Manipulate 3DH1 chk/fld file in Fourier space
Dear all, I have run a 3DH1 simulation and I need to set all Fourier coefficients to zero except those corresponding to a fundamental wavelength and its harmonics (e.g. keep only modes 0, +-5, +-10, etc and set all the rest to zero). I have the intuition that some combination of FieldConvert options can accomplish exactly this, but I am unsure of what is the best (or some) way to proceed. My guess is that the "homplane:wavespace:planeid=?" can be used to extract the individual Fourier modes, but I am unsure as to how the Fourier modes are sorted and which ones must be extracted to recover a chosen fundamental wavelength and its harmonics (both real and imaginary parts, or modulus and argument, or however the Fourier modes are stored). The only one I'm pretty sure of is mode 0 or mean mode. Once I get the individual fld files for each one of the modes, I was intending to use the "addfld:fromfld=mode1.fld:scale=1 mode1.xml mode2.fld newfile.fld" sequentially to build a solution from the chosen modes. The issue here is that the fields extracted with "homplane" seem to be 2D and would therefore need to be recast in 3DH1 form before proceeding to "addfld". How can this be done? Thanks in advance for your help Fer Mellibovksy
Hi Fer, I think your approach with first extracting modes and than summing it up is a good strategy. I have not done enything exactly as you describe, but I did similar stuff. So I am not 100% sure this will work, but here you go: You might need to process the result of homeplane before summing it up. I would try the following combination: 1. Use homplane to extract the modes I need and meanmode to get the mean mode from the flow. 2. Use each of the results of homeplane as an initial conditions for a simulations with zero time steps and set appropriate length LZ for each, using HomModesZ=4. This way you should get separate modes stretched to their respective wavelengths. You might need to use Linearized NS by setting <I PROPERTY="EvolutionOperator" VALUE="Direct"/> for modes other than mean (with mean mode you might need to set HomModesZ=2). Also make sure that what you get is this mode only. Simplest is to visualize it with Paraview or Tecplot. 3. Use homstretch module to reproduce each of the obtained modes the right number of times (mode that is half-length 2 etc.) such that all your results have the same length. 4. Sum it all up. Also have look at the Modal Energy filter. It might be helpful in understanding which modes are present. Also, once you run your simulation with only specific modes you should see energies of those modes higher than the rest. Regarding the numbering of modes. Have a look at page 48 of the manual. The modes are numbered from 0 up to N (not -N to N). So setting HomModesZ=2 gives you the mean mode, HomModesZ=4 (HomModesZ needs to be even) gives you first cos and sin, and so on. Hope this works, Cheers! Stan On pią, 2019-09-27 at 09:21 +0200, F Mellibovsky wrote:
Dear all,
I have run a 3DH1 simulation and I need to set all Fourier coefficients to zero except those corresponding to a fundamental wavelength and its harmonics (e.g. keep only modes 0, +-5, +-10, etc and set all the rest to zero). I have the intuition that some combination of FieldConvert options can accomplish exactly this, but I am unsure of what is the best (or some) way to proceed.
My guess is that the "homplane:wavespace:planeid=?" can be used to extract the individual Fourier modes, but I am unsure as to how the Fourier modes are sorted and which ones must be extracted to recover a chosen fundamental wavelength and its harmonics (both real and imaginary parts, or modulus and argument, or however the Fourier modes are stored). The only one I'm pretty sure of is mode 0 or mean mode.
Once I get the individual fld files for each one of the modes, I was intending to use the "addfld:fromfld=mode1.fld:scale=1 mode1.xml mode2.fld newfile.fld" sequentially to build a solution from the chosen modes. The issue here is that the fields extracted with "homplane" seem to be 2D and would therefore need to be recast in 3DH1 form before proceeding to "addfld". How can this be done?
