Professional Documents
Culture Documents
2019-Muhammed - (THESIS) - Lulea University-Spiral GB-multi Physics Modelling
2019-Muhammed - (THESIS) - Lulea University-Spiral GB-multi Physics Modelling
Modeling
Cone crushers are widely used in the mining, mineral processing and quarrying
segments of the industry to crush ores and large rocks. In such machinery, the
load to be carried is rather heavy and the motion is gyratory which creates a
need for a bearing set that can withstand such severe conditions. Sandvik AB
is a high-technology Swedish engineering group specialized in tools and tooling
systems for metal cutting, equipment, as well as tools and services for the min-
ing and construction industries. One of their products relevant to the mining
industry is the cone crusher which utilizes a 3-piece bearing set to carry thrust
load. This bearing can be classified as a Spiral Groove Bearing 1 , and it has been
incurred that it wears out rather quickly and is believed to be running under
mixed-lubrication conditions where the interfaces in the bearing-set are not fully
lubricated. The aim behind this thesis is to create a multiphysics model of this
bearing in order to understand deeply how it works and the reasons why it does
not perform as expected as well as to predict design improvements which can
improve the performance of the bearing-set, thus increasing its operating life.
It has been concluded that the bearing operates under severe mixed-lubrication
conditions and that the generation of a squeeze film is the only method by
which lubrication takes place due to the excessive depth of the grooves which
is needed to allow for an adequate amount of cold oil to flow into the grooves
and cool the interface as well as to accommodate for a considerable amount of
wear particles. In light of the results and insight gathered from the simulations,
possible design variations of the bearing which can be advantageous in terms
of mitigating asperity friction in the interfaces of the bearing are discussed and
tested.
I would like to dearly thank my supervisor Prof. Andreas Almqvist for the time
and devotion he offered me, which was crucial to my development and progress
in this thesis. Not only did he help a lot, but his contagious enthusiasm to this
field of science made me appreciate and enjoy working tirelessly on this thesis. I
would also like to thank my co-supervisors at Sandvik, namely: Patrik Sjöberg,
Jan A Johansson, Sonny Ek and Tatiana Smirnova for their support and for the
fruitful bi-weekly progress meetings.
To mum and dad, I owe you every success that I am living now, for you have
sacrificed a lot and worked so hard and against all odds to put me on track to
succeeding on the top-level abroad. With this thesis, and my graduation from
my M.Sc., I hope to have given you back at least a fraction of the happiness and
pride I want to make you feel throughout your lives for what you have sacrificed
for my sisters and I.
To Salma, my dearest love and best friend, your presence in my life is nothing
short of magical. Your love and support throughout this year were one of the
main reasons I was able to make it, for I have lived this year alone in what
seems like the furthest up-north a person from our side of the world can reach.
You were my motivator and the person I confined to in the darkest moments,
and were the person I shared my happiness with when things were right. I can
only imagine how difficult it would have been without you, and so I dedicate
this thesis to you.
Finally, I would like to thank the EACEA of the European Union for the gen-
erous funding of my degree as well as every professor and colleague I was lucky
to meet and work with during my time in Leeds, Ljubljana and Luleå.
Preface
The thesis has been developed under the supervision of Prof. Andreas Almqvist
at the Department of Engineering Sciences and Mathematics, Division of Ma-
chine Elements at Luleå University of Technology during the period 10.10.2018
to 01.07.2019.
1 Introduction 5
1.1 Cone Crushers . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2 Sandvik’s Cone Crusher Bearing Set . . . . . . . . . . . . . . . . 7
1.3 COMSOL Multiphysics . . . . . . . . . . . . . . . . . . . . . . . 8
1.4 Objectives and Motivation . . . . . . . . . . . . . . . . . . . . . . 9
1.5 Delimitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2 Literature Review 11
2.1 Simulations in Tribology . . . . . . . . . . . . . . . . . . . . . . . 11
2.1.1 Analytical Methods . . . . . . . . . . . . . . . . . . . . . 11
2.1.2 Continuum Methods . . . . . . . . . . . . . . . . . . . . . 12
2.1.3 Particle-based Methods . . . . . . . . . . . . . . . . . . . 14
2.3 Spiral Groove Bearing in Literature . . . . . . . . . . . . . . . . . 16
2.4 Multiphysics Models of Bearings in Literatu . . . . . . . . . . . . 19
2.5 Gap in Literature . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4 Results 42
4.1 Modelling ladder, part 1 . . . . . . . . . . . . . . . . . . . . . . 42
4.1.1 Analytical - Rayleigh Step-bearing . . . . . . . . . . . . . 42
4.1.2 Numerical - Rectangular Rayleigh Step-bearing Pad, Dif-
ferent Width:Length Ratios . . . . . . . . . . . . . . . . . 43
4.1.3 Numerical - Cylindrical (Almost-Quadratic) Pad . . . . . 44
4.1.4 Numerical - Cylindrical Pad, Different Circumferential
Widths . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
4.1.5 Numerical - Straight Groove Bearing (Shallow Grooves . 46
1
4.2 Modelling ladder, part 2 . . . . . . . . . . . . . . . . . . . . . . 47
4.2.1 Straight Groove Bearing - Rigid . . . . . . . . . . . . . . 48
4.2.2 Spiral Groove Bearing - Rigid . . . . . . . . . . . . . . . 49
4.2.3 Sandvik’s Piston-step - Rigid . . . . . . . . . . . . . . . . 50
4.2.4 Spiral Groove Bearing - Deformed . . . . . . . . . . . . . 51
4.2.5 Sandvik’s Piston-step - Deformed . . . . . . . . . . . . . 52
4.2.6 Validating the Implementation of Solid Mechanics Physics 53
5 Possible Design 54
5.1 Shallower Grooves . . . . . . . . . . . . . . . . . . . . . . . . . . 54
5.2 Hydrostatic Pressure . . . . . . . . . . . . . . . . . . . . . . . . 54
5.3 Larger Surface Area of Ridges . . . . . . . . . . . . . . . . . . . 55
5.3.1 Decreasing groove:ridge surface area ratio . . . . . . . . . 55
5.3.2 Decreasing number of grooves . . . . . . . . . . . . . . . 57
6 Discussion 59
6.1 Validation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
6.2 Sandvik’s Bearing vs. Parameterzied S.G.B . . . . . . . . . . . . 59
6.3 Rigid vs. Deformed Simulations . . . . . . . . . . . . . . . . . . 59
6.4 Design Changes . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
7 Conclusions 61
8 Future Work 62
9 References 63
2
Nomenclature
α Threshold constant in cavitation model
β Inclination angle between grooves and velocity vector in s.g.b
Threshold pressure in cavitation model [Pa]
Ω Piston-step area [m2 ]
ω Eccentric velocity [rad/s]
ρ Fluid density [kg/m3 ]
σ Stress [Pa]
τ Shear stress [Pa]
υi Poission ratio
ε Strain
aij , bi Flow factors
E Young’s Modulus [Pa]
E∗ Equivalent Young’s Modulus [Pa]
Fasp Asperity load [N]
Fhyd Hydrodynamic load [N]
h Separation (including surface deformation) [m]
hm Separation at point of maximum pressure [m]
h0 Rigid separation [m]
pasp Asperity pressure [Pa]
phyd Hydrodynamic pressure [Pa]
r1 Inlet radius of s.g.b [m]
r2 Outlet radius of s.g.b [m]
t Time [s]
u Velocity in x-direction (in Reynold’s eq) [m/s]
Deformation in x-direction (in Solid Mechanics) [m]
v Velocity in y-direction (in Reynold’s eq) [m/s]
Deformation in y-direction (in Solid Mechanics) [m]
w Velocity in z-direction (in Reynold’s eq) [m/s]
Deformation in z-direction (in Solid Mechanics) [m]
3
This page is intentionally left blank.
