the top region of the runway girder is subjected to a stress eld comprising local stress components induced by the concentrated load, i.e. local transverse pressure and local shear stress in addition to the direct and shear stress components due to global bending. The superposition of all aforementioned stress components (and stress ranges, resp.) during each single crane pass results in a so-called non-proportional cyclic stressing of the rail welds with alternating principal stress direction. Due to the variation of the principal stress direction the conventional fatigue evaluation procedures appear not to be reliable. To approach the complex issue two crane runway girders were tested under proportional multiaxial loading, i.e. under exural bending and stationary wheel load, both simultaneously pulsating. The paper presents the fatigue evaluation of these tests using different local concepts. 2011 Elsevier Ltd. All rights reserved.

Article history: Received 27 October 2010 Received in revised form 16 February 2011 Accepted 20 February 2011 Available online 23 February 2011 Keywords: Crane runway Crane rails Multiaxial fatigue Wheel loads Local concept

1. Introduction The vast majority of crane runways for top running overhead travelling cranes are made of steel, Fig. 1a. Approximately 70% of all crane runways are designed for light and moderate crane service and usually consist of a hot-rolled section (mild steel, steel grade S 235, S 355) with a crane rail fastened to the top ange by automatic llet welds, Fig. 1b. Continuous static systems are quite common, Fig. 2. At the point of wheel load application the top region of the runway girder (Fig. 1b) is generally subjected to a stress eld comprising local stress components induced by the concentrated load, i. e. local transverse pressure rz,local and local shear stress sxz,local, in addition to the global stress components rx and sxz due to global bending. For example the top region at location x of the continuous crane runway shown in Fig. 2 is subjected to tensile bending stress superposed by the local stresses due to wheel load application.

the stress history at particular locations of a two-span and single-span crane runway girder caused by a travelling crane assuming constant travelling velocity. Since the crane possesses two crane wheels each passage of the crane causes two wheel passages. Fig. 3a indicates that all stress components do not reach their amplitudes simultaneously for the two-span girder. In case of the single-span girder (Fig. 3b) the amplitudes of rx and rz,local appear at the same time despite that the frequency of rz,local is twice as large as for rx. The fact that the stresses rz,local and sxz,local have to be inherently out-of phase is illustrated in Fig. 3c. In summary, the stressing of both considered crane runway systems indicates a permanent shift of the principal stress directions. From that it can be concluded that crane runways are usually stressed non-proportionally. 1.2. Stress-rising effects of crane runways Basically, the nominal stresses of a crane runway differ from the real stressing for three reasons:

1.1. Proportional and non-proportional loading A stress history of two stress components rA and rB is referred to as proportional if it holds rA/rB = constant. That results in a constant direction of the principal stresses in time. In contrast, a stress history is called non-proportional if the direction of the principal stresses changes throughout the stress history. Fig. 3 illustrates
Corresponding author. Tel.: +49 711 685 66246.
E-mail address: u.kuhlmann@ke.uni-stuttgart.de (U. Kuhlmann). 0142-1123/$ - see front matter 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.ijfatigue.2011.02.010

1. The welding of the crane rail leads to material homogeneities, and residual stresses are induced by the welding procedure. 2. The shape of the weld toes and roots forms geometric notches amplifying the nominal stresses. 3. The concentrated load causes local stress peaks as described above. Altogether, the crane runway undergoes a complex multiaxial state of stress with several stress risers of different nature.

