Refinement of Mesh in FSI

[Arumugam ce19d044]

Dear All

I am working on a simulation of a shock tube problem using LSDYNA. Thin steel plates are placed in the shock tube and exposed to the shock loading. In the numerical simulation, the fluid mesh is 5 times larger than the steel plate element size, hence to make an actual account of the contact between the Fluid and Steel plate, *CONTROL_REFINE_ALE is used. But with the CONTROL_REFINE_ALE for the TYPE=2, mesh is not refined and for the Type = 0 and 5, mesh is refined and contact is not generated between the fluid and structure.

Please provide guidance on solving the issue.

Regards.

Arumugam

Ph.D Scholar 

Indian Institute of Technology Madras

India

[James Kennedy]

Dear Arumugam,

 

What version/Revision of LS-DYNA are you using?

 

I used the following

 

     |  Version : smp d R15                              |

     |  Revision: R15.0.2-1-g1e7cfa42b1                  |

 

With the job terminating with the following messages:

 

Beginning of keyword reader                                   03/05/25 11:44:16

 

*** Warning 10140 (KEY+140)

     Total of 2 empty *SET_SEGMENT

     (Check d3hsp for detail)

 

*** Warning 10302 (KEY+302)

     CHECKING MATERIAL INPUT Part ID= 4

     The default hourglass properties of the following

     ALE part are being modified to avoid unnecessary

     application of hourglass forces.

     Resetting coefficient qm=1.0e-6.

     PART ID 4 with

     MATERIAL ID 4 and

     EOS ID 0

     This is PART 3 in the order of input.

 

Memory required to process keyword     :     256810

Additional dynamic memory required     :    3346465

 

 

input of data is completed

 

initial kinetic energy = 0.00000000E+00

 

Memory required to begin solution      :     257K

Additional dynamically allocated memory:    3605K

                                   Total:    3861K

 

initialization completed

       1 t 0.0000E+00 dt 2.89E-07 flush i/o buffers            03/05/25 11:44:17

       1 t 0.0000E+00 dt 2.89E-07 write d3plot file            03/05/25 11:44:17

cpu time per zone cycle............         0 nanoseconds

average cpu time per zone cycle....         0 nanoseconds

average clock time per zone cycle..     16646 nanoseconds

 

estimated total cpu time          =         1 sec (       0 hrs  0 mins)

estimated cpu time to complete    =         0 sec (       0 hrs  0 mins)

estimated total clock time        =      3366 sec (       0 hrs 56 mins)

estimated clock time to complete  =      3365 sec (       0 hrs 56 mins)

termination time                  = 2.000E-02

 

*** Error 40509 (SOL+509)

     negative volume in solid element # 865503 cycle 11 time 3.0456E-06

 

*** Error 40509 (SOL+509)

     negative volume in solid element # 865144 cycle 11 time 3.0456E-06

 

*** Error 40509 (SOL+509)

     negative volume in solid element # 865323 cycle 11 time 3.0456E-06

 

*** Error 40509 (SOL+509)

     negative volume in solid element # 865344 cycle 11 time 3.0456E-06

 

*** Error 40509 (SOL+509)

     negative volume in solid element # 865504 cycle 11 time 3.0456E-06

 

*** Error 40509 (SOL+509)

     negative volume in solid element # 865164 cycle 11 time 3.0456E-06

 

*** Error 40509 (SOL+509)

     negative volume in solid element # 865324 cycle 11 time 3.0456E-06

 

*** Error 40509 (SOL+509)

     negative volume in solid element # 865363 cycle 11 time 3.0456E-06

 

*** Error 40509 (SOL+509)

     negative volume in solid element # 865523 cycle 11 time 3.0456E-06

 

*** Error 40509 (SOL+509)

     negative volume in solid element # 865523 cycle 11 time 3.0456E-06

 

*** Error 40509 (SOL+509)

     negative volume in solid element # 865523 cycle 11 time 3.0456E-06

 