4
1 Introduction
In this section, the spiral groove bearing used by Sandvik as well as their cone
crushers in which the bearings operate will be introduced. COMSOL Multi-
physics will be shortly introduced and finally, the objectives of the thesis and
its delimitations will be laid out.
Throughout the most of history, crushing of rocks and other materials happened
using manpower as force is concentrated on the material to be crushed from the
tip of a sledge hammer or other similar tools. Initially, most ore crushing and
sizing was carried out by hand and hammers at mines or by water powered
trip hammers in the small charcoal fired smithies and iron works typical of the
Renaissance period, after which explosives came into use. Later on, during the
industrial revolution of twentieth century, various forms of mechanical crushers
were developed, of which the cone crusher is one of the most widespread.
5
Figure 1: Schematic showing where the rocks/gravel enter the crushing cycle.
As we see in Figure 1, the rocks/gravel is feed from the top into the interface
between the mantle and the bowl which is lined with manganese alloys (14 to
22 percent manganese) that provide a certain balance between impact resis-
tance and abrasion resistance. The higher the manganese content, the higher
the abrasion resistance but the lower the impact resistance and vice-versa. The
eccentric shaft rotates with a gyratory motion and causes the mantle to rotate
as well, which causes the crushing action.
6
1.2 Sandvik’s Cone Crusher Bearing Set
In Figure 2, you can see the 3-piece step bearing set comprising of a mainshaft
step, step washer and a piston step. This bearing serves the crucial purpose
of supporting the heavy irregular thrust load and gyratory motion of the cone
crusher, therefore it has to have the appropriate design which produces the
hydrodynamic lift needed for a load bearing ability that is as high as possible
in order to prevent mixed or boundary lubrication from happening under such
demanding conditions which leads to wearing out and loss of function of the
bearing and therefore damage to the cone crusher.
In Figure 3 the bearing set used by Sandvik is shown more clearly. This bearing
can be classified as a logarithmic spiral groove bearing, which will be talked
about in detail in the literature review Section 2.3.
The bottom piece of the bearing (piston step) is stationary while the upper
piece (mainshaft step) moves in a similar manner and eccentric speed as the
main/eccentric shaft. The middle piece (step washer) moves in a gyratory mo-
tion at a speed that is unknown.
Figure 3: Sandvik’s bearing set showing the piston step and step washer.
7
1.3 COMSOL Mutliphysics
8
1.4 Objectives and Motivation
The main objective behind this thesis is to create a multiphysics model of Sand-
vik’s bearing on COMSOL in order to predict and understand how the bearing
operates. The physics/equations that would be coupled and solved are:
Another end goal of the thesis is to create user-friendly applications for both
Sandvik’s geometry and the parameterized geometry which can be used by Sand-
vik to change geometrical and other model parameters and run simulations eas-
ily.
As for the greater motivator, sustainablity has been a main motive in industry
for a while now, so an optimized bearing design that can operate for a longer
lifetime will certainly be more sustainable. If the design also leads to lower fric-
tion and wear rates, then unwanted material and energy loss will be eliminated
as well.
9
1.5 Delimitations
While creating the models, there was a need to make a number of justified as-
sumptions and simplifications in order to save time as well as computational
complexity of the models. These assumptions are:
2. The heating and thermal expansion of the interface due to friction as well
as the cooling of the interface due to the lubricant flow in the grooves is
neglected. This is, however, important to model in future work since the
frictional heating of the interface is expected to be significant and would
affect the performance of the bearing.
3. Only the lower interface between the piston step and the step washer is
modelled, since it was believed that modelling the upper interface as well
would be redundant as both interfaces are similar to a great extent.
4. The axial loading cycle that the bearing is subject to from the cone crusher
was approximated and simplified by a fourier series to make it easily con-
figurable and to facilitate convergence in COMSOL.
10
2 Literature Review
In Subsection 2.1, simulation methods utilised in the field of Tribology are in-
troduced. In Subsection 2.2, the Finite Element Method is elaborated in more
detail since it is pivotal to the work done in this project. In Subsection 2.3, the
various literature in which spiral groove bearings were developed are mentioned.
In Subsection 2.4, a number of multiphysics models created in COMSOL for dif-
ferent types of bearings are mentioned and finally in Subsection 2.5, the gap in
the literature is shortly spoken about.
Analytical methods are most widely used in the area of contact mechanics, and
the most famous theory in this area is the Hertzian theory that has been a
cornerstone in contact mechanics and tribology ever since it has been described
in [5]. The Hertzian theory solves the problem of two ideally smooth surfaces in
11
frictionless contact. The JKR 2 and DMT 3 models [6, 7] are two other famous
models for ideally smooth surfaces, in which the latter takes only the adhesion
in the area out of the contact zone into consideration, while the former does
the opposite. Rough contact has been studied by Greenwood-Williamson and
their model remains the most widely used, although it has a lot of inaccurate
assumptions [8]. Recent advancements in this field have been made in con-
tacts in the presence of sharp edges [9] and conformal configurations [10] as well
as studying of contact in the presence of anisotropic and functionally graded
materials, with varying friction coefficient along the interface in sliding and
partial slip conditions [11]. Finally, Majumdar and Bhushan have created on
of the most popular theories after the GW model for studying contact of rough
surfaces [12]. As for the contact of lubricated surfaces, analytical solutions also
exist in special cases such as infinitely-wide bearings as shown in Subsection 3.4.
