Anda di halaman 1dari 10

Solids and Structures (SAS) Volume 2 Issue 4, December 2013

www.seipub.org/sas

Computational Investigation of RollerBearing Premature-Failure in Horizontal-Axis


Wind-Turbine Gearboxes
M. Grujicic *1, S. Ramaswami, J.S. Snipes, R. Galgalikar, V. Chenna, R. Yavari
Department of Mechanical Engineering, Clemson University
Clemson, SC 29634, USA
gmica@clemson.edu

*1

Abstract
To make wind energy economical, wind-turbines are
required to operate with only regular maintenance for at
least twenty years. However, some key wind-turbine
components (especially the gear-box) often require
significant repair or replacement after only three to five
years in service. Consequently, the wind-energy cost and the
cost of ownership of the wind turbine are increased. To
bring the wind-energy cost down, durability and reliability
of gear-boxes have to be substantially improved. These goals
are currently realized using mainly laboratory and field-test
experimental approaches. While these empirical approaches
are valuable in identifying shortcomings in the current
design of the gear-boxes and the main phenomena and
processes responsible for the premature failure of windturbine gear-boxes, advanced computational engineering
methods and tools can not only complement these
approaches but also provide additional insight into the
problem at hand (and do so in a relatively short time). The
present work demonstrates the use of these methods and
tools, and discusses the benefits offered by them within the
context of wind-turbine gear-box roller-bearing prematurefailure.
Keywords
Wind Turbine; Gearbox Reliability; Roller-Bearing Fatigue Failure

Introduction
The present work addresses the problem of rollerbearing premature failure and the long-term reliability
of gear-boxes in wind (energy-harvesting) turbines.
Consequently, the concepts most relevant to this work
are: (a) wind-energy harvesting; (b) wind-turbine gearbox reliability; and (c) premature failure of windturbine gear-box roller-bearings.
Wind Energy Harvesting
Fossil-fuel reserve depletion, stricter environmental
regulations and the worlds ever-growing energy
needs have led to various renewable energy sources

deployed/utilized. Wind energy is one of the most


promising and the
fastest-growing
installed
renewable-energy production technologies.
A wind-turbine is essentially a converter of wind
energy into electrical energy. This energy conversion is
based on the principle having the wind drive a rotor,
thereby transferring power to an electrical generator.
To attain greater structural stability of the rotor and
high aerodynamic efficiency, the rotor is usually
constructed as a set of three aerodynamically-shaped
blades. The blades are (typically) attached to a
horizontal hub (which is connected to the rotor of the
electrical generator, via a gear-box/drive-train system,
housed within the nacelle). The assembly is placed on
a tower and the resulting energy converter is referred
to as the Horizontal Axis Wind Turbine (HAWT).
To reduce the energy-production cost, commercial
wind turbines have grown considerably in size over
the last 30 years. As the hub-height/rotor-radius
increases, the average wind-speed/wind-energycaptured increases. Consequently, fewer windturbines are required to generate the same energy,
which in turn leads to a reduced cost of operation. As
the rotor grows larger, the structural performance,
durability and dynamic-stability requirements become
more challenging, and it is not clear what ultimate
rotor diameter can be attained with the present design,
material and manufacturing technologies.
The blades and gear-box are perhaps the most critical
components/subsystems in current designs of wind
turbines. In our recent work (Grujicic et al., 2007ab,
2011a), the problem of structural integrity and
durability was investigated. The present work, on the
other hand, focuses on issues related to the
performance, reliability and failure-modes of gear-box
components.

47

www.seipub.org/sas

Solids and Structures (SAS) Volume 2 Issue 4, December 2013

Wind Turbine Gear-box Reliability

A schematic of a prototypical wind-turbine gear-box is


shown in Figure 1. The low-speed stage is a planetary
configuration with either spur or helical gears. The
planetary-gear carrier is driven by the wind-turbine
rotor, and the ring gear is stationary/reactionary, while
the sun pinion shaft drives the intermediate-speed
stage, and in turn, the high-speed stage (connected to
the rotor of the electric generator). Typically, the latter
two stages consist of helical gears. Predominantly,
failure initiation is observed in planet bearings,
intermediate-shaft bearings and high-speed-shaft
bearings.

