Modal contributions and effects of spurious poles in nonlinear subspace identification

Stabilisation diagrams have become a standard tool in the linear system identification, due to the capability of reducing the user interaction during the parameter extraction process. Their use in the presence of nonlinearity was recently introduced and it was demonstrated to be effective even in presence of non-smooth nonlinearities and high modal density. However, some variability of the identification results was reported, in particular concerning the quantification of the nonlinear effects, because of the presence of spurious modes, due to an over-estimation of the system order. In this paper the impact of spurious poles on the nonlinear subspace identification is investigated and some modal decoupling tools are introduced, which make it possible to identify modal contributions of physical poles on the nonlinear dynamics. An experimental identification is then conducted on a multi-degree-of-freedom system with a local nonlinearity and the significant improvements of the estimates obtained by the proposed approach are highlighted.