M. Euler, U. Kuhlmann / International Journal of Fatigue 33 (2011) 11181126

1119

Nomenclature Stresses rx, ry, rz normal stress components sxz, syz shear stress components Dr, Ds stress ranges Drc, Dsc fatigue strengths Indices of stresses a amplitude eq equivalent value due to DEH local due to wheel load mean mean value eff effective value mod modied nom nominal notch related to effective notch stress concept Other symbols as notch stress concentration factor d DMA fs F, G F0, F1 k Nf phase angle multiaxial damage sum size factor terms related to EESH single loads coefcient number of cycles to failure angular velocity normal stress probability of survival ctitious notch radius stress ratio distance plane angle time or notch deepth

x
p Ps

qref
R s h t

crane

crane wheels

(a) (b)

crane runway

crane runway

stresses of the member rather than on the complex local stresses at the crack-causing notch. Its simplicity makes it favourable for the practitioners while its limitation on only known and well-studied constructional details (detail categories) forms its greatest disadvantage. To apply the nominal stress concept to a multiaxial stress problem requires a failure hypothesis translating the multiaxial state of stress into an equivalent uniaxial one. This is necessary since all

crane wheel crane rail

xz

continuous fillet weld

xz

xz,local

(a)

s A B x

s = travelled distance t = elapsed time 2,00 m

crane runway girder

x
z,local
x

wheel B passing
x

wheel A passing
xz,local

s, t

Fig. 1. Overhead travelling crane on crane runways (a) and typical cross section of a crane runway for light and moderate crane service (b).
x

z,local

s, t

(b)
A x B

(c)

2,00 m

xz,local

s, t
wheel A passing

wheel B passing

z,local

s, t
Fig. 2. Multiple-span crane runway with two crane wheel loads.

1.3. Current global approaches and conventional theories of failure The nominal stress concept also known as classication method is the most commonly used global approach to assess fatigue problems. This method is called global since it refers to the nominal
Fig. 3. Stress history caused by one crane passage at location x = 4.00 m in case of a two-span runway (2 4.00 m) (a), at midspan location x = 2.00 m in case of a single-span runway (b) and three states of stress for particular wheel locations (c). Remark: global shear stress sxz is not considered for the sake of simplicity.

1120

M. Euler, U. Kuhlmann / International Journal of Fatigue 33 (2011) 11181126

strength parameters of the detail categories are obtained from uniaxial tests. Principally, the choice of an appropriate failure hypothesis has to account for the failure mechanism. From static tests on ductile materials such as mild steel it is known that the damage is primarily initiated by shear stress. That is why the distortion energy hypothesis (DEH) of von Mises is commonly employed. It is also proven that this hypothesis holds for ductile materials under cyclic loading [11]. Moreover, Gough and Pollard recognized that the results of combined stress fatigue tests of unwelded small-scale specimens (normal stress due to bending, shear stress due to torsion) are in close agreement with the von Mises hypothesis that has the form of an ellipse,

the same fatigue failure. As a rule the summation of the individual damages follows [3] according to Eq. (3) with Dr\ as nominal stress range perpendicular to the fatigue crack, Ds as nominal shear stress range and Drc\, Dsc as the corresponding fatigue strengths [3].

 DMA

Dr? Dr c ?

3

Ds Dsc

5 61 3

Dr Dr c

2

Ds Dsc

2 61 1

Because of the required exclusion of simultaneously occurring amplitudes Eq. (3) is not able to evaluate multiaxial loadings safely where the single stress components interact [2]. In summary, the conventional global approaches seem to be not promising to evaluate non-proportional loadings with out-of phase acting stress components such as shown in Fig. 3. 1.4. Local approaches and modern theories of failure The crack initiation is a local problem and requires local concepts. Local concepts can be based on structural stresses, notch stresses, notch strains as well as the crack propagation or stress intensity [10]. This paper focuses on the so-called effective notch stress approach. In contrast to the global methods that refer to the members nominal stress the notch stress method uses elastic stresses at weld notches, e. g. the weld toe. Since notch stresses are hardly measurable they are determined numerically for instance by a Finite Element Analysis (FEA). The appropriate modelling of the weld notches is the crucial point. The exact weld geometry can be obtained from measurements on microsections, Fig. 4. Alternatively, a concept of reference radii can be applied, e.g. qref = 1 mm according to Radaj and Neuber. Through the FEA the local stresses can be computed and nally transformed into an equivalent stress in terms of an appropriate failure hypothesis. In contrary of the notch strain approaches that account for elastic-plastic material behaviour in the vicinity of the notch (e.g. Neuber rule) the effective notch stress approach assumes fully elastic behaviour resulting in an apparent conict that the calculated stresses at the notches partially exceed the yield stress. But these calculated stresses do not become completely effective since the less stressed material surrounding the maximum stressed notch region induces a constraint leading to a multiaxial state of stress

