Data driven controller design via a frequency domain approach with application on structure control

This article introduces a frequency domain approach for designing a data-driven controller. Unlike the model-based control, it does not require a parametric model but only needs the input-output open loop test data for the purpose of spectral analysis. The resulting discrete frequency response is then utilized to give an optimal parameter design for a pre-defined structure controller, which is formulated as a non-convex optimization problem. The design approach is applied on a case study, an end-driving flexible beam with highly nonlinear dynamics. Simulations in finite element analysis (FEA) software show that the proposed controller can achieve similar performance as a model-based mixed sensitivity controller under the same design setup. In addition, it has the design flexibility of choosing the structure of the designated controller, which can largely decrease the controller order for implementation merits.