System identification using FRA and a modified MLMVN with arbitrary complex-valued inputs

A new architecture for the identification of lumped models of distributed parameter systems is proposed in this paper. It is assumed to use, by means of repeated measurements or intensive simulations, the Frequency Response Analysis (FRA) of the system or device under exam. The procedure can be used to extract the component values. A key element of this architecture is a multilayer neural network with multi-valued neurons (MLMVN), which is modified in the paper in order to use arbitrary complex-valued inputs. It is shown that this modification requires just a slight change in the MLMVN learning algorithm.