Thanks in advance for your help
Fer Mellibovksy
_______________________________________________ Nektar-users mailing list Nektar-users@imperial.ac.uk https://mailman.ic.ac.uk/mailman/listinfo/nektar-users
Thanks, Stan, I'm struggling at a couple of steps in the procedure 1) Extract mean mode with "Fieldconvert -m meanmode". The results is a 2D field 2) Extract the fundamental mode (mz) and its harmonics (n*mz) using "Fieldconvert -m homplane:planeid=2*(n*mz)" for the "cos" expansion "Fieldconvert -m homplane:planeid=2*(n*mz)+1" for the "sin" expansion The results are again 2D fields 3) Run the time-stepper using the individual modes as initial condition with number of steps set to 0, lz=Lz/mz (Lz is the spanwise length of the original file) and HomModesZ=4 (2 for the mean mode). Here I have the *first problem*: Does this not simply take the initial condition as a mean mode? How do I get the the sine and cosine spanwise modulations? Multiplying the input field by cos(2pi*z/lz) and sin(2pi*z/lz) for cos and sin components, respectively? Is it not possible to read multiple files in the InitialConditions section and create the desired reconstructed field by doing the cos and sin multiplications to the respective modes and adding up, all in just one timestepper run? *Second question* why do I need to use the linearised NS solver? Will the nonlinear term of the full NS solver pollute time step zero? 4) Use "Fieldconvert -m homstretch:factor=Lz/lz:ouput-points-hom-z=Mz". 5) Sum everything up with "FieldConvert -m addfld:fromfld=file1.fld:scale=1 ..." On 27/09/2019 10:27, Stanisław Gepner wrote:
Hi Fer,
I think your approach with first extracting modes and than summing it up is a good strategy. I have not done enything exactly as you describe, but I did similar stuff. So I am not 100% sure this will work, but here you go:
You might need to process the result of homeplane before summing it up. I would try the following combination:
1. Use homplane to extract the modes I need and meanmode to get the mean mode from the flow. 2. Use each of the results of homeplane as an initial conditions for a simulations with zero time steps and set appropriate length LZ for each, using HomModesZ=4. This way you should get separate modes stretched to their respective wavelengths. You might need to use Linearized NS by setting <I PROPERTY="EvolutionOperator" VALUE="Direct"/> for modes other than mean (with mean mode you might need to set HomModesZ=2). Also make sure that what you get is this mode only. Simplest is to visualize it with Paraview or Tecplot. 3. Use homstretch module to reproduce each of the obtained modes the right number of times (mode that is half-length 2 etc.) such that all your results have the same length. 4. Sum it all up.
Also have look at the Modal Energy filter. It might be helpful in understanding which modes are present. Also, once you run your simulation with only specific modes you should see energies of those modes higher than the rest.
Regarding the numbering of modes. Have a look at page 48 of the manual. The modes are numbered from 0 up to N (not -N to N). So setting HomModesZ=2 gives you the mean mode, HomModesZ=4 (HomModesZ needs to be even) gives you first cos and sin, and so on.
Hope this works, Cheers! Stan
On pią, 2019-09-27 at 09:21 +0200, F Mellibovsky wrote:
Dear all,
I have run a 3DH1 simulation and I need to set all Fourier coefficients to zero except those corresponding to a fundamental wavelength and its harmonics (e.g. keep only modes 0, +-5, +-10, etc and set all the rest to zero). I have the intuition that some combination of FieldConvert options can accomplish exactly this, but I am unsure of what is the best (or some) way to proceed.
My guess is that the "homplane:wavespace:planeid=?" can be used to extract the individual Fourier modes, but I am unsure as to how the Fourier modes are sorted and which ones must be extracted to recover a chosen fundamental wavelength and its harmonics (both real and imaginary parts, or modulus and argument, or however the Fourier modes are stored). The only one I'm pretty sure of is mode 0 or mean mode.
Once I get the individual fld files for each one of the modes, I was intending to use the "addfld:fromfld=mode1.fld:scale=1 mode1.xml mode2.fld newfile.fld" sequentially to build a solution from the chosen modes. The issue here is that the fields extracted with "homplane" seem to be 2D and would therefore need to be recast in 3DH1 form before proceeding to "addfld". How can this be done?
