Welcome to the Computational Modeling and Flow Physics Group at UC San Diego! We develop advanced numerical methods and data mining approaches to understand and predict complex, high-Reynolds-number aerospace flows. Specializing in modal decomposition for feature extraction and physical discovery, we distill flow structures into predictive reduced-order models for forecasting and optimizing turbulent flows.
We tackle aerospace challenges such as jet noise control, unsteady aerodynamics, aeroacoustics, transition, aero-optics, and hypersonics using physics-based modeling, while curiously exploring machine learning techniques to understand their potential. While emphasizing aerospace applications, our methods extend to complex flow phenomena in natural environments and have been successfully applied in atmospheric science and physical oceanography.
Current Projects
Modal Decomposition for Discovery and Modeling of Nonlinear Flow Physics in Open Cavity Flows
This project explores the use of Bispectral Mode Decomposition (BMD) for uncovering nonlinear flow phenomena and developing reduced-order models in turbulent open cavity flows. BMD, unlike traditional methods such as Proper Orthogonal Decomposition and Dynamic Mode Decomposition, provides direct quantitative insight into the nonlinear interactions governing flow dynamics. The work focuses on the complex interplay of shear-layer instabilities, acoustic pressure waves, and low-frequency oscillations at turbulent, technically relevant Reynolds numbers. Key objectives include applying BMD to experimental datasets and simulation results, extending the method for non-time-resolved data and improving its computational efficiency, and constructing reduced-order models using second-order Volterra series to capture dominant triadic interactions.
Machine Learning Control of Jet Noise
Airframe-integrated rectangular nozzles offer enhanced aerodynamic performance and simpler mechanical integration of thrust vectoring. To address aft-angle mixing noise generated by complex aero-acoustic processes, we utilize active flow control with localized arc filament plasma actuators. By deploying high-fidelity large-eddy simulations—including numerical models of these plasma actuators—we can accurately predict noise emissions. Combining human expert knowledge with machine learning techniques, we identify new control laws for jet noise mitigation. By combining human expertise with Deep Reinforcement Learning, we identify new control laws for jet noise mitigation.
Numerical Modeling and Reduced-Order Techniques for Multi-Fidelity Aerodynamic Optimization
As part of the Collaborative Center for Optimal Risk-quantified and Robust Design of Aerospace Vehicles (ACCORD), led by PI John T. Hwang, this project focuses on advancing unsteady flow simulations at different levels of fidelity to support multidisciplinary design optimization. Central to this effort is the use of multi-fidelity modeling, incorporating purely data-driven approaches, adjoint-based optimization, and spectral proper orthogonal decomposition for extracting dominant flow features. These techniques enable reduced-order modeling of unsteady aerodynamic phenomena, enhancing predictive capabilities and optimizing performance across a range of conditions.
Multi-Physics Optimization of Amplifier Heads for Next-Generation High-Average-Power Lasers
Effective thermal management is crucial for increasing the power output of high-energy, ultrashort pulse lasers used in fusion energy, high-energy-density science, industry, and medicine. This project focuses on optimizing gas-cooled laser amplifier heads, such as the HEATER experiment at Lawrence Livermore National Laboratory (LLNL), by addressing the complex interplay of heat transfer, aero-optics, and fluid mechanics in gas cooling systems. Utilizing Reynolds-Averaged Navier-Stokes and Large-Eddy Simulation methods, cooling efficiency is enhanced by sustaining turbulent flow through aerodynamic modifications like turbulators or boundary-layer trips. Unsteadiness and structural vibrations caused by flow separation are eliminated by refining diffuser designs and vane configurations. Another goal is to improve laser beam quality by reducing coherence in density fluctuations, generating more homogeneous turbulence to minimize light scattering while maintaining effective heat dissipation. The aim is to achieve a significant increase in extractable heat flux, thereby advancing the performance of high-power laser cooling systems.
Bispectral Mode Decomposition for Discovery of Triadic Interactions in Laminar-Turbulent Transition
Modal decomposition techniques are at the forefront of scientific discovery of flow physics from experimental and numerical data of technical and natural flows. Bispectral mode decomposition (BMD) is a recently developed method that extracts flow structures associated with triadic nonlinear interactions—the fundamental mechanism of energy transfer in turbulent flows. This project utilizes BMD to deepen the understanding of nonlinear flow phenomena during the laminar-turbulent transition in wall-bounded shear flows. The capabilities of BMD are demonstrated by applying it to the well-understood K- and H-type transition scenarios of the flat plate boundary layer. The method’s potential to estimate nonlinear transfer functions is explored.
Mesh-free Resolvent Analysis for Discovery of Large-Scale Coherent Structures
Resolvent, or input-output analysis, has revealed its potential to predict large-scale coherent structures in turbulent flows accurately and to facilitate a deep understanding of the physics of their generation. Considering its role as an essential emerging tool in turbulence research, novel numerical approaches are required that enable its application to technical and natural flows in complex geometries. To enable resolvent analyses of such flows, we implement a mesh-free approach based on radial basis functions that permit the adjustment of the accuracy to arbitrarily high order, significantly simplifies the discretization of complex domains and local grid refinement and works as a stand-alone tool that can be applied to flow data from different numerical or experimental sources.
Statistical Analysis and Numerical Modeling of Shoaling Internal Gravity Waves
Understanding small-scale ocean dynamics is essential for accurate hydrodynamic modeling and forecasting. We use a high-fidelity numerical simulation environment to investigate the development of coherent structures during the nonlinear evolution of shoaling internal gravity waves. By modeling nonlinear internal solitary waves and applying hydrodynamic stability theory, we aim to quantify the flow physics of spanwise instabilities in these phenomena. Our numerical approach enhances predictive capabilities for small-scale ocean dynamics and contributes to addressing challenges in flow-structure interactions and operational forecasting.
