DATE:

Friday, October 15, 2021

TIME:

3:30 PM

LOCATION:

GMCS 314

SPEAKER:

Qi Wang, Aerospace Engineering, San Diego State University

ABSTRACT:

Turbulence affects fluid flows across scales and application domains. The capabilities to accurately model and estimate information for turbulence is essential for understanding, predicting and controlling these chaotic flows. Two traditional approaches: laboratory experiments and numerical simulation can be combined to overcome their individual deficiencies. An approach that optimally infuses measurements from experiments into computational models is demonstrated, the realism and fidelity of computational models are increased and uncertainties mitigated. This data assimilation approach can be formulated as optimization problems, constrained by the nonlinear governing equations, and guided by gradients from their discrete adjoint. We demonstrate this adjoint-based optimization in two canonical scenarios: scalar source reconstruction and flow state estimation from limited measurements in a turbulent channel flow. Scalar source reconstruction focuses on finding the location of the source of passive pollutant transported in turbulent environments, which is an ill-posed problem with broad applications. Current study uses direct numerical simulation of both forward and adjoint scalar transport equations to identify the location and size of a source in turbulent channel flow based on sensor measurements downstream. We can reconstruct the spatial distribution of a steady source by doing forward and adjoint simulation repeatedly. The other application involves estimating the initial state of turbulent channel flow from surface measurements. The discrete adjoint operator is applied to perform a domain-of-dependence analysis of surface measurements. The analysis relies on evaluation of the Hessian matrix of the cost function, in the vicinity of the true solution, and analysis of the Hessian eigen-spectra. The leading eigenmodes correspond to the highest sensitivity of measurements to the flow; these modes are concentrated in the near-wall region upstream of the sensor location. Their structure explains the fundamental difficulty of estimating the flow state from wall measurements.

HOST:

Parag Katira

VIDEO: