On the Fatigue Life of Wind Turbine Structural Components

Axel Dombrowski, Milan Ristow, Shanmuga Priyan Subramanian

Germanischer Lloyd Industrial Services GmbH, Business Segment Wind Energy (GL Wind)
Steinhoeft 9, 20 459 Hamburg, Germany
Phone: + 49 40 36 149 7141
Fax: + 49 40 36 149 1720
shanmuga-priyan.subramanian@gl-group.com
Abstract: Modern wind turbines are characterised by high, complex loads and by relatively light weight,
optimised structures. The load bearing structural components in wind turbines are mostly made of large,
elaborately shaped iron castings and steel fabrications. Complex, varying loads call for analysis methods to
consider the different phase relationships between the loads acting on the components. These cause variations
in stress directions, usually termed as multiaxiality. One problem is that the limitation of brittle materials, such
as cast iron, to withstand multi-axial stresses is a little known problem outside the research centres. This paper
intends to focus on the effect of high fatigue stresses and multiaxiality on wind turbine components and
discusses the results of recent collaborative research work carried out at GL Wind.
1 Introduction
Fatigue failures of structural components under cyclical loading occur in distinct phases the crack initiation
phase, followed by the crack growth phase and then rupture.
The crack initiation phase is the life of the component up to the formation of a surface crack under fatigue
loading, the size of this crack being defined by the engineering crack criterion. The crack-propagation phase is
the remaining part of the life till the crack length reaches the critical length. The later part belongs to the science
of fracture mechanics where material properties and micro-structure (grain size, imperfections, etc.) play a major
role. The former, the crack initiation life is predicted using methods of damage accumulation which include
hysteresis properties of the material involved and stress or strain histories at highly stressed locations.
2 Fatigue in wind turbine machinery
Modern wind turbines are characterised by high, complex loads and by relatively light weight, optimised
structures. The load bearing structural components in wind turbines are mostly made of large, complexly shaped
spheroidal graphite iron castings and steel fabrications. Also because of remote operations, conservative safety
requirements and long life expectancy, a long crack initiation phase is necessary. The unpredictability of loads
also makes determination of remaining life during crack propagation phase very difficult. Hence, the prediction of
life up to the crack initiation is more relevant to these heavily loaded structural components. Based on this
assumption, suitable fatigue analysis procedures could be applied on the components to be investigated, to
determine the crack initiation life.
A stress based analysis can be divided to two parts: the determination of stresses under fatigue loading and
the determination of representative fatigue property of the material under consideration.
Characteristic
Fatigue Loads

FE Analysis
Stress Response

Material Data

SN Curves

Stress Hypothesis

Stress Time Histories

Cycle counting,
Mean Stress Corrections,
Knock Down Factors, etc.

Safety Concept

Damage Accumulation:
Life, Strength Reserves,
Degree of Utilisation

Figure 1: Fatigue Analysis Procedure

Germanischer Lloyd Industrial Services GmbH, Business Segment Wind Energy

Page 1 of 4

The diagram shows a typical fatigue analysis using the two above mentioned parts. The life of a component is
a function of these two parts. The stresses on the component are calculated using appropriate stress hypothesis
and then cycle counted to enable a comparison with applicable material data such as Stress-Life curves (SN
curves). Different safety and knock down factors are considered in these calculations before damage
accumulation is carried out to determine the component life and the degree of utilisation.
Component
Material Strength

50% P

97.5% P

5%

Safety Factors

50%

Probability Density

Component Stress
due to Loading

Stress

Figure 2: Safety Concept in Fatigue Calculations (based on FKM Guidelines, [2])

