- Nektar-users - imperial.ac.uk

Problem in using "NekPy" instead of "FieldConvert" utility
by Sina Asadi 22 Jan '25

22 Jan '25

******************* This email originates from outside Imperial. Do not click on links and attachments unless you recognise the sender. If you trust the sender, add them to your safe senders list https://spam.ic.ac.uk/SpamConsole/Senders.aspx to disable email stamping for this address. ******************* Dear all, I'm using "FieldConvert" utility (Shell command) multiple times in a python code and it is working well. Now I want to use NekPy modules instead to decrease running time of "FieldConvert" execution. But I ran into a problem and I can't solve it. I will give you a simple example: In a python code, when I run the following two similar "FiledConvert" shell commands, I expect to get two identical output and I get it correctly: !FieldConvert -f -m gradient mesh.xml fields.fld output1.fld !FieldConvert -f -m gradient mesh.xml fields.fld output2.fld And in NekPy when I run the equivalent codes twice in my .py code: from NekPy.FieldUtils import * import sys field = Field(sys.argv, force_output=True) InputModule.Create("xml", field, "mesh.xml").Run() InputModule.Create("fld", field, "fields.fld").Run() ProcessModule.Create("gradient", field).Run() OutputModule.Create("fld", field, output1).Run() InputModule.Create("xml", field, "mesh.xml").Run() InputModule.Create("fld", field, "fields.fld").Run() ProcessModule.Create("gradient", field).Run() OutputModule.Create("fld", field, output2).Run() I get two different outputs, output1.fld and output2.fld It seems that both "gradient" ProcessModule are used in the second output. I can't find a way to prevent the first "gradient" module from being used in the second output. Surely this problem have an easy solution. I would be grateful if you could help me solve this problem. Sina.

1 0

Gaussian Noise Generator in Quasi-3D Approach
by Selman BAYSAL 03 Jan '25

03 Jan '25

******************* This email originates from outside Imperial. Do not click on links and attachments unless you recognise the sender. If you trust the sender, add them to your safe senders list https://spam.ic.ac.uk/SpamConsole/Senders.aspx to disable email stamping for this address. ******************* Dear All,I am working to gain a better understanding of modeling turbulent flows. Some of the topics I am interested are if defining “small initial perturbations (noise)” is necessary in turbulent flow modeling, and how the noise affects the stability of the model. Therefore, I have conducted a literature review, but I have only been able to find a few studies* regarding the Gaussian Noise generator. It would be very helpful if I could review a few more studies on introducing noise particularly into Quasi-3D models. Are there any academic papers or tutorials you could recommend?Any recommendations would be greatly appreciated!Best regards,Selman BaysalR. A. in HydraulicsDept. of Civil Eng'gÇanakkale Onsekiz Mart University--ITU Civil Eng., BSc.ITU Coastal Sci. &amp; Eng., MSc.*Moxey et al. (2020) and &nbsp;Nektar++ tutorial

1 0

Re: [Nektar-users] Adding a passive scalar transport with a Coordinate Transformation
by Daniel Garcia 02 Jan '25

