You are on page 1of 10

Estimating Edge Contact Fatigue Damage in Wind Turbines Low Speed Shaft

G. Shatil1)
1)
MBN Technology, Bristol BS3 2AR, UK

ABSTRACT

A simplified analysis has been carried out to estimate the effect of collar pressure and
contact edge on fatigue performance of a typical wind turbine low speed shaft component.
Three interference fit levels were considered; high, medium and low in SG iron hollow shaft
under cyclic bending. A method was developed to assess the critical material crack length to
initiate fatigue cracking and the associated stress/load levels. This analysis was based on
two fracture criteria; the critical fatigue crack growth threshold value (∆Kth) and the critical
material through thickness fatigue limit length to propagate fatigue crack growth. It was found
that under extreme conditions fatigue cracks may initiate and hence a further fatigue crack
growth analysis is required. However, this could be avoided in the design stage.

INTRODUCTION

Stress analysis of contact in solids

This subject has been under investigation for over 100 years. A detailed analysis of the basic
stress theory of contact mechanics elasticity is given by Johnson [1]. Most of the analytical
theories were developed to establish solutions of a generalised or specific stress intensity
factor. Fig. 1 is a general description of edge contact in solids. More recently elastic-plastic
analysis was used and several theories were developed to establish the non-linear relation
between the material subjected to the contact loads and cyclic loads, for example see Refs
[2- 4;] those papers include analysis of fretting fatigue.

A detailed description of the contact and fretting fatigue process is given in ESDU 90031 [4]
including the material damage influenced by several factors such as type of loading surface
roughness and mean stress. The main fatigue issues are summarized below; (see also Fig.
1):
1. The stresses in the contact surface layer are the resultants of the applied alternating
loads, applied normal to the contact surface, and the contact pressure.
2. The combined action of those forces is to initiate a crack that is not normal to the
surface, but most commonly start at the contact edge.
3. Below the surface the radial stress is usually compressive. This compressive stress
slows down (retards) further growth of the initial crack. Further fracture depends on
the magnitude of alternating loads and mean load.
4. If the applied alternating stresses are sufficiently high the crack could propagate to
fracture.
5. The applied tensile mean stress may also overcome contact compressive stress and
will promote contact related crack growth.
P (x) Normal
traction
τ (X) Surface
traction
Fatigue O X
limit σ (x)
distance σx Cyclic field
l Stress
τxz
σZ

Z
.
Fig. 1: Stresses at the edge of solids in contact

Fatigue and fracture analyses of solids in contact

The analysis of fracture due to contact has received much attention in the past due to the
shortened crack initiation stage. Most past publications have calculated the stress intensity
factors (SIFs) analytically at the crack tip near the contact edge, Fig. 1.

Previous methods to calculate the SIFs commonly belong to one of the following. Traditional
methods used elasticity analysis to calculate solutions for a range of geometry problems [5].
More recent methods are described in Refs [6] and [7] for contact problems. They are used
to produce a general analytical solution by using the remote stress field that describes local
singularity at the tip. Analyses in Refs [8] and [9] are based on the assumption that it is
possible to find a similitude solution for the stresses in a large blunt notch and in the crack tip
of contact edge by assuming that at the contact edge a (blunt) notch is presented. This type
of analysis is useful because it reduces the singularity that appears at the contact edge with
the stress concentration at notch root. Fatigue assessment methods were developed using
this principle, Ref [7]. One of the challenges of using the LEFM approach to estimate contact
problems is how to account for the complex stress at the crack tip. Commonly fatigue crack
growth is taken into account by considering K1 (or fracture Mode 1) fracture. However, due to
the tangential slip and the normal pressure to the crack front, Mode 2 or shear fracture also
exits in contact problems. Hence, a solution of both, K1 and K2 is required. Recently it was
suggested that the two be combined into an equivalent SIF, [2 - 3] in order to measure
fatigue crack growth.

Outline of a simplified design procedure for integrity assessment of solids subjected


to contact, under complex fatigue load

The majority of the reviewed methods considered a solution that included a local estimation
of stresses, a form of elastic fracture mechanics analysis and correlation with material high
cycle fatigue properties to estimate the fatigue strength. Those considerations were
incorporated into a set of calculations that could be applied to components fatigue analysis
by using existing material data. This is outlined below:
1. Use a fairly detailed elastic FEA model of the contact region. Obtain from post-
processing and several critical load cases the local stresses along a path normal to
the contact edge. Stress outputs include Sx – axial, Sz – normal, and Sxz – shear
stresses. The Sf – friction stress is contained within the FEA contact model.
2. Calculate stress intensity factors (SIFs) along a crack path; K1 and K2. Taking into
account existing subsurface stress gradients up to an arbitrary distance under the
surface that extend well beyond the material initial stage cracking. At first estimation
10mm subsurface distance seems reasonable.
3. Use the concept of equivalent SIF [2] to calculate an equivalent SIF along the path
distance, for instance:

