Recursive Identification of Dynamic Systems with Time-Varying Memoryless Nonlinearities having Deadbands

An algorithm has been developed for the recursive identification of nonlinear systems from input and output measurement data. The systems are modelled by a memoryless nonlinearity which includes deadband behavior, in cascade with a discrete-time deterministic autoregressive moving average (DARMA) model. Both the nonlinearity and the DARMA model parameters may be time varying. The basic idea is to represent the nonlinearity as a shifted polynomial, where the shift (along the input axis) models the deadband. The parameters of the systems dynamics and the coefficients of the polynomial nonlinearity are adjusted to minimize a mean square error criterion. Digital computer simulations are included to demonstrate the feasibility of the technique.