Wind-turbine gear-box failure remains a major


problem to the wind-energy industry (Musial et al.,
2007). The root causes of failure in earlier designs were
associated with problems related to: (a) fundamental
design errors; (b) manufacturing deficiencies; and (c)
under-estimated operating loads. While these problems
have been mainly eliminated by developing and
applying design standards, and establishing good
manufacturing practices (ISO/IEC 81400-4:2005), gearboxes still generally fail to achieve their design-life
goal of twenty years. High failure rates, long
downtimes and the high repair cost have contributed
to: (a) increased wind-energy cost; (b) increased sales
price of wind-turbines due to higher warranty premiums;
and (c) a higher cost of ownership due to the need for
funds to cover repair after warranty expiration. To
make wind-energy more viable, its cost must be
brought back to a decreasing trend, which entails a
significant increase in long-term gear-box reliability.
The current understanding of the basic features and
processes/mechanisms of gear-box failure can be
summarized as follows (Musial et al., 2007): (a) failure
is not strongly related to differences in design, and
generally cannot be attributed to poor workmanship;
(b) failure is often caused by excessive and unexpected
(e.g. misalignment) loading conditions; (c) failure
usually appears to be initiated in excessively and
unfavorably loaded bearings. The resulting damageinduced loading conditions and the propagation of the
bearing-wear debris to the gears cause tooth wear and
gear misalignment (resulting in the final failure of the
gear-box); and (d) the essential features and
mechanisms of damage and failure appear not to
change with size of the wind-turbine.

FIG. 1 SCHEMATIC OF A PROTOTYPICAL WIND-TURBINE


GEAR-BOX. THE MAJOR COMPONENTS AND SUB-SYSTEMS
ARE IDENTIFIED. FAILURE TYPICALLY OCCURS WITHIN THE
(PLANET, INTERMEDIATE-SPEED SHAFT AND HIGH-SPEED
SHAFT) ROLLER-BEARINGS.

48

Premature Failure of Wind-Turbine Gear-box RollerBearings


Provided roller-bearings are: (a) properly maintained
and lubricated; and (b) not subjected to excessive and
unintended loading conditions, their service-life is
controlled by the material high-cycle fatigue (typically
within the bearing races/rings), commonly referred to
as roller-bearing contact fatigue (RCF) failure. The inservice cycling stresses arise from the repeated
exposure of the ring material to ring/roller-element
non-conformal contact stresses. Under well-lubricated/
clean-lubricant conditions, RCF is typically initiated
by subsurface-crack nucleation (in regions associated
with critical combinations of the largest shear stress
and the presence of high-potency microstructural
defects). During subsequent repeated loading, cracks
tend to advance towards the inner surfaces of the
raceways, leading to spall/fragment formation. Under
proper lubrication and normal loading conditions, the
roller-bearing service-life is generally well-predicted
by standard bearing-life calculation methods (ISO
15243:2004; SKF, 2013).
Roller-bearings in wind-turbine gear-boxes tend to fail
much earlier than expected. In addition, the
mechanism and the appearance of roller-bearing
prototypical premature-failure seem different from the
classic RCF failure. In the latter mode, the sub-surface
region contains dark and white bands as well as
chevron-shape cracks. (The visual appearance of RCF
failure is described in detail in ISO 15243:2004). In
premature-failure, the damaged region acquires a
characteristic White Etching Crack appearance, and is
initially localized at or slightly beneath the contact
surfaces. In addition to the chevron-shaped cracks, socalled butterfly white-etching cracks are also often
observed in RCF failure. These cracks are formed at
greater depths and normally associated with excessive
loading. By contrast, white-etch cracking in
premature-failure is believed to be a surface or near-

Solids and Structures (SAS) Volume 2 Issue 4, December 2013

www.seipub.org/sas

surface phenomenon (Stadler and Stubenrauch, 2013).


Specifically, it is believed that a combination of
disturbed bearing kinematics, unfavorable loading
and inadequate lubrication can lead to local tensilestress concentrations, at the root of surface asperities
and/or at inclusion/matrix interfaces near the surface.
For sufficiently high stress concentrations and the
number of loading cycles, surface and/or subsurface
cracks can nucleate. Due to proximity of the contact
surfaces, subsurface cracks can readily extend to these
surfaces (becoming surface cracks).

cracks, ultimately resulting in spallation, appears to be


associated with the operation of corrosion-cracking
mechanisms (Stadler and Stubenrauch, 2013) that are
related to hydrogen and lubricant-breakdown
products diffusing into the crack-tip region of the
raceway material. As a result, crack-propagation
failure is quite fast compared to classical RCF failure,
which takes place mainly in the subsurface region
(which is not accessible to these corrosion agents).

