Implicit: Contact treatment
THEORY: Implicit contact interfaces
Several contact interfaces are now available for use in implicit calculations. Since implicit jobs take large time steps, the interfaces can behave differently than in explicit simulations. These differences are described below:
*CONTACT_TIED_NODES_TO_SURFACE (type 2)
*CONTACT_TIED_SURFACE_TO_SURFACE (type 6)
These interfaces use the constraint method. Slave nodes are located within master segments during initialization, and then treated as dependent degrees of freedom during the simulation. This reduces the size of the global stiffness matrix. Presently, all six degrees of freedom are constrained, a significant difference from the explicit implementation. This will be changed to constrain only translational motion, and support for type 7 will be added for the six degree of freedom case.
*CONTACT_TIED_NODES_TO_SURFACE_OFFSET (type o2)
*CONTACT_TIED_SURFACE_TO_SURFACE_OFFSET (type o6)
These interfaces use the penalty method, and are similar to the tied interfaces. These were the first tied interfaces to be implemented for implicit LS-DYNA, and are also available in version 950.
All of these interfaces use the penalty method. For interfaces which allow gaps to open, an important new feature in version 960 is the gap flag IGAPF, found in the first field of the optional contact card "C". This parameter selects one of two options for treating situations where gaps are opening:
IGAPF = 1: This option can significantly improve convergence behavior, but can also produce "sticky" interfaces which resist opening of gaps. An indicator of sticking is a fluctuating energy norm during equilibrium iterations, while the displacement norm is well behaved. This option is the default for non-automatic contact interfaces.
IGAPF = 2: This option allows gaps to open freely, but can slow convergence in some problems.
*CONTACT_SURFACE_TO_SURFACE (type 3)
*CONTACT_NODES_TO_SURFACE (type 5)
*CONTACT_ONE_WAY_SURFACE_TO_SURFACE (type 10)
Outward normal vectors for shell elements are chosen at initialization, making these interfaces reliable when large time steps are used. Two formulations are available, depending on the choice of the shell thickness flag SHLTHK.
By default, thicknesses are ignored (SHLTHK=0) and the "old" type contact is used. These interfaces have generally poorer searching capability than the "new" types, so they should be avoided unless the model has been constructed to ignore thickness offsets. This option was implemented for implicit analysis in version 960.
When thickness offsets are considered (SHLTHK=1,2) the "new" interfaces are used. These have superior searching capability, and generally provide better convergence behavior.
*CONTACT_FORMING_SURFACE_TO_SURFACE (type m3)
*CONTACT_FORMING_NODES_TO_SURFACE (type m5)
*CONTACT_FORMING_ONE_WAY_SURFACE_TO_SURFACE (type m10)
Outward normal vectors for shell elements are chosen at initialization, making these interfaces reliable when large time steps are used. These are the oldest and most extensively tested of the implicit interfaces.
*CONTACT_AUTOMATIC_SURFACE_TO_SURFACE (type a3)
*CONTACT_AUTOMATIC_NODES_TO_SURFACE (type a5)
*CONTACT_AUTOMATIC_ONE_WAY_SURFACE_TO_SURFACE (type a10)
*CONTACT_AUTOMATIC_SINGLE_SURFACE (type 13)
Outward normal vectors for shell elements are chosen automatically as the simulation proceeds. This creates potential problems for implicit simulations, where large steps can "fool" the normal vector selection logic. For best results, take small time steps until contact is established. These interfaces were available in version 950, but required that the consistency flag be activated on *CONTROL_PARALLEL. This flag is not required in version 960.
Oscillating energy norm may indicate nodes are going in and out of contact.
*CONTACT_AUTO_MOVE is good to put parts into immediate contact (1st step more likely to converge).
RCFORC data should be correct for non-sticky contact (IGAP=2). Contrary to the 970 User's Manual, automatic contacts can be sticky. Use of "sticky" contact in implicit analysis (GAP=0 or 1) will sometimes produce misleading RCFORC data since the sticky component of contact force is not included in RCFORC. You will see meaningful RCFORC data if you turn sticky contact off (IGAP=2) in your model or if you lower the penalty scale factor for the contact, say, by 2 orders of magnitude, so that the sticky component of contact force becomes very small or zero. The value of SHLTHK affects, among other things, the default contact stiffness and so it follows that the sticky component of contact force is also affected by SHLTHK.