Approximation of Optimal Profile of Rolling Bearings Cylindrical Rollers
Approximation of Optimal Profile of Rolling Bearings Cylindrical Rollers
E-mail: [email protected]
Abstract. In this paper a method to find the best approximation of the optimal profile of the rollers of a cylindrical roller
bearing is presented. The purpose is defined as an optimization problem and solved by means of Evolutionary Algorithms.
To validate the optimization, the obtained profile was compared with the optimal one taking into account four significant
criteria. The results were excellent encouraging future research.
Keywords: Cylindrical roller bearings, Roller optimal profile approximation, Evolutionary Algorithms.
99 | P a g e
5th International Conference on Power Transmissions BAPT2016
100 | P a g e
5th International Conference on Power Transmissions BAPT2016
finally, by the chamfer profile. The abscissae of the and to associate to every j a corresponding distance dj
f,g
points S, T, and U will be denoted by s, t, and u, between the graphs of the functions f and g. In this
respectively. way an objective function can be easily constructed:
s < t < u < Lw / 2 (where Lw is
the total length of the roller) and R1 > R2 must be
satisfied. Note that if the circle of radius R1 does not
exist, the obtained profile is the old ZB cylindrical The simplest idea (and probably the most used for
roller profile. such approximation) is to consider:
The 2ZB profile is completely and uniquely defined
by four parameters: R1, R2, , and q. Obviously, R1, R2,
are strictly positive and q
that even when the above mentioned constrained are This kind of approach has a flaw for this particular
satisfied there is an infinite number of combinations case in which the function f has a very special shape,
each representing a 2ZB profile. especially in its last part, toward the roller side. In this
A certain 2ZB profile, described by certain four part of the curve the curvature radius decreases rapidly
parameters, R1, R2, , and q, has the following with the increasing of y conferring a special
equation: rounding to the roller. We tried to capture this
aspect by defining a special distance between the two
curves, measured along the normal nj to the graph of
the function f. Obviously at the very last part of the
curves (situation marked with * on Fig. 2) the normal
to the graph of the function f does not intersect the
graph of function g in a point whose abscissa is within
the interval [c, u]. In this case the distance will be
where t, s, u are easily calculated with R1, R2, , and q. measured vertically.
When q = 0 the circle of radius R1 is tangent to the The objective function can be now computed and it
roller generatrix (fully crowned roller). Note that was implemented in a standalone optimization
between these parameters and the parameters needed software based on a modified version of the CMGEA
for the profile machining there are very simple algorithm where a new mechanism for species was
mathematical relationships. introduced and the simulated binary crossover (SBX)
[1] and the simplex crossover (SPX) [14] were added
Optimal approximation of the optimal profile
to the pool of operators. The modified CMGEA
According to the already mentioned approach, the algorithm uses multiple search operators in each
equation of the optimal profile is: generation, an idea similar to MO-GA and SAMO-GA
[2]. The algorithm also employs a teaching-learning
where Kf and a are constants. scheme inspired by TLBO [11] [13] which simulates
two fundamental modes of learning: through a master
and interacting with other peers.
Evaluation of the performances of roller profiles
In order to evaluate the performances of different
roller profiles two possible profiles of a cylindrical
roller were taken into account:
Optimal profile (O) (considered as referential),
Optimal approximation profile (OA).
Also, as evaluation criteria, four important
parameters were taken in consideration for a bearing
equipped with rollers having a certain profile and
loaded with six different levels of the radial load:
Fig. 2 Optimal (f) and 2ZB (g) profiles Basic rating life of the bearing (according to
ISO 16281: 2008),
The goal of the optimization is to find a function g Modified rating life of the bearing (according to
whose graph is as close as possible to the graph of the ISO 16281: 2008),
optimal profile f (Fig. 2). The approximation is Maximum contact pressure between the most
performed along the interval [c, u], where c is the loaded roller and the inner ring raceway,
greatest integer for which the condition f(c m is Maximum von Mises stress in the inner ring
still fulfilled. beneath the contact surface between the most
It is natural to divide the interval [c,u] into several loaded roller and the inner ring raceway and the
subintervals of equal width: depth at this occur.
Case study
101 | P a g e
5th International Conference on Power Transmissions BAPT2016
I 0 0
102 | P a g e
5th International Conference on Power Transmissions BAPT2016
TABLE VIII
Loading case I (RICw
103 | P a g e
5th International Conference on Power Transmissions BAPT2016
104 | P a g e
5th International Conference on Power Transmissions BAPT2016
TABLE IX
Loading case II (RICw
results
Fig. 9 Loading case II: Basic rating life, % (relative to
the optimal profile O)
105 | P a g e
5th International Conference on Power Transmissions BAPT2016
Fig. 11 Loading case II: Maximum contact pressure, Fig. 13 Loading case II: Maximum von Mises
% (relative to the optimal profile O) stress, % (relative to the optimal profile)
TABLE X
Loading case III (RICw
results
106 | P a g e
5th International Conference on Power Transmissions BAPT2016
107 | P a g e
5th International Conference on Power Transmissions BAPT2016
of the bearing with rollers having the optimal O the behavior of the optimal profile O (the curves
profile). of any considered parameter for OA and O
;
Even in presence of normally residual radial
internal clearance inside the bearing and
misalignment of the axes of the bearing bore
and housing, in the same loading conditions, the
expected life of a bearing with rollers having
OA profile will be with only 2% ... 4% less than
the expected life of a similar bearing but with
optimally profiled rollers;
In the case of the OA profile, even the above
mentioned clearance and the misalignment are
presented, up to a bearing radial load of about
Cr the maximum von Mises stress in the
ring and roller material does not exceed 1700
MPa, fact that makes the bearing made of steel
with martensitic structure suitable for a large
range of mechanical applications. If the steel
structure is bainitic the load cam increases up to
Cr.
In all above situations the maximum contact
pressure between the most loaded roller and the
inner ring raceway remains at actable values
Fig. 17 Loading case III: Maximum von Mises
bellow 2500 MPa;
stress
In the same conditions a bearing with fully-
crowned rollers reaches only 2/3 from the
The maximum von Mises stress in the case of the performances of a similar bearing having rollers
OA with OA profile.
similar loading condition for the O profile, differences
All these make from the optimal approximation
being more or less constant. If the bearing radial load
OA profile of the ideally profile a very competitive
is lower than 130 kN the maximum von Mises stress is
one at very low manufacturing costs.
kept bellow 1700 MPa in the case of the O and OA
profiles. Moreover, for both O and OA profile, when ACKNOWLEDGMENT
the bearing load reaches about 130 kN the maximum
The authors would like to thank the RKB Group,
von Mises stress is located at 0.07 mm deep below the
the Swiss bearing manufacturer, for the permission to
contact surface, whilst the lower maxima of the
publish these results and the RKB staff for their great
maximum von Mises stress becomes major about 45
interest and support during the development of this
earlier in the fully-crowned roller case.
project.
III. CONCLUSIONS
Examining the simulation results in all three
assumed loading cases some important remarks and
conclusions can be drawn:
The optimal profile O is indeed the best from all
points of view assumed in this research;
Whatever is the criterion, the optimal
approximation profile OA follows constantly
108 | P a g e
5th International Conference on Power Transmissions BAPT2016
109 | P a g e
5th International Conference on Power Transmissions BAPT2016
ANNEX
TABLE AI
110 | P a g e
5th International Conference on Power Transmissions BAPT2016
TABLE AII
111 | P a g e
5th International Conference on Power Transmissions BAPT2016
TABLE AIII
112 | P a g e
5th International Conference on Power Transmissions BAPT2016
113 | P a g e