Once formed, cracks are infiltrated by the lubricant


which contains various additives and possibly
contaminants like water. Passage of the rolling
elements over the damaged area can have
hydrodynamic effects, leading to crack spreading and
branching. Newly formed clean-metal crack faces
readily react with the lubricant, causing the formation
of a chemically altered fracture-toughness-inferior
region at the crack tip. These changes, in turn, lead to a
transition from a purely mechanical-fatigue-cracking
regime to a corrosion-assisted fatigue-cracking regime.
The same reactions produce hydrogen, which diffuses
into the surrounding crack-tip region, primarily along
the grain boundaries. This (embrittling) process
reduces grain-boundary cohesion and promotes intergranular cracking. By contrast, in RCF failure, cracking
is predominantly trans-granular and tends to spread
along the bands associated with the maximum shear
stresses and strains.

In our recent work, the problem of gear-box failure


caused by gear-tooth bending-induced fatigue was
investigated computationally. However, as mentioned
earlier, even when the ultimate failure is localized
within the gears, the initial damage could be traced
back to the roller-bearings. Consequently, the main
objective of this work is to conduct a computational
investigation of the wind-turbine gear-box rollerbearing premature-failure, which is assumed to be
associated with hydrogen-assisted grain-boundary
embrittlement and the subsequent inter-granular
fracture. It is hoped that the use of the proposed
approach can: (a) provide additional insight into the
mechanisms responsible for the gear-box rollerbearing premature-failure; (b) assist in developing
new high-reliability gear-boxes; and (c) reduce the
design-loop cycle time and allow alternatives to be
assessed while building confidence in the proposed
solution.

The defining characteristics of the roller-bearing


premature-failure mode are:

Computational Procedure

(a) it preferentially occurs at the inner races/rings;


(b) the cracks nucleate predominantly at the race surfaces;
(c) the final damage is almost always associated with
heavy spallation of the inner-ring raceways;
(d) roller-bearing type/design does not appear to have
a first-order effect on the frequency and intensity of
occurrence of premature-failure;
(e) often, surface-crack initiation is associated with
improper lubrication or contamination/degradation of
the lubricant, or some unfavorable tribo-chemical
surface phenomena and processes. These conditions
generally lead to the changes in the contact surface
referred to as surface distress, which act as a
precursor to the surface-crack formation and include:
(i) discoloration and dulling of the surface; and (ii) the
presence of micro-spalls, micro-cracks or micro-pits;
and
(f) subsequent spreading and branching of the surface

Main Objectives

The problem of roller-bearing premature failure is


investigated using a conventional displacement-based
nonlinear-dynamics finite-element analysis (FEA) of
the type employed in our recent work (Grujicic et al.,
2010a, 2011b, 2012abc). To reduce the computational
cost, all the computational analyses are of a twodimensional plane-strain nature.
Problem Formulation
The basic problem analyzed involves the contact and
the structural response of a roller-bearing element and
one of the raceways. A close-up view of the transverse
section of a roller-bearing is displayed in Figure 2(a).
The contact-patch width (and hence the size of the
highly-stressed region within the contacting pair) is
quite small compared to the characteristic dimensions
of the roller-bearing element and the raceway. Also,
damage responsible for the roller-bearing premature
failure is preferentially located within the raceways.

49

www.seipub.org/sas

Taking all this into account, the problem is simplified


as follows: (a) it is reduced to a two-dimensional
plane-strain problem; (b) only a small region [marked
using a small rectangular white box in Figure 2(a)]
adjacent to the contact patch and located within the
(bottom) raceway is analyzed. The roller-element is
replaced by the surface-distribution of the Hertziantype contact-stresses, Figure 2(b). This distribution
moves along the top surface of the domain, as
indicated by the arrows; (c) crack spreading/branching
responsible for the premature failure is assumed to be
localized within the upper central portion of the
domain. Since premature failure of roller-bearings is
predominantly the result of an inter-granular fracture,
the upper central portion of the domain (referred to
hereafter as the granular sub-domain) is assigned a
granular structure. The material microstructure within
the granular sub-domain is approximated using twodimensional Voronoi cells. Poisson-type Voronoi cells
can realistically represent the topology of grain
microstructure and the statistics of grain size and
shape. Common edges of the adjacent Voronoi cells
are treated as grain boundaries; and (d) the remainder
of the computational domain is treated as a featureless
continuum.

Solids and Structures (SAS) Volume 2 Issue 4, December 2013

FIG. 2(a) CLOSE-UP VIEW OF THE TRANSVERSE SECTION OF A


ROLLER-BEARING; (b) GEOMETRICAL MODEL; AND (c)
MESHED MODEL OF THE TWO-DIMENSIONAL
COMPUTATIONAL DOMAIN ANALYSED.