The application of DEH for cyclic loaded structures with welded joints needs some adjustment to account for the material inhomogeneity in the vicinity of welds. In detail, the assessment is supposed to differentiate in terms of the fatigue strengths perpendicular to and parallel of the weld in form of Eq. (2) where Drjj, Dr\ are the nominal stress ranges jj weld and ? weld, Ds is the nominal shear stress range, and Drcjj, Drc\, Dsc are the corresponding fatigue strengths.

Drjj Drcjj

2

Dr? Dr c ?

2

  Drjj Dr? Ds 2 61 Drcjj Drc? Ds c

It appears for several combined loadings (bending/axial loading plus torsion) that the nominal stress concept in combination with DEH gives only reliable results for in-phase stressing, i.e. in case of simultaneously occurring amplitudes of the stress components and therefore xed principal stress directions. For out-of-phase loading (i.e. shifting principal stress directions) this approach might severely overestimate the actual fatigue resistance since the equivalent stress becomes lower expecting a smaller fatigue damage. That is opposed to the experience that non-proportional loading has been identied to be more damaging than proportional loading under comparable conditions. If the single stress components occur after each other in the course of time, the different stress components do not interact and therefore the damages done by the single stress components can be summed up provided all stress components contribute to

Fig. 4. Weld modelling for test girder No. 2: microsection of weld A (a and b), corresponding modelling in FEA (c), microsection of weld B (d and e) and corresponding modelling (f).

M. Euler, U. Kuhlmann / International Journal of Fatigue 33 (2011) 11181126

1121

with a higher resistance compared with that of the unrestrained uniaxial one (micro-support effect) [13]. As pointed out before the conventional failure hypotheses cannot assess the damage due to out-of-phase loading causing changing principal stress directions. A couple of new failure hypotheses tries to overcome this shortcoming. Basically, the existing approaches fall into two groups. While the rst group, the so-called integral approaches, presumes that the overall damage is the integration of the damages on different planes at a location, the critical plane approaches forming the second group expect that the plane with the largest alternating shear stress (i.e. critical plane) controls the fatigue failure and coincides with the preferential orientation of the early crack growth [1]. The hypothesis of Findley [4] belongs to the oldest critical plane approaches. It was originally derived for combined normal stress due to bending/axial loading and shear stress due to torsion and assumes that the fatigue failure is initiated in the plane with the maximum alternating shear stress sa. Additionally, it accounts for the inuence of the normal stress rmax on this plane by calculating an effective alternating shear stress seff,a, Eq. (4)

(a)

(b)

(c)

Fig. 5. Test set-up (a), cross section (b) and interface between crane rail and upper ange (c).

Dseff seff ;a sa k rmax 2

Note that sa is the shear stress amplitude (i.e. half the stress range) in contrary to rmax that is the sum of alternating and mean stress. For steel it holds k = 0.3 [4]. The effective alternating shear stress is increased by tensile (positive) normal stress and decreased by compressive (negative) normal stress. Moreover, Findleys hypothesis delivers reliable results under proportional but fails under nonproportional (out-of-phase) loading. The hypothesis of the effective equivalent stress of Sonsino is a further modern theory of failure [17]. It also expects that the fatigue failure is initiated by shear stress. In contrast to Findley it accounts additionally for the interaction of the shear stress in different planes initiating corresponding dislocations [16]. That is why this hypothesis is an integral plane oriented approach [15]. Its major scope is to evaluate out-of-phase loading, and its algebraic algorithm is described in [17] for example. Both hypotheses mentioned above will be applied for the particular situation of crane runways in the following.

= 178 N/mm

= 144 N/mm

Fig. 6. State of stress in terms of nominal stresses.