*** Error 40509 (SOL+509)

     negative volume in solid element # 865364 cycle 11 time 3.0456E-06

 

*** Error 40509 (SOL+509)

     negative volume in solid element # 865524 cycle 11 time 3.0456E-06

 

*** Error 40509 (SOL+509)

     negative volume in solid element # 865383 cycle 11 time 3.0456E-06

 

*** Error 40509 (SOL+509)

     negative volume in solid element # 865543 cycle 11 time 3.0456E-06

 

*** Error 40509 (SOL+509)

     negative volume in solid element # 865384 cycle 11 time 3.0456E-06

 

*** Error 40509 (SOL+509)

     negative volume in solid element # 865544 cycle 11 time 3.0456E-06

 

*** Error 40509 (SOL+509)

     negative volume in solid element # 865544 cycle 11 time 3.0456E-06

 

*** Error 40509 (SOL+509)

     negative volume in solid element # 865563 cycle 11 time 3.0456E-06

 

*** Error 40509 (SOL+509)

     negative volume in solid element # 865564 cycle 11 time 3.0456E-06

 

*** Error 40509 (SOL+509)

     negative volume in solid element # 865583 cycle 11 time 3.0456E-06

 

*** Error 40509 (SOL+509)

     negative volume in solid element # 865584 cycle 11 time 3.0456E-06

      11 t 3.0456E-06 dt-3.38E-07 write d3plot file            03/05/25 11:44:17

      11 t 3.0456E-06 dt-3.38E-07 write d3plot file            03/05/25 11:44:17

 

Sincerely,

James M. Kennedy

KBS2 Inc.

March 5, 2025

--
You received this message because you are subscribed to the Google Groups "LS-DYNA2" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ls-dyna2+u...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/ls-dyna2/CAG%2BwLY94W7qNw5SP4-sXdRD7HQ1KMTikJbVmzaBHT_CR%3DcMQVg%40mail.gmail.com.

[Arumugam ce19d044]

Dear Len

My understanding is that, lagrangian element needs to be finer compared to that of ALE element. See blow (from User's Manual)

  Number of Coupling Points. The number of coupling points, NQUAD × NQUAD, is distributed over the surface of each Lagrangian seg ment. Generally, 2 or 3 coupling points per each Eulerian/ALE element width is adequate. Consequently, the appropriate NQUAD values must be estimated based on the relative resolutions between the Lagrangian and ALE meshes. For example, if one Lagrangian shell element spans two ALE elements, then NQUAD for each Lagrangian segment should be 4 or 6. Alternatively, if two or three Lagrangian segments span one ALE element, then maybe NQUAD = 1 would be adequate. If either mesh compresses or expands during the interaction, the number of coupling points per ALE element will also change. The user must account for this and try to maintain at least two coupling points per each ALE element side length during the whole process to prevent leakage. Too many coupling points can result in instability, and not enough can result in leakage.   

As you have rightly mentioned, with TYPE = 0, 1, 5, multiple ALE parts are generated and it doesn't inherit the parent material initial conditions and contact.

Right now, I am trying with Shell elements and will update if there is any progress.

Regards

Arumugam

On Wed, Mar 5, 2025 at 10:23 PM <l...@schwer.net> wrote:

Thanks Arumugam for sending the TYPE=1 model.

I ran the model, but need to look at the model & “results” more closely.

I will be out of my office most of today.

 

The TYPE=1 model generates FOUR new parts: the original part in the define box, the new mesh inside the refine box, and two ALE parts representing the part in the define box and the rest of the driven part.

I am not sure what roll all these parts play when you try to use TYPE=3.

 

But for the TYPE=1 model, it looks as if the define box ALE part does not know about the ALE driven part; see below.

 

In the meantime, two suggestions:

1\ Your steel plate might be better modeled using shell elements?

2\ The ALE mesh needs to be finer than the Lagrange (steel) mesh. At a minimum 4 ALE elements to each Lagrange element. I assume this refinement of the ALE mesh is why you are using *CONTROL_REFINE_ALE.