K equivalent
=
2
K +K
1
2
2
(1)

4. Based on material properties, calculate the material critical subsurface fatigue limit
length [8]; using materials fracture threshold and endurance values:
2
 K th  (2)
l fatigue lim it =  /π

σ lim it 

5. Calculate the local equivalent crack lengths, ae (assuming no fretting occurs):

2
K e (3)
ae = /π
 
σ a 

Where σa is the nominal applied stress amplitude.

6. Compare ae with the lfatiguelimit by using the fatigue limit threshold, Fig. 2:

Calculated ae > lfatiguelimit Fatigue limit not exceeded (4)


Calculated ae ≤ lfatiguelimit Fatigue limit exceeded

If the calculated critical crack due to the stress is longer than that of material fatigue
limit it means that the stress is not high enough to propagate a crack at the critical
distance and vice versa.

Ke
Equivalent
SIF
[MPa√√mm]

∆ae =
f(Ktσa)

lfl ae a crack length


Calculated
[mm]

Fig 2: Calculation procedure (for illustration purpose only).


7. Recalculate ae by inserting a stress concentration factor Kt.σa, Fig. 2.

8. If Kt is above A – consider local structure safe under this extreme loading at this
point. If Kt is below A - this structural item could be subject to fatigue cracking, where
A is a design safety factor.

9. Calculate the final crack length using fracture toughness or geometry consideration:
2
 K 1c  (5)
a = /π
f 
σ a 
10. Estimate life by counting the number of cycles to obtain these final cracks using load
spectrum.

FINITE ELEMENT SIMULATION OF WIND TURBINE LOW SPEED SHAFT IN BENDING

A typical wind turbine tubular shaft with an interference pressure collar was loaded in
bending. The FEA axisymmetric model geometry, loading and mesh are shown in Fig. 3. A
bending moment of 3700 kNm was applied perpendicular to the X axes. In the region of the
contact edge a refined FEA mesh was used having a mesh size of about 0.25mm. Data
accumulation geometry boundaries for the analysis were +5mm and -10mm along the Z
(longitudinal) axes from the contact edge and 15mm along the y (radial or normal) axes for
the same Z coordinates from the shaft surface inward.

shaft cross section

600
[mm]

400
radial distance [m]

200

0
-1500 -1000 -500 0 500 1000 1500
-200

-400
Y
X
-600
axial distance [mm]
[m]
Z

Fig 3: Geometry and FEA model of the low speed shaft

MATERIAL AND APPLIED FRACTURE ANALYSIS

The material properties of commercial wind turbine low speed shaft materials were not
available. Instead, three typical commercial nodular cast iron (SG) materials were
considered, grades 370/17, 420/12 and 500/7. The relevant material properties were
obtained from material data (9) as shown below.
Alternating Fracture
Minimum Threshold Calculated
fatigue Toughness
Tensile ∆Kth fatigue
Grade strength ∆K1c
strength (MPa) (MPa√m) limit (mm)
(MPa) (MPa√m)
370/17 370 63 91 5.5 2.426

420/12 420 67 91 5.5 2.145

500/7 500 75 85 5.1 1.4718

Table 1: Mechanical Properties of three wind turbines casting nodular cast irons

Stress Intensity Factors (SIFs)

The following simplified approach was taken to estimate stress intensity factors of a potential
fatigue cracking in the shaft near the edge contact, using the FEA results:
1. The crack propagates perpendicular to the surface inwards.
2. The growth is in fracture Mode I.
3. The stress normal to the crack front is along the shaft longitudinal (Z) direction.
Considering those assumptions, stress intensity factors (SIFs) were derived as a
superposition of two cases, using text books analytical solutions:
a. K1a = 1.21518P√πa, Where P is the total stress/ geometrical area for a constant field
stress.
b. K1b = 0.43856P√πa, Where P is the total stress/geometrical area for a linearly decreased
subsurface stress field.
Hence at each point the total SIF is (a is the crack length):

K1 = K1a + K1b= 1.56018P√πa (6)

Using the FEA results, and the above solution, SIFs were calculated at the points of the
maximum stress through the thickness and are shown in Fig 4b. This Figure shows a
polynomial through thickness increase of SIF as function of distance from surface inwards.