Continuum methods are the basis of most commercial simulation softwares avail-
able, and are widely used in modeling large-scale components due to their ac-
curacy and small computational time [8]. The most notable and widely used
continuum methods are the Finite Element Method (FEM), Finite Difference
Method (FDM) and Finite Volume Method (FVM).
2 To introduce the effect of adhesion in Hertzian contact, Johnson, Kendall, and Roberts
created their theory of adhesive contact using a balance between the stored elastic energy and
the loss in surface energy. This model considers the effect of contact pressure and adhesion
only inside the area of contact.
3 This model, named after Derjuagin-Muller-Toporov is an alternative model for adhesive
contact where the contact profile is assumed to remain the same as in Hertzian contact but
with the addition of attractive interactions only outside the area of contact.
12
FDM is one of the simplest and oldest methods of solving differential equa-
tions, of which its one-dimensional formulation was known to the famous math-
ematician L.Euler as early as 1768. It was then extended to two dimensions by
C.Runge in 1908. However, serious development and practical application of this
method has started after the 1950s when computers have emerged which allowed
solving more complex problems in engineering and science using this method.
FDM is a method used more commonly on regular grids such as rectangular
ones that are not complex in shape, it takes the continuous form of a PDE
and replaces it with a set of discrete equations and then solves the equations.
This is usually done on regular grids and is more efficient, fast and accurate
than FEM in large scale computations such as astrophyiscal and meteorological
simulations given that they can fit into regular grids. However, FDM runs into
problems when having to deal with curved boundaries or complex geometries.
BEM is the most effective method for simulating contact problems in contact
mechanics or any problem in which we are a lot more interested in what is hap-
pening on the boundary (surface) than in the volume of the geometry. When
using BEM, one can mesh only the boundary and solve on it which save com-
putational resources. The main difference between BEM and the previously
mentioned continuum methods is that BEM only needs to solve for unknowns
on the boundaries, whereas FEM, FDM and FVM solve for unknowns in the
volume [14]. However, to formulate a BEM, a fundamental solution linking pres-
sure and vertical displacement in two orthogonal directions is needed, and such
solution exists for a limited number of cases and mainly under the assumption
that the solid can be locally considered as a flat half-space, making the BEM
more restrictive to use [15].
4 Computational fluid dynamics (CFD) is a branch of fluid mechanics which uses numerical
analysis and simulation to solve and analyze problems that involve fluid flow.
13
2.1.3 Particle-based Methods
Particle-based methods are used for simulation mainly on the atomic and mi-
croscales and most of them are based on Molecular Dynamics, they work by
discretizing the domain into particles (atoms) rather than meshes, as is the case
in continuum methods. The DEM 5 and MCA 6 methods may be the most
notable particle-based methods, the latter being a more recent method that
has proven to be effective and accurate in mesoscale simulations. This MCA
method combines the advantages of classical cellular automata, discrete element
methods and molecular dynamics. It assumes that the simulated system is dis-
cretised into a series of small elements of a finite size called movable cellular
automata [16]. The movements of those elements are simulated using Newton-
Euler equations of motion while including their pair relationship, many body
forces and bond forces and was created by Psakhie et al [17]. However, if a ma-
chine component is to be simulated, it would be made of billions of automata
making it computationally very expensive to simulate and this the most recent
problem that needs to be faced in this area. Currently, scientists in this field
want to examine the possibility of parallel-computing of MCA simulations, that
is to compute on multiple-processors to speed up the computational time and
be able to simulate larger scale components.
The Finite Element Method can be traced back to the appendix of a paper by
Courant in 1943 [18] in which the piecewise linear approximation of the Dirich-
let problem over a network of triangles was discussed. Many additions have
been made throughout the following years to include partitioning of domains,
boundary conditions and local polynomial approximations to come to the first
mature version of FEM in 1956 [19] in which an attempt was made at both
a local approximation of the PDE of linear elasticity and the use of assembly
strategies essential to FEM.
5 The Discrete Element Method (DEM) is a particle-scale numerical method for modeling
method based on the discrete concept. In the MCA approach, an object being modeled
is considered a set of interacting elements/automata. The dynamics of the set of automata
are defined by their mutual forces and rules for their interaction. This system exists and
operates in time and space and its evolution in time and space is governed by the equations
of motion.
14
The working principle of the Finite Element Method (FEM) is that it numeri-
cally approximates a given Partial Differential Equation related to some physical
phenomena intended to be studied on a domain (eg. viscous heating of a lubri-
cant film) by discretizing the domain into meshes (discussed in Subsection 3.2)
and numerically approximating the PDE locally on each element (mesh) and
adding them up to obtain the global approximation of the PDE on the whole
domain.
The FEM has gained wide popularity and approval in engineering during the
twentieth century and has been used in every conceivable area of engineering
that makes use of models of nature characterized by partial differential equa-
tions. The reason behind the popularity of the FEM is its flexibility which is
derived from these intrinsic characteristics of the FEM [20]:
These advantages make FEM the best numerical method to be used for approx-
imating the pressure distributions and surface deformation in Sandvik’s bearing
during this project. Out of all FEM-based tools, COMSOL was the best to
use due to its capability of carrying out multiphysics simulations on complex
geometries, as mentioned previously.
15
2.3 Spiral Groove Bearing in Literature
The first paper to give an account of a spiral groove bearing in the usual form,
with two opposing surfaces, one flat and the other grooved, was one from 1957,
written by Harris, Ford and Pantall [21]. However, grooved bearings are be-
lieved to have been the subject of an earlier paper written by Whipple in 1949
[22]. Using a theoretical approach, Whipple derived a formula for predicting
the pressure profile between horizontal strips, one of which has a parallel array
of straight grooves as shown in Figure 4, while employing various assumptions
to assure that the pressure distribution across the grooves can be regarded as
linear. In this paper, Whipple makes no attempt to apply his formula directly
to spiral groove bearings, but the formula was later used by other researchers
to obtain rough but useful estimates of the operating characteristics of spiral
groove bearings. Whipple’s analysis was then improved in accuracy and ex-
tended by researchers to become better applicable to the spiral groove bearing.
One of the most important pieces of literature that laid down a solid theory of
equations governing the operation of spiral groove bearings without employing
much of the assumptions used in [22] is a thesis by Muijderman produced in
Philips Research Laboratories in The Netherlands in 1964 [23].
16
The working principle of the spiral groove bearing is described in more detail
in Subsection 3.1. The pumping effect added to the restriction effect gives an
advantage to spiral groove bearings over other types of self-acting bearings in
that they produce a pressure distribution from the edge of the bearing to the
center while pressure at the center of the tilting pad thrust bearing, for example,
is zero. This contributes greatly to a higher load bearing ability for the same
size of the bearing. Therefore, air can be used as a lubricating medium and still
produce an acceptable load carrying capacity [24]. Since air is more environ-
mentally friendly than oil and grease and of course incurs less cost, it gives this
type of bearing a large prospect of being more widely applied in industry. A
quotation from [23] states the following:
“If the grooved disc, of 100 mm diameter is placed on the smooth one and ro-
tated in the right direction, a rotational speed of about 1 revolution per second
will be enough to separate the two discs by a layer of air about 11 micron thick,
which will then support the weight, 200 grf (1,96 Newtons), of the grooved disc.
This combination thus works as an air-lubricated thrust bearing.”
The logarithmic spiral groove bearing can also produce relatively high load car-
rying capacities for a smaller size of a bearing, which makes it widely used in
electrical appliances and since spiral groove bearings in general can produce
very high load carrying capacities for a given size of bearing compared to other
types of self-acting bearings, they can sometimes be the only design solution in
extreme cases such as huge thrust bearings under extreme load conditions in
turbines, or in very small bearings (a few mm in diameter) with large rotational
speeds.
The reason why grooves in this type of bearing are logarithmic, is that there
exists a certain optimum angle β between the tangent to the grooves and the
local velocity vector which produces the highest load carrying capacity and the
grooves being logarithmic ensures that this angle is kept constant throughout the
bearing surface. This optimum angle was found to be 20◦ for inward-pumping
logarithmic spiral groove bearings, according to recent calculations made in [39].
17
Figure 5 shows the piston step of Sandvik’s bearing as an example of a logarith-
mic spiral groove bearing. The grooves in the piston step of Sandvik’s bearing
are excessively deep as mentioned previously, which prevents the formation of a
full-film of lubricant due to restriction action and thus detrimentally affects the
load-carrying capacity of the bearing.
18
2.4 Multiphysics Models of Bearings in Literature
A few relevant multiphysics models created using COMSOL for different types
of bearings which I have found to be very useful are shown below:
19
3 Theoretical Basis and Methods
In this section, the theoretical basis of concepts and equations needed for simu-
lation of the physics in Sandvik’s bearing is laid out including the Navier-Stokes
equations, Homogenized Reynold’s equation, meshing, Solid Mechanics physics
and the mass-conserving cavitation model used to account for cavitation.
3. Thrust bearings and radial bearings are sub-classes of bearings that op-
erate using the restriction effect/action. Thrust bearings are those which
support mainly or exclusively thrust (axial) loads.
4. Spiral groove bearings are one sub-class of thrust bearings, while stepped
bearings (ex. Rayleigh step-bearing) and Mitchell bearings (ex. tilting
pad thrust bearings) are other sub-classes of thrust bearings.
5. Under the sub-class of flat spiral groove bearings, bearings with transverse
flow are the logarithmic spiral groove bearings and those with no trans-
verse flow are the herringbone spiral groove bearings.
20
If one follows the green boxes from the very top till the bottom of the figure,
the inward pumping logarithmic spiral groove bearing will be reached, which is
the type of bearing used by Sandvik in their cone crushers but with excessively
deep grooves.
21
Now to understand the restriction effect/action, let’s imagine the simple Rayleigh
step-bearing shown in Figure 7, when the lubricating oil is pushed in the direc-
tion shown in the figure it has to pass through a narrower gap (restriction) right
after the step which causes pressure to be build up in order to open up this gap
to balance the flow at the inlet and outlet. If the two surfaces are restrained by
an external force, then the pressure build-up will restrain the amount of fluid
trying to enter at the inlet and will push more fluid out at the outlet so that
flow is equal at the inlet and outlet, this is called the restriction action.
What causes pressure build-up in the logarithmic spiral groove bearing, for ex-
ample, is the restriction action and pumping action together, in which the pump-
ing action happens due to the angle between the tangent line to the grooves and
the local velocity vector of the opposing surface which is rotating. The direction
of rotation determines if the pumping will be inward or outward [28]. As the
lubricating fluid is pumped over the grooves, a narrow gap is faced similar to
the Rayleigh step-bearing which leads to pressure build-up. Inward pumping is
required to produce a positive pressure distribution and load carrying capacity.
It is very important to note that the grooves in the piston step and mainshaft
step of Sandvik’s bearing are excessively deep in order to allow for a larger vol-
ume of cold lubricant to be pumped into the grooves to prevent overheating of
the interface as well as to accommodate for a large amount of wear that happens
on the bearing surface during its operation, where the worn out material falls
into the grooves. Such deep grooves do not allow for hydrodynamic pressure
build-up by a mixture of restriction action and inward pumping as explained in
this subsection, but rather only by squeeze action.
22
In Figure 8, different types of self-acting bearings can be seen, of which the
logarithmic spiral groove bearing and herringbone spiral groove bearing are
sub-classes of the spiral groove bearing.
23
3.2 Meshing
24
3.2.1 Sandvik’s Geometry Mesh
In Figure 10 (a), the meshing of Sandvik’s piston-step is shown where the mesh
type is tetrahydral, mesh size is set to ’fine’ in COMSOL with 231514 domain
elements, 49966 boundary elements, and 8815 edge elements. It is expected
that setting ’fine’ mesh on COMSOL for this geometry would produce a much
larger number of meshes than for the parameterized geometry, which is needed
for greater detail in Sandvik’s geometry.
I have tried running both models with varying physics-controlled element sizes
in COMSOL ranging from ’fine’ to ’extremely fine’ and I found that the results
in the load sharing graphs converge to certain values as the element size is
smaller but without significant error when using larger element sizes in the
case of Sandvik’s geometry, therefore the physics-controlled mesh size ’fine’ was
chosen for Sandvik’s geometry to significantly reduce the computation time.
25
(a) Meshing on Sandvik’s geome- (b) Meshing on the parameterized
try geometry.
Since these equations are considerably complex, they have a very small number
of analytical solutions. It is relatively easy to solve these equations analytically
for laminar flows between two parallel plates or for flows in circular pipes but for
more complex geometries the equations ought to be approximated numerically
using finite element, finite volume, finite difference as well as spectral methods.
