Data-Driven Control-Oriented Modeling for Response of Fluidic Thrust Vectoring

Response to control input is of significance to the application of real-time active flow control (AFC). In this paper, a novel data-driven framework is used to discover the underlying physics of the dynamic response process of fluidic thrust vectoring (FTV), a typical application of AFC. In the proposed framework, sparse identification of a nonlinear dynamics (SINDy) algorithm is used to identify the governing equations of the flow control responses of sets of noisy measurement data. The clustering algorithm is then used to seek the generalized coefficients of basis functions for different sets of data, which improve the robustness of the model to noisy measurement data. First, a simulated mechanical system is used to validate the effect of the framework. To simplify the modeling, control performance and characteristics are investigated in a detailed manner. Then a dimensionless parameter [Formula: see text] based on the pressure coefficient is found to exhibit a linear relationship with the vector angle under different working conditions. This parameter is introduced in the proposed framework to model the dynamic process of response to control input. The obtained governing equations can describe the dynamic process accurately based on the validation of testing data. The form of the governing equation is rewritten and analyzed based on the control theory, revealing the physics of this process, which is significant to practical AFC implementation.