Graduate Seminar in Fluid Mechanics

November 8, 2019 @ 4:00 pm – 5:00 pm
132 Gilman Hall

4:10 pm Presentation

Optimization of LES Subgrid Models using Data Assimilation

 Presented by YIFAN DU (Adviser: Prof. Zaki)

The uncertainties of subgrid model constants limit the performance of large eddy simulations in various applications. In this effort, we seek to improve the LES model prediction by searching for the optimal distribution of model coefficient in turbulent channel flow. A cost functional is formed as the L2 norm of discrepancy between available statistical measurements and LES observations. The optimal distribution of the model coefficient is found by minimizing this functional using ensemble variational method. The LES predictions using the optimal coefficients represent the best achievable performance of the subgrid models. By comparing the associated statistics and existing data, the structural deficiency of various subgrid models can be revealed and assessed.

4:35 pm Presentation

In Vivo Time-Resolved Echo-PIV Measurement of Cardiovascular Flows During Extracorporeal Membrane Oxygenation”

 Presented by ZENG ZHANG (Adviser: Prof. Katz)

Veno-arterial Extracorporeal Membrane Oxygenation (VA-ECMO) is a life support technology that provides both respiratory and hemodynamic support by delivering oxygenated blood through arteries and draining deoxygenated blood from the veins. Therefore, VA-ECMO is often used as a salvage therapy for patients with heart and/or lung failure by offloading injured myocardium and thus favoring recovery. However, the choice of the optimal VA-ECMO flow rate (QECMO), which involves the tradeoff between maintaining sufficient end-organ perfusion and reducing the afterload of the native left ventricle, is challenging. Routinely, it is chosen by assessing the cardiovascular conditions with hemodynamics measurements, but the results are influenced by patient-related factors, and the implications are controversial. To promote the understandings of the direct ECMO-cardiac flow interactions, the current study applies an integrated and optimized PIV/PTV method to characterize the time-resolved velocity field inside aortic root of a Yorkshire pig with severe myocardial ischemia at QECMO from 4 to 1.5 L/min. This method is enhanced to address additional challenges including elongation of the bubble traces with increasing distance from the probe, and motion of the boundaries, especially the aortic valve. Modified blind deconvolution and active contouring are used to resolve these issues. The results show that at higher QECMO (4-3 L/min), the ECMO flow influences the systole by generating a higher maximum deceleration and forcing an earlier valve closing. The valve opening time is not obviously influenced. It is also observed that during systole a secondary vortex pair is generated at high QECMO while it is not observed at lower QECMO (2.5-1.5 L/min). As QECMO is decreased from 4 to 3 L/min, the cardiac output increases by 82%. A sharp drop of cardiac output happens at 2.5 L/min and then keeps decreasing as QECMO is reduced to 1.5 L/min. This trend, together with the abrupt drop in right ventricular pressure at QECMO =2.5L/min suggests a cardiac arrest. The results suggest that for the present conditions, an ECMO flow rate of 3 L/min is optimal.

Back to top