The Navier-Stokes set of equations are shown below in equation (2) and are
solved together with equation (1).
26
Continuity equation:
∂ρ ∂ρu ∂ρv ∂ρw
+ + + =0 (1)
∂t ∂x ∂y ∂z
X-Momentum:
∂ρu ∂ρu2
∂ρuv ∂ρuw ∂p 1 ∂τxx ∂τxy ∂τxz
+ + + =− + + + (2a)
∂t ∂x ∂y ∂z ∂x Rer ∂x ∂y ∂z
Y-Momentum:
Z-Momentum:
The Navier-Stokes equations and the continuity equation are also often solved
with the Energy equation if the fluid flow is non-isothermal. However, since an
isothermal solution is assumed in this thesis, the Energy equation is not written
here.
It is important to note that solving the Navier-Stokes equation for the case of
Sandvik’s bearing or any other flow problem would give more accurate and com-
prehensive numerical simulations than solving the Reynold’s equation because
then any turbulent flow can be modelled and also in the case of the Rayleigh
step-bearing, the flow in the gaps between successive pads can be modelled as
well, however it is very computationally expensive to do so and in most cases
parallel-computing on super computers is required which is not available to me,
and is also unnecessary given that the Reynold’s equation will give sufficient
and accurate simulations as well. The homogenized form of Reynold’s equation
used in this thesis to model mixed lubrication conditions in Sandvik’s bearing
is discussed in Subsection 3.5.
27
3.4 1D Reynold’s Equation and its Analytical Solution
Since the Rayleigh step-bearing in this case is infinitely wide, the pressure dis-
tribution is assumed to be constant in the direction of width since there will
be virtually no side-leakage, and thus the pressure gradient in this direction is
negligible. Equation (2) is the Reynold’s equation applicable to the infinitely
wide Rayleigh step-bearing in which the pressure gradient and velocity are neg-
ligible in the y and z directions and are accounted for only in the direction of
the length of the bearing (x-direction).
h3 dp
d u dh
= (3)
dx 12η dx 2 dx
dp h − hm
= 6ηu (4)
dx h3
where hm is the height at the maximum pressure point and h is the film thickness
at any given point along the bearing length. This equation should be integrated
to produce the pressure distribution.
28
3.5 Homogenized Reynold’s Equation
1. Newtonian fluid
2. No boundary slip
3. Incompressible flow
ρh3 ∂p ρh3 ∂p
∂ ∂ ω ∂ρh ∂ρh ∂(ρh)
+ = + + (5)
∂x 12η ∂x ∂z 12η ∂z 2 ∂x ∂z ∂t
where ω is the vector of the eccentric motion in the x and z directions. Finally,
∂(ρh)
∂t is the squeeze term. The left hand side of the equation pertains to the
Poiseuille flow and the right hand side to the Couette flow and the squeeze term.
29
The time-dependent 2D homogenized form of Reynold’s equation with assump-
tion of symmetric roughness is:
∂ ρa11 (h) ∂p ∂ ρa22 (h) ∂p ω ∂(ρb1 (h)) ∂ρb2 (h) ∂(ρh)
+ = + + (6)
∂x 12η ∂x ∂z 12η ∂z 2 ∂x ∂z ∂t
In Figure 11, the flow factors a11 can be seen interpolated against the separation
h, and as mentioned before they are obtained in a parameterized fashion from
a local scale roughness taken from a representative topology as in [31].
30
3.6 Applied Force
The thrust force which Sandvik’s bearing set is subject to is highly fluctuating
in a stochastic manner, and so a two-minute signal of the hydroset pressure
from the cone crusher was captured. When multiplying this pressure with the
hydroset piston area, the thrust force that the bearing set is subject to is ob-
tained. This force however, with its high fluctuations, will make it difficult for
COMSOL to capture it and will cause convergence error. To solve this problem,
a simple force function was selected, adapted and periodised from the force data
as shown in Figure 12.
31
The periodised force function is shown in Figure 13 for 1 [s], where the force
function is initialized by starting from a lower force to aid in convergence.
32
3.7 Force Balance
Since Sandvik’s bearing operates under mixed lubrication conditions, part of the
applied force is carried by the hydrodynamic pressure generated in the lubricant
and the rest is carried by the pressure caused due to asperity contact. In order
to calculate this force partitioning (load sharing), a force-balance equation is
posed as a Global ODE physics in COMSOL and solved. The equation is as
follows:
Z
Fhyd = 0.9 phyd (8)
Ω
and
Z
Fasp = 0.9 pasp (9)
Ω
The hydrodynamic and asperity pressures are integrated over the surface area
of the piston step and then multiplied by 0.9 to obtain the hydrodynamic and
asperity forces. Multiplying by 0.9 is for the reason that during eccentric mo-
tion of the step washer relative to the piston step, between only 83 % to 92
% of the piston step area is in contact due to the eccentric shift between the
step washer and the piston step and depending on the magnitude of the crusher
stroke. The percentage of area in contact area for the maximum and mini-
mum crusher stroke are shown in Table 1 below. The piston step area and area
in contact are calculated using COMSOL’s ’measure’ feature which is able to
measure the surface area of boundaries and objects. The multiplication factor
0.9 (pertaining to 90 %) is thus chosen as an intermediate value within the range.
33
In Figure 14, the area in contact between the piston step of the straight groove
bearing and the step washer can be seen shaded in purple while the area out of
contact is not shaded.
34
3.8 Squeeze Action
When looking at the homogenized form of Reynold’s equation (5), we can see
the squeeze term is ∂(ρh)
∂t . This term can be easily posed in COMSOL’s Co-
efficient Form Boundary PDE physics which is used to solve the equation and
leads to COMSOL being able to capture this squeeze effect in time-dependent
simulations.
35
3.9 Solid Mechanics
1. The step washer is assumed to be rigid and the elastic properties of the
piston step are taken as the equivalent Young’s modulus (E ∗ ) and equiv-
alent Poisson ratio (ν ∗ ) of the combined material of the step washer and
piston step, where:
1 1 − ν12 1 − ν22
= + (10)
E∗ E1 E2
and
σ = E.ε (11)
3. To finally achieve the coupling, the deformed separation between the step
washer and piston step at the interface is:
h = h0 + v (12)
36
3.10 Analytical Solution of Pocket Bearing in 1D
One of the analytical solutions that would be largely useful and relevant to the
case of spiral groove bearings is the pocket bearing which can be seen as a much
more simplistic form of a spiral groove bearing. Why this solution will be helpful
is because each groove can be seen as a pocket if sectioned, as shown in Figure 16.