02 Jan '25

I can't be sure as I've never adopted some methods that you are using. Once, I heard that a colleague had issues with mapping + homogeneous expansion, after all he ended up using a 3D mesh. He did not use a passive scalar, but his problems were related to the results, I'm not sure what exactly. Anyway, are you sure about your body force implementation? Are you aware it's being defined for the whole domain and every time step? Perhaps, you should test without the body force and/or for a specific region and not state the same values of w for T. Also, if doable, I suggest trying with a 3D mesh. Beyond that, I would not know what to suggest, sorry. Em qui., 2 de jan. de 2025 às 10:33, Antonio Mazzara < ant.mazzara(a)studenti.unina.it> escreveu: > Hi Daniel, > > Thank you very much for your response to my previous message. I greatly > appreciate the time you took to provide me with potential solutions. > > Unfortunately, I regret to inform you that despite following your > suggestions, the issue I originally reported still persists. I tried > different combinations of methods and even reduced the timestep. > > To assist you in further investigating the problem, I have attached part > of my configuration file for your review. I hope it can help identify any > potential inconsistencies or errors that might be contributing to the issue. > > Please let me know if there’s any additional information I can provide to > assist you. I remain at your disposal for further clarification or testing > if needed. > > Thank you again for your support and patience. > > Looking forward to your reply. > > Best regards, > Antonio > > Il 27/12/2024 18:20, Daniel Garcia ha scritto: > > Hello Antonio, how are you? > > Usually, I get that error when the simulation diverges, so try to monitor > the CFL to see if that is the case. If so, evaluate the possibility of > reducing the time step. > However, that could be the option of something wrong with the setup or the > combination of methods is not implemented. I'm suspicious about that > because the error would be completely different, I guess so... > > If you have solved the problem, please answer so it can be registered in > the forum. > > Att, > Daniel. > > > > > > > > Em seg., 23 de dez. de 2024 às 21:36, ANTONIO MAZZARA < > ant.mazzara(a)studenti.unina.it> escreveu: > >> This email from ant.mazzara(a)studenti.unina.it originates from outside >> Imperial. Do not click on links and attachments unless you recognise the >> sender. If you trust the sender, add them to your safe senders list >> <https://spam.ic.ac.uk/SpamConsole/Senders.aspx> to disable email >> stamping for this address. >> >> >> Dear all, >> >> When I run a 2.5D simulation with a coordinate transformation and an >> extra passive scalar transport, I get the following error: >> CG iterations made = 5001 using tolerance of 1e-09 (error = -nan, rhs_mag >> = 1) >> Fatal : Level 0 assertion violation >> Exceeded maximum number of iterations >> >> However, if I repeat the simulation without the passive scalar transport, >> it proceeds correctly. >> Has anyone else encountered the same problem? If so, would you be able to >> assist me in resolving it? >> Thank you in advance for youe help, >> >> Antonio. >> _______________________________________________ >> Nektar-users mailing list >> Nektar-users(a)imperial.ac.uk >> https://mailman.ic.ac.uk/mailman/listinfo/nektar-users >> >

1 0

Adding a passive scalar transport with a Coordinate Transformation
by ANTONIO MAZZARA 27 Dec '24

27 Dec '24

******************* This email originates from outside Imperial. Do not click on links and attachments unless you recognise the sender. If you trust the sender, add them to your safe senders list https://spam.ic.ac.uk/SpamConsole/Senders.aspx to disable email stamping for this address. ******************* Dear all, When I run a 2.5D simulation with a coordinate transformation and an extra passive scalar transport, I get the following error: CG iterations made = 5001 using tolerance of 1e-09 (error = -nan, rhs_mag = 1) Fatal : Level 0 assertion violation Exceeded maximum number of iterations However, if I repeat the simulation without the passive scalar transport, it proceeds correctly. Has anyone else encountered the same problem? If so, would you be able to assist me in resolving it? Thank you in advance for youe help, Antonio.

3 2

TKE dissipation rate estimation
by Alexander Schukmann 18 Dec '24

18 Dec '24

******************* This email originates from outside Imperial. Do not click on links and attachments unless you recognise the sender. If you trust the sender, add them to your safe senders list https://spam.ic.ac.uk/SpamConsole/Senders.aspx to disable email stamping for this address. ******************* Hello everyone, I would like to follow up on a post from Fri Apr 7 14:41:21 BST 2023 titled “Compute velocity gradient during runtime” that was made by another user. I am using the incompressible solver in Nektar++ 5.5.0 and need to calculate the Kolmogorov length scale for DNS analysis. For this, I rely on an estimation of the TKE dissipation rate which is defined using the fluctuation strain-rate tensor. In the post mentioned above, Ilteber suggested using the FieldConvert Checkpoints filter including the "gradient" module in order to extract field gradients. These could then be used to derive the fluctuation strain-rates and hence the dissipation in a post-processing step. However, in order to get a good estimate of the time-averaged dissipation, a sufficient amount of sampling points is needed. At this point, writing the whole domain seems quite excessive with regard to disk space. For this reason, I wonder whether there is maybe a clever way of combining the FieldConvert Checkpoints filter with e.g. the HistoryPoints filter in order to write only field gradients at specific points during runtime. I've seen some Nektar papers where the authors estimated the Kolmogorov length scale. Could anybody maybe comment on this? Thanks in advance! All the best Alex Sicher versendet mit [Proton Mail](https://proton.me/mail/home).

