In data-driven modelling, one estimates models for dynamical systems from data. In the general case, both input and output data are subject to disturbing noise. Moreover, when feedback is present for stabilizing or controlling the system, one cannot simply model the system from input and output data. The solution is to model the system from reference to input and from reference to output and afterwards applying a deconvolution for obtaining an input-output model. In this thesis, the deconvolution will be implemented for periodically time-varying systems. Periodically time-varying models are well suited to describe rotating machines (e.g. helicopters), and to analyze the linearized response of nonlinear systems around a periodic steady state solution. The algorithm will be applied to real-life measurements.
Prerequisites: good knowledge in system theory, signal processing, measurement, data-based modelling and Matlab.