Ph.D. Thesis Proposal by
(Advisor: Professor Vigor Yang)
“Data-Driven Feature Extraction and System Identification of Rocket Injector Dynamics”
Thursday, November 16th, 2:00-3:00 pm
Montgomery Knight, Room 317
For high-performance power generation and propulsion systems, such as those of airbreathing and rocket engines, physical experiments are expensive due to the harsh requirements of operating conditions. In addition, it is difficult to gain insight into the underlying mechanisms of the physiochemical processes involved because of the typical reliance upon optical diagnostics for experimental measurements. High-fidelity simulations can be employed to capture more salient features of the flow and combustion dynamics in engines. These computations, however, are often too expensive and time-consuming for design and development purposes.
To enable usage of modeling/simulation in the design workflow, the present study proposes a data-driven framework for modeling and analysis to facilitate decision making for combustor designs. Its core is a surrogate model employing a machine-learning technique called kriging, which is combined with data-driven basis functions to extract and model the underlying coherent structures from high-fidelity simulation results. This emulation framework encompasses key design parameter sensitivity analysis, physics-guided classification of design parameter sets, and flow evolution modeling for efficient design survey. A sensitivity analysis using Sobol’ indices and a decision tree are incorporated into the framework to better inform the model. This information improves the surrogate model training process, which employs basis functions as regression functions over the design space for the kriging model. The novelty of the proposed approach is the construction of the model through Common Proper Orthogonal Decomposition, allowing for data-reduction and extraction of common coherent structures. The accuracy of prediction of mean flow features for new swirl injector designs is assessed and the dynamic flowfield is captured in the form of power spectrum densities. This data-driven framework also demonstrates the uncertainty quantification of predictions, providing a metric for model fit. The significantly reduced computation time required for evaluating new design points enables efficient survey of the design space.
To further utilize simulation results, a data analytic methodology to characterize the complex nature of turbulent combustion is used to analyze the system dynamics. Comprehensive combustion stability analysis has long been sought after, as a good understanding of the coupling process would reduce the amount of testing and level of capital required for engine development. A vital component is the quantification of the distributed combustion response. The proposed methodology leverages high-fidelity large eddy simulation (LES) in combination with machine-learning techniques to quantify the spatial combustion response. This response is intended to serve as an acoustic source term in the generalized wave equation, which can be used to analyze the stability of complex propulsion systems. Treating the extracted coherent structures as time series signals, the combustion response can be deduced through autoregressive model selection, accounting for data sparsity, multicollinearity, and noise. The results show that acoustic-vortical dynamics is the dominant mechanism determining flame stabilization. This data-driven methodology quantifies the gain and phase relationship between flowfield variables and unsteady heat release. The methodology not only accounts for the distributed combustion response through incorporation of proper orthogonal decomposition (POD) analysis, but also uses the data to identify relevant time scales, replacing the need for forcing and focusing on intrinsic dynamics.
Dr. Vigor Yang
Dr. Joseph Oefelein
Dr. Lakshmi Sankar
Dr. C.F. Jeff Wu (ISyE)