(a)
PC

(b)
SC

2. Experimental investigation
SC

2.1. Test set-up In a research project of the Deutsche Ausschuss fr Stahlbau (DASt) under funding of AiF as part of the IGF-program of the German Federal Ministry of Economics and Technology (BMWi) the fatigue strengths of crane runway girders with particular focus on the load application were determined [6]. Only the two test girders Nos. 2 and 3 with continuous rail welds are going to be considered here. The girders were tested in a servo-hydraulic test rig with two axial actuators under load control. The testing frequency amounted to about 12.5 Hz. The load ratio was R  0.1. The applied load ranges measured DF0 = 460 kN and DF1 = 160 kN, Fig. 5a and b. All loads acted in phase with constant amplitudes. The bottom rail surface (Fig. 5c) was recessed to avoid any contact between the rail and the girder and to ensure that the load transfer from the rail to the top ange was exclusively realized by the rail welds. Hence, all loads acting on the top surface of the rail were transmitted into the top ange by the rail welds. The treated bottom rail surface was kept rough in order to simulate the defects of the mill scale. A more detailed description can be found in [6,7].

(c)

PC SC

primary crack secondary crack

Fig. 7. General failure mode: phase 1 (a), phase 2 (b) and phase 3 (c).

With the test set-up only an interaction of longitudinal normal stress rx and transverse pressure rz could be achieved at the point of wheel load application, Fig. 6. It did not allow the superposition of these stress components together with the local shear stress sxz,local, Fig. 3c.

2.2. Failure modes The observed failure mode comprised three phases, Fig. 7; see also Fig. 8. The so-called primary crack was initiated at the bottom

1122

M. Euler, U. Kuhlmann / International Journal of Fatigue 33 (2011) 11181126

(a)
rail flange

wheel

(b)

wheel rail

flange

view

view

Fig. 8. Failure of test girder No. 2 (a) and No. 3 (b).

surface of the rail right below the wheel load (phase 1). After the rail through-crack the rail welds failed in a ductile manner (phase 2). Subsequently, the crack propagated into the top ange leading to two crack growth centres (phase 3). Presumably, during phase 2, secondary cracks to each side of the rail developed right beneath the wheel load along the weld toes, see Fig. 7b. These cracks were hardly detectable. Their fracture plane was oriented normal to that of the primary crack. The number of load cycles at crack initiation in the rail was about 700,000 for test girder No. 2 and about 640,000 for test girder No. 3. Note that in [6] and [7] the load cycles to failure are 762,000 and 700,000 since the failure criterion was rail through-cracking instead of crack initiation there.

3. Numerical investigation 3.1. Finite element analysis The fully elastic notch stresses were computed by a FEA using the software package ANSYS Version 11. Basically, the stress

determination was subdivided into two steps. Firstly, a global model composed by solid elements (solid45) with a coarse mesh was set up comprising the tested crane runway girder between the supports together with the crane wheel taking advantage of the symmetry, Fig. 9a. The external bending moments were simulated by pairs of equivalent forces. The wheel load at midspan was introduced through the crane wheel. The interface between crane wheel and crane rail were modelled by contact elements (contac174, target170). Secondly, a 3-D submodel was created in order to rene the results of the global model, Fig. 9b. The exact geometry of the welds was obtained by an optical measurement, see Fig. 4. Moreover, the weld toes and the weld root were rounded off with a ctitious radius qref = 1 mm according to Radaj and Neuber, Fig. 9c. Due to the irregular shape of the submodel tetrahedral solid elements (solid92) were used. The thickness of the submodel amounted to 6 mm. The outer surfaces of the submodel were subjected to the boundary conditions caused by the global model. Consequently, unavoidable disturbances occurred there due to the interpolation between both models. Therefore, only the stresses at the centre

(a)

(b)
I II VI VII

Node ... II VI VII

yz

...
... ... ... ...

-41 ... -215 -226

-712 ... -446 -343

155 ... 305 273

FE stresses

(c)

(d)

upper weld toe

lower weld toe

weld root

Fig. 9. Contours of global model (a) and of submodel (b), meshed weld of submodel (c) and paths of stress determination at the centre section of submodel (d).

