2. Define OOP Occupant Kinematic Boundary Conditions

Why Boundary Conditions Must Be Explicit

In a reclined occupant finite element (FE) model, the pre-impact posture is not a single geometry — it is a set of kinematic states that must be prescribed before the restraint system and the crash pulse are activated. If those states are left implicit (for example, by letting the model settle under gravity alone), the resulting posture will drift toward whatever equilibrium the contact definitions happen to produce. That drift is not repeatable, and it silently changes belt-to-body routing, pelvis angle, and head-to-headrest clearance — the very quantities that later cases in this series will measure.

The task in this case is therefore narrow and mechanical: define the out-of-position (OOP) initial kinematic state as an explicit, auditable set of boundary conditions, apply them to the baseline 45° reclined model, and then prove that the prescribed posture is actually reached without the constraints themselves altering the occupant's response. The distinction matters. A prescribed rotation that is held by a rigid constraint during the impact phase is an artificial constraint; a prescribed rotation that positions the body and is then released is a legitimate initial condition.

Decomposing the OOP State into Independent Parameters

An OOP posture is best described by a small set of independent kinematic parameters, each with a defined reference frame. Adopt the vehicle coordinate system throughout: +X forward, +Y to the occupant's left, +Z upward, with rotations right-handed about each axis. Every parameter below is expressed in that frame, and the same convention is used in the model, the table, and the figure.

• Pelvis rotation — pitch about Y (anterior/posterior pelvic tilt), roll about X, and yaw about Z. Reclined submarining cases are dominated by posterior pelvic pitch, so this is the primary degree of freedom.
• Torso lean — the angle between the thoracic spine axis and the seat back plane, decomposed into lateral lean (roll) and forward slump (pitch).
• Limb placement — hip flexion/abduction, knee flexion, and shoulder abduction, which determine whether the arms interact with the belt or the side bolster.
• Head/neck — a small relative rotation that follows from torso lean; it should be derived, not independently prescribed, to avoid over-constraining the cervical spine.

Each parameter needs three attributes to be usable: a target value, a tolerance band, and a settling strategy (how the model reaches the target and what happens to the constraint afterward).

Prescribing the State in the Baseline Model

The baseline model from the previous case is a validated 45° reclined occupant with defined materials and contacts. To impose the OOP state, use a two-phase approach rather than a single static offset.

Phase A — kinematic positioning. Apply prescribed rotations to the pelvis and torso reference nodes and prescribed translations to the limb joints, ramping them over a short positioning interval (for example 50 ms) with a smooth step function. During this phase, disable gravity and the crash pulse, and keep belt elements in a slack, non-load-bearing state so they do not resist the repositioning.

Phase B — release and settle. At the end of Phase A, remove the prescribed-motion constraints and allow a brief settling interval (for example 30 ms) under gravity only. The occupant should hold the target posture within tolerance; any large drift indicates that the prescribed state was not in equilibrium with the contact definitions.

# Pseudocode for the two-phase boundary-condition setup
# (parameter names are illustrative; map to your solver's keyword syntax)

define_coordinate_system(name="VEH", x_forward, y_left, z_up)

# --- Phase A: kinematic positioning ---
prescribe_rotation(node=pelvis_ref, axis=Y, target=-18 deg, ramp=50 ms, profile=smoothstep)
prescribe_rotation(node=pelvis_ref, axis=X, target= 4 deg,  ramp=50 ms, profile=smoothstep)
prescribe_rotation(node=torso_ref,  axis=Y, target= 12 deg, ramp=50 ms, profile=smoothstep)
prescribe_translation(node=hip_L, dof=flexion,  target=95 deg, ramp=50 ms)
prescribe_translation(node=hip_R, dof=flexion,  target=95 deg, ramp=50 ms)
prescribe_translation(node=knee_L, dof=flexion, target=70 deg, ramp=50 ms)
prescribe_translation(node=knee_R, dof=flexion, target=70 deg, ramp=50 ms)
gravity = OFF
belt_state = SLACK

# --- Phase B: release and settle ---
release_all_prescribed_motion(at=50 ms)
gravity = ON
settle_interval = 30 ms
belt_state = TENSIONED_AT_REST

# --- Verification gates ---
assert posture_error(pelvis_pitch) <= 2 deg
assert posture_error(torso_lean)   <= 3 deg
assert posture_error(hip_flexion)  <= 5 deg
assert constraint_reaction_force  == 0 during impact phase

The final assertion is the one that catches the most common failure: a prescribed motion that was never released. If the reaction force at the pelvis reference node is non-zero once the impact phase begins, the model is being held in place by the boundary condition, and every downstream injury metric is contaminated.

Reading the Prescribed State Against the Model

The figure below shows how the four parameter groups map onto the occupant, where each is measured, and the two-phase release logic that separates a legitimate initial condition from an artificial constraint.

Verifying the Prescribed Posture Is Genuine

Verification has two independent parts, and both must pass before the model is used downstream.

Posture accuracy. After Phase B, extract the pelvis pitch, torso lean, and hip flexion from the reference nodes and compare them to the targets. The tolerance bands in the table below are deliberately tight enough to catch a mis-specified axis but loose enough to absorb contact compliance. A systematic offset in one direction usually means the prescribed rotation was applied about the wrong node or the wrong axis sign.

Constraint independence. Confirm that no prescribed-motion constraint remains active during the impact phase. The cleanest check is to query the reaction force at each reference node at the first impact increment; it must be zero. A non-zero reaction means the occupant is being held, and the posture is artificial regardless of how accurate it looks.

Common Failure Modes and Their Causes

• Posture drifts during settling. The prescribed state is not in equilibrium with the contact definitions — typically the seat cushion or belt is pushing the pelvis back. Reduce belt pre-tension during Phase A or adjust the cushion stiffness, then re-settle.
• Angles overshoot the target. The ramp is too short for the joint stiffness. Lengthen the positioning interval or use a smoother profile.
• One limb does not reach its target. The joint is hitting a contact (bolster, door trim) before the prescribed angle. Check the limb placement against the interior geometry; the OOP state may be physically unreachable in that direction.
• Reaction force non-zero at impact. The release step was omitted or applied to the wrong node set. Verify that every prescribed-motion definition is deactivated at the end of Phase A.

With the OOP state defined, applied, and verified, the baseline model now carries a reproducible pre-impact posture. The next step in the assessment chain is to vary the seat back angle and observe how this posture couples with the restraint system — but that is a separate task, and it depends on the boundary conditions established here being trustworthy.