Combined identification of parameters and nonlinear characteristics based on input-output data
Christian Hintz, Martin Rau, Dierk F. Schroder · 2000
We present an identification method for systems consisting of a linear part with unknown parameters and an unknown nonlinearity (systems with an isolated nonlinearity). A structured recurrent neutral network is used to identify the unknown parameters of the known signal flow chart. The isolated nonlinearity is approximated by a feedforward neural network, which is part of the structured recurrent neural network. The novelty of this approach is the simultaneous identification of the parameters of the linear part and the nonlinearity. The structure of the recurrent network results from prior structural and parameter knowledge.