Development of a rubber V-belt finite element model : traineeship report

This paper describes how to make a finite element model of a scooter CVT. This finite element model will be made in ABAQUS. To check the results from the ABAQUS model an experimental CVT test setup is used. While making a model, different in fluences will be studied and simplified to make a simple but effective model for studying rubber belt CVTs. First some material test are done to achieve the material properties of the belt. The properties of the pulleys and boss are achieved from the manufacturer. From the measurements can be concluded, that the belts material is anisotropic and viscoelastic in the longitudinal direction. Because the belt will be simulated as isotropic and the material behavior in the longitudinal direction is most important, an isotropic viscoelastic model should be used. The reason for choosing isotropic instead of anisotropic is due to its complexity and lack of time. The other parts will be simulated as analytical rigid due to its 200 times larger Young's modulus. The simulation is split into 5 steps, it starts with a model where forces or stresses are not present, converted into a model where the pulleys are rotating and the belt is clamped. The elements used for the model are 20-node quadratic bricks with reduced integration. To see the difference between the effects, different models are made. The details that are researched in this paper are: - The geometry of the modeled belt - Material behavior (viscoelastic/elastic) - Misalignment of the pulley sets - Symmetric/Asymetric model To see the difference and compare the results of the experiments, 6 models are made. After studying the results of the experiments and the models, it became clear that the in fluence of the torque on the effciency has the same behavior in the simulation as in the experiments. Only the in uence of the speed on the effciency is not visible in the results. To see the effects of the different details which are investigated, the mean effciency is compared between the models. The highest in uence on the effciency is the belt shape. The in uence of the cogs on the effciency was 3.7% on average. The in fluence of viscoelastic and misalignment were respectively 0.89% and 0.33%, which are remarkably low. In the model with cogs, only simple square cogs are used to compare against a model without cogs. Probably when a exact copy of the belt shape was used even a larger effciency increase was noticed. The difference between a symmetric or asymmetric model in mean effciency is small, only 2%. When a model without misalignment, clearance and exact pulley positions is suffcient a symmetric model is preferred because of its lower computational costs.