Finite-Element Analysis
1) Geometrical Model
The geometrical model/computational domain is
depicted in Figure 2(b). The model comprises a
rectangular region consisting of two distinct subdomains: (a) a granular sub-domain which is
allowed to undergo inter-granular cracking; and (b)
the surrounding region which is allowed to
undergo deformation but not damage or cracking.
The sizes of the overall domain (10b 7b) and the
granular sub-domain (4b 4b) are defined in terms
of the (contact-pressure-dependent) contact-patch
half-width b. The Voronoi cell (i.e. grain) size is
selected in such a way that the granular subdomain contains a representative statistical sample
of the material microstructure.
2) Meshed Model
The featureless portion of the computational
domain is discretized using four-node, first-order,
quadrilateral plane-strain continuum elements.
Each Voronoi cell within the granular sub-domain
is meshed using three-node, first-order, triangular
plane-strain continuum elements. These elements
are constructed by connecting the adjacent vertices
of each Voronoi cell with its centroid. The grain
boundaries are represented using four-node twodimensional traction/separation cohesive-zone
finite elements which enable modeling of the
normal- and/or shear-traction induced grainboundary decohesion (i.e. inter-granular cracking).
A close-up of the meshed model used is depicted in
Figure 2(c). As discussed later, in order to model
hydraulic loading of the partially or fully
debonded grain-boundaries, the cohesive elements
are assigned additional degrees of freedom (i.e. the
pore-pressure).

50

Solids and Structures (SAS) Volume 2 Issue 4, December 2013

3) Computational Algorithm
All calculations are based on a transient, displacement-based, purely Lagrangian, conditionally
stable, ambient-temperature explicit finite-element
algorithm.
4) Initial Conditions
At the beginning of the analysis, the computational
domain is assumed to be stationary and stress-free.
5) Boundary Conditions
The following boundary conditions were applied:
(a) all the degrees of freedom of all the nodes along
the bottom face are fixed; (b) to account for the
repeated loading of the raceway by consecutive
rolling elements, periodic boundary conditions are
applied along the left and right sides of the domain;
(c) a moving Hertzian-pressure profile is applied,
as distributed loading, along the top edge of the
domain, in a periodic sense; and (d) a periodic
hydraulic (i.e. pore-pressure) loading, synchronized
with the passage of the Hertzian-pressure profile, is
also applied to all the cohesive elements extending
to the top edge of the granular sub-domain. This
portion of the loading is used to represent
hydraulic loading of the cracks extending to the top
surface of the raceway, by the pressurized
infiltrating lubricant subjected to the piston-like
action of the passing roller-element.
6) Contact Interactions
Inter-granular fracture is modeled as the process of
degradation and ultimate failure of the grainboundary cohesive-zone elements. Once a cohesivezone element has failed, it is removed from the
model. Thereafter, the faces of the corresponding
triangular elements of the adjacent grains are
prevented from interpenetration by activating a
penalty-type contact algorithm between them. In
addition, tangential interaction of the contacting
elements is accounted for using a modified
Coulomb friction law (Grujicic et al., 2012d).
7) Material Models
While the RCF failure is generally assumed to be
preceded
by
extensive
localized
plastic
deformation, premature roller-bearing failure is
typically considered to be stress-driven/controlled.
Consequently: (i) the (continuum AISI 52100 steel)
material within the entire computational domain is
assumed to be of an isotropic, linearly-elastic
character; and (ii) its elastic response is defined in

www.seipub.org/sas

terms of the Youngs modulus (E=210 GPa) and the


Poissons ratio (=0.3).
The constitutive model for the grain-boundary
material also had to be defined. As mentioned
earlier, grain boundaries are represented using
traction/separation cohesive-zone elements. Within
the present formulation, the grain-boundary
separation vector and the associated traction vector
each contain two (one normal and one tangential)
components. To account for the interactions
between the normal and tangential responses of the
grain boundary, each traction component is taken
depending on the two separation components. The
essential feature of these relations is that as the
separation increases, the corresponding traction
first increases, then a peak is reached and
subsequently decreases towards zero (indicating a
complete loss of cohesion). Furthermore,
separation of one type causes additional
degradation in the cohesion of the other type. For
the cases of pure normal separation, i.e. tangential
displacement jump U t = 0 , and the pure tangential
separation, i.e. normal displacement jump U n = 0 ,
the normal traction, Fn , and the tangential traction,

Ft are defined as:


U
Fn (U n , U=
0)
= exp (1.0 ) max n
t
n

Un
exp

(1)

U2
U
2 exp (1.0 ) max t exp t2 (2)

t
t
Examination of Eqs. (1)(2) reveals that four material
parameters are associated with the grain-boundary
cohesive-zone material model: (a) max , the normalF=
0,=
Ut )
t (U n

separation strength; (b) max , the tangentialseparation strength; (c)