Fatigue Crack Propagation Threshold and Critical Crack Length

The values of the fatigue crack propagation thresholds used in the analysis are shown in
Table 1. The fatigue limit crack lengths were calculated using equation 2 for each of the
materials as follows:

STRUCTURAL INTEGRITY ANALYSIS OF A TYPICAL LOW SPEED SHAFT

One of the unknowns in this analysis was the actual applied collar interference pressure.
Initial FEA simulations of the shaft have shown axial stress at the edge of the collar of about
1200 MPa which appears unrealistically high. Further FEA simulation applied half of that
initial interference pressure for which the maximum axial stress was approximately 500 MPa
and subsequent verification analysis applied a very low IF pressure to assess the effect of
the IF stress on the shaft behaviour, Fig. 5.

Finite Element Analysis Results

Initial FEA results have shown that critical stresses developed in the shaft area that in the
proximity of the edge of the collar. A refined FEA model was subsequently used at this area
and the data was analysed in terms of the shaft stress field on the surface at a distance of +/-
25mm from the contact in the axial direction and also into subsurface at a distance up to
10mm in the radial direction. The FEA results are shown in Fig 5 for the subsurface and
surface stresses.

a. Contact edge through thickness axial stress b. Contact edge through thickness SIFs
160

Stress Intensity Factor


1100
Low IF

900
MPa

120 Med IF
stress MPa

Low IF
High IF
Stress

700
Med IF
80
High IF
Axial

500
Axial

40
300

100 0
0 2 4 6 8 10 12 14 0 2 4 6 8 10 12 14
Depth
Depth mm
mm
Depth mm

Fig. 4: Maximum axial stress and stress intensity factor vs. thickness distance at the contact edge

High stresses and gradient with high IF and much lower stress and gradient for the lowest IF
are shown in Fig. 5. The highest surface stresses are at the contact for the highest IF. Under
the contact, stress increases linearly with IF level. At the subsurface the maximum stress is
under the contact, but at a much lower level and at the free subsurface the stress is changing
very little, as to be expected.

1200 250
a. Surface axial stress
b. 15mm under surface axial stress
1000 under contact
Contact Edge 200
Axial stress MPa

Axial stress MPa

800
Contact Edge
150

600 Free surface


Under contac t
100
400

50
200
Free surface
Low IF Low IF Med IF High IF
Med IF High IF
0 0
0 0.25 0.5 0.75 1 0 0.25 0.5 0.75 1
Pressure Fit level Pressure Fit level

Fig. 5: Change of axial stress with collar pressure level at 3 locations: at the free surface away from
contact; at the contact edge and under the contact away from the edge.
Subsurface limit load analysis using critical crack length

Using the short cracks subsurface limit approach proposed above, an approximated fatigue
limit length through the thickness of 2mm for the SG cast iron was approximated from Table
1. The FEA result of maximum stress value at the contact edge location with 2mm depth was
used to calculate a critical stress intensity factor for each IF level simulation. By using the
critical SIFs, fatigue limit crack lengths as a percentage of the applied load were calculated;
assuming a linear relationship between the applied load, the maximum stress and the
maximum SIF.

Fig. 6 shows application of the critical crack length method to the FEA simulations. It should
be noted that this analysis is limited to a linear elastic fracture. The analysis shows that the
analysis is highly dependent on the assumed IF in the collar and critical conditions. The
fatigue crack propagations limits are reached at about 35%, 17% and 7% of the applied load
for the low, medium and high IF pressure respectively. This could mean that fatigue limit was
exceeded and preceded by fatigue crack growth.

∆Kth) limit analysis


Fatigue critical SIF threshold (∆

In addition to the analysis above, sensitivity to fatigue cracking was investigated in the
contact region by comparing the subsurface and surface SIFs at locations along the axial
direction, with approximated material fatigue threshold SIF of 5.5 MPa√m as follows:

∆K max < ∆K th no cracking


(6)

∆K max ≥ ∆K th possible cracking

Surface cracks were assumed as a ‘Penny shape’ fracture, for intervals of 0.25mm along the
surface and a range of -5 to +10 mm distance from the contact edge. For this analysis the
FEA surface stresses and at 0.25mm subsurface were used to interpolate the corresponding
SIFs. The analysis has shown similar trends to the critical subsurface analysis with
dependency on the initial collar pressure, distance to the contact edge and sensitivity to sub-
surface stresses.

