75 Prediction of the Hysteresis and Creep Energy Loss with FEA within Filled Rubber Materials As Applied to Power Transmission Components

Wednesday, October 10, 2012: 2:00 PM
Room 203-204 (Duke Energy Center)
Sven Burghardt, Dipl.-Ing.1, James Brighton, EngD2 and Kim Blackburn, PhD2, (1)Centre of Automotive Technology, Cranfield University, Bedfordshire, United Kingdom, (2)Centre for Automotive Technology, Cranfield University, Bedford, United Kingdom
The prediction of energy dissipation from filled rubber components undergoing complex loading cycles is an important part of the design process if optimized components and minimized energy consumption are required. The feasibility of commercial FE codes for such a purpose is currently limited without the development of user defined subroutines which impose computational performance and development requirements putting them outside many practical applications.

A method for the practical application and fit of an appropriate built-in model is presented, based on an arbitrary filled elastomer, and a simplified physical testing approach.  This characterizes the equilibrium and cyclic deformation response of the material, primarily focussing on hysteresis losses.Several constitutive implementations are reviewed and discussed concerning their suitability and applicability. For the hyperelastic response, the strain energy density functions of Ogden, Arruda-Boyce, and Marlow have been employed. To add time dependence, the hysteresis model of Bergström and Boyce is discussed as well as the numerously applied Prony series expansion.

It has been concluded that a combination of the hyperelastic Marlow model combined with linear viscoelasticity can provide reasonable results within a common range of engineering applications, when appropriate fitting and post-processing is performed. Over a wide range of strain amplitudes, from 15 to 45%, a remaining error of less than 5% could be achieved, whereas for lower strain amplitudes the loss prediction will be less consistent due to nonlinear effects of the loss factor. Currently, a maximum error of 20% is detected, putting the method within a useful spectrum for an engineering problem.