M. Euler, U. Kuhlmann / International Journal of Fatigue 33 (2011) 11181126

1123

Table 1 Calculated notch stresses at the welds A and B [N/mm2]. Remark: the values of weld B in parentheses. Weld root Weld toe low 138 (136) 82 (96) 15 (14) 0 (0) 29 (33) 0 (0) Weld toe up 27 (32) 226 (160) 712 (609) 0 (0) 305 (252) 0 (0)

4. Fatigue evaluation of the girder tests 4.1. Objective As mentioned before the primary crack was obviously initiated at the rough bottom rail surface simulating the mill scale with its defects. In practice the greatest defect at the tensile stressed bottom rail surface will cause the primary crack. The question arises whether the secondary crack is a consequence of the growing primary crack or whether the initiation of the secondary crack is independent of the primary one. So is it possible that a rail with awless surface fails through the secondary crack, especially, in case of moving wheel loads, when the secondary crack can grow in x-direction? Of course the growth of both cracks will interact after exceeding a certain crack size. Concerning the crack initiation and early crack growth, however, both cracks might be independent of each other. In detail, the evaluation is supposed to answer two questions: 1. Are the notch stresses at the location of the secondary crack in the uncracked rail sufciently high to be the sole reason for the secondary crack? 2. Is it possible to reveal any inuence of the longitudinal stresses rx on the secondary crack? If so, there might be an interaction between the crack initiation of the primary crack mainly driven by longitudinal stresses and that of the secondary crack. 4.2. Hypothesis of ndley Due to the outcome of the tests only planes at the upper weld toe have to be considered making an angle with the y-axis, compare Fig. 11a and b. Eqs. (6) and (7) hold for the stresses at such planes.

rx ry rz sxy syz sxz

239 (235) 291 (249) 827 (748) 0 (0) 312 (277) 0 (0)

section of the submodel were taken for the evaluation, Fig. 9d. The Youngs modulus of the girder and the wheel was 2 105 N/mm2. The girders strains were recorded during the cyclic testing at particular locations, e.g. at the ange to web junction at the loading section etc. These measurements served for the validation of the generated model. The FE strains of the global model and the measured strains show a very good t. See for more details in [6]. The assessment of the local stresses was primarily carried out along paths at the weld notches. The FE stresses at 7 nodes along the paths within the weld toes and at 43 nodes within the weld root were recorded, Fig. 9d. The maximum stress components per path were determined and combined for the assessment regardless if they occur on the same node along the path or not, compare Fig. 9d. The computed FE stresses are printed in Table 1. Because of the smaller weld toe angle (Fig. 4) weld A exhibits higher stresses. A selection of the most important stress histories of the upper weld toe of weld A is given by Eq. (5). It can be seen that the stress ratio equals R = 0.

ry t ry;mean ry;a sin x t 113 113 sin x t rz t rz;mean rz;a sin x t 356 356 sin x t syz t syz;mean syz;a sin x t 152 152 sin x t

rn ry sin2 h rz cos2 h 2 syz sin h cos h sn ry rz sin h 2 syz cos2 h sin2 h

6 7

Interestingly, the component rx of the lower weld toe is in good correspondence with the nominal stress, Fig. 6. At the upper weld toe and the weld root being closer to the point of loading the component rx differs from the nominal value. This is a strong onset for a multiaxial state of stress.

The effective shear stress range Dseff,a can be computed based on Eq. (4). According to Findley [4] rmax is referred to as the maximum (positive) normal stress at the considered plane. From the stress history in Fig. 11c it can be seen that the greatest positive normal stress becomes rn,max = 0 in all planes. Subsequently, the normal stress has no inuence on the effective shear stress seff, Table 2. In

(a)
18

(d)
26 weld A

(e)
24 weld A

(b)

5 mm

(c)
24

(f)
20

(g)
24

weld B

weld B

Fig. 10. Crack surfaces together with the early secondary crack growth orientation of test girder No. 2 (ac) and of test girder No. 3 (dg).

