| A nonlinear identification technique is presented which obtains accurate models for nonlinear systems with little or no a priori knowledge of the type of nonlinearity involved. Given discrete time-domain measurements and an assumed model, which contains what is known of the system, the Minimum Model Error estimation algorithm (MME) is used to obtain estimates of the dynamic model error and the state trajectory. A correlation technique is employed to select the best functional basis for the dynamic correction from an extensive function library. A least-squares fit is performed to find the respective coefficients of the model error function(s).;Examples are presented using data from digital and analog simulations, as well as from laboratory experiments. Trigonometric, algebraic, exponential, hysteretic and time lag nonlinearities, and mixes thereof, are identified. The tests demonstrate that the method is robust with respect to prior ignorance of the nonlinear system model, robust for short measurement record length, and robust regardless of initial conditions. It works well in the presence of noise and is demonstrated for systems up to eighth order. |