Thanks in advance for your help
Fer Mellibovksy
_______________________________________________ Nektar-users mailing list Nektar-users@imperial.ac.uk https://mailman.ic.ac.uk/mailman/listinfo/nektar-users
Hi Fer, As I wrote, I have not exactly tried this solution myself. On wto, 2019-10-01 at 09:15 +0200, Fer Mellibovsky wrote:
Thanks, Stan,
I'm struggling at a couple of steps in the procedure
1) Extract mean mode with "Fieldconvert -m meanmode". The results is a 2D field 2) Extract the fundamental mode (mz) and its harmonics (n*mz) using "Fieldconvert -m homplane:planeid=2*(n*mz)" for the "cos" expansion "Fieldconvert -m homplane:planeid=2*(n*mz)+1" for the "sin" expansion The results are again 2D fields 3) Run the time-stepper using the individual modes as initial condition with number of steps set to 0, lz=Lz/mz (Lz is the spanwise length of the original file) and HomModesZ=4 (2 for the mean mode). Here I have the first problem: Does this not simply take the initial condition as a mean mode? Well, seems it does. But you can try with: <I PROPERTY="ModeType" VALUE= "SingleMode" />
How do I get the the sine and cosine spanwise modulations? Multiplying the input field by cos(2pi*z/lz) and sin(2pi*z/lz) for cos and sin components, respectively? Not sure. This part is something I have not tried. Sorry. Possibly fieldfromstring ?
Is it not possible to read multiple files in the InitialConditions section and create the desired reconstructed field by doing the cos and sin multiplications to the respective modes and adding up, all in just one timestepper run? Not that I know of. You might write your own processing routine for FieldConvert that would do this. It is not hard, possibly you could just copy and modify something that is there already. Second question why do I need to use the linearised NS solver? Will the nonlinear term of the full NS solver pollute time step zero? I was thinking that this will let you prevent the initial condition from being the mean mode. But this is apparently achievable with SingleMode. So seems you can use whichever suits you.
Cheers, Stan
4) Use "Fieldconvert -m homstretch:factor=Lz/lz:ouput-points-hom- z=Mz". 5) Sum everything up with "FieldConvert -m addfld:fromfld=file1.fld:scale=1 ..."
On 27/09/2019 10:27, Stanisław Gepner wrote:
Hi Fer,
I think your approach with first extracting modes and than summing it up is a good strategy. I have not done enything exactly as you describe, but I did similar stuff. So I am not 100% sure this will work, but here you go:
You might need to process the result of homeplane before summing it up. I would try the following combination:
1. Use homplane to extract the modes I need and meanmode to get the mean mode from the flow. 2. Use each of the results of homeplane as an initial conditions for a simulations with zero time steps and set appropriate length LZ for each, using HomModesZ=4. This way you should get separate modes stretched to their respective wavelengths. You might need to use Linearized NS by setting <I PROPERTY="EvolutionOperator" VALUE="Direct"/> for modes other than mean (with mean mode you might need to set HomModesZ=2). Also make sure that what you get is this mode only. Simplest is to visualize it with Paraview or Tecplot. 3. Use homstretch module to reproduce each of the obtained modes the right number of times (mode that is half-length 2 etc.) such that all your results have the same length. 4. Sum it all up.
Also have look at the Modal Energy filter. It might be helpful in understanding which modes are present. Also, once you run your simulation with only specific modes you should see energies of those modes higher than the rest.
Regarding the numbering of modes. Have a look at page 48 of the manual. The modes are numbered from 0 up to N (not -N to N). So setting HomModesZ=2 gives you the mean mode, HomModesZ=4 (HomModesZ needs to be even) gives you first cos and sin, and so on.
Hope this works, Cheers! Stan
On pią, 2019-09-27 at 09:21 +0200, F Mellibovsky wrote:
Dear all,
I have run a 3DH1 simulation and I need to set all Fourier coefficients to zero except those corresponding to a fundamental wavelength and its harmonics (e.g. keep only modes 0, +-5, +-10, etc and set all the rest to zero). I have the intuition that some combination of FieldConvert options can accomplish exactly this, but I am unsure of what is the best (or some) way to proceed.
My guess is that the "homplane:wavespace:planeid=?" can be used to extract the individual Fourier modes, but I am unsure as to how the Fourier modes are sorted and which ones must be extracted to recover a chosen fundamental wavelength and its harmonics (both real and imaginary parts, or modulus and argument, or however the Fourier modes are stored). The only one I'm pretty sure of is mode 0 or mean mode.
