Contact Definition Error Resolution — FEA

Classic inverted normal problem. In FEA, contact detection is directional — the solver checks whether slave nodes penetrate the master surface, and that detection depends on which way the surface normal points. If the normals are pointing inward (into the solid) instead of outward (away from the surface), the solver thinks the surfaces are on the wrong side of each other and never detects contact. Plot the surface normals in your pre-processor and verify they point outward from each body.

Absolutely. The master surface resists penetration — slave nodes are not allowed to penetrate master facets, but the master surface can penetrate slave nodes if the mesh is coarser than the slave. So the rule is: assign the stiffer body or the coarser mesh as the master. If you have a rigid tool pressing on a deformable part, the rigid surface is always master. For two deformable bodies: coarser mesh or higher stiffness = master. Getting this backwards leads to slave nodes punching through the master without resistance.

This is contact chattering — one of the most frustrating convergence issues in FEA. It happens when the contact state (open vs. closed) is right on the edge of changing, and the solver can't find a consistent state that satisfies equilibrium. The most common cause is aggressive time increments: the load jump in a single increment is large enough that the optimal contact state changes, but Newton-Raphson iterations can't find the transition point. Reducing the maximum increment size is the first fix to try.

At that point, stabilization is your main tool. In Abaqus, add *CONTACT CONTROLS, STABILIZE — this introduces a small viscous damping force at the contact interface that prevents the surface from oscillating open and closed. The stabilization energy should be small compared to the strain energy; check ALLSD vs ALLIE in field output to confirm. The alternative is switching the contact formulation from penalty to augmented Lagrangian, which enforces exact zero-penetration rather than a stiffness approximation, and often resolves chattering by removing the "slack" that penalty methods introduce.

CAD assemblies that "look" correctly positioned often have tiny geometric interferences — less than 0.1 mm in a 1-meter part — that are invisible in CAD but huge in the FEA contact algorithm. This happens especially at corners, fillets, and bosses. Also, when you mesh the geometry, the mesh surface is a piecewise linear approximation that may protrude slightly beyond the exact CAD surface. You have a few options: adjust the geometry in CAD to add a small clearance, use Abaqus *CONTACT INITIALIZATION to allow the solver to automatically adjust the initial state, or specify a contact interference treatment.

For intentional interference fits, use *CONTACT INTERFERENCE in Abaqus. You ramp the interference from full overlap to zero over the first load step — this simulates the press-fit process and deposits the correct residual stress. For purely numerical small overclosures you don't want physically, use *SURFACE BEHAVIOR, PRESSURE-OVERCLOSURE=EXPONENTIAL to soften the contact at small penetrations, or set a small initial clearance adjustment in the contact property.

This is a classic penalty stiffness problem. In the penalty contact formulation, contact pressure equals penalty stiffness times penetration depth. If the penalty stiffness is very large (Abaqus default scales it to 10× the underlying element stiffness), even a tiny numerical penetration of 0.0001 mm can produce gigantic contact pressures. The solver "converges" because the residuals balance globally — but the contact pressure is a post-processing artifact of the penalty parameter, not a physically meaningful stress.

Three approaches. First, switch to augmented Lagrangian or Lagrange multiplier contact — these enforce exact no-penetration without a penalty spring, so the contact pressure is a true constraint force. Second, if you need to stay with penalty contact, use "softened contact" via *SURFACE BEHAVIOR, PRESSURE-OVERCLOSURE=TABULAR and define a realistic pressure-closure relationship from Hertz theory or test data. Third, make sure your mesh is fine enough at the contact zone — artificially coarse meshes concentrate contact forces on too few nodes, inflating the apparent pressure.

Friction actually makes convergence harder, not easier. The friction algorithm must track whether each contact point is in the stick state (tangential force below friction limit) or slip state (tangential force at friction limit, relative motion occurring). This stick-slip transition is a non-smooth discontinuity — the tangential stiffness changes sign as you cross the friction limit — and Newton-Raphson struggles with discontinuities. What you're seeing is nodes cycling between stick and slip states from iteration to iteration.

Not necessarily — the physics requires whatever friction coefficient is real. The fix is usually algorithmic. First, reduce the slip tolerance: *CONTACT CONTROLS, SLIP TOLERANCE in Abaqus limits how much "slip" is allowed in the stick state. Second, reduce the time increment — smaller steps mean smaller tangential displacement jumps that are easier to classify. Third, use the "Lagrange friction" formulation (exact stick, no micro-slip) instead of the default penalty-based friction: it enforces the friction limit exactly and eliminates the ambiguity that causes oscillation.

Shells are geometrically 2D — the nodes live at the mid-plane, but physically the shell has thickness extending above and below. By default, Abaqus treats the contact surface as the mid-plane (SPOS and SNEG refer to the top and bottom surface respectively). If you define contact on SPOS (top face) of shell A and SNEG (bottom face) of shell B, the contact happens at the actual physical surfaces and the thickness offset is accounted for correctly. If you define contact on the reference surface of both, the shells will overlap by a combined half-thickness before contact activates.

Bolt/pin contact in shells is tricky because the shell edge doesn't have a real 3D surface. The most robust approach is to use a coupling constraint at the bolt hole — DSECANT coupling in Abaqus connects the bolt load to the shell nodes around the hole perimeter without requiring explicit surface contact geometry. For the clamping force (preload), use a connector element with an initial length change rather than contact. If you genuinely need bearing stress around the hole, switch the local region to a 3D solid mesh and use shell-to-solid coupling at the transition boundary.