Dynamics of rotors supported on fluid-film bearings using neural networks

Abstract

This paper deals with the determination of shaft centre response and oil-film forces of a rotating rotor-bearing assembly. The successive points in a long time history of the rotor-centre motion during a transient vibration period have been identified. Fluid-film forces are generally influenced by several design variables. The calculation of these forces is not straightforward because these equations of motion of the system contain non-linear terms. Initially the most influential parameters are identified. A supervised multi-layer neural network model is then trained with the input and output data using the back-propagation algorithm. The response characteristics and fluid-film forces are derived as the outputs of the neural network for different conditions of bearing parameters. The results are compared with the usual solution techniques.