The solution is then shown in Figure 17. It is important to note that a cavita-
tion model has been included in order to omit the negative pressures caused by
cavitation in the area a to z shown in Figure 16.
Figure 17: Analytical and FDM solutions of the pocket bearing in 1D [34].
37
It can be noted that the pressure drops from 0 to a, then is constant at zero
(where cavitation should be happening) between a and z, then the pressure rises
sharply after z where it reaches maximum at b (at the step), then drops again
after the step in a non-linear manner. In a spiral groove bearing, there are
typically more than one of these pockets (grooves) in a given radial direction so
the pressure distribution should look like that in Figure 17 but repeated with
the same number of times as there is grooves while the magnitude of maximum
pressure drops with every groove starting from the inner edge till reaching the
outer edge of the bearing. However, this is for the case when grooves are shallow
enough to allow for hydrodynamic pressure build-up due to wedge/restriction
action.
38
3.11 Cavitation Model
1, p - pa > 0
f = 0, p - pa <-β (13)
p−pa 3 p−pa 2
1 − 2( β ) − 3( β ) , -β ≤ p − pa ≤ 0
where
f +α
ρ = ρ0 (14)
1+a
f +α
η = η0 (15)
1+a
where f is the degree of saturation of the fluid and it varies with the fluid
pressure as seen in equation (12), β is the threshold pressure and α is a threshold
constant that takes a value approaching zero. Then in the areas where pressure
is less than −β the value f is set to zero and in the area where pressure is between
−β and zero this is the transient and so on. The threshold pressure β is set to
9000 [Pa] and α is set to 0.0001.
39
3.12 Validation
1. Modelling ladder, part 1: In this part, the grooves are shallow enough
to allow for hydrodynamic pressure build-up due to restriction action and
is shown in Figure 18. Pure rotational motion of the counter-surface is
also assumed rather than eccentric motion.
40
2. Modelling ladder, part 2: In this part, the grooves are excessively deep
as in Sandvik’s piston step geometry and is shown in Figure 19. The ec-
centric motion of the counter-surface (ie. step-washer) is also considered.
41
4 Results
In this section, the results from the simulations of the Rayleigh step-bearing
and the spiral groove bearing according to their respective modelling ladders
are shown and discussed.
In this subsection the solutions following the validation modelling ladder shown
in Figure 18 are laid out.
The analytical solution was computed on MATLAB using the script found in
the Appendix, assuming an infinitely wide 2D Rayleigh step-bearing. The pres-
sure distribution (showing the maximum pressure as the analytical solution) is
shown in Figure 20.
42
4.1.2 Numerical - Rectangular Rayleigh Step-bearing Pad, Different
Width:Length Ratios
It is clear here that the closer the width:length ratio is from being infinitely
wide, the closer the maximum pressure value becomes of that of the analytical
maximum pressure solution shown in Figure 20 and the closer the width becomes
to the length, the smaller this maximum pressure value becomes due to side
leakage. This proves that the analytical solution does not suffice when we add
another dimension (width) and that the numerical solution of some form of
the 2D Reynold’s equation is required for more accurate approximations of the
pressure distribution. It is also clear that the pressure maximum is at the step,
as seen in the analytical solution.
43
4.1.3 Numerical - Cylindrical (Almost-Quadratic) Pad
This pad is built in polar coordinates, using the simple assumption that the
further away the pad is from the center of the bearing and the smaller the angle
is between the left and right sides of the pad, the more rectangular it will look
like..The pressure distribution schematic is shown in Figure 22 and the maxi-
mum pressure value is shown in Table 3.
Parameter Value
Max. pressure 2.78 [M P a]
It is notable that the maximum pressure value here pertains to the maximum
pressure value of the rectangular Rayleigh step-bearing pad (1:1 width:length
ratio) since the width:length ratio in this cylindrical (almost-quadratic) pad is
almost 1:1.
44
4.1.4 Numerical - Cylindrical Pad, Different Circumferential Widths
Next, the cylindrical Rayleigh step-bearing pad was modelled while varying the
circumferential width and keeping the surface area constant at 0.002722 m2 .
The schematics of the pressure distributions are shown in Figure 23 and the
maximum pressure values are shown in Table 4.
Table 4: Maximum pressure and load carrying capacity values for the three
variants.
It is notable that the more circumferentially wide the bearing pad is, the more
side leakage will be present which will lead to a lower maximum pressure value
and a lower load carrying capacity but also for the typical width, the maximum
pressure value pertains approximately to the 1:1 width:length ratio rectangular
pad solution.
45
4.1.5 Numerical - Straight Groove Bearing (Shallow Grooves)
Figure 24: Load sharing in the rigid straight groove bearing (shallow grooves) -
load in [N].
(a) Hydrodynamic pressure at 0.14 [s] (b) Asperity pressure at 0.14 [s] - pres-
- pressure in [Pa]. sure in [Pa].
Figure 25: Rigid straight groove bearing (shallow grooves) pressure schematics.
46
If one section (groove + ridge) is taken out of this straight groove bearing sur-
face, the pressure distribution can be seen to resemble that of the cylindrical
Rayleigh step-bearing pad shown in Figure 23 in that the maximum pressure
occurs on the step (groove-ridge boundary) due to the restriction/wedge action
but it is most similar to the pressure distribution expected from a pocket bear-
ing as shown in Figure 17 where the pressure is zero in the grooves but spikes
up rapidly to reach a maximum at the groove-ridge boundary then falls down
non-linearly over the ridge, and thus the pressure distribution obtained in this
simulation can be deemed as sensible.
In this subsection the solutions following the validation modelling ladder shown
in Figure 19 are laid out. In this modelling ladder, the grooves are deep just
like in Sandvik’s piston step. Due to the depth of the grooves, a hydrodynamic
pressure build-up due to the restriction action will not be present, but rather
only due to a squeeze action over the ridges as explained previously. For each
variant, the load sharing graph will be shown for 0.2 [s] of the loading cycle
and also a graphic representation of the hydrodynamic pressure build-up due to
squeeze action as well as a graphic representation of the asperity pressure.
47
4.2.1 Straight Groove Bearing - Rigid
The first variant in this modelling ladder is the most simple one, with parame-
terized straight grooves and a rigid surface (ie. no coupling of Solid Mechanics
physics). The load sharing graph can be seen in Figure 26 and the graphic
representation of the hydrodynamic and asperity pressures taken at 0.14 [s] are
shown in Figure 27.