=
n U=
max ) ; and (d)
n ( Fn

=
t U=
max ) . These parameters are set to be
t ( Ft
their typical values for the standard roller-bearing
material AISI 52100 steel, as specified in Xu and
Needleman (1993), as follows:
2.5 1010 m ,
max = 1.4 GPa , max = 1.6 GPa , =
n
and =
1.0 109 m .
t
8) Hydrogen-Induced Grain-Boundary Embrittlement
Following Stadler and Stubenrauch (2013), it is
assumed that premature failure of wind-turbine
gear-box roller-bearings is a manifestation of
hydrogen-induced grain-boundary embrittlement
i.e., reductions in max and max as a result of
hydrogen segregation to the grain-boundaries. The
extent of these reductions as a function of the
51

www.seipub.org/sas

degree of hydrogen segregation to the grainboundaries was assessed using the first-principle
calculation results reported by Du et al. (2011, 2012).
Details of this procedure, as well as another kinetic
Monte Carlo-based procedure used to quantify the
effective hydrogen grain-boundary diffusivity and
the kinetics of hydrogen segregation to the grain
boundaries have been reported by Grujicic et al.
(1997). However, the results of these procedures
are utilized in the present work to model
hydrogen-enhanced grain-boundary cracking, the
process which is assumed to be responsible for the
premature failure of the roller-bearings.
To include the effect of grain-boundary embrittlement into the present finite-element analysis of the
roller-bearing premature-failure, the following
procedure should be implemented at each
computational step:
(a) since the process is assumed to be diffusioncontrolled, a fixed hydrogen concentration is
assumed to exist at each crack-tip. This
concentration corresponds to the condition of equal
hydrogen chemical-potential in the lubricant
located in the crack-tip wake and at the adjacent
crack-faces. Then, using the information about the
effective grain-boundary diffusivity of hydrogen,
the concentration profiles and the associated
average concentrations of hydrogen along the (still
bonded) grain-boundaries connected to the cracktip are calculated;
(b) the information obtained in (a) is used to impart
the hydrogen-induced embrittling effect to the
grain-boundaries in question. This was done by

reducing max and max in accordance with the


attendant hydrogen concentration at the grainboundaries;

(c) loading induced by the passage of the rollerelements over the computational domain of the
raceway is then used within the finite-element
analysis to further (mechanically) degrade the
grain-boundary elements. Two components of this
loading are included: (i) surface-type loading
associated with the passage of the Hertzianpressure profile over the top edge of the
computational domain; and (ii) hydraulic loading
associated with the pressurization of the lubricant,
residing within the surface-originating crack(s),
due to the passage and the piston-like action of the
roller-element over the crack mouth; and
(d) the progress of a spread-out and branched

52

Solids and Structures (SAS) Volume 2 Issue 4, December 2013

crack-front is monitored in order to identify the


instant when one of the fronts arrives at the
raceway free surface. This instant is then identified
as the moment of failure of the roller-bearing.
To most accurately determine the initiation and
evolution of damage within the computational
domain, the procedure described above should be
applied at each computational step. However, due
to a very large number of loading cycles required
to produce a spall, a jump-in-cycles approach was
implemented. This procedure involves the
following steps: (i) first, an integer value is
assigned to the number of cycles to be skipped; (ii)
the analysis is then performed over a number of
consecutive time steps required to complete one
loading cycle, i.e. one passage of the Hertzian
contact-pressure over the top edge of the
computational domain; (iii) then the clock is
advanced by the time equal to the product of the
cycle-jump and the total duration of the justcompleted loading cycle; (iv) the time obtained in
(iii) is used to update the grain-boundary hydrogen
concentration, and the normal and shear cohesive
strengths; and (v) the next loading cycle is applied,
and the sequence of steps (i)(v) repeated.
9) Computational Tool
The problem is analyzed using an explicit solution
algorithm implemented in ABAQUS/Explicit. Since
the dynamic explicit finite-element formulation is
only conditionally stable, attention had to be taken
to ensure that the time increments during the
analysis did not exceed the critical time increment
(Grujicic et al., 2008, 2010b).
Results and Discussion
Premature-failure of wind-turbine gear-box rollerbearings is generally assumed to be initiated by one of
the surface-distress processes. Typically, these
processes result in surface-cracking. The granular
subdomain is assumed to contain two surface cracks
from the onset. It is then investigated how the
presence of these cracks and the absence/presence of
hydrogen-induced grain-boundary embrittlement
affect the temporal evolution and spatial distribution
of the inter-granular damage/failure.
All the results presented were obtained under the
following loading and environmental conditions: (a)
Hertzian peak-pressure of 3 GPa; (b) contact-patch
half-width b = 100 m; (c) the intra-crack lubricant
hydraulic-pressure of 375 MPa; and (d) crack-tip

