Research

With the advancement of modern high-performance computing, high-fidelity modeling, like LES and hybrid RANS/LES, has become powerful tool to provide valuable insights into underlying combustion physics especially at high-pressure, high-temperature and high Reynolds number conditions for realistic problems, for example, rocket and gas turbine engines. These kinds of insights are very difficult to access through the current state-of-art modeling and experimental techniques. The first part of my research focuses on using LES and hybrid RANS/LES approaches to simulate laboratory rocket and gas turbine engines, which are properly scaled based on full-scale engines and accompanied with companion experiment.

Reduced-Order Modeling (ROM) has been demonstrated to be powerful technique in providing accurate & efficient predictions for aerodynamics and aeroelasticity applications. The main idea is to project a high-order Partial Differential Equations (PDE) system onto a test function basis (e.g. POD basis) and reduce the PDE to a smaller Ordinary Differential Equations (ODE) system, thus significantly enhancing the computational efficiency while preserving the fidelity of the high-order model. However, adaption of such technique to turbulent reacting flow requires considerable amount of research efforts, which motivated my work. The second part of my research is dedicated to the Air Force funded Center of Excellence (CoE) on Multi-Fidelity Modeling of Rocket Combustion Dynamics, with the joint collaboration between University of Michigan, MIT, Purdue and University of Wisconsin-Madison. The center aims at developing a multi-fidelity modeling methodology with integration of ROM to predict the stability characteristics of the full-scale rocket engine. For example, given a nominal engine configuration, designers can use the methodology to: a) Efficiently characterize combustion dynamics in O(days) on small cluster; b) Explore effect of parametric changes on Quantities of Interest (QoIs).

Conventional closure models were developed based on assumptions regimes and conditions that may not be applicable for practical engine devices. To perform high-fidelity modeling for realistic problems which exhibits a much wider range of combustion behaviors, more comprehensive turbulent combustion closure models are desirable. The third part of my research focuses on using data-driven modeling for turbulent combustion closure model development is to use the appropriately designed Direct Numerical Simulation (DNS) datasets to inform the estimation of model parameters and forms through inverse modeling.