From antonios at mech.kth.se Fri May 15 11:22:38 2009 From: antonios at mech.kth.se (Antonios Monokrousos) Date: Fri, 15 May 2009 18:22:38 +0200 Subject: [Nek5000-users] perturbation mode In-Reply-To: References: <49886BEF.6010002@mech.kth.se> Message-ID: <4A0D96CE.20202@mech.kth.se> Hello! I have a question about the pertubation mode of the code. First I computed the base flow using the full non-linear solver and letting it run until I get a steady solution. This worked fine. Then I construct a perturbation which I add on the steady solution I got before (base flow) with a very small amplitude, around 0.1%. The solution, after I subtract the base flow looks good. (I'm comparing to a different code). Next I try to use perturbation mode since I want to use the exact linearised Navier-Stokes and the problems appear. I have the base flow and the perturbation in two different restart files, which I specify in .rea file like this: ------------------------------------------------------ 2 PRESOLVE/RESTART OPTIONS ***** meanflow.u ic_pert_field.fld ------------------------------------------------------ I activate the perturbation mode (option 31 in .rea). I damp both the solution of the base-flow and of the pertubation. I can see that the code is solving for the base-flow (which it shouldn't) but also has as initial condition the one that is supposed to be for the The simulation blows-up after a few time steps. I'm not sure I use the perturbation mode in the right manner. Thanks in advance for any feedback or help. Antonios -- --------------------------------------------------------------------------------------------- Antonios Monokrousos Department of Mechanics email: antonios at mech.kth.se Royal Institute of Technology (KTH) phone: +46 8 790 68 76 Osquars Backe 18 telefax: +46 8 796 98 50 SE-100 44 Stockholm, SWEDEN From fischer at mcs.anl.gov Fri May 15 11:50:39 2009 From: fischer at mcs.anl.gov (Paul Fischer) Date: Fri, 15 May 2009 11:50:39 -0500 (CDT) Subject: [Nek5000-users] perturbation mode In-Reply-To: <4A0D96CE.20202@mech.kth.se> References: <49886BEF.6010002@mech.kth.se> <4A0D96CE.20202@mech.kth.se> Message-ID: Hi Antonios, I'll look into this. One key thing is that the perturbation code is designed to solve for perturbations about an evolving base flow, so that one can compute Lyapunov exponents.... You should be able to override this by continually resetting the base flow to the desired steady state base flow. We could also likely modify the code to work in the mode that you desire (i.e., by simply turning off the evolution of the base flow). Paul On Fri, 15 May 2009, Antonios Monokrousos wrote: > Hello! > I have a question about the pertubation mode of the code. > > First I computed the base flow using the full non-linear solver and > letting it run until I get a steady solution. This worked fine. > Then I construct a perturbation which I add on the steady solution I got > before (base flow) with a very small amplitude, around 0.1%. The > solution, after I subtract the base flow looks good. (I'm comparing to a > different code). > > Next I try to use perturbation mode since I want to use the exact > linearised Navier-Stokes and the problems appear. > I have the base flow and the perturbation in two different restart > files, which I specify in .rea file like this: > ------------------------------------------------------ > 2 PRESOLVE/RESTART OPTIONS ***** > meanflow.u > ic_pert_field.fld > ------------------------------------------------------ > I activate the perturbation mode (option 31 in .rea). > > I damp both the solution of the base-flow and of the pertubation. I can > see that the code is solving for the base-flow (which it shouldn't) but > also has as initial condition the one that is supposed to be for the > The simulation blows-up after a few time steps. > > I'm not sure I use the perturbation mode in the right manner. > Thanks in advance for any feedback or help. > Antonios > > > -- > --------------------------------------------------------------------------------------------- > Antonios Monokrousos > Department of Mechanics email: antonios at mech.kth.se > Royal Institute of Technology (KTH) phone: +46 8 790 68 76 > Osquars Backe 18 telefax: +46 8 796 98 50 > SE-100 44 Stockholm, SWEDEN > _______________________________________________ > Nek5000-users mailing list > Nek5000-users at lists.mcs.anl.gov > https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users > From antonios at mech.kth.se Mon May 18 02:49:09 2009 From: antonios