Adaptive identification algorithm of nonlinear output frequency response function of multivariable system

Because of the large calculated amount and the existence of matrix inversion, it can not meet the online identification demand that identifying nonlinear out frequency response function (NOFRF) based on least square batch algorithm. In order to implement online identification of NOFRF of multivariate nonlinear system, a frequency domain variable step size normalized LMS adaptive identification algorithm is proposed on the combination of variable step size and normalization conceptions. The step size is updated according to the real-time error so that the convergence rate and estimation error are both meet the demands. Simulation experiment result indicates that the proposed algorithm has fast convergence rate, small estimation error and good robustness. Compared to traditional batch algorithm, the proposed algorithm has better timeliness that can meet the online identification demand.