search

HiFiLES / HiFiLES-solver

High Fidelity Large Eddy Simulation Solver
Other
173 stars 131 forks source link

Mapping from existing solution to high-order scheme #36

Open sherry413 opened 9 years ago

sherry413 commented 9 years ago

Hi,

I am going to test a backward facing step flow with LES. I thought it may be helpful to use an existing RANS result as the initial condition. However, I found the RANS result is solved by the finite volume mesh and does not fit the solution points for CPR scheme. (The RANS model in your code only support 2D) Do you have any ideas how to perform the mapping? Is simply linear interpolation decrease the accuracy? Thank you.

Sherry

jrbull commented 9 years ago

Sherry,

Thank you for your interest in HiFiLES. Since all you need is an initial condition, linear interpolation should be adequate. You can insert interpolated data into a restart file. The restart file can be generated by running the backward facing step for 1 timestep with restart_dump_freq = 1.

A word of advice from one who spent a long time running this case in a different code: you need to use realistic inlet conditions to get good results. I used mean velocity profiles from DNS of channel flow (data available from various databases e.g. ERCOFTAC). Ideally, you should use a synthetic turbulence inlet condition but we haven’t written one for HiFiLES yet. Instead I suggest one of 3 methods to set the inlet flow: (1) specify a velocity profile as an arbitrary 4th order polynomial (see input_template), (2) hard-code the interpolated RANS/DNS data into the initial conditions routine (see eles::set_ics). Let us know how you get on! Regards, Jonathan

Postdoctoral Scholar Dept of Aero/Astro Engineering Stanford University

On Feb 11, 2015, at 1:39 PM, sherry413 notifications@github.com wrote:

Hi,

I am going to test a backward facing step flow with LES. I thought it may be helpful to use an existing RANS result as the initial condition. However, I found the RANS result is solved by the finite volume mesh and does not fit the solution points for CPR scheme. (The RANS model in your code only support 2D) Do you have any ideas how to perform the mapping? Is simply linear interpolation decrease the accuracy? Thank you.

Sherry

— Reply to this email directly or view it on GitHub.

sherry413 commented 9 years ago

Hi Jonathan,

Thank you for your reply. I check the code with your advice. You mentioned to add some fluctuation at the inlet boundary but I do not quite understand the methods you suggest. They are all working on the initial boundary condition (set_ics is only called at beginning and will not be called for restart case). Besides, the DNS data is about mean not instantaneous so cannot be used to generate fluctuation. Do I need to modify the inlet conditions rather than initial conditions? Thank you.

Sherry

mlopez14 commented 8 years ago

Sorry we let your question slip through the cracks. Are you still interested in an answer?