Figure 26: Load sharing in the rigid straight groove bearing - load in [N].
(a) Hydrodynamic pressure at 0.14 [s] (b) Asperity pressure at 0.14 [s] - pres-
- pressure in [Pa]. sure in [Pa]
48
4.2.2 Spiral Groove Bearing - Rigid
The second variant in this modelling ladder is the parameterized spiral groove
bearing with a rigid surface (ie. no coupling of Solid Mechanics physics). The
load sharing graph can be seen in Figure 28 and the graphic representation of
the hydrodynamic and asperity pressures taken at 0.14 [s] are shown in Figure
29.
Figure 28: Load sharing in the rigid spiral groove bearing - load in [N].
(a) Hydrodynamic pressure at 0.14 [s] - (b) Asperity pressure at 0.14 [s] - pres-
pressure in [Pa]. sure in [Pa].
49
4.2.3 Sandvik’s Piston-step - Rigid
The third variant in this modelling ladder is Sandvik’s actual piston-step geom-
etry with a rigid surface (ie. no coupling of Solid Mechanics physics). The load
sharing graph can be seen in Figure 30 and the graphic representation of the
hydrodynamic and asperity pressures taken at 0.14 [s] are shown in Figure 31.
(a) Hydrodynamic pressure at 0.14 [s] - pres- (b) Asperity pressure at 0.14 [s] - pres-
sure in [Pa]. sure in [Pa].
50
4.2.4 Spiral Groove Bearing - Deformed
The fourth variant in this modelling ladder is the parameterized spiral groove
bearing with a deformed surface (ie. Solid Mechanics physics coupled). The
load sharing graph can be seen in Figure 32 and the graphic representation of
the hydrodynamic and asperity pressures taken at 0.14 [s] are shown in Figure
33.
Figure 32: Load sharing in deformed spiral groove bearing - load in [N].
(a) Hydrodynamic pressure at 0.14 [s] - (b) Asperity pressure at 0.14 [s] - pressure
pressure in [Pa]. in [Pa].
51
4.2.5 Sandvik’s Piston-step - Deformed
The fifth variant in this modelling ladder is Sandvik’s actual piston-step geom-
etry with a deformed surface (ie. Solid Mechanics physics coupled). The load
sharing graph can be seen in Figure 34 and the graphic representation of the
hydrodynamic and asperity pressure taken at 0.14 [s] are shown in Figure 35.
(a) Hydrodynamic pressure at 0.14 [s] - (b) Asperity pressure at 0.14 [s] - pres-
pressure in [Pa]. sure in [Pa].
52
4.2.6 Validating the Implementation of Solid Mechanics Physics
Figure 36: Load sharing for different Young’s Modulus of bearing material.
53
5 Possible Design Changes
In this section, the insight gathered from the simulation results in the previous
section is used to predict and test possible advantageous design changes which
can mitigate undesirable asperity contact in Sandvik’s bearing-set during its
operation.
If we were to compare, for example, the results for the Straight Groove Bearing
in 4.1.5 where the grooves are shallow and in 4.2.1 where the grooves are exces-
sively deep, it would be immediately clear that shallow grooves (in order of a
few multiples the oil film thickness over the ridges) allow for full-film generation
due to the restriction action discussed previously and this leads to complete
mitigation of undesirable asperity contact. Such full-film generation is not pos-
sible when the grooves are excessively deep as in the current design Sandvik’s
bearing-set. The reason why the grooves were chosen to be excessively deep in
the first place is to allow for a larger volume of cold oil to flow in the grooves and
cool down the heat generated due to asperity friction over the ridge interface,
and thus it may be wise to think of ways to enhance the squeeze action over
the ridges while keeping the depth of the grooves. However, if shallow grooves
were used, will the full-film generation of the lubricant reduce heating of the
interfaces to an extent that a large volume of cold oil flow in the grooves will
not be needed any more? This is an important question to investigate in fu-
ture work where the thermal effects will be included in the model. Also, it is
important to note that in cone crushers, the operating conditions in terms of
load and motion are extreme and may make impossible for a full-film of lubri-
cant to be generated, in which case shallow grooves will be a great disadvantage.
If the lubricating oil is pumped into the interface through the center of the
piston-step using an adequate amount hydrostatic pressure, this can separate
the piston-step and step-washer surfaces from coming into contact. However, if
we rely only on hydrostatic pressure and it ceases to exist momentarily due to
a technical fault, the piston-step and step-washer surfaces will crash in a way
that can destroy the interface.
54
5.3 Larger Surface Area of Ridges
If we were to choose to only enhance the squeeze action over the ridges, then
increasing the surface area of the ridges on which squeeze action occurs while
decreasing the surface area of the grooves can be a good choice. However, the
total volume of oil that can flow in the grooves has to be kept constant in order
to not reduce the cooling capability in the bearing. The increase in surface area
of the ridges can be done by editing geometric parameters in the model in the
following manner:
And therefore, these two possibilities were tested and the results are shown in
5.3.1 and 5.3.2.
Herein, 5 different variants of the parameterized spiral groove bearing are mod-
elled, with different groove depths and groove:ridge surface area ratios. The
geometric parameters of the different variants are shown in Table 5 where G:R
is the groove:ridge surface area ratio, GD is the groove depth. The current (de-
fault) design parameters are as in variant 2. The hydrodynamic and asperity
load shares for all variants are shown in Figure 37. The results were truncated
at 0.18 [s] because some of the simulations faced trouble in convergence beyond
this point in time.
55
(a) Hydrodynamic load share.
Figure 37: Load sharing for different groove:ridge surface area ratios.
56
5.3.2 Decreasing number of grooves
Herein, 5 different variants of the parameterized spiral groove bearing are mod-
elled, with different groove depths and number of grooves. The geometric pa-
rameters of the different variants are shown in Table 6 where GN is the number
of grooves and GD is the groove depth and are set such that the total groove
volume is kept constant.
57
(a) Hydrodynamic load share.
’
(b) Asperity load share.
58
6 Discussion
In this section, the results shown in the previous section as well as the insight
gained about the behaviour of Sandvik’s bearing will be discussed.
6.1 Validation
If the modelling ladder is followed step by step, it would be clear that the sim-
ulation results are logical such that every simulation result behaves as expected
in light of the simulations that comes before it in the ladder. However, the log-
ical links between successive simulations may not be clear by simply looking at
the results and so they will be discussed in more detail in Subsections 6.2 and 6.3.