1 0

Re: [Nektar-users] Noise forcing
by Daniel Garcia 18 Dec '24

18 Dec '24

In the case of a body force, you have to create the body force inside the "forcing" flag outside "conditions", then make a function for it inside "conditions". Have you tried to put the "noise forcing" flag outside the "conditions"? I think that the place is there. If that doesn't work, I can share with you how to set the body force with a random number to trigger turbulence, that would work for your case, I guess. Anyway, check if there is an example/tutorial using the noise forcing method and check its xml file. Em qua., 18 de dez. de 2024 às 10:46, Alexander Schukmann < alexander.schukmann(a)protonmail.com> escreveu: > Hello Daniel, > > thank you for your suggestions and sorry for the late reply. The user > guide doesn't mention any additional flag to be set for the noise forcing, > but I am not sure whether it is implemented for the incompressible solver. > How would I look that up? > > Thanks again and all the best > Alex > > Sicher versendet mit Proton Mail <https://proton.me/mail/home>. > > Daniel Garcia <ribeiro.danielg(a)gmail.com> schrieb am Montag, 16. Dezember > 2024 um 19:10: > > Hello Alexander. > > I was not aware of this forcing method. However, towards triggering > transition, I've successfully used the random number generator before. To > do so, you would use the body force as the forcing type and then create the > function based on your desired body force. Perhaps they work in a similar > way. In any case, check if that "Noise" forcing type does not require > another flag on the XML file or if it's implemented using your adopted > solver/set up. > > Cheers, > > Daniel. > > Em seg., 16 de dez. de 2024 às 16:35, Alexander Schukmann < > alexander.schukmann(a)protonmail.com> escreveu: > >> This email from alexander.schukmann(a)protonmail.com originates from >> outside Imperial. Do not click on links and attachments unless you >> recognise the sender. If you trust the sender, add them to your safe >> senders list <https://spam.ic.ac.uk/SpamConsole/Senders.aspx> to disable >> email stamping for this address. >> >> >> Hello everyone, >> >> I would like to follow up on a post from Wed Oct 13 02:46:37 BST 2021 >> titled “Noise Force Specification” that was made by another user. I am >> using the incompressible solver in Nektar++ 5.5.0 and trying to trigger the >> transition of a flow between two parallel plates (2.5D case) using noise >> forcing. However, it seems that the forcing is not activated at all, since >> the velocity component along the homogeneous spatial direction is zero up >> to machine epsilon. Therefore, no superposed disturbance is recognizable. >> >> I use the following lines within the CONDITIONS section of my session >> file: >>  >> <FORCING> >> <FORCE TYPE="Noise"> >> <WHITENOISE> 0.005 </WHITENOISE> >> <UPDATEFREQ> 10 </UPDATEFREQ> >> <NSTEPS> 5000 </NSTEPS> >> </FORCE> >> </FORCING> >> >> The value for the WHITENOISE parameter was also used by Stan back then. I >> tried varying it, with no effect. Does anyone know whether this is this the >> variance or the amplitude? >> >> Back then, the user wrote: *I tried your suggestions, but came up with >> the same results (a laminar flow profile without any "noise").* >> >> However, there doesn't seem to be a reply to that. It looks like the >> problem still exists. >> >> All the best, >> Alex >> >> >> >> _______________________________________________ >> Nektar-users mailing list >> Nektar-users(a)imperial.ac.uk >> https://mailman.ic.ac.uk/mailman/listinfo/nektar-users >> > >

1 0

Questions about Nektar++
by Gao 18 Dec '24

18 Dec '24

******************* This email originates from outside Imperial. Do not click on links and attachments unless you recognise the sender. If you trust the sender, add them to your safe senders list https://spam.ic.ac.uk/SpamConsole/Senders.aspx to disable email stamping for this address. ******************* Hello, everyone I have two questions about using Nektar++. First, When I change the initial value to the std file of the previous time step, then the instantaneous result (5.chk, 6.chk...) can come out, but the averge field result didn't appear. Is this normal? And the average field file is output only when the calculation is complete, is the mean field file at this point the result of a continuation run? Second, I am trying to monitor aerodynamic forces in the CompressibleFlowSolver. But it seems that the FilterAeroForces supports only the incompressible solver. Is there any other way to get the aerodynamic force over time? Thanks a lot. With best regards