at mech.kth.se (Antonios Monokrousos) Date: Mon, 18 May 2009 09:49:09 +0200 Subject: [Nek5000-users] perturbation mode In-Reply-To: References: <49886BEF.6010002@mech.kth.se> <4A0D96CE.20202@mech.kth.se> Message-ID: <4A1112F5.7090503@mech.kth.se> Hi Paul, Thank you for you answer. As I see it the best it to have a switch to deactivate the solver for the base flow. This would make the computation cheaper. Is that implemented? Can one just add an if statement in the nek_solve subroutine of the form: ---------------------------------------- if(jp.ne.0) then call nek_advance call prepost (.false.,'his') ... ... end if ---------------------------------------- Would something like that work? Antonios Paul Fischer wrote: > > Hi Antonios, > > I'll look into this. > > One key thing is that the perturbation code is designed > to solve for perturbations about an evolving base flow, > so that one can compute Lyapunov exponents.... > > You should be able to override this by continually resetting > the base flow to the desired steady state base flow. > > We could also likely modify the code to work in the mode > that you desire (i.e., by simply turning off the evolution > of the base flow). > > Paul > > > On Fri, 15 May 2009, Antonios Monokrousos wrote: > >> Hello! >> I have a question about the pertubation mode of the code. >> >> First I computed the base flow using the full non-linear solver and >> letting it run until I get a steady solution. This worked fine. >> Then I construct a perturbation which I add on the steady solution I got >> before (base flow) with a very small amplitude, around 0.1%. The >> solution, after I subtract the base flow looks good. (I'm comparing to a >> different code). >> >> Next I try to use perturbation mode since I want to use the exact >> linearised Navier-Stokes and the problems appear. >> I have the base flow and the perturbation in two different restart >> files, which I specify in .rea file like this: >> ------------------------------------------------------ >> 2 PRESOLVE/RESTART OPTIONS ***** >> meanflow.u >> ic_pert_field.fld >> ------------------------------------------------------ >> I activate the perturbation mode (option 31 in .rea). >> >> I damp both the solution of the base-flow and of the pertubation. I can >> see that the code is solving for the base-flow (which it shouldn't) but >> also has as initial condition the one that is supposed to be for the >> The simulation blows-up after a few time steps. >> >> I'm not sure I use the perturbation mode in the right manner. >> Thanks in advance for any feedback or help. >> Antonios >> >> >> -- >> --------------------------------------------------------------------------------------------- >> >> Antonios Monokrousos >> Department of Mechanics email: antonios at mech.kth.se >> Royal Institute of Technology (KTH) phone: +46 8 790 68 76 >> Osquars Backe 18 telefax: +46 8 796 98 50 >> SE-100 44 Stockholm, SWEDEN >> _______________________________________________ >> Nek5000-users mailing list >> Nek5000-users at lists.mcs.anl.gov >> https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users >> -- --------------------------------------------------------------------------------------------- Antonios Monokrousos Department of Mechanics email: antonios at mech.kth.se Royal Institute of Technology (KTH) phone: +46 8 790 68 76 Osquars Backe 18 telefax: +46 8 796 98 50 SE-100 44 Stockholm, SWEDEN From fischer at mcs.anl.gov Mon May 18 06:47:24 2009 From: fischer at mcs.anl.gov (Paul Fischer) Date: Mon, 18 May 2009 06:47:24 -0500 (CDT) Subject: [Nek5000-users] perturbation mode In-Reply-To: <4A1112F5.7090503@mech.kth.se> References: <49886BEF.6010002@mech.kth.se> <4A0D96CE.20202@mech.kth.se> <4A1112F5.7090503@mech.kth.se> Message-ID: Hi Antonios, I think it would work if you were to simply comment out the call to "call fluid" in drive1.f and then run your perturbation simulation. Do you have a small test case on which you can try this? I'll also try to make a small test here, but unfortunately am very busy with visitors this week. Paul On Mon, 18 May 2009, Antonios Monokrousos wrote: > Hi Paul, > Thank you for you answer. As I see it the best it to have a switch to > deactivate the solver for the base flow. This would make the computation > cheaper. Is that implemented? Can one just add an if statement in the > nek_solve subroutine of the form: > > ---------------------------------------- > if(jp.ne.0) then > call nek_advance > call prepost (.false.,'his') > ... > ... > end if > ---------------------------------------- > > Would something like that work? > > Antonios > > > > Paul Fischer wrote: >> >> Hi Antonios, >> >> I'll look into this. >> >> One key thing is that the perturbation code is designed >> to solve for perturbations about an evolving base flow, >> so that one can compute Lyapunov exponents.... >> >> You should be able to override this by continually resetting >> the base flow to the desired steady state base flow. >> >> We could also likely modify the code to work in the mode >> that you desire (i.e., by simply turning off the evolution >> of the base flow). >> >> Paul >> >> >> On Fri, 15 May 2009, Antonios Monokrousos wrote: >> >>> Hello! >>> I have a question about the pertubation mode of the code. >>> >>> First I computed the base flow using the full non-linear solver and >>> letting it run until I get a steady solution. This worked fine. >>> Then I construct a perturbation which I add on the steady solution I got >>> before (base flow) with a very small amplitude, around 0.1%. The >>> solution, after I subtract the base flow looks good. (I'm comparing to a >>> different code). >>> >>> Next I try to use perturbation mode since I want to use the exact >>> linearised Navier-Stokes and the problems appear. >>> I have the base flow and the perturbation in two different restart >>> files, which I specify in .rea file like this: >>> ------------------------------------------------------ >>> 2 PRESOLVE/RESTART OPTIONS ***** >>> meanflow.u >>> ic_pert_field.fld >>> ------------------------------------------------------ >>> I activate the perturbation mode (option 31 in .rea). >>> >>> I damp both the solution of the base-flow and of the pertubation. I can >>> see that the code is solving for the base-flow (which it shouldn't) but >>> also has as initial condition the one that is supposed to be for the >>> The simulation blows-up after a few time steps. >>> >>> I'm not sure I use the perturbation mode in the right manner. >>> Thanks in advance for any feedback or help. >>> Antonios >>> >>> >>> -- >>> --------------------------------------------------------------------------------------------- >>> >>> Antonios Monokrousos >>> Department of Mechanics email: antonios at mech.kth.se >>> Royal Institute of Technology (KTH) phone: +46 8 790 68 76 >>> Osquars Backe 18 telefax: +46 8 796 98 50 >>> SE-100 44 Stockholm, SWEDEN >>> _______________________________________________ >>> Nek5000-users mailing list >>> Nek5000-users at lists.mcs.anl.gov >>> https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users >>> > > > -- > --------------------------------------------------------------------------------------------- > Antonios Monokrousos > Department of Mechanics email: antonios at mech.kth.se > Royal Institute of Technology (KTH) phone: +46 8 790 68 76 > Osquars Backe 18 telefax: +46 8 796 98 50 > SE-100 44 Stockholm, SWEDEN > From fischer at mcs.anl.gov Tue May 26 09:53:44 2009 From: fischer at mcs.anl.gov (Paul Fischer) Date: Tue, 26 May 2009 09:53:44 -0500 (CDT) Subject: [Nek5000-users] Nek5000 In-Reply-To: <43779.161.3.6.143.1243349237.squirrel@arcon.univ-st-etienne.fr> References: <56893.161.3.6.143.1241713100.squirrel@arcon.univ-st-etienne.fr> <36064.161.3.6.143.1242307789.squirrel@arcon.univ-st-etienne.fr> <59113.161.3.6.143.1243325180.squirrel@arcon.univ-st-etienne.fr> <43779.161.3.6.143.1243349237.squirrel@arcon.univ-st-etienne.fr> Message-ID: Dear Emmanuel, This doesn't make sense to me... it should all come as part of the svn repo. The key is to use the makenek script, which will do the right thing by setting the correct flags. What compiler is being used? I will certainly be attending the conference in honor of Michel and would be happy to visit you, schedule permitting. Best regards, Paul On Tue, 26 May 2009, emmanuel.leriche at univ-st-etienne.fr wrote: > dear Paul, > > at the moment, we do not have yet access to BGP. > > Alex is trying to compile on its LINUX laptop; it seems that every time > there is missing routines, that Alex could find on your website, and add > one by one... > > I guess that you will be at Michel's conference mid-February 2010 ? if > yes, would you like that I arrange an invitation to give a seminar at the > Center for Health Engineering of the school of Mines in Saint-Etienne, > http://www.emse.fr/spip/-CIS-.html?lang=en ? > St-Etienne is 300km far from Lausanne, let's say 3H30 by train. > > regards, > Emmanuel > >> Dear Emmanuel, >> >> Is he having difficulties even on BG/P ?? >> >> Paul > > >