Solids and Structures (SAS) Volume 2 Issue 4, December 2013

hydrogen concentration of 4 x 103 at. %. For the


loading conditions (a) and (b), the values chosen are
consistent with the normal operating conditions of an
intermediate-speed-shaft (ISS) type of bearing in a
prototypical 5 MW HAWT gear-box. The value for the
loading condition (c) was obtained in a separate
analysis in which the piston-like action of the rollerelement onto the intra-crack lubricant, with a
pressure-dependent bulk modulus, was investigated.
In the absence of any experimental data, the
environmental condition (d) is set equal to the nominal
surface solubility limit of hydrogen in BCC iron.

www.seipub.org/sas

initially-induced surface cracks;


(b) after a sufficient number of Hertzian contactpressure passages, the newly-formed cracks arrive
to the top edge of the granular subdomain, yielding
the formation of relatively thin spalls/flakes; and

For clarity, the results corresponding to the absence


and presence of the grain-boundary embrittling effects
are presented and discussed separately.

(c) subsequent loading of the computational


domain by the passing Hertzian contact-pressure
profile produces no additional damage. It should
again be recalled that no attempt was made to
model RCF failure. Consequently, under the
present conditions, one cannot exclude the
possibility of additional damage due to the
interplay of local plasticity and shear localization.
However, modeling of these phenomena is beyond
the scope of the present work.

1) In the Absence of Grain-Boundary HydrogenEmbrittlement

2) In the Presence of Grain-Boundary HydrogenEmbrittlement

Spatial distribution of the grain-boundary


cohesive-zone elements in which both the normal
and the shear cohesive-strengths are degraded by
at least 90% relative to their initial values, after the
25-, 50-, 75- and 100-millionth passage of the
Hertzian contact-pressure profile over the top edge
of the computational domain, in the absence of
hydrogen-embrittlement effects, is shown in
Figures 3(a)(d). Examination of these results
reveals that:

Spatial distribution of the grain-boundary


cohesive-zone elements in which both the normal
and the shear cohesive-strengths are degraded by
at least 90% relative to their initial values, due to a
combined effect of hydrogen-embrittlement and
excessive loading after the 25-, 50-, 75- and 100millionth passage of the Hertzian contact-pressure
profile over the top edge of the computational
domain, is shown in Figures 3(a)(d). Examination
of these results and their comparison with the
corresponding results displayed in Figures 2(a)(d)
reveals that:

(a) the repeated passage of the Hertzian contactpressure first causes some branching of the two

FIG. 3 SPATIAL DISTRIBUTION OF THE GRAIN-BOUNDARY COHESIVE-ZONE ELEMENTS IN WHICH BOTH THE NORMAL AND THE
SHEAR COHESIVE-STRENGTHS ARE DEGRADED BY AT LEAST 90% RELATIVE TO THEIR INITIAL VALUES, AFTER THE: (A) 25-; (B) 50-;
(C) 75-; AND (D) 100-MILLIONTH PASSAGE OF THE HERTZIAN CONTACT-PRESSURE PROFILE OVER THE TOP EDGE OF THE
COMPUTATIONAL DOMAIN, IN THE ABSENCE OF GRAIN-BOUNDARY HYDROGEN-EMBRITTLEMENT EFFECTS.

53

www.seipub.org/sas

Solids and Structures (SAS) Volume 2 Issue 4, December 2013

FIG. 4 SPATIAL DISTRIBUTION OF THE GRAIN-BOUNDARY COHESIVE-ZONE ELEMENTS IN WHICH BOTH THE NORMAL AND THE
SHEAR COHESIVE-STRENGTHS ARE DEGRADED BY AT LEAST 90% RELATIVE TO THEIR INITIAL VALUES, DUE TO A COMBINED
EFFECT OF THE GRAIN-BOUNDARY HYDROGEN-EMBRITTLEMENT AND EXCESSIVE LOADING AFTER THE: (A) 25-; (B) 50-; (C) 75-;
AND (D) 100-MILLIONTH PASSAGE OF THE HERTZIAN CONTACT-PRESSURE PROFILE OVER THE TOP EDGE OF THE
COMPUTATIONAL DOMAIN.