1124

M. Euler, U. Kuhlmann / International Journal of Fatigue 33 (2011) 11181126

(a)
z

(b)
+
y

n yz n

zy z

+stress

(c)

390
n,a n,a n

-stress

-459

n,mean = n,max

-229,5

=0

Fig. 11. Orientation of critical planes (a), stresses at any plane with angle h (b) and stress history of the components rn and sn of the plane with h = 20 (c).

addition, a modied Dseff,mod is determined that is received if using the mean stress rn,mean instead of the maximum positive normal stress rn,max. The orientation of the critical plane in the rst case amounts to 20 and in the second case to 30. That is in good agreement with the measurements, compare Fig. 10. The effective shear stress ranges can be transformed in equivalent normal stress ranges according to DEH by applying Eq. (8).

Dreq

q p 3 Ds2 3 Dseff eff

4.3. Effective equivalent stress hypothesis (EESH) The loading of the test girders did not exhibit any phase shift (d = 0), consequently, the computation of the effective equivalent stress according to EESH simplies solely to (10) since the second and third term of (9) become unity.

q d90 2 F d G1 90 Dreq d Dreq d 0 F d 0 q 2 2 Dreq d 0 r2 x ry rx ry fs 3 Dsxy fs

9 10 11

less stressed surrounding material [14]. That explains why a member under uniform normal stress has a shorter fatigue life than the same member under bending stress of the same magnitude. The size effect is expressed by Eq. (11) with req,axial and req,shear as the supportable effective equivalent stress with a survival probability, Ps, of 50% under pure bending/axial loading and pure shear according to DEH. The exact computation of the size effect factor fs requires statistically proven SN curves for the considered constructional detail under pure bending/axial loading and pure shear. Since such specic information is not available for the welded crane rail this factor has to be estimated. In [12] llet welds under pure shear stress Dsnom due to torsion were tested, Table 3. The determined supportable shear stress ranges indicate primarily weld root failure. Thus, these values are conservative concerning the fatigue strength of the weld toe. The supportable notch stress ranges Dsnotch and the corresponding equivalent stress ranges Dreq according to DEH can be computed as given by Eqs. (12) and (13). The stress concentration is estimated based on the notch stress concentration factor as of Neuber [9] for a shallow notch, Fig. 12b.

Dreq;axial Dreq;shear

Therefore, the effective equivalent stress depends only on the stress ranges and the factor fs. This factor captures the so-called microsupport size effect. The size of the maximum stressed material volume has a signicant inuence on the fatigue life. If only a small volume is subjected the maximum stress due to a steep stress gradient, the fatigue life will be longer due to the support effect of the

Table 3 Fatigue strength [N/mm2] at Nf = 2 106 under shear loading, llet welds, automatic welding, R = 1, for different probability of survival Ps [12]. Ps (%) 50 97.7

Dsnom acc. to [12]

140 118

Calculated Dsnotch 280 236

Calculated Dreq,DEH 484 409

Table 2 Notch stresses at the upper weld toe of weld A at planes of different angle h [N/mm2]. Remark: the values of weld B in parentheses. h () 90 226 0 305 305 237 0 712 0 305 305 91 15 527 0 386 386 228 20 459 0 390 390 (337) 675 (583) 252 25 392 0 382 382 264 30 326 0 363 363 265 (237) 459 (410) 35 266 0 333 333 253 90 226 0 305 305 237

rn,min rn,max Ds n Dseff req Dseff,mod req,mod

M. Euler, U. Kuhlmann / International Journal of Fatigue 33 (2011) 11181126

1125

(a)

(b)
t

t/ = 2

(a)

from torsion from bending

max

= 2.0 t/ = 2 t

(c)

(b)
p

(c)

max

= 3.33 p

Fig. 12. Stress concentration of moderately curved notches (a), under pure shear [9, p. 32] (b) and under pure tension [9, p. 141] (c). Fig. 14. Typical test set-up with combined direct stress due to bending and shear stress due to torsion (a), crack opening modus III (b) and modus I (c).

