Data-driven mean-field modeling for complex wake dynamics -- Towards automatable data-driven ROM

Increasing high-quality flow data are generated by simulations and experiments with advanced measurement technology and high-performance computers. Data-driven methods and machine learning (ML) techniques are thus bringing new opportunities and challenges to the current research paradigm in the era of big data.

In this talk, we pave the way to automatable Reduced-Oder Modeling (ROM) using first principles and ML techniques. We establish a benchmark configuration, fluidic pinball, for important dynamical features of wake flows, and give deep insight into the nonlinear dynamics of generic bifurcations and instabilities in fluid mechanics. We separately propose the first-principles-based [1,2] and data-centric [3] strategies to model unsteady flows resulting from successive bifurcations. The key enablers are mean-field theory, sparse calibration [4], and network science [5].

These data-driven ROMs hold the promise to automate the model reduction of complex dynamic systems and have broad applications in industry.