[FMDenaro]

Your geometry is quite complex ...honestly, I would not advice to perform the spectral analysis. What is your goal in computing the spectra? you will get (expecially at low r+) spectra very different from the the classical ones, no inertial regions will be evident... You can hope to get some inertial part in the spectrum only near the centerline. What is more, near the wall you need a DNS-like grid resolution so that checking near the walls you have not clear response about the LES quality.
Finally, what kind of results (experimental/numerical) you have for comparisons?

I suggest to compute only the mean variables.

[Jomid]

Actually, I will have to compare capability of different turbulence models including LES, RANS and a hybrid model in capturing the hydrodynamics (e.g. flow pattern) and heat transfer results both in boundary layer and mean flow area. For heat transfer, I have some relevant experimental results to compare with the result of the turbulence models but regarding flow patterns there is nothing similar to this, that’s why I chose to compare a few turbulence models and use energy spectrum as a means to evaluate turbulence energy at a range of wavenumbers and/or frequencies. Essentially, I want to use Energy Spectrum to see how mesh refinement near wall is important in the correct prediction of heat transfer which it seems to be a key in determination of flow pattern. My goal in computing the spectra is to see if LES is computationally worth to try or not.
I have used my code and got the attached spectra. The problem with this is that whatever number of cells I use for energy spectrum, whether it is a cross-section or a domain with 1 million cells, at low frequencies I get very limited number of data points.
Do you think the energy spectrum and the method you explained can help me in this?
Thanks in advance for your reply

[FMDenaro]

Originally Posted by Jomid

Actually, I will have to compare capability of different turbulence models including LES, RANS and a hybrid model in capturing the hydrodynamics (e.g. flow pattern) and heat transfer results both in boundary layer and mean flow area. For heat transfer, I have some relevant experimental results to compare with the result of the turbulence models but regarding flow patterns there is nothing similar to this, that’s why I chose to compare a few turbulence models and use energy spectrum as a means to evaluate turbulence energy at a range of wavenumbers and/or frequencies. Essentially, I want to use Energy Spectrum to see how mesh refinement near wall is important in the correct prediction of heat transfer which it seems to be a key in determination of flow pattern. My goal in computing the spectra is to see if LES is computationally worth to try or not.
I have used my code and got the attached spectra. The problem with this is that whatever number of cells I use for energy spectrum, whether it is a cross-section or a domain with 1 million cells, at low frequencies I get very limited number of data points.
Do you think the energy spectrum and the method you explained can help me in this?
Thanks in advance for your reply

sorry to say that you are out of the right way...LES and RANS are not different turbulence model but different formulations in which you compute different variables. Furthermore, in RANS you cannot compute any high order statistics such as energy spectra...
I suggest to open a specific thread for your questions

[Jomid]

Thank you very much for your comment
It's true that in RANS we cannot compute energy spectra, however in DES we can do it as DES uses LES formulation in mean flow area and RANS formulation in boundary layer which is why I would like to use Energy Spectrum for LES and DES only. I do not need to compute Energy Spectrum in RANS as RANS is contributing in boundary layer result of DES.
Actually, I have shown that the heat transfer result of DES is the best in terms of compliance with experimental ones. Now using Energy spectrum, I just need to investigate that whether we can rely on the result of relatively cheaper hybrid models i.e. DES rather than using much more expensive LES as they give more reasonable results.
IN SUM, IF THE RESULT OF “DES” AND “LES” ARE SIMILAR, THIS WILL TELL US THAT THERE IS NO NEED TO USE “LES” AND WE CAN USE “DES” AND SAVE COMPUTATIONAL RESOURCE!
This is to some extent new area. However, since the concept of energy spectrum is quite new for me, before going through this, I have to ensure that I am correctly computing the energy spectrum particularly in terms of the procedure and the matlab code I am using.
Any idea about this would be appreciated

[FMDenaro]

but in DES, as you wrote, near the wall you switch to RANS formulations, so it is no longer meaningful to compute the energy spectra...

Comparing LES and DES can have some merits but I think you should consider DES and LES as different tools to compute different things... I do not agree that you can demonstrate that DES and LES can produce similar results. You could take advantage of DES if you requires to compute mainly zero-th order statistics and you do not need of great details in the flow field.

[juliom]

Without any homogeneous direction, how are you planning to compute the FFT? Like I said, you need a homogeneous direction, otherwise you need to apply an special procedure which is very cumbersome, it is not as straight forward as just doiung FFT in Matlab.

Another important aspect is that in the boundary layer, the flow is highly anysotropic, then again the issue will be the FFT. If I am not mistaken Professor Denaro used an average over planes at a specific y+, what I do not remember is if the y+ was above the boundary layer. Maybe I am VERY wrong, but within the BL the phenomenon is highly dissipative due to viscous forces then, there is no cascade. Like I said, MAYBE I AM VERY WRONG WITH THIS STATEMENT. !!. (I would liek to read Professor Denaro reply)
I have only computed the Energy Spectrum for periodic BC. In your case without any homogeneous direction you need an special approach.