Based on FKM Guidelines [2], the concept of safety for fatigue is accordingly represented in the above
diagram. It vividly tells that even component stress due to loading (with a low probability of occurrence), needs to
be below a material strength value with high probability of survival, after consideration of all applicable knock
down and correction factors. The difference between these stress values is the safety margin and should at least
cover the fatigue safety factor requirements determined by the consequences of failure and the probability of its
occurrence.
3 Determination of material properties
Complexly varying loads call for fatigue analysis methods to consider the different phase relationships
between the loads acting on the components. These can cause variations in stress directions, usually termed as
multiaxiality. The problem is that the limitations of brittle materials, such as cast iron, to withstand multi-axial
stresses are a little known problem outside the research centres.
Fatigue strength of materials, such as SN curves for spheroidal graphite cast iron, is experimentally
determined on specially prepared laboratory test specimens and usually defined only for uniaxial stress states.
The uncertainties involved in quantifying this property are covered using statistical parameters such as probability
of failure, scatter, slope & knee of SN curve, level of confidence and the number of tests. Influence of surface
roughness, technical defects, wall thickness, mean stress and residual stress, notch effects are also appropriately
considered as knock down factors to determine the component material strength under fatigue loads for uniaxial
stress states. These factors cover all related known risks and a careful combination of these factors is necessary
to theoretically optimise load carrying capacity of the component.
This is the generally accepted procedure to derive component SN curves since it has been financially unviable
to conduct accelerated tests on full size components to gain the necessary material data.
4 Determination of component stresses
Because of the increased complexity of the component geometry and varying load scenarios, the local stress
approach in combination with detailed FE models and load time histories is the most widely used method in this
industry.
The finite element calculations are based on CAD models of component geometries with linear, isotropic
material properties and multiple loading points. The load histories for the components are extracted from multibody simulations of a wind turbine model for various fatigue loading scenarios. Linear superposition of the loadcoefficients (unit load stresses or transfer functions) determined from FE computations are multiplied with load
histories to get local stress tensor histories at critical locations. This is the generally accepted procedure for linear
calculations but if non-linear boundary conditions are to be observed, for e.g. contact surfaces of bearings etc.,
then modified approaches are necessary.
It is to be noted that reliable fatigue material properties exist predominantly for uniaxial stress states. Several
hypotheses also exist to enable conversion of component stresses to an equivalent uniaxial stress. There also
Germanischer Lloyd Industrial Services GmbH, Business Segment Wind Energy

Page 2 of 4

exist separate methods such as the critical plane approach to deal with multiaxial stress states caused by non
proportional loading conditions.
To help understand this, structural components of a 2 MW wind turbine are used as an example in this study.
Fatigue loads in combination with FE models of the components were used to compute the stress histories at
critical locations. Based on this, a local stress state analysis was performed to analyse the multi-axial behaviour
of the stresses at critical locations and investigate the degree of multiaxiality (i.e. change in principal stress
directions during a loading sequence).
M
U
M

U
U: Uniaxial Stress during loading
M: Multiaxial Stress during loading

Figure 3: Locations selected for Local Stress State Analysis

The results show that at several locations in the hub there exist multiaxial stress states. At these locations, it
was observed that though the directions of principal stresses change greatly, the magnitude of the stresses
remain low. The main frame on the other hand, has several locations with uniaxial stress states. Following are
plots showing samples of typical uniaxiality and multiaxiality.

Figure 4: Sample results of Local Stress State Analysis

The procedure for fatigue analysis needs to consider the influence of these changing principal stress
directions by applying appropriate stress hypothesis and quantify its detrimental effect on fatigue strength of the
chosen material. This is critical since the used materials show an additional cyclical stress hardening effect with
varying principal stress directions. Fatemi has even indicated that out of phase stresses are more damaging than
equivalent in-phase stresses in the low cycle region [7].
The appropriate stress hypothesis is to be selected based on several factors:
- Uniaxial or multiaxial stress state
Germanischer Lloyd Industrial Services GmbH, Business Segment Wind Energy

