autocorrelations for integral scales
 

Hi all,
Can anybody tell me how to extract the integral scales in fluent to determine the optimized extent of the computational domain for an LES simulation (to use periodic BC)?
I heard it would be done on a coarser mesh using a RANS model. any other details?
Best regards

hi,

i dont think fluent has a built in correlation function...you would either need to write a UDF to do it...or even better, i would recommend you export the data you need (x,y,z positions and u,v,w velocities i suppose) and do it in something like matlab which has loads of built in functions.

akour

thanks
 

thank you akour, I think the mean of the product <u(x)U(x+xi)> necessitates a simulation in time. what do you think?
Best wishes

hi,

to calculate a two point correlation, i would do it in space...but ONLY along planes that are statistically homogenous, and only on a domain that is statistically stationay in time (you need to make sure your unsteady simulation has reached steady state b4 you do this calculation).

regarding averaging in space, if you have a cube surrounded by periodic boundaries, then you can do the correlations in any direction you want (since the domain is fully isotropic), if however you introduce two walls...you can only do two point correlations along 2D planes parallel to the walls. i.e. the average <u(x)*u(x+dx)> can only be on planes/lines/volumes that are statistically homogenous. you can also calculate the integral timescale, by choosing a single control voluume and correlating it with its value at different points in time (again, the solution must be steady for this)

hope that helps
akour

thanks again
 

many thanks Akour, it should be ok