Parallel Wiener-Hammerstein Identification: A Case Study

In the field of nonlinear system identification, a parallel Wiener-Hammerstein model offers a good balance between clarity and complexity. To identify such a model, a two-step method exists, where each step has been studied in detail. A full identification scheme combining the two steps while keeping track of noise influence is the next step forward. Therefore, we applied the exisiting techniques to a specific computer-simulated case study. This document presents the total identification scheme, from start to finish, and compares different models.Errors of the final identified model are small, in the order of 0.7% relative to the output signal, when using a signal-to-noise ratio of 1%. This means the found model is able to describe the measured data accurately. The next in this research is to apply the two-step weighted approach on real-life measured data.