Application to load Data of wind turbine generators

The fracture analysis was applied to data obtained from a 2MW class of commercial wind
turbines load-duration-distribution spectrum as a full histogram of load distribution is
unavailable, Ref 11, Figure 7. A linear relationship was assumed between the applied load
and the stress developed at the shaft contact, considering that the simulated stress at the
contact edge is related to cyclic stress. The FEA analysis medium level of interference result
was used to apply a fatigue crack growth analysis, Equation 7, using typical blocks of cycles
from the load duration distribution shown in Figure 7.

m m
(1− ) (1− )
a f 2
− a i 2
Nf = (m ≠ 2) (7)
(
C β∆σ π ) m m
(1 − )
2
where m = 1.1855; C = 3.10-7 m/cycles (10) and geometry shape β = 1.0498 (constant).
Substituting maximum stress of 379.5 MPa from the medium interference FEA analysis in the
FCG (equation 7) has shown that if a 2mm crack exists in the structure it is likely to
propagate through the shaft thickness subject to the 1000 cycles of maximum load in 20
years duration (Fig 7). Hence design modifications should be considered. However, this is
conservative as it is considered that the cyclic axial stress is the result of the (constant)
interference pressure and the applied cyclic load.

5 5

4
SG 370 a. Low IF 4
SG 370
SG 420 SG 420
SG 500 SG 500
Fatigue limit length mm

Fatigue limit length mm


2mm crack 2mm crack
3 3

2 2

1 1
b. Medium IF

0 0
0 10 20 30 40 50 60 0 10 20 30 40 50 60
% of applied load % of applied load

SG 370
4 SG 420
SG 500
Fatigue limit length mm

2mm crack
3
c. High IF

0
0 10 20 30 40 50 60
% of applied load

Fig 6: Fatigue critical crack vs. % of applied load for 3 collar pressures.

CONCLUDING COMMENTS

Fairly simple fatigue fracture crack limit and crack threshold analyses were applied to
estimate possible fatigue crack initiation and growth in an SG iron wind turbine low speed
shaft at the edge of a pressure contact. The analysis indicated that the high localised
stresses in this region should be considered in design to avoid possible fatigue damage.
However, more investigation is required to establish the role of the contact pressure in the
fatigue process.
Applied Shaft Moment for Cycles Duration at 3 Wind speeds

120

8.5 m/s
100
10 m/s
Applied/maximum moment %

15 m/s
80

60

40

20

0
1.0E+03 1.0E+04 1.0E+05 1.0E+06 1.0E+07 1.0E+08 1.0E+09
Fatigue cycles duration in 20 years

Fig 7: 2MW class wind turbine low speed shaft typical 20 years load distribution at 3 average
wind speeds.

ACKNOWLEGMENT
This project was carried out in collaboration with Garrad Hassan Ltd, Bristol, UK.

REFERENCES

[1] Johnson K. L. :
Contact mechanics (1985), Cambridge University Press
[2] Edwards, R.E. de los Rios, E.R. Brown, M.W. Rodopoulos, C.A. Yates, J.R. Levers,
A.:
Introducing fretting fatigue to the fatigue damage map (2003), Engineering Integrity,
Vol 14, Num 1, 20-25
[3] Ciavarella M. Demelio G.:
A review of analytical aspects of fretting fatigue, with extension to damage
parameters, and application to dovetail joints (2001), International Journal of Solids
and Structures 38, 1791-1811
[4] ESDU 90031.: Fretting Fatigue
[5] J. Dundurs and M. S. Lee,:
Stress concentration at a sharp edge in contact problems (1971), Journal of Elasticity,
vol. 2, no 2, June 1972
[6] Giannakopoulos A.:
Aspects of equivalence between contact mechanics and fracture mechanics:
theoretical connections and a life prediction methodology for fretting-fatigue;
Giannakopoulos (1998), Acta Materialia , Vol 46 , Iss 9 , 2955 - 2968
[7] Ciavarella M. and Dini D.:
A refined CLNA model in fretting fatigue using asymptotic characterization of the
contact stress field, (2005), Fatigue and Fracture of Engineering Materials &
Structures V 28 Iss 12, 1099 - 1112
[8] El Haddad S., Smith K.N. and Topper T.H.:
Prediction of non propagating cracks (1979), Engineering Fracture Mechanics, 11,
573– 584
[9] Robert Taylor & Co.:
Engineering data on nodular cast irons SI units. (1974)
[10] Marahleh G., Knede A. R. I. and Al-Gousous S.:
Fatigue crack propagation in SGI and CGI (2005), J of Applied Sciences 5 (6), 1004 -
1011
[11] Bernd N., Andreas A., Peter D., Rainer G.:
Load data analysis for wind turbine gearboxes (2007), Germanischer Lloyd
WindEnergie GmbH Publications.

You might also like