LS Dyna

LS Dyna

Non-Reflecting Boundary Condition in SPH

    • MohammadMahdi
      Subscriber

      Hello. In order to absorb and eliminate artificial waves coming from boundaries in impact modeling, non-reflecting boundary condition (NRBC) is suggested. However, as I've seen so far in my models, NRBC does not work well for SPH nodes except for the fixed SPH nodes. I mean, non-fixed SPH nodes are deactivated when using BOUNDARY_SPH_NON_REFLECTING to simulate an impacted target. Is there anything else playing the same role as NRBC to absorb artificial rebounded waves from boundaries in non-fixed side SPH nodes? FYI, I usually fix the bottom SPH nodes, and the side SPH nodes making my target are free. Thanks in advance for sharing your thoughts.

    • tslavik
      Ansys Employee
      With BOUNDARY_SPH_NON_REFLECTING, particle movement should be minimal in that area of the domain. If particles start going across the plane, then bad things will happen. WeÔÇÖve only seen it used in soil modeling, where a non-reflecting plane is defined quite far away from the ÔÇ£actionÔÇØ, just to absorb some of the pressure waves coming to the edge of the domain. These would be low pressure waves; the feature was not designed for high pressure waves, and certainly not shocks.
    • MohammadMahdi
      Subscriber
      Thanks for the clarification. You're right, the movements of the boundary SPH nodes are not minimal since the material is ductile epoxy. So, what do you suggest in my case to absorb artificial waves coming from boundary SPH nodes? The artificial boundary waves advance the material failure. I need to somehow eliminate this artificial effect. Thanks.
    • tslavik
      Ansys Employee
      Here's an untested approach that might work. Transition to solid Lagrangian finite elements with DEFINE_ADAPTIVE_SOLID_TO_SPH (ICPL=1, IOPT=0). Several rows of solid elements should extend away from SPH part. Use BOUNDARY_NON-REFLECTING on the margin of the solid element part.
    • MohammadMahdi
      Subscriber
      Do you mean like the picture below? So, since the extended solid elements have the same material as the SPH nodes, what do you suggest for the contact between them? If the impact is only between the impactor and SPH nodes, is still DEFINE_ADAPTIVE_SOLID_TO_SPH required in order to eliminate the artificial boundary waves? Thanks in advance for sharing your thoughts with me in this regard.


    • tslavik
      Ansys Employee
      Please see the image below where a square plate is modeled with hexahedral finite elements, SPH particles and a transition layer between them. The outer red part (PID=1) is made of standard Lagrangian finite elements, green (PID=100) treated with DEFINE_ADAPTIVE_SOLID_TO_SPH, and SPH particles comprise the center section (PID=11). The second image below is from the LS-DYNA d3plot results and show the inactive SPH particles embedded in PID=100. Here is the corresponding keyword input data:
      *PART
      $# title
      finite elements
      $#pidsecidmideosidhgidgravadpopttmid
      11100000
      *SECTION_SOLID
      $#secidelformaet
      100
      *PART
      transition
      1001100000
      *DEFINE_ADAPTIVE_SOLID_TO_SPH
      100032012011
      *PART
      $# title
      SPH
      $#pidsecidmideosidhgidgravadpopttmid
      1133
      *SECTION_SPH
      $#secidcslhhminhmaxsphinideathstart
      31.2000000.2000004.0000000.0000.0000.000




    • MohammadMahdi
      Subscriber
      Thank you for the suggestion, I'll try it.
      However, I have a question about the suggested transition layer... Why do we need the transition layer and what would happen if we eliminate this layer and directly make a contact between the core SPH phase and surrounding finite elements?
    • tslavik
      Ansys Employee
      Give it a try and let us know how it works out.
    • MohammadMahdi
      Subscriber
      May I know which contact types you suggest for first, between the core SPH and the transition part and second, between the transition part and outer finite elements?
    • tslavik
      Ansys Employee
      No contact interface is required. The nodes on the face of the transition part will be common with the nodes on the outer finite element part. The SPH particles in the transition part will be "owned" by the SPH core part. Make a simple model to test this arrangement - one finite element, one transition element and a handful of SPH particles. You could paste the contents that entire model in this chat if you have trouble.
    • MohammadMahdi
      Subscriber
      I just tried what you said and did not work. In my opinion, since the transition part is a separate part, a contact algorithm has to be defined between the transition part and the core SPH part. The same scenario is required between the transition part and the outer finite elements. I'm gonna try Tied-Node-to-Surface contact card for both, between the core SPH and the transition, and between the transition part and the outer finite elements. I'll let you know if it works.
    • MohammadMahdi
      Subscriber
      As I mentioned before, a contact card is required between the core SPH and the transition part. I used Tied-Node-to-Surface for both, the contact between the core SPH and the transition part and between the transition part and the outer finite elements. The Non-Reflecting-Boundary-Condition was also applied on the outer segments of the outer finite elements. it is working now somehow. I mean, with your suggested method, I'm able to somehow apply NRBC on the target now; however, I'm not sure if the Tied-Node-to Surface is an appropriate select for the contact between the mentioned parts? Please let me know what you think and find the new scenario below:

    • tslavik
      Ansys Employee
      The nodes of the outer FE part should be "merged" to the nodes of the transition part. That is, a common node should be shared by each part. In that way, there is no need for the tied contact. Make a simple model to test this arrangement - one finite element, one transition element and a handful of SPH particles. You could paste the contents that entire model in this chat if you have trouble.
    • MohammadMahdi
      Subscriber
      I merged/shared the nodes of the outer FE part to the nodes of the transition part and it worked. So, as you mentioned, there is no need for a tied contact between the outer FE part and the transition part treated by *DEFINE_ADAPTIVE_SOLID_TO_SPH. However, the inner SPH and the transition parts do not still recognize each other, though CONT in *CONTROL_SPH is 0 and particle approximation is defined. In your suggested keyword input data above, why the section ID assigned to *DEFINE_ADAPTIVE_SOLID_TO_SPH is 201? Shouldn't it be equal to the section ID of the core SPH part?
    • tslavik
      Ansys Employee
      Yes, that's a good idea. Please try switching IPSPH and ISSPH to match the ID's of your SPH part. It would be helpful to make a simple model to test this arrangement - one finite element, one transition element and a handful of SPH particles. You could paste the contents that entire model in this chat if you have trouble.
Viewing 14 reply threads
  • You must be logged in to reply to this topic.