Once I get the individual fld files for each one of the modes, I was intending to use the "addfld:fromfld=mode1.fld:scale=1 mode1.xml mode2.fld newfile.fld" sequentially to build a solution from the chosen modes. The issue here is that the fields extracted with "homplane" seem to be 2D and would therefore need to be recast in 3DH1 form before proceeding to "addfld". How can this be done?
Thanks in advance for your help
Fer Mellibovksy
_______________________________________________ Nektar-users mailing list Nektar-users@imperial.ac.uk https://mailman.ic.ac.uk/mailman/listinfo/nektar-users
******************* 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 sent a post some time ago requesting for help on how to set a few Fourier modes to zero on a 3DH1 fld file using FieldConvert (and possibly also the time stepper). I have been trying repeatedly but I haven't succeeded. Has anyone done something similar? My impression is that it shouldn't be too difficult to modify FieldConvert to do so. As a matter of fact, the addfields module can scale an fld file so I guess it could easily set scale by zero only the desired Fourier modes. Any help would be welcome. Cheers On 01/10/2019 12:46, Stanisław Gepner wrote:
Hi Fer,
As I wrote, I have not exactly tried this solution myself.
On wto, 2019-10-01 at 09:15 +0200, Fer Mellibovsky wrote:
Thanks, Stan,
I'm struggling at a couple of steps in the procedure
1) Extract mean mode with "Fieldconvert -m meanmode". The results is a 2D field 2) Extract the fundamental mode (mz) and its harmonics (n*mz) using "Fieldconvert -m homplane:planeid=2*(n*mz)" for the "cos" expansion "Fieldconvert -m homplane:planeid=2*(n*mz)+1" for the "sin" expansion The results are again 2D fields 3) Run the time-stepper using the individual modes as initial condition with number of steps set to 0, lz=Lz/mz (Lz is the spanwise length of the original file) and HomModesZ=4 (2 for the mean mode). Here I have the first problem: Does this not simply take the initial condition as a mean mode? Well, seems it does. But you can try with: <I PROPERTY="ModeType" VALUE= "SingleMode" />
How do I get the the sine and cosine spanwise modulations? Multiplying the input field by cos(2pi*z/lz) and sin(2pi*z/lz) for cos and sin components, respectively? Not sure. This part is something I have not tried. Sorry. Possibly fieldfromstring ?
Is it not possible to read multiple files in the InitialConditions section and create the desired reconstructed field by doing the cos and sin multiplications to the respective modes and adding up, all in just one timestepper run? Not that I know of. You might write your own processing routine for FieldConvert that would do this. It is not hard, possibly you could just copy and modify something that is there already. Second question why do I need to use the linearised NS solver? Will the nonlinear term of the full NS solver pollute time step zero? I was thinking that this will let you prevent the initial condition from being the mean mode. But this is apparently achievable with SingleMode. So seems you can use whichever suits you.
Cheers, Stan
4) Use "Fieldconvert -m homstretch:factor=Lz/lz:ouput-points-hom- z=Mz". 5) Sum everything up with "FieldConvert -m addfld:fromfld=file1.fld:scale=1 ..."
On 27/09/2019 10:27, Stanisław Gepner wrote:
Hi Fer,
I think your approach with first extracting modes and than summing it up is a good strategy. I have not done enything exactly as you describe, but I did similar stuff. So I am not 100% sure this will work, but here you go:
You might need to process the result of homeplane before summing it up. I would try the following combination:
1. Use homplane to extract the modes I need and meanmode to get the mean mode from the flow. 2. Use each of the results of homeplane as an initial conditions for a simulations with zero time steps and set appropriate length LZ for each, using HomModesZ=4. This way you should get separate modes stretched to their respective wavelengths. You might need to use Linearized NS by setting <I PROPERTY="EvolutionOperator" VALUE="Direct"/> for modes other than mean (with mean mode you might need to set HomModesZ=2). Also make sure that what you get is this mode only. Simplest is to visualize it with Paraview or Tecplot. 3. Use homstretch module to reproduce each of the obtained modes the right number of times (mode that is half-length 2 etc.) such that all your results have the same length. 4. Sum it all up.
Also have look at the Modal Energy filter. It might be helpful in understanding which modes are present. Also, once you run your simulation with only specific modes you should see energies of those modes higher than the rest.
