Non-linear Collision

InqInq
I am looking to model a collision.  Although the structure I want to model is completely different, a real-world intuitive example might be a car's rear-end with bumper backing into a rigid brick wall.  I have looked at the PipeClip.liml example.  It has most of the requirements I will need including Nonlinear 3D, frictionless support and contact elements.  I see it uses displacement = f(time) to move the parts together.  I would like the movement function to be based on F = ma.  IOW, set an initial speed and mass and let the reaction with the brick wall to slow the "car" down to a velocity of zero.  I would need to gather accelerations, max stresses during the interaction.  The "car" would be geometrically non-linear.  It would not necessarily get into buckling, but would be great if that could be modeled also.  

Is there some way to mimic this scenario?  

Thanks.

Comments

  • VictorVictor
    Hello Inq

    The more powerful way is to use Nonlinear Dynamic Response 3D as the Analysis type. But there are a couple of limitations:
    • The solution doesn't have acceleration so you might have to calculate that from differences in velocities between time steps by exporting the Table to a spreadsheet
    • It can also be difficult to set an initial velocity so you may need to accelerate the object from rest with an applied force. You can use "gravity" for that to avoid causing deformation.

    Another way, if you don't need dynamic effects like vibration, might be to use Nonlinear Static 3D treating the model as being in the object's accelerating reference frame. The deceleration would be a ramped Gravity load pressing it against the wall.
  • InqInq
    Hello Victor,

    Thank you for replying.  It's been a long time since we last traded messages (2014).  I couldn't recover my old logon and thus created this one.  No real loss... I might as well be a different person.  B)

    Let me elaborate on what I want to achieve.  I will be sharing my work on this thread as I go along, so I'm not trying to hide the actual model.  For now, I just think anyone that has been near a car understands the concepts intuitively of car bumpers.  Probably first hand.  :#   There are three phases that are common.
    1. No damage - There is an initial car velocity (say 5 mph) that the bumper will have very large, non-linear deflections as it absorbs energy decelerating (again... non-linearly caused by the integration of boundary forces at each time steps) the car to 0 mph.  In the real world, the bumper will spring out back to normal.  No harm, no foul.
    2. Controlled damage - The next case (say 10 mph) absorbs as much energy in the "No damage" zone and then some planned structure starts crushing.  Things are damaged, but people walk away... usually cussing.  In well designed cars, these sacrificial pieces are easily replaced.  I've seen where some cars use expanded bead styrofoam.  Others use specially crimped steel tubes that buckle in a controlled and precise manner.
    3. All hell breaks loose - Higher still (say 30 mph) primary structure starts buckling and bending.  Hopefully absorbing energy and keeping the occupants unharmed, but there is no expectation of structural re-use.

    I want to design my part in this 1st phase.  I will iterate the design (mostly geometrically) to achieve the highest initial velocity without permanent damage.  I won't design for a phase 2 component and the onset of phase 3 will most likely occur due to buckling.

    I think the Nonlinear Dynamic Response 3D sounds like it will fit the bill to handle the analysis determination of the deceleration forces.  The two caveats you mention: the spreadsheet work and initial velocity doesn't sway me.  What I thought I had to do is far more intensive. 

    Although I'm not concerned about the vibrations, I am trying to wrap my head around how to use the results (stresses, boundary forces, etc) of the finished large deflection state at zero speed to feed into another analysis to predict onset of structural failure due to buckling.  I know this type simulation is possible https://youtube.com/watch?v=_HpSaFSnQoY, but I'm unsure of how much is built-in to FEA programs and how much has to be handled outside the program... and more importantly... how close I can get with Mecway and what steps I need to perform outside of Mecway either with spreadsheets or even programming.
  • VictorVictor
    Oh, I see who you are now. Hello again!

    CCX allows you to chain several analyses together in steps which sounds like what you're asking for.

    But I don't think you need to do that because nonlinear dynamic response already includes buckling and plastic deformation. You may also have to add some imperfection in the geometry to ensure buckling initiates if everything is too symmetrical. You can add additional contacts on areas where you anticipate self-contact might occur.
  • MikeMcMullenMikeMcMullen
    I will probably be looking into this as well, once I get my FEM skills up to speed. Besides looking into better resistance models from the impacted side of things, I am interested in the fall off of force effects with distance and intervening mass, especially as structures soften in their non-linear zones, and the effects on adjacent areas of the structure. The codes and recent research have data of quasi static forces that can be used for initial design, but these are only appropriate within a short time frame and close to the load application due to inertial resistance.

    As far as your problem, while I have not done it, in both actual physical tests and computer modeling for bridge and roadway rail tests a pendulum arrangement is often used to generate the contact velocity. Bogies are also used, but they require modeling of functioning wheels, winches, etc.

    In rock fall modeling, which I have done in the deep past, gravity can be used directly with a sliding or rolling rock on a slope.

    Please keep this forum posted on your progress as you are quite a bit ahead of me and I find the advice here useful, especially for similar problems, as well as any sample input provided to see how the problem was coded. Fortunately I won't be needing self contact, but I expect you will for higher velocities. There is a fair amount in our crash tests of cars and PU trucks.

  • rwhirleyrwhirley
    If you set up the entire nonlinear dynamic problem in Mecway (for CCX) and output the .inp file for CCX, I think it isn't hard to edit in a block of data to specify an initial velocity for all "car" nodes. This will be much easier than accelerating it with a force, although in the end either should work.

    If your mesh is sufficiently fine to capture the buckling modes of interest, and you include nonlinear (elastoplastic, for example) materials, then in my experience buckling behavior will be accurately represented.

    Expect to need *much* smaller timesteps than you might predict. Best to set them as something something like the time for the "car" to move 5% of an element length, and set this as a max time step or the implicit nonlinear solver will struggle with capturing the collision. You can adjust once you get things running.

    Just a few thoughts. Good luck!
    -Robert
Sign In or Register to comment.

Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!

Sign In Register