[FMDenaro]

Just as example of the effects in the BL, I suggest to check the existing DNS database for the plane channel flow, you can plot the spectra computed from y+=5 up to the final y+ in the centerline.

For the pipe flow with inflow/outflow BC.s I can only think of the azimuthal periodic direction to perform the FFT. That can be done for several r+ values.

[FMDenaro]

http://torroja.dmt.upm.es/turbdata/a...ARD-AR-345.pdf

[juliom]

Professor Denaro I am wondering why from y+ = 5 ?
I found this data set (From JHTDB Jhons Hopkins University)
http://turbulence.pha.jhu.edu/docs/README-CHANNEL.pdf

And they start from y+ = 10.11. Although the BL theoretically starts at y+ = 10^-1.
I did not find the Energy spectrum for y+ = 5. [age 22 shows the energy spectrum for decaying isotropic but there is no information about y+.

Thanks

[FMDenaro]

Here the file for Re_tau=590
--------------------------------

# The data in this file was extracted from a direct numerical
# simulation of fully developed plane turbulent channel
# flow. Particulars are listed below. Note that some of the statistical
# quantities provided in this data set should be zero due to
# (statistical) symmetries in the flow. These quantities are reported
# here none-the-less as an indicator of the quality of the statistics.
#
# Authors: Moser, Kim & Mansour
# Reference: DNS of Turbulent Channel Flow up to Re_tau=590, 1998,
# To appear in Physics of Fluids. For an up-to-date reference,
# check http://www.tam.uiuc.edu/Faculty/Moser/channel
# Numerical Method: Kim, Moin & Moser, 1987, J. Fluid Mech. vol 177, 133-166
# Re_tau = 587.19
# Normalization: U_tau, h
# Description: Streamwise (x) One-dimensional spectra of velocity and pressure at indicated y-location
# Filename: chan590/spectra/chan590.xspec.5
# Date File Created: Dec 15, 1998
#
#---End of Header---
#
# nx = 384, Re = 587.19, y = 0.00910, y+ = 5.34
#
# k_x E_uu E_vv E_ww E_pp
#
0.0000e+00 4.8677e-01 7.3746e-04 3.1778e-02 7.0881e-02
1.0000e+00 7.0215e-01 1.4500e-03 6.2682e-02 2.3846e-01
2.0000e+00 4.7409e-01 1.3949e-03 5.1455e-02 3.2417e-01
3.0000e+00 3.5188e-01 1.3415e-03 4.4875e-02 2.9949e-01
4.0000e+00 2.7658e-01 1.2603e-03 4.0961e-02 2.8689e-01
5.0000e+00 2.2510e-01 1.2183e-03 3.7255e-02 2.8152e-01
6.0000e+00 1.7738e-01 1.1344e-03 3.2866e-02 2.5634e-01
7.0000e+00 1.4569e-01 1.0900e-03 3.0263e-02 2.4699e-01
8.0000e+00 1.2165e-01 1.0304e-03 2.8255e-02 2.2439e-01
9.0000e+00 1.0114e-01 9.6823e-04 2.5458e-02 2.1876e-01
1.0000e+01 8.4106e-02 9.2857e-04 2.3482e-02 2.0114e-01
1.1000e+01 7.1388e-02 8.7910e-04 2.1290e-02 1.9000e-01
1.2000e+01 6.0809e-02 8.2953e-04 1.9604e-02 1.8255e-01
1.3000e+01 5.0356e-02 7.9098e-04 1.7633e-02 1.6890e-01
1.4000e+01 4.2725e-02 7.4605e-04 1.6014e-02 1.5990e-01
1.5000e+01 3.6569e-02 7.0777e-04 1.4619e-02 1.4636e-01
1.6000e+01 3.0415e-02 6.5993e-04 1.3107e-02 1.4004e-01
1.7000e+01 2.6607e-02 6.2925e-04 1.1880e-02 1.3496e-01
1.8000e+01 2.2589e-02 5.9556e-04 1.0700e-02 1.2242e-01
1.9000e+01 1.9396e-02 5.5873e-04 9.6193e-03 1.1596e-01
2.0000e+01 1.6928e-02 5.2358e-04 8.7356e-03 1.0753e-01
2.1000e+01 1.4498e-02 4.8802e-04 7.7565e-03 9.8067e-02
2.2000e+01 1.2531e-02 4.6103e-04 6.9862e-03 9.3582e-02
2.3000e+01 1.0857e-02 4.3033e-04 6.3188e-03 8.5594e-02
2.4000e+01 9.4042e-03 4.0534e-04 5.6644e-03 8.0041e-02
2.5000e+01 8.1042e-03 3.7157e-04 5.0116e-03 7.3135e-02
2.6000e+01 7.1631e-03 3.5501e-04 4.5860e-03 6.6692e-02
2.7000e+01 6.2495e-03 3.3296e-04 4.1227e-03 6.2546e-02
2.8000e+01 5.4729e-03 3.0941e-04 3.6361e-03 5.6763e-02
2.9000e+01 4.8112e-03 2.9443e-04 3.3155e-03 5.3208e-02
3.0000e+01 4.2172e-03 2.7643e-04 2.9866e-03 4.9016e-02
3.1000e+01 3.7448e-03 2.5769e-04 2.6943e-03 4.5041e-02
3.2000e+01 3.3581e-03 2.3840e-04 2.4075e-03 4.0982e-02
3.3000e+01 2.9010e-03 2.2343e-04 2.1376e-03 3.7559e-02
3.4000e+01 2.6024e-03 2.1021e-04 1.9445e-03 3.4767e-02
3.5000e+01 2.2999e-03 1.9617e-04 1.7712e-03 3.1904e-02
3.6000e+01 2.0397e-03 1.8416e-04 1.6053e-03 2.9137e-02
3.7000e+01 1.8222e-03 1.6890e-04 1.4076e-03 2.6407e-02
3.8000e+01 1.6154e-03 1.6111e-04 1.2876e-03 2.4381e-02
3.9000e+01 1.4566e-03 1.4863e-04 1.1656e-03 2.2701e-02
4.0000e+01 1.2817e-03 1.3935e-04 1.0570e-03 2.0868e-02
4.1000e+01 1.1475e-03 1.3000e-04 9.5525e-04 1.8841e-02
4.2000e+01 1.0321e-03 1.2203e-04 8.6132e-04 1.7284e-02
4.3000e+01 9.2468e-04 1.1393e-04 7.8134e-04 1.5694e-02
4.4000e+01 8.3123e-04 1.0677e-04 7.1226e-04 1.4539e-02
4.5000e+01 7.5287e-04 9.8968e-05 6.3962e-04 1.3240e-02
4.6000e+01 6.6653e-04 9.3964e-05 5.8171e-04 1.2580e-02
4.7000e+01 6.0718e-04 8.6908e-05 5.2961e-04 1.1238e-02
4.8000e+01 5.4665e-04 8.0992e-05 4.8060e-04 1.0192e-02
4.9000e+01 4.8737e-04 7.5507e-05 4.2996e-04 9.3533e-03
5.0000e+01 4.4300e-04 7.1003e-05 3.9332e-04 8.7898e-03
5.1000e+01 4.0404e-04 6.5889e-05 3.5513e-04 7.9470e-03
5.2000e+01 3.6416e-04 6.1596e-05 3.2402e-04 7.3591e-03
5.3000e+01 3.3914e-04 5.8084e-05 2.9695e-04 6.7352e-03
5.4000e+01 3.0306e-04 5.3403e-05 2.6481e-04 6.0785e-03
5.5000e+01 2.7508e-04 5.0137e-05 2.4720e-04 5.6861e-03
5.6000e+01 2.5068e-04 4.7240e-05 2.2704e-04 5.2177e-03
5.7000e+01 2.2614e-04 4.4509e-05 2.0631e-04 4.8133e-03
5.8000e+01 2.0891e-04 4.1698e-05 1.8892e-04 4.4269e-03
5.9000e+01 1.9047e-04 3.9094e-05 1.7215e-04 4.1172e-03
6.0000e+01 1.7435e-04 3.6144e-05 1.5639e-04 3.7227e-03
6.1000e+01 1.6073e-04 3.4212e-05 1.4492e-04 3.4810e-03
6.2000e+01 1.4632e-04 3.1654e-05 1.3212e-04 3.1699e-03
6.3000e+01 1.3519e-04 3.0230e-05 1.2327e-04 3.0240e-03
6.4000e+01 1.2310e-04 2.8223e-05 1.1238e-04 2.7089e-03
6.5000e+01 1.1300e-04 2.6631e-05 1.0376e-04 2.4846e-03
6.6000e+01 1.0388e-04 2.4704e-05 9.5026e-05 2.3083e-03
6.7000e+01 9.5280e-05 2.3128e-05 8.6977e-05 2.1057e-03
6.8000e+01 8.7613e-05 2.1749e-05 8.0863e-05 1.9817e-03
6.9000e+01 8.1442e-05 2.0535e-05 7.4932e-05 1.8326e-03
7.0000e+01 7.4580e-05 1.9431e-05 6.8693e-05 1.7172e-03
7.1000e+01 6.8803e-05 1.8208e-05 6.3291e-05 1.5678e-03
7.2000e+01 6.3551e-05 1.7139e-05 5.8342e-05 1.4505e-03
7.3000e+01 5.9172e-05 1.6116e-05 5.3113e-05 1.3603e-03
7.4000e+01 5.4715e-05 1.5209e-05 4.9413e-05 1.2589e-03
7.5000e+01 5.0458e-05 1.4235e-05 4.5676e-05 1.1524e-03
7.6000e+01 4.6685e-05 1.3391e-05 4.2305e-05 1.0612e-03
7.7000e+01 4.3323e-05 1.2639e-05 3.9150e-05 1.0033e-03
7.8000e+01 4.0060e-05 1.1836e-05 3.6375e-05 9.2910e-04
7.9000e+01 3.7622e-05 1.1327e-05 3.4123e-05 8.8803e-04
8.0000e+01 3.4639e-05 1.0679e-05 3.1650e-05 8.1565e-04
8.1000e+01 3.2264e-05 1.0013e-05 2.9132e-05 7.5344e-04
8.2000e+01 3.0218e-05 9.5845e-06 2.7475e-05 7.1789e-04
8.3000e+01 2.7711e-05 8.9389e-06 2.5402e-05 6.5972e-04
8.4000e+01 2.6007e-05 8.4585e-06 2.3815e-05 6.1628e-04
8.5000e+01 2.4058e-05 7.9957e-06 2.2214e-05 5.7103e-04
8.6000e+01 2.2364e-05 7.5818e-06 2.0635e-05 5.3522e-04
8.7000e+01 2.1000e-05 7.0770e-06 1.9392e-05 4.9412e-04
8.8000e+01 1.9780e-05 6.7502e-06 1.7918e-05 4.6404e-04
8.9000e+01 1.8429e-05 6.4426e-06 1.6917e-05 4.3999e-04
9.0000e+01 1.7219e-05 6.0757e-06 1.5629e-05 4.0728e-04
9.1000e+01 1.6055e-05 5.7466e-06 1.4641e-05 3.7996e-04
9.2000e+01 1.4983e-05 5.4652e-06 1.3730e-05 3.5627e-04
9.3000e+01 1.4141e-05 5.1654e-06 1.2679e-05 3.3413e-04
9.4000e+01 1.3082e-05 4.8073e-06 1.1718e-05 3.0813e-04
9.5000e+01 1.2301e-05 4.6075e-06 1.1028e-05 2.9221e-04
9.6000e+01 1.1579e-05 4.3297e-06 1.0329e-05 2.7304e-04
9.7000e+01 1.0928e-05 4.1132e-06 9.7421e-06 2.5589e-04
9.8000e+01 1.0236e-05 3.8821e-06 9.0691e-06 2.3784e-04
9.9000e+01 9.5424e-06 3.6620e-06 8.4866e-06 2.2286e-04
1.0000e+02 8.9781e-06 3.4996e-06 8.0324e-06 2.1077e-04
1.0100e+02 8.4399e-06 3.3357e-06 7.5608e-06 1.9938e-04
1.0200e+02 7.9277e-06 3.1384e-06 7.0644e-06 1.8797e-04
1.0300e+02 7.4211e-06 2.9760e-06 6.5906e-06 1.7576e-04
1.0400e+02 6.9055e-06 2.8030e-06 6.1246e-06 1.6445e-04
1.0500e+02 6.5328e-06 2.6984e-06 5.8598e-06 1.5585e-04
1.0600e+02 6.1172e-06 2.5451e-06 5.4419e-06 1.4571e-04
1.0700e+02 5.8129e-06 2.4325e-06 5.0964e-06 1.3734e-04
1.0800e+02 5.3972e-06 2.2698e-06 4.7949e-06 1.2681e-04
1.0900e+02 5.1195e-06 2.1779e-06 4.5444e-06 1.2234e-04
1.1000e+02 4.7945e-06 2.0653e-06 4.2918e-06 1.1490e-04
1.1100e+02 4.5207e-06 1.9615e-06 4.0315e-06 1.0768e-04
1.1200e+02 4.3010e-06 1.8915e-06 3.8147e-06 1.0223e-04
1.1300e+02 4.0233e-06 1.8007e-06 3.5859e-06 9.5903e-05
1.1400e+02 3.8182e-06 1.7192e-06 3.4277e-06 9.1794e-05
1.1500e+02 3.5614e-06 1.6386e-06 3.2349e-06 8.6643e-05
1.1600e+02 3.3578e-06 1.5750e-06 3.0409e-06 8.2884e-05
1.1700e+02 3.1612e-06 1.4835e-06 2.8694e-06 7.6888e-05
1.1800e+02 3.0056e-06 1.4259e-06 2.7111e-06 7.2716e-05
1.1900e+02 2.8152e-06 1.3374e-06 2.5492e-06 6.7494e-05
1.2000e+02 2.6753e-06 1.2769e-06 2.4019e-06 6.4247e-05
1.2100e+02 2.5185e-06 1.2179e-06 2.2813e-06 6.0695e-05
1.2200e+02 2.3738e-06 1.1610e-06 2.1594e-06 5.7533e-05
1.2300e+02 2.2486e-06 1.0986e-06 2.0827e-06 5.3550e-05
1.2400e+02 2.1131e-06 1.0502e-06 1.9639e-06 5.1195e-05
1.2500e+02 2.0181e-06 9.9854e-07 1.8571e-06 4.8240e-05
1.2600e+02 1.8976e-06 9.5849e-07 1.7496e-06 4.5700e-05
1.2700e+02 1.8086e-06 9.2472e-07 1.6572e-06 4.3500e-05
1.2800e+02 1.7302e-06 8.8388e-07 1.5772e-06 4.1542e-05
1.2900e+02 1.6338e-06 8.3598e-07 1.4932e-06 3.9058e-05
1.3000e+02 1.5386e-06 8.0059e-07 1.4173e-06 3.6583e-05
1.3100e+02 1.4569e-06 7.6606e-07 1.3353e-06 3.5044e-05
1.3200e+02 1.3847e-06 7.3581e-07 1.2806e-06 3.3010e-05
1.3300e+02 1.3123e-06 7.0351e-07 1.2070e-06 3.1387e-05
1.3400e+02 1.2615e-06 6.7313e-07 1.1553e-06 2.9996e-05
1.3500e+02 1.1985e-06 6.4078e-07 1.0909e-06 2.8410e-05
1.3600e+02 1.1331e-06 6.1772e-07 1.0438e-06 2.7214e-05
1.3700e+02 1.0806e-06 5.8983e-07 9.9583e-07 2.5638e-05
1.3800e+02 1.0286e-06 5.6619e-07 9.4855e-07 2.4381e-05
1.3900e+02 9.8193e-07 5.4082e-07 9.0605e-07 2.3155e-05
1.4000e+02 9.3108e-07 5.1985e-07 8.6172e-07 2.2133e-05
1.4100e+02 8.8516e-07 4.9532e-07 8.1669e-07 2.1061e-05
1.4200e+02 8.4380e-07 4.7678e-07 7.8716e-07 2.0018e-05
1.4300e+02 8.0868e-07 4.6138e-07 7.5662e-07 1.9273e-05
1.4400e+02 7.7870e-07 4.4248e-07 7.2814e-07 1.8397e-05
1.4500e+02 7.4175e-07 4.2699e-07 6.8819e-07 1.7566e-05
1.4600e+02 6.9963e-07 4.0681e-07 6.5463e-07 1.6552e-05
1.4700e+02 6.6360e-07 3.9271e-07 6.3445e-07 1.5778e-05
1.4800e+02 6.3972e-07 3.7703e-07 6.0724e-07 1.5100e-05
1.4900e+02 6.1073e-07 3.6100e-07 5.8283e-07 1.4344e-05
1.5000e+02 5.8419e-07 3.4421e-07 5.5774e-07 1.3500e-05
1.5100e+02 5.6239e-07 3.3274e-07 5.3594e-07 1.2896e-05
1.5200e+02 5.3505e-07 3.1891e-07 5.1632e-07 1.2293e-05
1.5300e+02 5.1559e-07 3.1033e-07 4.9735e-07 1.1774e-05
1.5400e+02 4.8871e-07 2.9560e-07 4.7905e-07 1.1132e-05
1.5500e+02 4.6982e-07 2.8584e-07 4.6178e-07 1.0627e-05
1.5600e+02 4.4842e-07 2.7498e-07 4.4163e-07 1.0201e-05
1.5700e+02 4.2745e-07 2.6609e-07 4.2491e-07 9.7151e-06
1.5800e+02 4.1298e-07 2.5870e-07 4.1095e-07 9.4280e-06
1.5900e+02 3.9862e-07 2.5282e-07 3.9634e-07 9.0834e-06
1.6000e+02 3.8252e-07 2.4350e-07 3.8357e-07 8.6855e-06
1.6100e+02 3.6654e-07 2.3362e-07 3.6931e-07 8.2760e-06
1.6200e+02 3.5327e-07 2.2458e-07 3.5540e-07 7.9233e-06
1.6300e+02 3.3960e-07 2.1574e-07 3.4141e-07 7.6037e-06
1.6400e+02 3.2596e-07 2.0954e-07 3.3184e-07 7.3586e-06
1.6500e+02 3.1143e-07 1.9932e-07 3.2121e-07 7.0208e-06
1.6600e+02 2.9962e-07 1.9151e-07 3.0833e-07 6.7341e-06
1.6700e+02 2.8740e-07 1.8364e-07 2.9664e-07 6.4871e-06
1.6800e+02 2.7825e-07 1.7868e-07 2.8746e-07 6.2790e-06
1.6900e+02 2.6678e-07 1.7357e-07 2.8022e-07 6.0944e-06
1.7000e+02 2.5732e-07 1.6696e-07 2.7059e-07 5.8683e-06
1.7100e+02 2.5044e-07 1.6137e-07 2.6263e-07 5.7329e-06
1.7200e+02 2.3994e-07 1.5527e-07 2.5264e-07 5.5782e-06
1.7300e+02 2.2939e-07 1.4889e-07 2.4264e-07 5.4110e-06
1.7400e+02 2.2215e-07 1.4440e-07 2.3472e-07 5.3021e-06
1.7500e+02 2.1587e-07 1.3909e-07 2.2751e-07 5.1609e-06
1.7600e+02 2.0528e-07 1.3269e-07 2.1859e-07 5.0125e-06
1.7700e+02 1.9764e-07 1.2807e-07 2.0921e-07 4.9274e-06
1.7800e+02 1.9027e-07 1.2304e-07 2.0282e-07 4.8120e-06
1.7900e+02 1.8409e-07 1.1863e-07 1.9516e-07 4.6875e-06
1.8000e+02 1.7790e-07 1.1269e-07 1.8240e-07 4.5709e-06
1.8100e+02 1.6880e-07 1.0674e-07 1.7547e-07 4.4690e-06
1.8200e+02 1.6359e-07 1.0068e-07 1.6842e-07 4.3526e-06
1.8300e+02 1.5743e-07 9.8913e-08 1.6321e-07 4.3314e-06
1.8400e+02 1.5154e-07 9.3774e-08 1.5525e-07 4.2553e-06
1.8500e+02 1.4602e-07 8.9167e-08 1.4721e-07 4.2371e-06
1.8600e+02 1.4042e-07 8.6446e-08 1.4023e-07 4.1870e-06
1.8700e+02 1.3778e-07 8.5241e-08 1.3871e-07 4.1521e-06
1.8800e+02 1.3391e-07 8.0606e-08 1.3295e-07 4.0963e-06
1.8900e+02 1.3273e-07 8.0084e-08 1.3148e-07 4.1555e-06
1.9000e+02 1.3007e-07 7.9807e-08 1.2678e-07 4.2340e-06
1.9100e+02 1.3927e-07 8.6560e-08 1.3778e-07 4.4073e-06

[Jomid]

Originally Posted by FMDenaro

but in DES, as you wrote, near the wall you switch to RANS formulations, so it is no longer meaningful to compute the energy spectra...

Comparing LES and DES can have some merits but I think you should consider DES and LES as different tools to compute different things... I do not agree that you can demonstrate that DES and LES can produce similar results. You could take advantage of DES if you requires to compute mainly zero-th order statistics and you do not need of great details in the flow field.

What you are saying is true. But here, we are dealing with a compromise between accuracy and computational costs. Let me clarify this with a simple example; for my case, in order to meet all LES meshing criteria, we should have a mesh with at least 200 million cells which is because of very small turbulence scale at the boundary layer and viscous sub-layer. We know that this is not cost effective which is why using LES in pipes is not popular so we are restricted because of ultra fine mesh requirement near the wall.
Essentially, wall-induced turbulence flow (in a helical pipe) is one of my concerns. I want to see how wall can affect mean flow behavior. The main questions are:
1. Can we massively save computational costs by reducing the cell number, e.g. as small as 10 million, and make sure that DES will give us good results? (Please be advised that in that number of cells LES had unreasonable results which was even worse than RANS results. So I wonder if energy spectrum might be able to help me to find what the potential reason is for getting high discrepancy in the experimental results and LES results)
2. Why the result of DES are more realistic than LES? To answer that I should manage to compare the energy spectrum obtained from LES with the one obtained from DES as DES had much better results than LES.
3. Can we see the cascade in either of these to models? (the last picture I attached proves that cascade can be seen).

But the more important question is how I can perform energy spectrum in a helical pipe with non-periodic BC as most people deal with turbulence flow in channels. However, I think having the energy spectrum code for a straight pipe should be fine to compare with my results.

Regards

[Jomid]

Originally Posted by juliom

Jess1, look at Professor Denaro script, this is everything you need.

http://www.cfd-online.com/Forums/mai...tml#post565863

But in that code the energy spectrum is computed based on the velocity field by computing the kinetic energy. Your case is different though, because you are using the two point correlation. I can share with you my code (that I wrote based on prof Denaro script, I need to acknowledge his contribution) based on Popes' book.

Dear juliom
Can I ask you to share your code with me? I would like to give it a try and see how I can manage to find a way for my problem.
Thanks in advance

[FMDenaro]

Originally Posted by Jomid

What you are saying is true. But here, we are dealing with a compromise between accuracy and computational costs. Let me clarify this with a simple example; for my case, in order to meet all LES meshing criteria, we should have a mesh with at least 200 million cells which is because of very small turbulence scale at the boundary layer and viscous sub-layer. We know that this is not cost effective which is why using LES in pipes is not popular so we are restricted because of ultra fine mesh requirement near the wall.
Essentially, wall-induced turbulence flow (in a helical pipe) is one of my concerns. I want to see how wall can affect mean flow behavior. The main questions are:
1. Can we massively save computational costs by reducing the cell number, e.g. as small as 10 million, and make sure that DES will give us good results? (Please be advised that in that number of cells LES had unreasonable results which was even worse than RANS results. So I wonder if energy spectrum might be able to help me to find what the potential reason is for getting high discrepancy in the experimental results and LES results)
2. Why the result of DES are more realistic than LES? To answer that I should manage to compare the energy spectrum obtained from LES with the one obtained from DES as DES had much better results than LES.
3. Can we see the cascade in either of these to models? (the last picture I attached proves that cascade can be seen).

But the more important question is how I can perform energy spectrum in a helical pipe with non-periodic BC as most people deal with turbulence flow in channels. However, I think having the energy spectrum code for a straight pipe should be fine to compare with my results.

Regards

Your questions mix many issues... LES means nothing if you do not specify the type of tools (filtering, discretization, SGS model). If you used Fluent I am not so suprised that the LES solution can have some lack. Have you set the unbounded second order discretization with dynamic SGS model?

However, to study interation between wall turbulence and outer flow you need a good resolution. What about the dimension and characteristic velocity of your problem? How many grid nodes you used in the BL for DES?

Finally, I would suggest the administrators to shift these discussions in a new post.

[FMDenaro]

Originally Posted by Jomid

Actually, I will have to compare capability of different turbulence models including LES, RANS and a hybrid model in capturing the hydrodynamics (e.g. flow pattern) and heat transfer results both in boundary layer and mean flow area. For heat transfer, I have some relevant experimental results to compare with the result of the turbulence models but regarding flow patterns there is nothing similar to this, that’s why I chose to compare a few turbulence models and use energy spectrum as a means to evaluate turbulence energy at a range of wavenumbers and/or frequencies. Essentially, I want to use Energy Spectrum to see how mesh refinement near wall is important in the correct prediction of heat transfer which it seems to be a key in determination of flow pattern. My goal in computing the spectra is to see if LES is computationally worth to try or not.
I have used my code and got the attached spectra. The problem with this is that whatever number of cells I use for energy spectrum, whether it is a cross-section or a domain with 1 million cells, at low frequencies I get very limited number of data points.
Do you think the energy spectrum and the method you explained can help me in this?
Thanks in advance for your reply

This spectra does not help ... first, you see the lack in the statistical samples appearing with the oscillations; second, it seems you are in a largely viscous region. What about the distance from the inlet and from the wall of this spectrum? I suggest to superimpose the slope -5/3 in a double logarithmic scale.

[Jomid]

Originally Posted by FMDenaro

Your questions mix many issues... LES means nothing if you do not specify the type of tools (filtering, discretization, SGS model). If you used Fluent I am not so suprised that the LES solution can have some lack. Have you set the unbounded second order discretization with dynamic SGS model?

However, to study interation between wall turbulence and outer flow you need a good resolution. What about the dimension and characteristic velocity of your problem? How many grid nodes you used in the BL for DES?

Finally, I would suggest the administrators to shift these discussions in a new post.

Dear Filippo
Thanks very much for your help
For momentum spatial discretization I use bounded central differencing as Fluent suggests. I assume unbounded second order discretization means second order upwind! Is my assumption right? If yes, DO YOU SUGGEST ME TO USE SECOND ORDER UPWIND IN THAT CASE?

For SGS model I use WALE, as proposed in a paper that WALE has more balance in cases like my case. Do you think using WALE model is the reason for the overestimated results?

Regarding the mesh resolution, I am using the same mesh for all turbulence models to have a fair comparison. The viscous sub-layer is resolved by 2-3 cells in the direction normal to the wall (r+=1.2). However theta+ is much bigger (7+) and in stream-wise direction (y+=16). The Re number is 4700.

[Jomid]

Originally Posted by FMDenaro

This spectra does not help ... first, you see the lack in the statistical samples appearing with the oscillations; second, it seems you are in a largely viscous region. What about the distance from the inlet and from the wall of this spectrum? I suggest to superimpose the slope -5/3 in a double logarithmic scale.

The lack in the statistical samples is my main concern especially in low frequencies. But sadly I do not know what is source of the problem. To solve the issue I did, I increased the cell numbers to nearly 1 million, the result was almost the same. I even excluded information related to samples near wall (r+<3) the result did not improve. Any solution for this would as valuable as gold.
But could you please let me know why do you say that I am in a largely viscous region?
The result are from a cross section which is very far from the inlet before outlet (suppose that we have 10 locations (cross-sections) in the helical pipe (divided on an equal interval), this result are from 9th location. The cross section gives all the velocity values (u,v,w) excluding wall velocities which are zero for approximately 3400 cells.
I agree with you. I was thinking that after finalizing the current challenge, my next task would be fitting the -5/3 kolmogorov slope.
Look forward to hearing your opinion.
Cheers

[FMDenaro]

Originally Posted by Jomid

Dear Filippo
Thanks very much for your help
For momentum spatial discretization I use bounded central differencing as Fluent suggests. I assume unbounded second order discretization means second order upwind! Is my assumption right? If yes, DO YOU SUGGEST ME TO USE SECOND ORDER UPWIND IN THAT CASE?

