Maarten Witters

Black-box identification of a continuously variable semi-active damper

This paper proposes a procedure to identify a nonlinear black-box simulation model of a continuously variable, semi-active damper for a passenger car. A neural network based output-error model structure is used to describe the complex, highly nonlinear dynamic behaviour of the electro-hydraulic damper. The used identification procedure automatically selects the regression variables by means of a state of the art technique for the model structure selection. Experimental validation of the proposed method shows that the obtained models are able to accurately simulate the dynamic damper behaviour.

Optimal experiment design for the black box identification of a semi-active damper

Abstract This paper discusses the design of an optimal excitation for the experimental identification of a continuously variable, semi-active damper of a passenger car. The applied excitations are multisine signals of which the phases are optimized to obtain a uniform coverage of the achievable working range. Based on the obtained experimental data, a neural network based output error model of the damper is identified using a state of the art iterative procedure that includes the automated model structure selection and the parameter estimation. It will be shown that models identified using the data from the optimized experiments are considerably more accurate than those identified using conventional random phase multisine excitations.