2. Dummy model architecture in PAM-CRASH

A Hybrid III or THOR dummy in PAM-CRASH is not a single finite element mesh. It is an assembly of many independent parts, each assigned a specific mechanical role, and held together by a small set of kinematic and structural definitions. Understanding this architecture is the prerequisite for everything that follows: positioning, belt routing, contact setup, and injury-signal extraction all depend on knowing which parts are rigid, which deform, and how they are joined. This section explains the four building blocks of the architecture — rigid bodies, joints, deformable segments, and connection definitions — and how they are organized for Hybrid III and THOR.

Why the dummy is decomposed at all

An occupant dummy must reproduce two different physical behaviours simultaneously. Its skeleton must transmit and distribute inertial loads like a stiff articulated linkage, and its soft regions must absorb energy and produce measurable deformation like flesh and bone. A single deformable mesh cannot do both efficiently: making the whole body deformable would require an impractically fine mesh and very small stable time steps, while making everything rigid would eliminate the very deformation that injury criteria are derived from.

The standard solution is a hybrid architecture. Mass-dominant, stiff regions — the head shell, pelvis, upper and lower limbs, and the rigid internal components of the thorax — are modelled as rigid bodies. Regions whose deformation is the measurement of interest — the thorax, abdomen, and neck — are modelled as deformable segments. The two are joined by joints that constrain relative motion to the anatomical degrees of freedom, and by connection definitions that transfer load between rigid and deformable parts.

Rigid bodies

In PAM-CRASH a rigid body is defined by a part or part set whose nodes are constrained to move as one, with the mass and inertia lumped at a centre of gravity. The essential inputs are the part identifier, the material law (typically a rigid material with density), and the initial velocity. Because the nodes cannot deform, the stable time step is not limited by the rigid parts, which is the main computational reason for using them.

For a Hybrid III 50th percentile model, the head, upper and lower arms, upper and lower legs, pelvis, and the metallic internal components of the thorax and lumbar spine are rigid. For THOR, the head, pelvis, and limbs are likewise rigid, but the thorax contains additional rigid internal structures — the sternum plate and rib mounting hardware — that interact with the deformable rib cage. Each rigid body carries its own centre of gravity and inertia tensor, and these must match the certification values for the dummy build being used. A common error is to leave the default density on a rigid part, which shifts the total mass and changes the measured accelerations without producing any obvious solver warning.

Joints

Joints define the allowed relative motion between two rigid bodies. PAM-CRASH provides kinematic joint types — spherical, revolute, cylindrical, translational, and universal — each removing a specific set of degrees of freedom. The dummy uses them anatomically:

  • Neck: a spherical or universal joint at the occipital condyle, allowing flexion, extension, and lateral bending while resisting axial rotation according to the neck spring definition.
  • Shoulder: a spherical joint at the glenohumeral centre, with rotational stiffness representing the shoulder capsule.
  • Elbow and knee: revolute joints with a single flexion–extension axis, plus a rotational spring and stop to represent ligament resistance and hyperextension limits.
  • Hip: a spherical joint at the femoral head, with stiffness in flexion, abduction, and rotation.
  • Lumbar spine: a series of joints between vertebral bodies, each with rotational stiffness, which together produce the characteristic lumbar flexion response.

Each joint is defined by the two connected bodies, the joint location in the global or local coordinate system, the joint type, and — where applicable — a rotational or translational stiffness curve. The stiffness curve is what makes the joint behave like a dummy rather than a frictionless mechanism. For example, a knee joint with a flexion stop at approximately 120° and a hyperextension stop near −5° prevents the non-physical configurations that would otherwise occur during belt loading.

Joint locations must be defined consistently with the dummy's local coordinate system. If a joint is placed at the wrong node or in the wrong reference frame, the limb will rotate about the wrong axis, and the resulting kinematics will look plausible in a single frame but will be wrong throughout the event. Always verify joint axes against the dummy drawing before running a full simulation.

