Identification of a Wiener–Hammerstein system using an incremental nonlinear optimisation technique

A method for black-box identification of a Wiener–Hammerstein system is described and applied to a set of Benchmark data originally presented at the 15th IFAC Symposium on System Identification. An incremental nonlinear optimisation procedure is used, which is able to avoid local minima, thus enabling the solution to converge to the global minimum. The use of a dual-polynomial to describe the static nonlinearity allows the number of parameters needed to be significantly reduced compared with the case if a single polynomial is utilised; this also improves robustness against extrapolation errors. The overall approach requires a relatively small number of parameters.