Sie sind auf Seite 1von 9

Engineering Failure Analysis 7 (2000) 189197

www.elsevier.com/locate/engfailanal

Prediction of fatigue failure in a camshaft using the crack


modelling method
G. Wang a,*, D. Taylor a, B. Bouquin a, J. Devlukia b, A. Ciepalowicz b
a

Department of Mechanical Engineering, Trinity College, Dublin, Ireland


b
Rover Group, Gaydon Test Centre, Warwick, UK
Received 30 April 1999; accepted 26 June 1999

Abstract
Camshafts made of grey cast iron and used in Rover vehicles were tested under cyclic bending and torsion and
modelled using FE. A new technique known as crack modelling was used to predict the fatigue limit. The method
uses a linear elastic nite element analysis to derive an equivalent stress intensity factor (K ) for stress concentrations
in components. K is calculated without introducing a crack into the component: the stress eld around the
maximum stress point is examined and compared to that for a standard centre-cracked plate. This component was a
challenge for the technique because it involved a blunt notch and local surface eects. Fatigue limits were
successfully predicted for two dierent designs and two loading modes, but some inaccuracies remain which suggest
modications to the theory. # 2000 Elsevier Science Ltd. All rights reserved.
Keywords: Fatigue limit; Stress concentrations; Finite element analysis; Surface layers; Cast iron

1. Introduction

Fatigue failure in components usually initiates at stress concentrations: geometric features such as
holes, grooves and corners, and despite some local plasticity, high-cycle fatigue behaviour is essentially a
linearelastic problem. Traditional methods show poor accuracy when applied to high-gradient stress
concentrations (e.g. sharp notches and small notches) and to materials of low notch-sensitivity (e.g. cast
irons). The crack modelling approach deals with this problem by modelling the notch as a crack. This
allows the calculation of an equivalent stress-intensity factor (K ), enabling standard fracture mechanics
* Corresponding author. Tel.: +353-1-6081976; fax: +353-1-6795554.
E-mail address: wangg@tcd.ie (G. Wang).
1350-6307/00/$ - see front matter # 2000 Elsevier Science Ltd. All rights reserved.
PII: S 1 3 5 0 - 6 3 0 7 ( 9 9 ) 0 0 0 1 5 - 1

190

G. Wang et al. / Engineering Failure Analysis 7 (2000) 189197

methodology to be used. Fatigue is assumed to occur if the cyclic value of K exceeds the crack
propagation threshold.
Previous papers [1,2] have reported the application of this method to components of complex
geometry, which were subjected to bending and, recently [3], torsion. The present paper considers a
component which presented two challenges for the method. Firstly, failure occurred from a relatively
blunt notch, whose stress-concentration factor, Kt, was approximately 2, so it was not clear whether this
type of notch could be usefully modelled using fracture mechanics; secondly, the as-cast component was
known to have a hard surface layer which appeared to improve its fatigue strength. It was therefore
necessary to extend the crack modelling method to include this eect.

2. Experimental details
Tables 1 and 2 show the material properties and composition of the material, which was a typical
grey cast iron. The threshold stress intensity factor at R=1 was found previously [4] to be
DKth=15.94 MPa m1/2.
Fig. 1 shows a general view of the camshaft component, which contained a number of geometric
features. The component was clamped at one end as shown, and loaded either in bending or torsion
(both at R=1), to produce failure at a chosen location, as shown in the detail in Fig. 1. Two designs
were tested: the rst had an as-cast notch, 1 mm deep, 1.5 mm root radius; in the second design,
machining was used to remove the cast surface, deepening the notch to 1.75 with 1.63 mm root radius.
The Kt factors for these notches were 2.1 in bending (for both notch depths) and 1.6 for the as-cast
notch in torsion. The S/N data for the material was obtained using standard hourglass specimens loaded
in axial tension/compression at R=1. Vickers hardness was measured using macro- and micro-indents,
to record its variation as a function of distance from the surface.

3. Results
Figs. 24 show test results in the form of Wohler curves, plotting applied load range for the
component tested in bending (Fig. 2), torque range for the torsion testing (Fig. 3) and stress range for
the plain specimens (Fig. 4). The material fatigue limit, dened at 107 cycles, was 190 MPa. The results
Table 1
Mechanical properties of grey cast iron
Young's modulus
Poisson's ratio
Yield stress
Ultimate strength

170 GPa
0.29
202 MPa (s0.2)
249 MPa

Table 2
Composition of grey cast iron (% weight)
C

Mn

Si

Ni

Cu

Cr

3.3

0.09

1.5

1.8

0.07

0.2

0.03

0.05

G. Wang et al. / Engineering Failure Analysis 7 (2000) 189197

191

Fig. 1. Camshaft component, showing loading and detail of the notch which caused failure.

showed a degree of scatter which is typical for this material, giving variations on the load (or stress) axis
of 1020% from the mean.
Fig. 5 shows hardness results; it is clear that, despite some scatter, the micro-indent hardness of the
as-cast component was higher near the surface, especially within 1 mm, whereas this was not the case
for the machined component or for the macro-indent hardness readings. The average Vickers-hardness
on the as-cast surface was 395.5. The average value in the bulk was 326.6. This dierence appeared to
arise due to a change in the amount and morphology of the cementite phase; it was thus detected by the

Fig. 2. Camshaft bending test results: (a) as-cast notch; (b) machined notch.