Dreq d 0

q 2262 7122 226 712 1:00 3 3052 700 N=mm2 at weld B

823 N=mm2

4.4. Comparison The comparison of all determined equivalent stress ranges Dreq is illustrated in Fig. 15 together with a part of the scatter band for R = 0 according to [13] (compare Fig. 13) assuming an averaged number of load cycles of the girder tests before failure, Nf = 670,000. The EESH and the original formulation of Findleys hypothesis estimate a shorter fatigue life (50% propability) than it was recorded in the tests and give conservative results, consequently. The EESH results suggest that the crack initiation occurred slightly earlier. Because the secondary cracks grew under compression (crack closing) they were very hard to detect. That is why an earlier crack initiation as recorded might be possible. Also the factor fs being an important input parameter has to be estimated for the considered detail. For that reason the information for an assessment based on EESH might not be sufcient. 5. Conclusions Crane runways are subjected to a complex multiaxial state of stress due to geometric and metallurgic notch effects and the introduction of concentrated loads. The amplitudes of the stress components do not occur simultaneously resulting in an alternating direction of the shear stress that has been recognised to control the crack initiation. To investigate this complex issue two girders with in-phase interaction of the global bending rx and the local wheel pressure rz,local were tested. The goal of this paper is to compute the corresponding equivalent notch stress ranges according to DEH applying two different local approaches. On this basis the questions of Section 4.1 can be answered: 1. The notch stresses at secondary cracks location can be considered as sufcient to be its sole reason. 2. The calculated longitudinal stress rx that induces the primary crack is quite low at the location where the secondary crack is initiated (upper weld toe). Altogether, the conclusion can be supported that the initiation of the secondary crack is independent of that of the primary crack.

Fig. 13. Haigh diagram with mean stress inuence in fatigue tests under pure axial load or bending [13].

Dsnotch as Dsnom p Dreq;shear 3 Dsnotch

12 13

Fig. 13 summarizes the outcome of an extensive experimental analysis of different chord-web junctions under pure bending or axial loading accounting for the mean stress inuence [13]. The test results for weld toe failure and weld root failure fall into the same scatter band of this analysis. Unfortunately, it is unclear what failure hypothesis is used in [13] whether maximum principal stress or equivalent stress of DEH. Nevertheless, it can be stated that Fig. 13 is in close correspondence to the allowable notch stress ranges for R = 0.5 proposed in [18] that amount to Drc = 225 N/ mm2 in terms of DEH and Drc = 200 N/mm2 in terms of prinicipal stresses. Since the mean stress inuence in Fig. 13 is quite pronounced and Table 3 only holds for R = 1, the estimated size effect factor fs is computed for R = 1 expecting no signicant R-sensitivity of fs:

494 fs  1:00 484

For comparison the tube-to-plate junctions in [17] have a comparable loading and a size factor fs = 1.39. Finally, the effective equivalent stress range at the upper weld toe (weld A) can be computed using Eq. (10):

1126

M. Euler, U. Kuhlmann / International Journal of Fatigue 33 (2011) 11181126

Fig. 15. Scatter band of test results [13] for stress ratio R = 0, compare also Fig. 13.

