Bilinear Realization from I/O Data with NNs

We present a method that connects a well-established nonlinear (bilinear) identification method from data in the time domain with the advantages of neural networks (NNs). The main challenge for fitting bilinear systems is the accurate recovery of the corresponding Markov parameters from the input and output measurements. Afterward, a realization algorithm similar to that proposed by Isidori can be employed. The novel step is that NNs are used here as a surrogate data simulator to construct input-output (i/o) data sequences from a single experiment. Then, classical realization theory is used to build an interpretable bilinear model that can further optimize engineering processes through robust simulations and control design.