Deformable segments

Deformable segments are the parts whose deformation produces the injury-relevant signals. In Hybrid III these are the thorax (a deformable chest with a rib cage and damping material), the abdomen, and the neck. In THOR the deformable set is larger and more detailed: the rib cage is modelled with individual ribs, the abdomen is a separate deformable region, and the neck is a deformable column. These parts use elastic–plastic or viscoelastic material laws, and their mesh density is chosen so that the deflection under a 250 mm chest-bar loading is captured accurately.

The key modelling decisions for deformable segments are the material law, the mesh density, and the contact definition with the rigid parts. A thorax modelled with a linear elastic material will rebound too quickly and under-predict chest deflection; a viscoelastic or elastic–plastic law with a defined yield stress is required to reproduce the plateau and residual deformation seen in certification tests. The mesh must be fine enough that the deflection measurement point — the chest potentiometer location — sits on a node whose displacement is representative of the local deformation.

Connection definitions

Connections are what turn a collection of parts into a working dummy. PAM-CRASH offers several mechanisms, and choosing the right one is critical:

Connection type Typical use in the dummy Key consideration
Kinematic joint Neck, shoulder, elbow, knee, hip, spine Removes DOF; requires stiffness curve for realistic resistance
Tied contact Rigid head shell to deformable neck top Must not over-constrain the deformable part; check for artificial stiffness
Sliding contact Rib cage against sternum plate (THOR) Friction coefficient affects load distribution
Spring / damper element Lumbar spine stiffness, neck resistance Curve must match certification data
Spot weld / rigid link Internal rigid components to shell Use only where no relative motion is intended

The most common connection error is over-constraining a deformable segment. If a tied contact is applied around the entire perimeter of the thorax where it meets the rigid spine box, the thorax cannot deform freely and chest deflection is under-predicted. The correct approach is to tie only the region that is physically bonded and to allow the remaining nodes to slide or remain free.

Architecture overview

The following diagram summarizes how the four building blocks are arranged for a Hybrid III thorax–neck–head assembly, showing which parts are rigid, which deform, and how they are connected.

Hybrid III Upper Body: Rigid / Deformable / Joint Architecture Head Shell Rigid body · accelerometer mount Neck Joint (spherical) Neck Column Deformable · viscoelastic Thoracic Joint Thorax (rib cage + damping) Deformable · elastic–plastic · chest deflection output Spine Box Rigid body · load path tied Pelvis Rigid body lumbar joint Legend Rigid body Deformable segment Joint Connections kinematic joint tied contact

Practical verification of the assembly

Before running any sled or barrier simulation, verify the assembled dummy in a short settling run. Apply gravity for 50–100 ms with no external load and check three things: the total mass reported by the solver matches the dummy build specification (Hybrid III 50th percentile is approximately 77.7 kg), the joint locations do not drift, and the deformable segments show no residual deformation. If the total mass is wrong, a rigid part density is incorrect. If a joint drifts, its stiffness curve is missing or its reference frame is inconsistent. If the thorax shows residual deformation after settling, the material law has yielded under gravity alone, which indicates an unrealistically low yield stress.

Once the settling run is clean, apply a known chest-bar loading — for example, a 250 mm diameter bar displacing the chest by 50 mm at 1 m/s — and confirm that the chest deflection output at the potentiometer location falls within the certification band. This single check validates the rigid–deformable–joint chain end to end, and it is far more informative than inspecting the model tree.

Keep a written record of the rigid body masses, joint stiffness curves, and material parameters for each dummy build. When a model is exchanged between teams or updated to a new dummy revision, these values are the first thing to reconcile, and a mismatch here will contaminate every downstream injury result.

The architecture described here — rigid bodies for mass and stiffness, joints for anatomical motion, deformable segments for injury measurement, and explicit connections between them — is the same for Hybrid III and THOR. The difference lies in the level of detail: THOR adds more deformable ribs, a separate abdomen, and additional internal rigid structures, which increases both fidelity and the number of connections that must be verified. Mastering the four building blocks on Hybrid III provides the foundation for working with THOR and with any other occupant model in PAM-CRASH.