PhD Thesis Defense: Yixing Li

Title: High-fidelity numerical simulation and emulation of bi-fluid swirl injector flow and combustion dynamics

Location: MK317

Date & Time: Monday, April 29 at 3:30pm

Advisor: Dr. Vigor Yang

Abstract:

Injectors are critical components of combustion devices in liquid-fueled propulsion systems. By controlling the atomization and mixing of propellants, injectors can affect combustion efficiency, dynamic characteristics, and engine life cycle. This work conducts a comprehensive study of the gas-centered liquid-swirl coaxial (GCLSC) injectors, operating at supercritical conditions. The study is composed of two parts. The first part investigates flow and combustion dynamics of GCLSC injectors based on high-fidelity large eddy simulations (LES), and the second part presents a high-fidelity emulation framework for the prediction of spatiotemporally evovling flow and combustion in a significantly shorter turnaround time.

For the first part of the study, LES simulations are conducted to study supercritical fluid flow dynamics and combustion characteristics of GCLSC injectors. Gaseous oxygen is axially injected into the center post at a temperature of 687.7K, while kerosene is tangentially introduced into the coaxial annulus at a temperature of 492.2K. The operating pressure is 25.3 MPa, well above the thermodynamic critical points of the propellants involved. Based on LES results, For non-reacting flows, detailed flow physics and structures are identified, followed by comprehensive analyses of mechanisms controlling key dynamic characteristics. These mechanisms include vortex shedding near the fuel injection slit, the shear layer instability in the recess region, and vortical expansion and merging in the taper region. For reacting flows, the flow field is categorized into four regions: propellant injection, flame initialization, flame development, and intensive combustion. Detailed flow structures and the flame evolution in each region are elaborated in detail. Moreover, the effects of the recess length on mixing, flow dynamics and combustion dynamics are investigated.

The second part presents a high-fidelity data-driven emulation framework, which utilizes training data from LES and enables flow field emulation in reasonable turnaround time. The framework employs common kernel-smoothed proper orthogonal decomposition (CKSPOD) as the surrogate model, which is able to extract dominant coherent flow structures through hadamard-based POD and kriging, and reconstruct them to predict the flow field of a new case. Significant improvements, including common grid interpolation and physics-based conditions, are incorporated to the this framework to accommodate the prediction of complicated mixing and combustion dynamics. In the current study, CKSPOD utilizes LES results of GCLSC injectors as training data, and recess length is chosen as the varying design parameter. Detailed evaluations of the predicted flow fields are carried out, and the current framework is able to capture both intantaneous and time-averaged flow fields with high accuracy. Moreover, the improved CKSPOD presents uncertainty quantification (UQ) of the predicted flow field, providing a metric for model fit. The proposed framework is further extended to injector design and optimization, based on the objective functions of mixing efficiency and injector wall thermal protection. The current framework is able to select the optimal case The proposed method significantly reduces the computational time for efficient survey of the design space, and will serve as a promising tool in the early design stages.

Committee:

Dr. Vigor Yang, Advisor, AE, Georgia Institute of Technology

Dr. Timothy Lieuwen, AE, Georgia Institute of Technology

Dr. Lakshmi Sankar, AE, Georgia Institute of Technology

Dr. Joseph Oefelein, AE, Georgia Institute of Technology

Dr. C.F. Jeff Wu, ISyE, Georgia Institute of Technology

Dr. Xingjian Wang, AE, Georgia Institute of Technology