Video: Large Eddy Simulations of Impinging Supersonic Jets: A Close Look at the Unsteadiness
APS Gallery of Fluid Motion 2023
Embedded Files
In-house PINNs code @ CFDLAB Technion with bells and whistles
Shimon Pisnoy, Hemanth Chandravamsi, Steven H. Frankel
A PyTorch based GPU accelerated PINNs code developed from scratch @ CFDLAB Technion.
Methods implemented:
Fourier feature mapping
Random weight factorization
Sin and SOS activation functions
Scheduler
Adam + L-BFGS optimizer switch
Curriculum training
Loss balancing
Casuality training
Batch training
Latin hyper space collocation points
Strict periodic & dirchlet boundary conditions
Efficient implicit audio signal learning through SOS activation + sinusoid amplification
All methods referenced and implemented from this excellent paper: https://arxiv.org/pdf/2308.08468
Poster presented at Research day (Dec 2022), Faculty of Mechanical Engineering, Technion Israel Institute of Technology. [PDF]
Multi-component triple point problem
Credits: Sean Boker, Amareshwara Sainadh Chamarthi, Hemanth Chandravamsi and Steven Frankel (CFD Lab Technion - Israel)
No, this is not a Kraken. This is the multi-component compressible triple point problem (single-phase), which is a three-state, two-dimensional Riemann problem with different materials (He-air). An important feature in this result is rich vortices generated at material interfaces due to the Kelvin-Helmholtz instability. The simulation was performed on a grid size of 3584 x 1536. The figure is showing the density gradient contours.
In-house Numerical Geometric Shock Dynamics Solver
Evolution of a shock front as it get diffracted by a wedge. Dotted line indicates the shock-shock interaction trajectory.
Euler equations solver
Hemanth Chandravamsi, Amareshwara Sainadh Chamarthi, Natan Hoffmann and Steven Frankel (CFD Lab Technion - Israel)
Mach 2 supersonic flow over cylinder (2D)
Schardin's test
A vortex convecting through a dynamically deforming mesh - YouTube link
Supersonic Jet Noise
Large Eddy Simulations to study the Physics of Supersonic Jets and the Noise
Mach 1.35 under expanded supersonic screeching jet
Mach number (top) and Density (bottom) contours
Impinging sonic jet
NPR = 4, TR = 1, Re = 60,000
Density and pressure fluctuation field around an impinging jet
Visualizing the Mach stem pulsation mode in jets impinging on a dented wall
Visualizing the Mach stem pulsation mode in jets impinging on a dented wall
Hemanth Chandravamsi, Dinesh Kumar Eswaran and Steven H. Frankel
pulses-video.mp4Farfield noise prediction using Neural Networks
Project: predicting farfield noise from the nearfield input using Neural Networks...
keywords: Image Outpainting, PINNS...
Turbulent inflow generation
Digital filtering is being used to generate realistic turbulent inflow with prescribed inflow turbulent stats (Reynolds stresses).
Solving Maxwell's equations using high order Padé schemes
Gal Shaviner, Hemanth Chandravamsi, and Steven H. Frankel
Pulse
Homework assignments form Prof. Frankel's CFD course 2021-2022
Scalar conservation laws 101
Gaussian hill advecting in space
Swirling scalar flow-1
Swirling scalar flow-2
Zalesak discontinuous slotted disk, a smooth hump and a cone advection in space
Kurganov-Petrova-Popov (KPP) rotating shock wave problem
2-D Burgers
1-D Burgers equation
Lax Fredrichs scheme
Lax Wendroff method
SPBSAT scheme
Temporal evolution of signed wall distance field for a dendrite grain configuration. Green lines show the evolution of grain boundary. (arXiv: https://arxiv.org/pdf/2111.09859)
Google Sites
Report abuse