NX Glue Source/Target Convention

Engineering Tools — NX Glue Source/Target Convention

The principle

When NX / Simcenter Nastran glued connections are used to represent an adhesive or compliant bondline as a controlled spring layer rather than a near-rigid weld, the resulting bondline stiffness depends on which side of the pair is the source. Source/target is not cosmetic. Two glue property cards with identical numbers can produce different effective stiffnesses if their source/target choices disagree.

This article documents the convention to follow so the source side is deterministic, the penalty factors mean what they look like they mean, and the bondline behaves the same everywhere it appears in the model. Face sheet to honeycomb core is used as the worked example because it is the most common case where readers run into this problem, but the rule generalises to every glued pair where the two sides have meaningfully different modulus, mesh density, or material definition.

Setup prerequisites

This guidance assumes a glue setup that targets an actual physical adhesive stiffness (`Ea / ta` and `Ga / ta`), not the default stiff-glue preset that NX picks for kinematic enforcement. In FEMAP's Define Connection Property dialog that typically means:

Auto Penalty Factor: unchecked. With Auto on, NX picks the penalty stiffness for you and your Normal / Tangential entries are scale factors against that auto value (default = 100, i.e. 100× the auto stiffness). With Auto off, the entries drive the stiffness directly and the formulas below apply.
Penalty Factor Units: pick one mode and use it consistently. The formulas in this article assume the Scale Factor mode against NX's `E / t` auto basis (PENN, PENT semantics). In Absolute mode the entered value is the stiffness per unit area directly and the formula collapses to `Kn = Ea / ta` with no further algebra.
Glue Factor: leave at 1.0 unless intentionally scaling. This is a top-level multiplier applied AFTER the per-direction factors and will rescale everything if you forget about it.

If the default setup is left in place (Auto Penalty Factor on, Normal = Tangential = 100, Scale Factor mode), NX produces a stiff glue tuned for kinematic enforcement rather than a physical adhesive stiffness, and the rest of this article does not apply.

Why source/target matters

NX glue is more robust than early contact implementations — source/target reversal is less fragile than it once was for basic projection and search behaviour — but several effects still depend on which side is the source:

Penalty stiffness scaling depends on source-side modulus and thickness. The auto-penalty basis is essentially `E_source / t_source`. Reversing source and target swaps the basis, and any penalty factor you calculated against one side then produces the wrong stiffness against the other.
Material modulus is often ambiguous on one side. Orthotropic, honeycomb-effective, or PCOMP-laminate moduli are not single numbers — through-thickness vs. in-plane vs. smeared matters. Pick the side where modulus is unambiguous as the source.
Shell offset and thickness handling can change interface behaviour. Source/target reversal can affect projection, offset interpretation, and where the solver effectively places the connection plane.
Traction sign conventions and result interpretation can change. Peel/tension and shear results are easier to read when source and target are standardised across the model.
Free-edge stress peaks are sensitive. Edge peel and shear peaks can move or change magnitude depending on whether the edge is represented as source or target.

Source-side selection rule

For any glued pair where the two sides differ in modulus, mesh density, or material clarity, prefer the source side that is:

Unambiguous in modulus — a single, well-defined Es you can write down. Isotropic metallics, PCOMP face sheets evaluated through classical lamination theory, or any side where one number represents the bond-controlling stiffness.
The controlling structural member — the side whose stiffness actually controls how load enters the bondline. Usually the face sheet, doubler, or skin rather than the core, potting, or substrate.
The finer / smaller patch if there is a clear mesh density mismatch. This is the legacy contact rule and it agrees with the bondline rule in most common geometries.

Common applications — recommended source side

Interface	Source	Target	Es to use
Upper face sheet to core	Upper face sheet	Core	Upper face sheet in-plane modulus
Lower face sheet to core	Lower face sheet	Core	Lower face sheet in-plane modulus
Doubler / patch to base skin	Doubler	Base skin	Doubler in-plane modulus
Stiffener flange to skin	Flange	Skin	Flange modulus
Composite to metallic plate	Whichever side is unambiguous and controls load path (usually the metallic plate)	The other	Source side material modulus
Metallic insert / potting to core	Insert (well-defined)	Core	Insert material modulus
Two composites of different layup	Whichever has the more controlled / better-characterised in-plane stiffness	The other	Source side CLT smeared modulus

Recommended naming convention

Encode source and target in the connector title so it is visible without opening the dialog:

  SRC_<source-region>__TGT_<target-region>

Worked examples:

  SRC_FS_TOP__TGT_CORE_TOP
  SRC_FS_BOT__TGT_CORE_BOT
  SRC_DOUBLER__TGT_SKIN
  SRC_FLANGE_LH__TGT_SKIN
  SRC_INSERT_05__TGT_CORE_REGION_3

The Connector Title Renamer tool can apply this format automatically across many connectors at once, pulling the master / slave region titles and the property title from each connector to build the name.

