In this paper, a two-stage method for identificationof the sandwich systems with hysteresis is proposed. In the firststage of this method, a monotonic and independentlyincremental signal is used to degenerate the hysteresis into astatic function 'with one-to-one mapping. Thus, the sandwichsystem with hysteresis can be transformed into a systemdescribed by the combination of the transformed variables andthe separated parameters based on the key term separationprinciple. Therefore, the parameters of the'linear subsystemsare estimated by the Recursive General IdentificationAlgorithm (RGIA) since the transformed model is of the formcontaining linear coefficients. Then, in the second stage, basedon the obtained linear subsystem models, the, sandwich systemwith hysteresis is re-excited by a sine signal with attenuatedamplitude. In this case, the hysteresis subsystem can beidentified using neural networks based on the expanded inputspace method. Finally, the corresponding experimental resultsare presented.
展开▼