192

G. Wang et al. / Engineering Failure Analysis 7 (2000) 189197

Fig. 3. Camshaft torsion test results (as-cast condition).

micro-indents, which were always placed in the matrix (pearlite/cementite) and not in the graphite
nodules. The average value for the machined condition was 319.3, which is not signicantly dierent
from the as-cast bulk value.

Fig. 4. Plain specimen bar tension/compression test results.

G. Wang et al. / Engineering Failure Analysis 7 (2000) 189197

193

Fig. 5. Vickers-hardness of camshaft: (1) micro hardness on the machined surface; (2) macro hardness on the machine surface; (3)
micro hardness on the as-cast surface; (4) macro hardness on the as-cast surface.

4. Finite element analysis


ANSYS software was used to create a nite element model; a quarter model was used (Fig. 6) because
the component is axially symmetrical apart from the cam lobes which do no carry much stress. The
bending and torsion loading was treated as a symmetrical or asymmetrical load, respectively. The
models were meshed using 3-D parabolic tetrahedral elements. Element size close to the notch was
about 0.75 mm.
A linear elastic analysis was used. In each case the highest stress occurred at the notch tip. Fig. 7
shows the rst principal stress plotted as a function of distance, r, measured from the notch tip along a

Fig. 6. The nite element model, showing in detail the mesh around the notch tip. The arrow shows the line on which the stress
distance curve was measured.

194

G. Wang et al. / Engineering Failure Analysis 7 (2000) 189197

Fig. 7. FEA results: sr curves for each case at the experimental fatigue limit.

straight line drawn in the direction perpendicular to the rst principal stress direction at the tip (see Fig.
6). The loading in each case corresponds to the experimental fatigue limit of the component.
5. Prediction of fatigue limit
5.1. The crack modelling method
The crack modelling method has been described in detail elsewhere [1,2,5]. It is based on the
Westergaard equation which gives the stress-distance (sr ) function for a crack of length aw in tension
at a stress sw in an innite plate:
s sw =1 aw =aw r2 1=2 :

The values of sw and aw in Eq. (1) can be varied to obtain a best t with the sr curve taken from a
component (as in Fig. 7). Then the appropriate K value is given by:
p
2
K sw paw :
This method assumes that if the equivalent stress intensity range, DK, found by applying Eq. (2), is less
than the threshold value of the material, fatigue failure will not occur. A post-processor was written in
Visual Basic to interface with ANSYS software, allowing the value of K to be calculated automatically.
Previous publications [2,3,5] have discussed the choice of direction on which to measure the sr curve
and the start and end points of the curve.
Table 3 shows predictions of the fatigue limit of the component (expressed in terms of applied load
range); here the crack modelling method is compared with predictions made simply by using the
maximum stress at the notch (i.e. the hot-spot stress). It is clear that the use of maximum stress
underestimates the performance of the component considerably. It is possible to modify this hot-spot
prediction by using empirical factors which depend on root radius (e.g. Peterson [6]) or on local stress
gradient (e.g. Siebel and Stieler [7]). Such factors are commonly used in industry but are unreliable
because they are highly material-dependent and not mechanistically based. Another commonly used
method is the `local-strain approach' in which performance is based on notch-root plastic strain. This
also gives improved predictions but requires a lot of specialised material data such as cyclic stressstrain
curves and elasticplastic FEA. The crack modelling method is advocated instead of the above methods
because it requires relatively little data and a simple elastic FE analysis, and because it is grounded in

G. Wang et al. / Engineering Failure Analysis 7 (2000) 189197

195

Table 3
Prediction of fatigue limit: without considering the surface eect
Load

Exp. data

Bendinga (kN)
Bendingb (kN)
Torsiona (Nm)
a
b

1.50
3.10
510

Prediction using maximum stress

Crack modelling prediction

Dsmax (MPa)

DP

Error %

DKmax (MPa m1/2)

DP

Error %

408.00
672.00
360.00

0.70
0.88
269.17

53.43
71.73
47.22

13.10
23.40
16.20

1.83
2.11
501.81

21.68
31.88
1.60

Camshaft with machined surface.


