Shock absorber dynamic behavior modelling taking into account a "memory - effect" in adams software

Abstract

To achieve valid results of vehicle dynamic simulation, it is essential to have an accurate vehicle model. In order to improve overall model accuracy, accurate description of different vehicle components is required. Results of testing of the shock absorber obtained on a test rig (damping force vs. relative velocity curve) show that system exhibits non-linearity and hysteresis. Moreover, curves of different shape are obtained when using different stimulation frequency. Such a behavior indicates that system possesses a "memory - effect", i.e. system output can be represented as a result of a current as well as of past inputs. Since no conventional mathematic tools are available for accurate description of non-linear dynamic systems, empirical modeling techniques have been developed, such as using of artificial neural networks, in order to obtain precise model of such systems for use in vehicle dynamic simulation. This paper introduces a way to implement use of artificial neural network to describe dynamic behavior of shock absorber in ADAMS software for dynamic simulation, taking into account a "memory - effect" present by such behavior. Due to lack of standard tools in ADAMS both for working with artificial neural networks and for use of the past results of the current analysis in current iteration step, custom user - defined subroutine had to be developed for making ADAMS - Solver to be able to solve the problem by using this approach. Finally, dynamic analysis of a quarter - vehicle has been carried out. Analysis results have been compared to those obtained by using conventional approach to shock absorber modeling.