An identification method for continuous-time transfer functions based on nonlinear optimization

In this paper, we propose an identification method for continuous-time transfer function models, where sampled input-output data is directly used. To obtain the ARX model of a system, the derivatives of input-output signals are needed, and are given as output of some filters. The identification method is argued under the assumption that the measurement noise is independent of the input-output signals and that covariance of the noise is known. This identification method differs from the traditional methods based on the least-square technique, because the measurement noise is taken into consideration explicitly and the identification problem can be formulated as an optimization with a nonlinear constraint. The solution for this optimization is presented. The effectiveness of the method is verified through numerical simulations.