For SGS model I use WALE, as proposed in a paper that WALE has more balance in cases like my case. Do you think using WALE model is the reason for the overestimated results?

Regarding the mesh resolution, I am using the same mesh for all turbulence models to have a fair comparison. The viscous sub-layer is resolved by 2-3 cells in the direction normal to the wall (r+=1.2). However theta+ is much bigger (7+) and in stream-wise direction (y+=16). The Re number is 4700.

First, I am talking about the central second order ubounded scheme. Despite the automatic switching to bounded when you active LES, you can force the unbounded version by command. I do not remember exactly the command, but @sbaffini can help.

Second, I suggest to do the LES first with the dynamic modelling. What is more, I strongly suggest to see the solution without any SGS model to check the real effect of what you are modelling.

Third, is not correct to compare RANS, DES, LES on the same computational grid. Each formulation has its grid requirement. Usually, for wall confined flows we use at least 3-4 nodes in y+<1, the the grid in enlarged toward the centerline. But the resolution in homogenous directions produces a quite large mesh size, I mean it is quite common to work with h+ = O(20).

Finally, searching for a computational compromise can drive you to test also the wall-modelled LES.

[FMDenaro]

Originally Posted by Jomid

The lack in the statistical samples is my main concern especially in low frequencies. But sadly I do not know what is source of the problem. To solve the issue I did, I increased the cell numbers to nearly 1 million, the result was almost the same. I even excluded information related to samples near wall (r+<3) the result did not improve. Any solution for this would as valuable as gold.
But could you please let me know why do you say that I am in a largely viscous region?
The result are from a cross section which is very far from the inlet before outlet (suppose that we have 10 locations (cross-sections) in the helical pipe (divided on an equal interval), this result are from 9th location. The cross section gives all the velocity values (u,v,w) excluding wall velocities which are zero for approximately 3400 cells.
I agree with you. I was thinking that after finalizing the current challenge, my next task would be fitting the -5/3 kolmogorov slope.
Look forward to hearing your opinion.
Cheers

The law -5/3 is only a theoretical results for the inertial cascade in a homogeneous isotropic field. You are very far from such a case. The -5/3 is only an arbitrary reference to see the difference with the inertial cascade.

Statistical sample is needed....that means you need to perform the ensemble average between several cross sections and at different times. But, again, the spectra has sense only performing the FFT in the azimuthal direction.

[Jomid]

Originally Posted by FMDenaro

The law -5/3 is only a theoretical results for the inertial cascade in a homogeneous isotropic field. You are very far from such a case. The -5/3 is only an arbitrary reference to see the difference with the inertial cascade.

Statistical sample is needed....that means you need to perform the ensemble average between several cross sections and at different times. But, again, the spectra has sense only performing the FFT in the azimuthal direction.

For sure I will try what you said and report the result here.

[sbaffini]

I also suggest to open a different thread as this has gone far away from the original one which, if i am allowed, is not what you need.

In the meanwhile, i would add:

1) As i wrote you in a previous post, your helical geometry does not necessarily precludes the use of periodicity. And this has nothing to do with the secondary motion, just with the way peirodicity is implemented.

Indeed, i suspect that the driving pressure gradient should be everywhere tangential to the pipe axis to be meaningful. And this would require a purposely written UDF, but is quite simple to do. However, there are also people doing it in the classical way:

http://iopscience.iop.org/article/10...1/1/012025/pdf

that is, just simple as using periodicity and fixing the mass flow rate.

2) Let's say you manage to have periodicity. You can compute spatial spectra, even using the Prof. Denaro's script, but that requires some fantasy from your side and a grid similar to the one in the paper above. Then you can write a journal file which will create monitoring points distributed along fictitious planes at constant r+. When your simulation is done you will need to read all those points at once and assemble your matrix of values as if it was extracted from a true plane and give it to the spectra routine.

3) Honestly, i recognize it might be difficult for a novice to follow route (1). So let's say that you don't have periodicity. Then you simply can't use any spatial spectra, no matter what. The only thing you can do is what they do in the paper above: temporal spectra in selected points. You pick up some relevant point and activate the monitoring of the variables there. After that you just need to do a simple FFT of the temporal signal. If you have a certain number of those points close enough you can also try to have a meaningful information from the correlation of their signals.

4) I would add that your grid does not look necessarily bad for LES. Maybe you need more cells in the viscous layer.

5) Adding from what Filippo just wrote, the (undbounded) central scheme is just there if you have LES as viscous model. You just need to pick it up. Note that it is just called central scheme. at this point however, there might be other issues with your numerical setting that limited your LES on this grid.