(a) the repeated passage of the Hertzian contactpressure and the operation of the attendant grainboundary hydrogen-embrittlement processes cause
extensive and deep spreading and branching of the
two initially-induced surface cracks;
(b) ultimately, some of the newly-created cracks
mutually connect, forming a large spall/fragment;
(c) the size of this spall, Figure 3(d), is at least an
order of magnitude larger than the one observed in
the no-hydrogen-embrittlement case, Figure 2(d).
This finding suggests that while in the absence of
hydrogen-embrittlement effects, surface-distress
can induce only very shallow craters into the
raceways, affecting somewhat the roller-bearing
performance, in the presence of the grain-boundary
embrittling effects, the raceways may experience a
major damage, resulting in the formation of large
and deep craters, and large-sized fragments. In this
case, the functionality of the roller-bearing element
may be compromised, as well as the functionality
and structural integrity of the adjacent gears
(should any of the large fragments propagate to the
nearby gear-box stage and get caught between the
teeth of meshing gears).
Since a wind-turbine can occasionally experience
unexpectedly high loading (due to the stochastic
nature of the wind), under such conditions, the
Hertzian peak contact-pressure may acquire

54

substantially higher values. Despite the generally


short duration of this excessive loading, exposure of
the bearing to such loading is expected to shorten the
service-life of the bearing. To demonstrate this effect,
the 25-millionth passage of the Hertzian contactpressure is arbitrarily selected to be associated with an
increased peak contact-pressure. Then the effect of the
excess peak contact-pressure on the service-life

FIG. 5 EFFECT OF THE EXCESS PEAK CONTACT-PRESSURE,


APPLIED DURING THE 25-MILLIONTH PASSAGE OF THE
HERTZIAN CONTACT-PRESSURE PROFILE, ON THE ROLLERBEARING SERVICE-LIFE (DEFINED AS THE NUMBER OF
HERTZIAN CONTACT-PRESSURE PROFILE PASSAGES NEEDED
TO PRODUCE A COMPLETE SPALL).

Solids and Structures (SAS) Volume 2 Issue 4, December 2013

(defined as the number of Hertzian contact-pressure


profile passages needed to produce a complete spall)
is investigated. The results of this investigation are
displayed in Figure 5. It is seen that the roller-bearing
service-life is a fairly sensitive function of the extent of
excessive loading. For example, when the Hertzian
peak contact-pressure is increased by 2 GPa, the
service-life has been reduced by 35%. It should also be
noted that the results displayed in Figures 35 are
found not to be significantly affected by the size of the
granular sub-domain, as long as the number of grains
within this sub-domain is sufficiently large (> ca. 1500).

www.seipub.org/sas

decrease their service-life.


REFERENCES

Du, Y. A., Ismer, L., Rogal, J., Hickel, T., Neugebauer, J., and
Drautz, R. First-principles study on the interaction of H
interstitials with grain boundaries in - and -Fe. Phys.
Rev. B 84 (2011): 144121.
Du, Y. A., Rogal, J., and Drautz, R.. Diffusion of hydrogen
within idealised grains of bcc-Fe: A kinetic Monte Carlo
study. Phys. Rev. B 86 (2012): 174110.
Grujicic, M., Zhao, H., and Krasko, G. L. "Atomistic

Summary and Conclusions

Simulation of Effect of Impurities (111) Grain

The problem of premature-failure of the wind-turbine


gear-box roller-bearings has been investigated
computationally using a finite-element analysis and
the relevant results obtained in companion kinetic
Monte Carlo investigations of hydrogen-diffusion in
BCC iron, its segregation to the grain boundary and
the resulting hydrogen-induced grain-boundary
embrittlement.

Boundary Fracture in Tungsten", Journal of Computer

Since the main mode of material damage and failure is


of an inter-granular character, the portion of the
computational domain closest to the rollingelement/raceway contact had to be represented as a
collection of grains and the adjoining grain-boundary
elements. Then, grain-boundary embrittlement was
modeled as degradation of the normal and shear
strengths of the grain-boundary material (represented
using cohesive-zone traction vs. separation finiteelements). Also, damage was assumed to be initiated
at the rolling-element/raceway contact-surface, as a
result of the operation of the surface-distress processes.
The results obtained reveal that in the absence of
hydrogen embrittlement, the initial damage remains
localized to the shallow sub-surface region and
ultimately results in the formation of thin spalls/flakes
and shallow surface craters.
In sharp contrast, in the presence of grain-boundary
hydrogen-embrittlement, the initial damage tends to
spread and branch deep into the contacting materials.
After a sufficient number of loading cycles, cracks
associated with different initial surface-cracks connect,
the formation of a large spall/fragment and deep
surface craters was caused.
In the presence of grain-boundary hydrogenembrittlement effects, occasional excessive loading
experienced by the roller-bearings may substantially

Aided Materials Design 4 (1997): 183192.


