Abstract
The design of an exhaust system is defined, among others, by vehicle dimensions and hanger positions. In preliminary design steps, the objectives are material selection and exhaust topology according to hangers positions. So the dynamic response of the model is function of these parameters. Due to comfort and exhaust durability reasons, sometimes, it may be necessary to modify some frequencies of vibration modes. Mass-damper method is analysed as solution to change normal modes of an exhaust system. The mass damper design involves several CAD and CAE analyses in order to accommodate parameter values to technical product specifications and computational time may quickly increase.
This work is focused in the development of an alternative methodology based on a neural network to analyse the exhaust normal modes, capable to run the study cases in less time than FEA. The position, mass and stiffness of the exhaust system correspond to the input for the neural network. Outputs are frequencies and eigenvectors of the new normal modes in the selected frequency range.
Keywords: Exhaust system, Neural network, Normal modes Subject groups: A5