Page 3 of 4

- Brittle or ductile material

- Phase relationships between applied loads
- Tensile, compressive or shear loading
- Type of stress concentration (for e.g., fillets and small radii, seam welds, spot weld, etc.)
The following is a summary of widely used stress hypotheses.
Conventional maximum stress theory (Tresca) is not a good parameter for the multiaxial fatigue life
predictions [7]. This method has not been included in the comparative-analysis and has been mentioned just as
information.
Von-Mises or Equivalent stress approaches reduce the multiaxial stress or strain states to a single scalar
value that can be used to calculate fatigue damage just as for a uniaxial state of stress. These approaches are
exact if the state of stress is uniaxial, and are generally very good if the state of stress is proportional or nearly
proportional [6]. It also correlates well with tests on ductile materials [7].
Maximum principal stress theory successfully predicts fatigue life in brittle material for in-phase loading [7].
Critical plane approach proves to be reasonably accurate in predicting fatigue life for both in-phase and out-ofphase loading [7].To determine the critical plane, the multiaxial stress state at a particular point is estimated.
Then, this multiaxial stress state is resolved to particular planes (orientations) in the material to compute the
normal and shear stresses and strains acting on the plane. Fatigue damage for these resolved stress histories is
computed based on a rainflow counting and an appropriate damage parameter. When all of the calculations for
the planes are complete, the plane with the highest calculated damage is the critical plane, and determines the
fatigue life of the structure at that location. Although there is an infinite number of planes at a particular point,
candidate critical planes are commonly taken to be ten degrees apart, because it has been observed that fatigue
damage does not vary much for planes less than ten degrees apart. Furthermore, if the damage is to be
calculated on a free surface of a structure, the number of candidate planes can be greatly reduced [6].
Absolute Maximum Principal Stress approach does not represent the real local stress-time history at the
location but the calculated damage is, according to the academicians, representative of the experimental results
and this has been accepted as de facto standard, although the approach neglects or nullifies the mean stress
corrections. If an SN curve specifically developed for the component under representative loading is not available,
statistically determined or analytically determined uniaxial SN curves for the material are used. This is accepted
but care has to be taken to convert multiaxial stresses at hot spot locations into representative uniaxial stresses
(using appropriate hypotheses), before cycle counting and comparison with a SN curve. Absolute maximum
principal stress approach is one such hypothesis. It gives conservative results and it covers the uncertainties
involved in combining multiaxial stress states with uniaxial material properties.
5 Conclusion
The paper presented briefly the fatigue analysis procedure, the concept of safety and the generally accepted
method for the determination of material properties. The main aim of this paper was to throw some light on to the
determination of equivalent component stresses that are representative and valid enough, in combination with
uniaxial material properties, for use in fatigue computations. Samples of uniaxial and multiaxial stresses were
shown to encourage designers to investigate the local stress states thoroughly before deciding on a stress
hypothesis. Interpretation of such local stress state investigations also reveal significant information that could be
used for further optimising the component geometries.
6 References
[1] Germanischer Lloyd Rules and Guidelines, IV Industrial Services, Part 1 Guideline for the
Certification of Wind Turbines, Edition 2003 with Supplement 2004
[2] FKM Guideline, Analytical Strength Assessment of Components in Mechanical Engineering, 5th, revised
Edition, 2003, VDMA Verlag
[3] Haibach, E., Betriebsfestigkeit, 2nd Edition, 2002, VDI Verlag
[4] Gudehus, H., Zenner, H., Leitfaden fr eine Betriebsfestigkeitsberechnung, 4th Edition, 1999, Verlag
Stahleisen, Dsseldorf
[5] Bishop, N.W.M., Sherratt, F., Finite Element Based Fatigue Calculations, 2000, NAFEMS Ltd.
[6] LMS-FALANCS Theory Manual, LMS Durability Technologies GmbH, 2000.
[7] Fatemi, A., Socie, D.F., A Critical Plane Approach to Multiaxial Fatigue Damage including Out-Of-Phase
Loading Fatigue Fract. Engng. Mater. Struct. Vol. 11, No. 3, pp. 149-165, 1988.
Germanischer Lloyd Industrial Services GmbH, Business Segment Wind Energy

Page 4 of 4