2 4

Noise forcing
by Alexander Schukmann 16 Dec '24

16 Dec '24

******************* This email originates from outside Imperial. Do not click on links and attachments unless you recognise the sender. If you trust the sender, add them to your safe senders list https://spam.ic.ac.uk/SpamConsole/Senders.aspx to disable email stamping for this address. ******************* Hello everyone, I would like to follow up on a post from Wed Oct 13 02:46:37 BST 2021 titled “Noise Force Specification” that was made by another user. I am using the incompressible solver in Nektar++ 5.5.0 and trying to trigger the transition of a flow between two parallel plates (2.5D case) using noise forcing. However, it seems that the forcing is not activated at all, since the velocity component along the homogeneous spatial direction is zero up to machine epsilon. Therefore, no superposed disturbance is recognizable. I use the following lines within the CONDITIONS section of my session file:  <FORCING> <FORCE TYPE="Noise"> <WHITENOISE> 0.005 </WHITENOISE> <UPDATEFREQ> 10 </UPDATEFREQ> <NSTEPS> 5000 </NSTEPS> </FORCE> </FORCING> The value for the WHITENOISE parameter was also used by Stan back then. I tried varying it, with no effect. Does anyone know whether this is this the variance or the amplitude? Back then, the user wrote: I tried your suggestions, but came up with the same results (a laminar flow profile without any "noise"). However, theredoesn't seem to be a reply to that. It looks like the problem still exists. All the best, Alex

1 0

Restarting the solution from previous time-step
by Nande, Ishan A. 28 Nov '24

28 Nov '24

******************* This email originates from outside Imperial. Do not click on links and attachments unless you recognise the sender. If you trust the sender, add them to your safe senders list https://spam.ic.ac.uk/SpamConsole/Senders.aspx to disable email stamping for this address. ******************* Hello, I am solving Linearized Euler Equations using the AcousticSolver on the cluster and I have encountered an issue. I am running initial calculations using a few time steps. The last time-step's field files are provided as an initial condition for the solution sampling phase of calculations. Let's say that the initial run's total simulation time is t_init and the same for next phase of simulations is t_sampling. These are the key-points of my setup : 1. In my initial run's session file the FinTime is set as t_init. For the sampling run, the FinTime is set as t_init + t_sampling 2. I am running the calculations in parallel and hence, the last time-step's folder( which has .chk in it's name) from the initial run is supplied to the InitialConditions field in the session of sampling run. 3. The time step for initial and sampling runs is the same. With the previously outlined setup, I have found that while the sampling phase of simulation run's for the desired total time of t_init + t_sampling , it gives the following error: Fatal : Level 0 assertion violation m_timestep < 0 I have replicated the exact same setup on one of the example cases and it results in the same error. What does this error mean ? Can someone please comment on what could be going wrong ? Regards, Ishan PGR

2 2

Element u_11 is 0 from dsptrf
by Alexander Schukmann 18 Nov '24

18 Nov '24

******************* This email originates from outside Imperial. Do not click on links and attachments unless you recognise the sender. If you trust the sender, add them to your safe senders list https://spam.ic.ac.uk/SpamConsole/Senders.aspx to disable email stamping for this address. ******************* Hello all, when I run a 2.5D simulation using the incompressible NS solver the following error message appears with e.g. 480 CPUs after reading the initial state file, so the run immediately crashes: Fatal : Level 0 assertion violation ERROR: Element u_11 is 0 from dsptrf However, when switching to a lower CPU count such as 128 CPUs, it works. I have a region of local P-refinement in my domain, where the polynomial order is 7. Outside this region I coarsened the spatial resolution to P = 3 and suspect that it has something to do with this. Why does this happen? All the best Alex Sicher versendet mit [Proton Mail](https://proton.me/mail/home).

2 1