open TELEMAC-MASCARET The mathematically superior suite of solvers

Hi,

I work with telemac3d in version v6p2.
I work with 2 configurations: one run with 19 plans et one run with 25 plans. The difference between these configurations is the plans repartition. With the first configuration, my simulation is ok and my results are good. With the second repartition, my run crashes before two days of simulation with this error:
RETOUR DE CVDF3D POUR V
APPEL DE CVDF3D POUR W
TRID3D : SYSTEME NON DEFINI
PRECONDITIONNEMENT 17 IMPOSSIBLE

PLANTE : ARRET DU PROGRAMME APRES ERREUR
SORTIE DE PVM : APPEL DE P_EXIT

SORTIE DE MPI

In the first configuration:
ZPLANE%R(1)=-9.0D0
ZPLANE%R(2)=-8.5D0
ZPLANE%R(3)=-8.0D0
ZPLANE%R(4)=-7.5D0
ZPLANE%R(5)=-7.0D0
ZPLANE%R(6)=-6.5D0
ZPLANE%R(7)=-6.0D0
ZPLANE%R(8)=-5.5D0
ZPLANE%R(9)=-5.0D0
ZPLANE%R(10)=-4.5D0
ZPLANE%R(11)=-4.0D0
ZPLANE%R(12)=-3.5D0
ZPLANE%R(13)=-3.0D0
ZPLANE%R(14)=-2.5D0
ZPLANE%R(15)=-2.0D0
ZPLANE%R(16)=-1.5D0
ZPLANE%R(17)=-1.0D0
ZPLANE%R(18)=-0.5D0
ZPLANE%R(19)=-0.0D0

In the second configuration:
ZPLANE%R(1)=-9.0D0
ZPLANE%R(2)=-8.7D0
ZPLANE%R(3)=-8.5D0
ZPLANE%R(4)=-8.2D0
ZPLANE%R(5)=-7.8D0
ZPLANE%R(6)=-7.4D0
ZPLANE%R(7)=-7.2D0
ZPLANE%R(8)=-6.8D0
ZPLANE%R(9)=-6.4D0
ZPLANE%R(10)=-6.0D0
ZPLANE%R(11)=-5.6D0
ZPLANE%R(12)=-5.2D0
ZPLANE%R(13)=-4.8D0
ZPLANE%R(14)=-4.4D0
ZPLANE%R(15)=-4.0D0
ZPLANE%R(16)=-3.6D0
ZPLANE%R(17)=-3.2D0
ZPLANE%R(18)=-2.8D0
ZPLANE%R(19)=-2.4D0
ZPLANE%R(20)=-2.0D0
ZPLANE%R(21)=-1.6D0
ZPLANE%R(22)=-1.2D0
ZPLANE%R(23)=-0.8D0
ZPLANE%R(24)=-0.4D0
ZPLANE%R(25)=-0.0D0

Someone knows this error? The modification of the plans repartition can explain this error?
In need some help

Thanks in advance
Florian

I precise that i must have more than 25 plans in my configuration

Hello,

Preconditioning 17 is an exact solution on every vertical, so it could be that such a linear system on the vertical has no solution. You can try another preconditioning, like 2, instead.

Regards,

JMH

Jean-michel,

In my steering files, i don't define preconditioning 17.
I put:

PRECONDITIONNEMENT POUR LA DIFFUSION DES TRACEURS =34;34

PRECONDITIONNEMENT POUR LA DIFFUSION DES VITESSES =34

PRECONDITIONNEMENT POUR PPE =34

PRECONDITIONNEMENT POUR LA DIFFUSION DU K-EPSILON =34

PRECONDITIONNEMENT POUR LA PROPAGATION =2

What is the preconditioning I need to change? all?

Florian

Hello,

34 is 2*17, which means that you use both 2 and 17, so change all 34 into 2...

Preconditioning 17 is just to speed up.

Regards,

JMH

Hi Jean Michel

I changed the preconditioning into 2.
But now my run crash with this error:

RETOUR DE SOURCE
APPEL DE MESH_PROP
RETOUR DE MESH_PROP
APPEL DE CVDF3D POUR U
MURD3D: ITERATION NO. REACHED 100 , STOP.
ALFA = 1.39374016659295913E-002
GUILTY POINT = 1
PLANTE : ARRET DU PROGRAMME APRES ERREUR
SORTIE DE PVM : APPEL DE P_EXIT

SORTIE DE MPI

I search on the forum and i saw that this type of error is caused by the SCHEME FOR ADVECTION OF VELOCITIES. I put 1 and in change into 4. What is your opinion about this error?

Thanks

Florian

Hello,

Yes, this was probable, the run is crashing and preconditioning 17 was the first to see it, so you delayed a bit the crash. You should now analyse where and why it crashes, it is probably due to large velocities somewhere, for some reason.

Regards,

JMH

I work on 32 processors. The only solution to see where is the hig velocities is to work in sequential mode?

I must analyse the NUZ parameter or the UV parameter?

Hello,

You can rerun the computation until shortly before the crash, and then look normally at the results, look at the velocities, which are normally U, V and W.

Regards,

JMH