Comparing the load sharing graph for Sandvik’s bearing (deformed) in Figure
34 and that of the parameterized spiral groove bearing in Figure 30, it is notable
that the squeeze action in Sandvik’s bearing is more prominent and thus asperity
contact is less. This is simply due to a larger area of ridges in Sandvik’s bearing.
A hypothesis for why the the deformed bearings seem to perform worse (more
asperity contact), is that as the ridges deform, they face maximum deformation
at the center of the ridge and the step washer (opposite surface) moves towards
the ridge with an amount that is proportional to the deformation. It was no-
ticed that as the step washer moves towards the deformed ridges, its surface gets
really close to the sides of the ridge, and so higher asperity pressure is noticed
on the sides of each ridge. This can be clear when looking at Figure 33 (b) and
Figure 35 (b).
59
6.4 Design Changes
The design changes proposed in Section 5 may well and truly be useful in miti-
gating asperity friction on the interfaces in Sandvik’s bearing, however a number
of factors are not considered in the simulations and should be considered before
applying such changes. These factors include:
2. The distance between grooves in the current design of the piston-step was
tailored to allow the whole surface of the piston-step to be exposed to the
lubricating oil in 1 gyratory cycle of the opposing surface (step-washer),
and so the effect of changing this distance has to be considered and prefer-
ably tested experimentally.
4. For each proposed design change, the effects of temperature rise of the
surfaces in the interface on the performance of the bearing has to be
considered. This is important to be able to determine whether excessively
deep grooves are really needed, as their function is to allow for a sufficient
volume of lubricant to be pumped through them and therefore cool the
interface.
60
7 Conclusions
When looking at the load sharing graphs obtained from both models in Section
4, which describe how the applied thrust load is partitioned between hydro-
dynamic and asperity loads in the lubricated interfaces of the bearing, it can
be concluded that Sandvik’s bearing runs under severe mixed-lubrication con-
ditions. This is believed to be, as explained previously, due to the excessive
depth of the grooves which make it impossible for a full-film of lubricant to be
generated due to a restriction action and therefore, the only method by which
a lubricant film is generated in the interfaces of the bearing-set is the squeeze
action. The pressure sustained in the lubricant film due to squeeze action is not
sufficient to fully support the thrust load that the bearing-set is subject to and
so asperities come into contact.
In Section 5, several possible design changes which can either enhance the
squeeze action or lead to full-film lubrication have been discussed and tested.
In Subsection 6.4, the factors that has to be considered before applying these
design changes are mentioned. Overall, the objectives of the thesis have been
met, although improvements have to be made in order to have models that
produce very realistic and reliable results. These improvements to be made are
discussed in the following section.
61
8 Future Work
In future work, the models for Sandvik’s bearing as well as the parameterized
S.G.B are to be developed such that they include thermal effects, and are able
to run simulations for many loading cycles rather than just one. The simulation
results may also be validated experimentally if time and resources allow. In
this manner, the models will be more realistic and will take into consideration
almost all physical phenomena that Sandvik’s bearing is subject to.
62
9 References
[4] Vakis AI, Yastrebov VA, Scheibert J, Nicola L, Dini D, Minfray C et al.
Modeling and Simulation in Tribology Across Scales: an Overview. Tribology
International. 2018 Sep;125:169-199.
[5] Hertz H. Über die Berührung fester elastischer Körper. Journal für die reine
und angewandte Mathematik (Crelle’s Journal). 1882;1882(92).
[6] Johnson K, Kendall K, Roberts A. Surface Energy and the Contact of Elastic
Solids. Proceedings of the Royal Society A: Mathematical, Physical and Engi-
neering Sciences. 1971;324(1558):301-313.
[8] Ciavarella M. The generalized Cattaneo partial slip plane contact problem.
I—Theory. International Journal of Solids and Structures. 1998;35(18):2349-
2362.
[10] Ballard P. Steady sliding frictional contact problem for a 2d elastic half-
space with a discontinuous friction coefficient and related stress singularities.
Journal of the Mechanics and Physics of Solids. 2016;97:225-259.
[11] Greenwood JA, Williamson JBP. Contact of nominally flat surfaces. Proc
R Soc A 1966;295:300–19.
63
[13] Sjodin B. What’s The Difference Between FEM, FDM, and FVM? [Inter-
net]. Machine Design. 2016 [cited 12 January 2019]. Available from: https://
www.machinedesign.com/fea-and-simulation/what-s-difference-between-fem-fdm-
and-fvm
[15] Stanley HM, Kato T. An FFT-based method for rough surface contact. J
Tribol 1997;119:481-5.
[19] Turner MJ, Clough RW, Martin HE, Topp LJ. Stiffness and Deflection Anal-
ysis of Complex Structures. Journal of the Aeronautical Sciences. 1956;23(9):805-
823.
[22] R.T.P. Whipple, ”Herringbone Pattern Thrust Bearings”, AERE T/M 29,
1949.
64
[25] Borras X. Multiphysics Modelling of Spring-Supported Thrust Bearings for
Hydropower Applications. MSc Thesis - Luleå University of Technology [Inter-
net]. 2012;. Available from: http://ltu.diva-portal.org/smash/record.jsf?pid=div
a2(%)3A1032708dswid=6032
[28] Bhushan B. Modern tribology handbook. Boca Raton, FL: CRC Press;
2001.
[30] What Are the Navier-Stokes Equations? [Internet]. Comsol.se. 2012 [cited
25 December 2018]. Available from: https://www.comsol.se/multiphysics/navier-
stokes-equations
65
[36] Efficiently Mesh Your Model Geometry with Meshing Sequences [Inter-
net]. COMSOL Multiphysics. 2016 [cited 14 March 2019]. Available from:
https://www.comsol.com/blogs/efficiently-mesh-your-model-geometry-with-me
shing-sequences
[37] What are the basics of gas lubricated seals? (Part One) [Internet]. Pumps
Systems. 2014 [cited 5 April 2019]. Available from: https://www.pumpsandsyste
ms.com/topics/seals/what-are-basics-gas-lubricated-seals-part-one
66
Appendix
In order to enable Sandvik to use the models and easily test the effects of de-
sign changes, user-friendly applications have been created out of the models for
Sandvik’s bearing as well as for the parameterized S.G.B. The graphical inter-
face of the application created from the parameterized S.G.B model is shown in
Figure 39 as an example.
67