Investigation of the relation between grinding process parameters and created curvature and roughness of the pushbelt element saddle surface for optimizing the pushbelt element-ring contact in Continuously Variable Transmissions

Abstract

In a Continuously Variable Transmission
(CVT), the pushbelt is the torque transmitting part consisting of around 400
steel elements which are held together by steel ring packs. Relatively high
friction losses of the operating pushbelt are associated with the element
saddle surface on which the steel rings are lying. The pushbelt runs between
two pulleys which are variable in diameter and, therefore, slip between the
elements and the steel rings occurs, which leads to these friction losses.

In order to increase the efficiency and
durability of the pushbelt, the shape and the surface roughness of the element
saddle has to be adjusted such that the friction is reduced. The pushbelt
element saddle surface is difficult to reach since it is situated in a small slot
of the element, and in the production process, it is currently not specially
treated after fine blanking.

The requirements in order to achieve a
friction reduction in the pushbelt's element-ring contact were specified in
terms of average surface roughness R_a along the long side of the
element saddle, called the horizontal R_a, and the radius across the
short side of the element saddle, called the short saddle radius (SSR).

A grinding machine was specially built for
testing the feasibility of belt grinding for this purpose and to be able to
produce pushbelt elements with the required adjustments for pushbelt efficiency
tests. The grinding machine uses two grinding belts to process the left and
right saddle of the elements which are arranged in a row on a semi-circle in an
element holder.

In this research the relation between
grinding process parameters and created curvature and roughness of the pushbelt
element saddle surface was examined. Furthermore, it was investigated which are
feasible process conditions in order to achieve the required grinding result,
and how accurately the saddle surface characteristics can be predicted in
dependency of the machine parameters.

For this purpose, machine tests were
executed and the ground element saddles were measured with the aid of a white
light interferometer in order to evaluate the achieved response variables which
are the surface roughness R_a and the SSR. The measurements were
analyzed in order to describe the relationship between machine parameters and
obtained grinding result. Furthermore, the measurements were used to fit an
empirical model to the data, and the applicability of a polynomial and also an
exponential model was investigated.

In preliminary machine tests, it was found
that it is necessary to keep the elements in a certain distance in the element
holder in order to achieve a stable grinding result. Furthermore, after the
preliminary tests, the grinding belt grit size was specified and it was decided
to hold the grinding speed constant during the final experimental design. The
chosen predictor variables for the grinding result in the final experimental
design were the grinding time and the pressure which is used in the pneumatic
cylinder in the machine in order to bring the grinding belt to tension. A full
factorial experimental design with nine duplication experiments was used, so in
total 18 experiments, where the factors time and pressure were varied on three
levels.

It was concluded that the applied pressure,
so the tension force in the grinding belt, is the dominating factor for the
achieved size of the SSR. An increased pressure decreases the size of the
achieved SSR and also the variation of the response variables of the samples is
decreased at higher pressures. The achieved horizontal R_a on the
saddle is affected by both factors, time and pressure, however, the grinding
time appears to have more influence. The empirical exponential model and also
the polynomial model give a similar prediction for the grinding result within
the experimental factor range, but beyond that range, it was found that the
exponential model is able to make a better prediction. However, the predicted
machine parameters from the exponential model, which would be necessary in
order to achieve the desired grinding result according to the requirements, are
not feasible. The needed pressure would be too low for the grinding belt to
make sufficient contact with the saddle surface and the predicted grinding time
would be very high.

Within feasible machine conditions, the achieved
SSR is always lower than specified in the requirements and decreases rapidly
with increasing pressure. Additionally, the lowest roughness R_a
which can be achieved according to the exponential model is slightly higher
than the requirement. This is consistent with the performed experiments and it
was concluded that with the current machine configuration, the requirements
cannot be achieved.

It is recommended to execute further
experiments with an element holder having an increased diameter since with the
current holder, although the element's saddle was positioned at the desired
radius, the achieved radius across the saddle surface was smaller after
grinding. Furthermore, it is recommended to perform further experiments with
different kinds of grinding belts having finer grain sizes for achieving a
lower surface roughness or having structured abrasives for less variation and a
better predictability in the grinding result.