Camshaft in as-cast form.

the well-known methods of fracture mechanics. In this case the crack modelling prediction was an
improvement over the maximum-stress prediction in all three cases, but signicant errors remained,
especially for the as-cast case in bending. Table 4 shows a revised prediction in which a correction factor
was introduced to allow for the surface eect noted above.
This correction was made by using the methodology of Murakami and Endo [8,9], who proposed a
generalised method for the analysis of small surface defects and notches. They proposed that, for steels,
the fatigue limit could be predicted by assuming that the notch was a crack: the size parameter they
used was `area' dened as the area of the defect projected normal to the stress axis. They proposed
equations for the fatigue limit and threshold for notches in all steels, as follows:
p
Ds 1:43Hv 120 area1=6 ,

p
DKth 3:3 103 Hv 120 area1=3 :

Here Hv is the Vickers micro-hardness. Since in a case such as the present one where the notch
geometry is kept constant and the hardness changes, both the threshold and the fatigue limit should
increase in proportion to (Hv+120) which, given the hardness values quoted above, implies an increase
of a factor of 1.15 for the as-cast case. This increase applies to the maximum-stress method as well as to
the crack-modelling method.

Table 4
Prediction of fatigue limit: including the surface eect
Load

Bendinga (kN)
Bendingb (kN)
Torsionb (Nm)
a
b

Exp. data

1.50
3.10
510

Camshaft with machined surface.


Camshaft in as-cast form.

Prediction using maximum stress

Crack modelling prediction

Dsmax (MPa)

DP

Error %

DK (MPa m1/2)

DP

Error %

408.00
672.00
360.00

0.70
1.01
309.54

53.43
67.49
39.31

13.10
23.40
16.20

1.83
2.43
577.09

21.68
21.66
13.15

196

G. Wang et al. / Engineering Failure Analysis 7 (2000) 189197

6. Discussion
As Table 4 shows, this method, along with the correction for surface hardness, has the eect of
bringing all the predictions to within about 20% of the experimental values. This is quite satisfactory,
since one would expect errors of at least 10% in the experimental results, the FE analysis and the basic
material data. However, it is interesting to note that the prediction for torsion loading is relatively nonconservative, whereas that for bending (as-cast) is conservative. This implies that the method
overestimates the fatigue strength in torsion, and this problem is one which merits further investigation.
This analysis has also demonstrated that, in some cases, the crack-modelling method can be applied
to very blunt notches. Previous papers [5,10] discussed the range of validity of the method: following
Smith and Miller [11] it is assumed that this fracture-mechanics approach will be valid for relatively
sharp notches, with the maximum-stress method being valid for blunter ones. However, for a material
such as cast iron, which has a particularly low notch-sensitivity, the crack-modelling method will have a
greater range of validity, being preferred for many of the stress concentrations which are found on
components.
7. Conclusions
1. The crack-modelling method, which predicts fatigue limits of stress concentrations using a fracturemechanics approach, was able to predict the behaviour of an automotive camshaft component in two
dierent design cases under both bending and torsion loading.
2. Failure occurred from a very blunt notch (Kt=1.62.1) but even so the crack-modelling method gave
better predictions than a method using the maximum notch-root stress. This implies that the method
is suitable for a very wide range of features in materials of low notch-sensitivity, such as this grey
cast iron.
3. A simple method based on hardness measurements was successful in allowing for the existence of a
hardened surface layer. This provides some support for the approach of Murakami and Endo.

Acknowledgements
The authors are grateful to Enterprise Ireland and Materials Ireland for funding provided to one of
the authors (G. Wang).
References
[1] Taylor D. Crack modelling: a technique for the fatigue design of components. Engineering Failure Analysis 1996;3(2):12936.
[2] Taylor D, Ciepalowicz AJ, Rogers P, Devlukia J. Prediction of fatigue failure in a crankshaft using the technique of crack
modelling. Fatigue Fract Engng Mater Struct 1997;20(1):1321.
[3] Taylor D, Zhou W, Ciepalowicz AJ, Devlukia J, Mixed-mode fatigue from stress concentrations: an approach based on
equivalent stress intensity, Int J Fatigue 1999;21:1738 (in press).
[4] Taylor D, Hughes M, Allen D. Notch fatigue behaviour in cast irons explained using a fracture mechanics approach. Int J
Fatigue 1996;18(7):43945.
[5] Taylor D, Lawless S. Prediction of fatigue behaviour in stress-concentrators of arbitrary geometry. Eng Fract Mechanics
1996;53:92939.
[6] Peterson S. Notch sensitivity. In: Sines G, Waisman JL, editors. Metal fatigue. New York: McGraw-Hill, 1959. p. 293306.

G. Wang et al. / Engineering Failure Analysis 7 (2000) 189197

197

[7] Siebel E, Stieler M. Dissimilar stress distributions and cyclic loading. Z Ver Deutsch Ing 1955;97:12131.
[8] Murakami Y, Endo M. Eects of hardness and crack geometries on DKth of small cracks emanating from small defects. J Soc
Mater Sci Japan 1986;35(395):9117.
[9] Murakami Y, Kodama S, Konuma S. Quantitative evaluation of defects of non-metallic inclusions on fatigue strength of high
strength steels. Int Journal of Fatigue 1989;11:2918.
[10] Taylor D, O'Donnell M. Notch geometry eects in fatigue: a conservative design approach. Engineering Failure Analysis
1994;1:27587.
[11] Smith RA, Miller KJ. Prediction of fatigue regimes in notched components. Int J Mech Sci 1978;20:2016.