Questions about the grid and the Step size in LES

medaouarwalid · July 15, 2017, 16:56

hi, thanks to members of this forum i am learning every day about LES more than ever, my thread has 2 parts

1- i have more or less understand the strategy of the LEs and what the filter is about, but until now i didnt find a clear answer to how to choose the step size of the grid or the N number of noeds to use it in the software, all answer were like this one : ''An LES computational grid only needs a dx,dy,dz (=dL) small enough to resolve the large scale flow structures. Any structures smaller than this are passed on to the subgrid scale (SGS) model" yes it is understood , but how to determine the size of the large scale flow structures if there is no availble mesurement devices that allow to mesure those structures . For DNS as example, it is more clair, the kolmogorove length scales can be calculated so it is possible to calculate the N number of points of the grid but for LES i did not find any formula or rule.

2- for the time step, i found in lot of answers this rule that says "time step< deltax/u " where deltax is the smallest cell size and u is the velocity. Where did it comes from?

cordially

BlnPhoenix · July 16, 2017, 09:24

1) I suggest reading up on the SGS model you are using, since this is the driver that the uncertainty in LES stems from.

2) You cite the well discussed Courant Criterion which imposes and upper limit for time steps in explicit solvers. It is a good indicator for a time step size but it comes from the stability of explicit solvers not from the accuracy of flow field description. This means: In LES you need to make sure that the time step actually temporally resolves the Eddies you have filterd with the mesh size, this in return is not guaranted by Courant Criterion alone. You can calculate the time step for a DNS and use this and be on the safe site of things. Or you can use a larger time step, which give you a stable and converged solution in each time step and do a sensitivity study afterwards to decide what an increase in temporal resolution actually contributes to the studied flow field property. Or you come up with a good estimate of the frequencies of the large eddies and derive a time step from that. For the last one i'm not able to tell you how you would do it, since the full energy spectrum is not knowable a priori or measured in many cases.

July 15, 2017, 16:56	Questions about the grid and the Step size in LES	#1
medaouarwalid Senior Member dilaw meda Join Date: Jun 2017 Location: algeria Posts: 145 Rep Power: 8	hi, thanks to members of this forum i am learning every day about LES more than ever, my thread has 2 parts 1- i have more or less understand the strategy of the LEs and what the filter is about, but until now i didnt find a clear answer to how to choose the step size of the grid or the N number of noeds to use it in the software, all answer were like this one : ''An LES computational grid only needs a dx,dy,dz (=dL) small enough to resolve the large scale flow structures. Any structures smaller than this are passed on to the subgrid scale (SGS) model" yes it is understood , but how to determine the size of the large scale flow structures if there is no availble mesurement devices that allow to mesure those structures . For DNS as example, it is more clair, the kolmogorove length scales can be calculated so it is possible to calculate the N number of points of the grid but for LES i did not find any formula or rule. 2- for the time step, i found in lot of answers this rule that says "time step< deltax/u " where deltax is the smallest cell size and u is the velocity. Where did it comes from? cordially

July 16, 2017, 09:24		#2
BlnPhoenix Senior Member Join Date: Aug 2014 Location: Germany Posts: 292 Rep Power: 13	1) I suggest reading up on the SGS model you are using, since this is the driver that the uncertainty in LES stems from. 2) You cite the well discussed Courant Criterion which imposes and upper limit for time steps in explicit solvers. It is a good indicator for a time step size but it comes from the stability of explicit solvers not from the accuracy of flow field description. This means: In LES you need to make sure that the time step actually temporally resolves the Eddies you have filterd with the mesh size, this in return is not guaranted by Courant Criterion alone. You can calculate the time step for a DNS and use this and be on the safe site of things. Or you can use a larger time step, which give you a stable and converged solution in each time step and do a sensitivity study afterwards to decide what an increase in temporal resolution actually contributes to the studied flow field property. Or you come up with a good estimate of the frequencies of the large eddies and derive a time step from that. For the last one i'm not able to tell you how you would do it, since the full energy spectrum is not knowable a priori or measured in many cases. medaouarwalid likes this. Last edited by BlnPhoenix; July 16, 2017 at 11:27.