Title: Neuro-leaf spring
Authors: A. Ghazi Zadeh, A. Fahim, M. EI-Gindy
Addresses: University of Ottawa, Department of Mechanical Engineering, 770 King Edward, Ottawa K1N 6N5, Canada. ' University of Ottawa, Department of Mechanical Engineering, 770 King Edward, Ottawa K1N 6N5, Canada. ' Pennsylvania Transportation Institute, The Pennsylvania State University, 201 Research Office Building, University Park, PA 16802, USA
Abstract: A recurrent neural network is taught to emulate a leaf spring that is typically employed in the suspension system of trucks. Leaf springs are known to have nonlinear and hysteresis behaviour. This makes their mathematical formulation difficult and susceptible to a considerable amount of estimation errors. Analysis of the vehicle|s dynamic behaviour is heavily reliant on the accurate determination of the suspension forces. It is shown that the recurrent neural network is able to emulate the leaf spring behaviour very accurately after it is taught with a set of input output data points. In order to generate the teaching data points an analytical model of the leaf spring is used. The performance of the developed neural network emulator is also evaluated in the time and frequency domains and compared to those of the analytical model.
Keywords: leaf springs; neural networks; trucks; suspension system; vehicle modelling; vehicle dynamics; suspension forces; leaf spring model.
International Journal of Heavy Vehicle Systems, 2000 Vol.7 No.4, pp.317 - 335
Published online: 16 Sep 2004 *Full-text access for editors Access for subscribers Purchase this article Comment on this article