Regarding the numbering of modes. Have a look at page 48 of the manual. The modes are numbered from 0 up to N (not -N to N). So setting HomModesZ=2 gives you the mean mode, HomModesZ=4 (HomModesZ needs to be even) gives you first cos and sin, and so on.
Hope this works, Cheers! Stan
On pią, 2019-09-27 at 09:21 +0200, F Mellibovsky wrote:
Dear all,
I have run a 3DH1 simulation and I need to set all Fourier coefficients to zero except those corresponding to a fundamental wavelength and its harmonics (e.g. keep only modes 0, +-5, +-10, etc and set all the rest to zero). I have the intuition that some combination of FieldConvert options can accomplish exactly this, but I am unsure of what is the best (or some) way to proceed.
My guess is that the "homplane:wavespace:planeid=?" can be used to extract the individual Fourier modes, but I am unsure as to how the Fourier modes are sorted and which ones must be extracted to recover a chosen fundamental wavelength and its harmonics (both real and imaginary parts, or modulus and argument, or however the Fourier modes are stored). The only one I'm pretty sure of is mode 0 or mean mode.
Once I get the individual fld files for each one of the modes, I was intending to use the "addfld:fromfld=mode1.fld:scale=1 mode1.xml mode2.fld newfile.fld" sequentially to build a solution from the chosen modes. The issue here is that the fields extracted with "homplane" seem to be 2D and would therefore need to be recast in 3DH1 form before proceeding to "addfld". How can this be done?
Thanks in advance for your help
Fer Mellibovksy
_______________________________________________ Nektar-users mailing list Nektar-users@imperial.ac.uk https://mailman.ic.ac.uk/mailman/listinfo/nektar-users
Hi Fer, Sorry I have been completely behind on emails. I am not aware that this feature has been implemented still. When you have a 3DH1 field there is a member variable called m_planes which is an array of expansion list 2d. So in principle zeroing some of these planes coefficient space would do what you want. There is not obviously a clean place to put this so ideally it would be put into a new module with FieldConvert. Cheers, Spencer. On 28 Feb 2020, at 10:15, F Mellibovsky <fernando.mellibovsky@upc.edu<mailto:fernando.mellibovsky@upc.edu>> wrote: ******************* 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 sent a post some time ago requesting for help on how to set a few Fourier modes to zero on a 3DH1 fld file using FieldConvert (and possibly also the time stepper). I have been trying repeatedly but I haven't succeeded. Has anyone done something similar? My impression is that it shouldn't be too difficult to modify FieldConvert to do so. As a matter of fact, the addfields module can scale an fld file so I guess it could easily set scale by zero only the desired Fourier modes. Any help would be welcome. Cheers On 01/10/2019 12:46, Stanisław Gepner wrote: Hi Fer, As I wrote, I have not exactly tried this solution myself. On wto, 2019-10-01 at 09:15 +0200, Fer Mellibovsky wrote: Thanks, Stan, I'm struggling at a couple of steps in the procedure 1) Extract mean mode with "Fieldconvert -m meanmode". The results is a 2D field 2) Extract the fundamental mode (mz) and its harmonics (n*mz) using "Fieldconvert -m homplane:planeid=2*(n*mz)" for the "cos" expansion "Fieldconvert -m homplane:planeid=2*(n*mz)+1" for the "sin" expansion The results are again 2D fields 3) Run the time-stepper using the individual modes as initial condition with number of steps set to 0, lz=Lz/mz (Lz is the spanwise length of the original file) and HomModesZ=4 (2 for the mean mode). Here I have the first problem: Does this not simply take the initial condition as a mean mode? Well, seems it does. But you can try with: <I PROPERTY="ModeType" VALUE= "SingleMode" /> How do I get the the sine and cosine spanwise modulations? Multiplying the input field by cos(2pi*z/lz) and sin(2pi*z/lz) for cos and sin components, respectively? Not sure. This part is something I have not tried. Sorry. Possibly fieldfromstring ? Is it not possible to read multiple files in the InitialConditions section and create the desired reconstructed field by doing the cos and sin multiplications to the respective modes and adding up, all in just one timestepper run? Not that I know of. You might write your own processing routine for FieldConvert that would do this. It is not hard, possibly you could just copy and modify something that is there already. Second question why do I need to use the linearised NS solver? Will the nonlinear term of the full NS solver pollute time step zero? I was thinking that this will let you prevent the initial condition from being the mean mode. But this is apparently achievable with SingleMode. So seems you can use whichever suits you. Cheers, Stan 4) Use "Fieldconvert -m homstretch:factor=Lz/lz:ouput-points-hom- z=Mz". 