Accelerated simulation of reactive flows using deep neural networks
Project Details
Program
Mechanical Engineering
Field of Study
Mechanical/aerospace engineering, computational modeling, machine learning
Division
Physical Sciences and Engineering
Faculty Lab Link
Center Affiliation
Clean Combustion Research Center
Project Description
Computational fluid dynamic (CFD) simulations of chemical reacting flows demand excessive computational time in order to solve a large number of highly nonlinear chemical reaction source terms. The project aims to develop an algorithm for accelerated computations by developing high fidelity reduced-order data-based chemical kinetics solver using autoencoder and neural network algorithm. The basic framework has been developed and the new project will apply the algorithm for renewable fuel applications over a wide range of thermodynamic conditions and assess the fidelity and performance enhancement of the new algorithm. The student will gain understanding of various modern machine learning tools in engineering applications with hands-on experience of programming and implementation.
About the Researcher
Hong Im
Professor, Mechanical Engineering
Affiliations
Education Profile
- Ph.D., Mechanical and Aerospace Engineering, Princeton University, 1994
- M.S., Mechanical Engineering, Seoul National University, 1988
- B.S., Mechanical Engineering, Seoul National University, 1986
Research Interests
a€‹Professor Ima's research interests are primarily fundamental and practical aspects of combustion and power generation devices using high-fidelity computational modeling. Recent research topics combustion characteristics of high hydrogen content fuels, advanced modeling of sooting flames, modeling of mixed-mode combustion in modern engines, dynamics of turbulent premixed flame propagation, turbulent flame stabilization, spray- and particle-laden flows and combustion, plasma and electric field effects on flamesA and combustion of low-grade fuels.Selected Publications
- Galassi, R.M., Ciottoli, P.P., Sarathy, S.M., Im, H.G., Paolucci, S., Valorani, M., 2018, ""Automated Chemical Kinetic Mechanism Simplification with Minimal User Expertise,"" Combustion and Flame, 197, 439-338.
- Hernandez Perez, F.E., Mukhadiyev, N., Xu, X., Sow, A., Lee, B.J., Sankaran, R., Im, H.G., 2018, ""Direct Numerical Simulations of Reacting Flows with Detailed Chemistry Using Many-core/GPU Acceleration,"" Computers & Fluids, 173, 73-79.
- An, Y., Jaasim, M., Raman, V., Hernandez Perez, F.E., Im, H.G., Johansson, B., 2018, ""Homogeneous Charge Compression Ignition (HCCI) and Partially Premixed Combustion (PPC) in Compression Ignition Engine with Low Octane Gasoline,"" Energy, 158, 181-191.
- Song, W., Tingas, E.A., Im, H.G., 2018, ""A Computational Analysis of Methanol Autoignition Enhancement by Dimethyl Ether Addition in a Counterflow Mixing Layer,"" Combustion and Flame, 195, 84-98.
- Belhi, M., Lee, B.J., Bisetti, F., Im, H.G., 2018, ""A Computational Study of the Effects of DC Electric Fields on Non-premixed Counterflow Methane-Air Flames,"" J. Physics D: Applied Physics, Special Issue on Multiscale Modeling of Nonequilibrium Plasma Discharges, 50 (49), 494005.
Desired Project Deliverables
Implementation of the code modules, simulations and analysis of the results for assessment of fidelity and performance. Upon successful outcome, a conference and/or journal paper publication is expected.
Recommended Student Background
Computer programming (C++, Fortran, etc.)
Fluid mechanics
Thermodynamics
Heat transfer