Consistency aspects of Wiener-Hammerstein model identification in presence of process noise

Paper in proceeding, 2016

The Wiener-Hammerstein model is a block-oriented model consisting of two linear blocks and a static nonlinearity in the middle. Several identification approaches for this model structure rely on the fact that the best linear approximation of the system is a consistent estimate of the two linear parts, under the hypothesis of Gaussian excitation. But, these approaches do not consider the presence of other disturbance sources than measurement noise. In this paper we consider the presence of a disturbance entering before the nonlinearity (process noise) and we show that, also in this case, the best linear approximation is a consistent estimate of underlying linear dynamics. Furthermore, we analyse the impact of the process noise on the nonlinearity estimation, showing that a standard prediction error method approach can lead to biased results.