Grujicic, M., Pandurangan, B., Huang, Y., Cheeseman, B. A.,
Roy, W. N., and Skaggs, R. R. Impulse Loading
Resulting from Shallow Buried Explosives in Watersaturated Sand, Journal of Materials: Design and
Applications 221 (2007a): 21-35.
Grujicic, M., Pandurangan, B., Zecevic, U., Koudela, K. L.,
and Cheeseman, B. A.. Ballistic Performance of
Alumina/S-2

Glass-Reinforced

Polymer-Matrix

Composite Hybrid Lightweight Armor Against Armor


Piercing (AP) and Non-AP Projectiles. Multidiscipline
Modeling in Materials and Structures 3 (2007b): 287312.
Grujicic, M., Arakere, G., He, T., Bell, W. C., Cheeseman, B.
A., Yen, C. F., and Scott, B. A ballistic material model
for cross-plied unidirectional ultra-high molecularweight

polyethylene

fiber-reinforced

armor-grade

composites. Materials Science and Engineering A 498


(2008): 231241.
Grujicic, M., Marvi, H., Arakere, G., Bell, W. C., Haque, I.
The Effect of Up-armoring the High-Mobility Multipurpose Wheeled Vehicle (HMMWV) on the Off-road
Vehicle Performance, Multidiscipline Modeling in
Materials and Structures 6 (2010a): 229256.
Grujicic, M., Pandurangan, B., He, T., Cheeseman, B. A., Yen,
C. F., Randow, C. L.. Computational investigation of
impact energy absorption capability of polyurea coatings
via deformation-induced glass transition, Materials
Science and Engineering A 527 (2010b): 77417751.
Grujicic, M., and Bell, W. C. A Computational Analysis of
Survivability of a Pick-Up Truck Subjected to Mine

55

www.seipub.org/sas

Detonation

Loads,

Solids and Structures (SAS) Volume 2 Issue 4, December 2013

Multidiscipline

Modeling

in

Materials and Structures, 7 (2011a): 386-423.

International

Organization

for

Standardization,

Wind

Turbines Part 4: Standard for Design and Specification

Grujicic, M., Bell, W. C., Pandurangan, B., and Glomski, P. S.


Fluid/Structure Interaction Computational Investigation
of the Blast-Wave Mitigation Efficacy of the Advanced

of Gear-boxes, ISO/IEC 81400-4:2005, ISO Geneva,


Switzerland, 2005.
International

Organization

for

Standardization,

ISO

Combat Helmet, Journal of Materials Engineering and

15243:2004, Rolling bearings - Damage and failures -

Performance, 20 (2011b): 877893.

Terms,

Grujicic, M., Arakere, G., Hariharan, A. and Pandurangan, B.

characteristics

and

causes,

ISO

Geneva,

Switzerland, 2004.

A Concurrent Product-development Approach for

Musial, W. D., Butterfield, S., and McNiff, B. Improving

Friction-stir Welded Vehicle-underbody Structures,

Wind Turbine Gearbox Reliability. Paper presented at

Journal of Materials Engineering and Performance 21

the European Wind Energy Conference, Milan, Italy,

(2012a): 437449.

May 710, 2007.

Grujicic, M., Arakere, G., Hariharan, A., and Pandurangan, B.

SKF brochure. Bearing Failures and Their Causes.

Two-level Weld-Material Homogenization for Efficient

Publication Number 401 E. Accessed August 8, 2013.

Computational Analysis of Welded Structure Blast

http://www.skf.com/group/knowledge-centre/aptitude-

Survivability. Journal of Materials Engineering and

exchange/articles/index.html.

Performance 21 (2012b): 786796.

Stadler, K., and Stubenrauch, A. Premature Bearing

Grujicic, M., Pandurangan, B., Bell, W. C., and Bagheri, S.


Shock-wave

Attenuation

and

Energy-dissipation

Energy Laboratory. Accessed on August 8, 2013.

Potential of Granular Materials, Journal of Materials

www.nrel.gov/wind/grc/pdfs/12_premature_bearing_fail

Engineering and Performance, 21 (2012c): 167179.

ures.pdf.

Grujicic, M., Arakere, A., Pandurangan, B., Yen, C.-F., and

Xu, X. P., and Needleman, A. Void nucleation by inclusion

Cheeseman, B. A. Process Modeling of Ti-6Al-4V Linear

debonding

Friction

Simulation in Materials Science and Engineering 1 (1993):

Welding

(LFW),

Journal

of

Materials

Engineering and Performance 21 (2012d): 20112023.

56

Failures in Industrial Gear-boxes. National Renewable

111132.

in

crystal

matrix.

Modelling

and