D. Vetturi and A. Magalini (Italy)
modeling, dynamic modeling, simulation, numericalmethods, neural networks, computer aided design
This contribution deals with some numerical methods implemented in a code of automatic calculus to simulate the dynamic (and also the static) behavior of rubber-to metal devices widespread in the automotive industry. The elastomeric components of these are modeled by concentrated parameters particle based models. These are calibrated using data pointed out by experimental tests performed on the used materials: viscoelastic properties are assigned by a numerical method based on a sort of back propagation neural network which works on sets of experimental data to estimate runtime (dynamically during simulations) proper values of (complex) Young's and shear modules depending on the operative conditions in terms of strain amplitude, frequency, temperature and static pre-strain (constituting a multidimensional domain on which the neural network computes). The application of these models on real devices takes to the individuation of concentrated parameters mechanical systems. Simulations are performed on them in the frequency dominion by the adoption of the Newton-Raphson method to solve the dynamic equations of the motion.
Important Links:
Go Back