Glue penalty factor calculation

For an adhesive bondline modelled as a Winkler-type spring layer, the desired physical stiffness per unit area is:

  Normal stiffness     Kn = Ea / ta
  Tangential stiffness Ks = Ga / ta

Where:

Ea = tensile / normal modulus of the adhesive
Ga = shear modulus of the adhesive
ta = adhesive / bondline thickness

NX's auto-penalty basis is approximately K_basis = E_source / t_source. With Penalty Factor Units = Scale Factor and PENN as the user input, the effective normal penalty stiffness behaves as:

  Kn_NX  ≈ PENN × (Es / ts)

Where:

Es = modulus of the source-side elements
ts = characteristic source-side thickness (typically source shell t)

Matching the physical bondline stiffness gives the factor entries:

  Normal Factor      PENN = (Ea × ts) / (Es × ta)
  Tangential Factor  PENT = (Ga × ts) / (Es × ta)

Note on the source-side thickness term. It is easy to drop ts from the formula and write PENN = Ea / (Es × ta). Don't — that form is dimensionally wrong (PENN must be dimensionless; the dropped-ts form has units of 1 / length) and can give an order-of-magnitude stiffness error depending on the source thickness in model units. If the source side is a thin face sheet (ts ≈ 0.02 in model units), dropping ts inflates the entered factor by 50×. Always verify the formula's units before keying numbers in, and confirm with the coupon check below.

If Penalty Factor Units is set to Absolute instead of Scale Factor, the entered value is the stiffness directly and the formula collapses to:

  Normal Factor (absolute)      = Ea / ta
  Tangential Factor (absolute)  = Ga / ta

Source-side properties drop out of the user input in Absolute mode, but source/target still matters for projection, offset, and result interpretation per the list earlier.

Modulus choices that catch people

Composite (PCOMP) source side. Use the in-plane smeared modulus from classical lamination theory:

  Es = A11 / t_laminate

where A11 is the in-plane extensional stiffness from the laminate [A] matrix. This represents how the laminate transmits in-plane load into the bondline. Do not substitute the outermost ply's E11 or a through-thickness modulus — both will be wrong.

Honeycomb / orthotropic core as target. Even though core is the target in the recommended convention, the auto-penalty basis may use a blended source/target value depending on solver internals. The core's relevant modulus for cross-checks is direction- dependent: through-thickness E33 controls peel, in-plane G13 / G23 controls shear. Picking the wrong one for a target-side sanity check silently produces the wrong reference.

Two-composite case. When both sides are PCOMP and both are well-characterised, pick the side whose layup is most stable across the model (less local variation in stacking sequence). Consistency across instances of the same conceptual joint matters more than which laminate is mathematically "stiffer".

Model QA checks

A model-audit utility is recommended for any model with more than a handful of glue connectors. Per-connector checks worth automating:

Source elements belong to the expected face sheet / flange / doubler / insert property, layer, or material set.
Target elements belong to the expected core / skin / base / panel set.
Glue property factors were calculated using the intended source modulus (compare the modulus baked into the factors to the actual Es of the source-side elements).
Search distance is reasonable for the bondline gap, offset, and the model's length units. A search distance many multiples of the adhesive thickness will still find pairs but is permissive about mismatched offsets.
Shell offset and thickness settings are intentional, not inherited defaults.

Flag any connector where:

Source side does not match the expected source set per the convention table.
Source-side modulus disagrees with the modulus used in the factor calculation.
Title naming convention (`SRC_xxx__TGT_xxx`) doesn't match the actual source/target assignment.

Treat reversed source/target connectors as a real analysis issue, not bookkeeping noise — they directly change the effective solved glue stiffness, often by orders of magnitude when source and target materials differ greatly.

Verification coupon

Before relying on a production model, verify the glue setup with a single coupon:

  source side  /  adhesive glue  /  target side

Run two unit-displacement cases:

Unit normal displacement of the source side
Unit shear displacement of the source side

Recover the glue reaction force, divide by the bonded area, and compare against:

  F / A = (Ea / ta) × Δ_normal   (target physical normal stiffness)
  F / A = (Ga / ta) × Δ_shear    (target physical shear stiffness)

This single check confirms the entered glue factors produce the intended bondline stiffness, including the source-side thickness term. Run the coupon once per distinct (source modulus, source thickness, adhesive properties) combination — usually only a handful of variants per real model.