 

 

From: Arumugam ce19d044 <ce19...@smail.iitm.ac.in>
Sent: Tuesday, March 4, 2025 8:14 PM
To: L...@schwer.net
Subject: Re: [LS-DYNA2] Refinement of Mesh in FSI

 

Dear Len

 

License detail is given below.

 

 

With TYPE=0,1 generates refined mesh with Part 4. So the same has been included in SET_PART_LIST and mentioned in CLIS.

 

Attached is the updated file with TYPE=1.

 

Regards

Arumugam

 

 

On Wed, Mar 5, 2025 at 8:08 AM <l...@schwer.net> wrote:

1\ I changed the parameter TYPE=1 on * CONTROL_REFINE_ALE but did not see a refined mesh; see attached modified file.

 

2\ What version of LS-DYNA are you using? I used R15 from March 2024

 

3\ Why does your model reference an ALE Part ID 4 ?

 

 

From: Arumugam ce19d044 <ce19...@smail.iitm.ac.in>
Sent: Tuesday, March 4, 2025 4:18 PM
To: L...@schwer.net
Subject: Re: [LS-DYNA2] Refinement of Mesh in FSI

 

Hi Len

 

As per your suggestion and keyword manual, I have tried the following.

 

CONTROL_REFINE_ALE with TYPE=3 doesn't create any refined mesh near the lagrangian part and "negative volume in solid #" error occurs.

 

CONTROL_REFINE_ALE with TYPE=0,1,5 creates refined mesh but contact is not established even though the CLIS card is defined.

 

In the model, lagrangian mesh is finer than the ALE, where  a single ALE element is in contact with 5 lagrangian elements which results in "negative volume error". To overcome this, I am using "CONTROL_REFINE_ALE".

 

Thanks

Arumugam

 

On Mon, Mar 3, 2025 at 9:58 PM <l...@schwer.net> wrote:

From the User Manual Volume I description of the keyword *CONTROL_REFINE_ALE

 

First Card Definition Only and TYPE. If only the 1st card is defined, only

TYPE = 0, 1, and 5 can be defined.

 

The input file you provided uses TYPE=3, with all default (zero) values on Cards 2 and 3. You need to provide some non-zero vales on these two cards.

               --len

 

From: ls-d...@googlegroups.com <ls-d...@googlegroups.com> On Behalf Of Arumugam ce19d044
Sent: Monday, March 3, 2025 4:45 AM
To: ls-d...@googlegroups.com
Subject: [LS-DYNA2] Refinement of Mesh in FSI

 

Dear All

--

[l...@schwer.net]

Arumugam --  I think you misread the “advice” you cited:

 

“For example, if one Lagrangian shell element spans two ALE elements, …”

 

If ONE shell spans TWO ALE cells, then the ALE mesh is finer than the Lagrange mesh. The reason for the finer ALE mesh is the Eulerian formulation is only first order accurate while the Lagrange element formulation is second order accurate.

 

In my experience, a MINIMUM of four ALE cells are need for each corresponding Lagrange segment, with NINE being preferred.

 

                              --len

[Arumugam ce19d044]

Thanks Len for the clarification.

Have run multiple cases on models with only Eulerian elements. With TYPE=0,1,5, refined mesh is created at the desired zone, however the child elements didn't inherit the parent elements initial conditions. As a result, the shock wave in the refined mesh region is not as expected.

As per your suggestion, using shell elements with TYPE=2,3,4, here, the model is running into negative volume error in advection cycle. This error points to the elements in contact with the shell elements. 

Regular mesh without *CONTROL_REFINED_MESH card, the analysis is running ok. But the problem is the thickness of the tested sample (1mm / 2mm), for which very fine mesh is required and it takes more than 90 hours.

Please suggest if there is any alternative or option to reduce the computational time if possible.

Thank you

Arumugam