5) Sum everything up with "FieldConvert -m addfld:fromfld=file1.fld:scale=1 ..." On 27/09/2019 10:27, Stanisław Gepner wrote: Hi Fer, I think your approach with first extracting modes and than summing it up is a good strategy. I have not done enything exactly as you describe, but I did similar stuff. So I am not 100% sure this will work, but here you go: You might need to process the result of homeplane before summing it up. I would try the following combination: 1. Use homplane to extract the modes I need and meanmode to get the mean mode from the flow. 2. Use each of the results of homeplane as an initial conditions for a simulations with zero time steps and set appropriate length LZ for each, using HomModesZ=4. This way you should get separate modes stretched to their respective wavelengths. You might need to use Linearized NS by setting <I PROPERTY="EvolutionOperator" VALUE="Direct"/> for modes other than mean (with mean mode you might need to set HomModesZ=2). Also make sure that what you get is this mode only. Simplest is to visualize it with Paraview or Tecplot. 3. Use homstretch module to reproduce each of the obtained modes the right number of times (mode that is half-length 2 etc.) such that all your results have the same length. 4. Sum it all up. Also have look at the Modal Energy filter. It might be helpful in understanding which modes are present. Also, once you run your simulation with only specific modes you should see energies of those modes higher than the rest. Regarding the numbering of modes. Have a look at page 48 of the manual. The modes are numbered from 0 up to N (not -N to N). So setting HomModesZ=2 gives you the mean mode, HomModesZ=4 (HomModesZ needs to be even) gives you first cos and sin, and so on. Hope this works, Cheers! Stan On pią, 2019-09-27 at 09:21 +0200, F Mellibovsky wrote: Dear all, I have run a 3DH1 simulation and I need to set all Fourier coefficients to zero except those corresponding to a fundamental wavelength and its harmonics (e.g. keep only modes 0, +-5, +-10, etc and set all the rest to zero). I have the intuition that some combination of FieldConvert options can accomplish exactly this, but I am unsure of what is the best (or some) way to proceed. My guess is that the "homplane:wavespace:planeid=?" can be used to extract the individual Fourier modes, but I am unsure as to how the Fourier modes are sorted and which ones must be extracted to recover a chosen fundamental wavelength and its harmonics (both real and imaginary parts, or modulus and argument, or however the Fourier modes are stored). The only one I'm pretty sure of is mode 0 or mean mode. Once I get the individual fld files for each one of the modes, I was intending to use the "addfld:fromfld=mode1.fld:scale=1 mode1.xml mode2.fld newfile.fld" sequentially to build a solution from the chosen modes. The issue here is that the fields extracted with "homplane" seem to be 2D and would therefore need to be recast in 3DH1 form before proceeding to "addfld". How can this be done? Thanks in advance for your help Fer Mellibovksy _______________________________________________ Nektar-users mailing list Nektar-users@imperial.ac.uk<mailto:Nektar-users@imperial.ac.uk> https://mailman.ic.ac.uk/mailman/listinfo/nektar-users _______________________________________________ Nektar-users mailing list Nektar-users@imperial.ac.uk<mailto:Nektar-users@imperial.ac.uk> https://mailman.ic.ac.uk/mailman/listinfo/nektar-users Spencer Sherwin FREng, FRAeS Head of Aerodynamics Section, Director of Research Computing Service, Professor of Computational Fluid Mechanics, Department of Aeronautics, s.sherwin@imperial.ac.uk<mailto:s.sherwin@imperial.ac.uk> South Kensington Campus, Phone: +44 (0)20 7594 5052 Imperial College London, Fax: +44 (0)20 7594 1974 London, SW7 2AZ, UK http://www.imperial.ac.uk/people/s.sherwin/
participants (4)
-
F Mellibovsky
-
Fer Mellibovsky
-
Sherwin, Spencer J
-
Stanisław Gepner