Findleys hypothesis estimates obviously the fatigue life best. The computed orientation of the critical plane corresponds very well with the measured early crack growth direction. Nevertheless, the results of the presented modied formulation of Findleys hypothesis suggest that the complete neglect of the compressive mean stress might be not reasonable. The EESH results estimate that the crack initiation occurred earlier. As the secondary cracks grew under compression (crack closing) they were very hard to detect. That is why an earlier crack initiation as recorded might be possible. On the other hand the lack of fatigue strength values under pure axial loading and pure shear for the considered constructional detail has not allowed a reliable application of the EESH, yet. Since the existing fatigue strengths for shear loading [12] were primarily determined under torsion leading to crack opening modus III (Fig. 14a,b) their usage for shear loadings that do not arise from torsion might be questionable [8]. It is known that combined stresses solely induced by bending lead to crack opening modus I, Fig. 14c. Further testing especially for out-of-phase loading is necessary. In this case the secondary crack growth might be accelerated due to an interaction with the local shear stress sxz,local caused by the load application [5]. In order to deepen the understanding of the wheel load introduction the authors plan tests on crane runway girders under moving (reciprocating) single loads for the near future. References
[1] Bckstrm M, Marquis G. Interaction equations for multiaxial fatigue assessment of welded structures. Fatigue Fract Eng Mater Struct 2004;27:9911003. [2] ECCS-TC 6 Fatigue: recommendations for the fatigue design of steel structures. ECCS-Publication No. 43; 1987. [3] Eurocode 3: Design of Steel Structures Part 19: Fatigue; 2005.

[4] Findley WN. A theory for the effect of mean stress on fatigue of metals under combined torsion and axial load or bending. Trans ASME: J Eng Indus 1959;81:3016. [5] Kuhlmann U, Euler M. Beitrag zur Ermdungsfestigkeit von Schienenschweinhten bei Kranbahnen. Stahlbau 2009;78:1709. [6] Kuhlmann U, Euler M. Kranbahntrger Wirtschaftliche Bemessung und Konstruktion robuster Radlasteinleitungen. Research report 5/2007 des Deutschen Ausschusses fr Stahlbau (DASt), Stahlbau Verlags- und Service GmbH, Dsseldorf; 2007. [7] Kuhlmann U, Euler M. Multiaxial fatigue behaviour of hot-rolled crane runway beams with welded block rail. In: Fatigue design 2007, Senlis, workshop W1.3; 2007. p. 18. [8] Maddox SJ, Razmjoo GR. Interim fatigue design recommendations for llet welded joints under complex loading. Fatigue Fract Eng Mater Struct 2001;24:32937. [9] Neuber H. Kerbspannungslehre. 3. Auage. Berlin, Heidelberg: SpringerVerlag; 1985. [10] Radaj, D., Sonsino C.M. Ermdungsfestigkeit von Schweiverbindungen nach lokalen Konzepten. Fachbuchreihe Schweitechnik, Band 142. Verlag fr Schweien und Verwandte Verfahren, DVS-Verlag, Dsseldorf; 2000. [11] Radaj D. Gestaltung und Berechnung von Schweikonstruktionen Ermdungsfestigkeit. Fachbuchreihe Schweitechnik, Band 82. Deutscher Verlag fr Schweitechnik (DVS) GmbH, Dsseldorf; 1985. [12] Seeger T, Olivier R. Ertragbare und zulssige Schubspannungen schwingbeanspruchter Schweiverbindungen. Stahlbau 1987;8:2318. [13] Seeger T. Grundlagen fr Betriebsfestigkeitsnachweise. In: StahlbauHandbuch, Teil 1B. Stahlbau-Verlagsges, mbH Kln; 1996. [14] Siebel E, Stieler M. Ungleichfrmige Spannungsverteilung bei schwingender Beanspruchung. VDI-Z. Band 97, vol. 5; 1997. p. 1216. [15] Sonsino CM. Multiaxial fatigue assessment of welded joints recommendations for design codes. IntJFat 2009;31:17387. [16] Sonsino CM. Multiaxial fatigue of welded joints under in-phase and out-ofphase local strains and stresses. IntJFat 1995;17:5570. [17] Sonsino CM. Schwingfestigkeit von geschweiten Komponenten unter komplexen elasto-plastischen, mehrachsigen Verformungen. Abschlussbericht, Europische Kommission, Generaldirektion Wissenschaft, Forschung und Entwicklung, Schriftenreihe Technische Forschung Stahl, Report No. EUR 16024 DE (1997), Luxembourg; 1997. [18] Sonsino CM. A consideration of allowable equivalent stresses for fatigue design of welded joints according to the notch stress concept with the reference radii rref = 1.00 and 0.05 mm. Weld World 2009;34:R6475.