2. Occupant CAE model landscape

Credibility assessment for occupant CAE models cannot be performed on a single artifact in isolation. The occupant model, the restraint system, the seat and interior structures, and the vehicle-level boundary conditions form a coupled simulation chain in which each link contributes its own uncertainty. Before any validation corridor can be defined, the analyst must first map the model landscape: what kinds of occupant representations exist, how restraint models are constructed and coupled, and how the simulation chain escalates from a sled-level rig to a full-vehicle environment. The credibility-critical interfaces are the joints between these elements — the places where a validated component can still produce an unvalidated system response.

Occupant model classes and their credibility characteristics

Occupant representations fall into two broad families: mechanical dummy models and human body models (HBMs). Anthropomorphic test device (ATD) models — Hybrid III 50th percentile male, THOR-50M, Hybrid III 5th percentile female, Q-series child dummies — are finite element representations of physical hardware. Their geometry and mass distribution are defined by the physical dummy, and their credibility argument rests on the correlation between the model and the hardware it represents. Because the physical dummy itself has been characterized extensively (certification tests, pendulum impacts, component-level stiffness measurements), the model can be validated against a well-defined reference. The key credibility question for an ATD model is not "does it behave like a human" but "does it reproduce the hardware response under the loading conditions of interest."

Human body models — THUMS, GHBMC, SAFER HBM, and their variants — represent human anatomy with deformable organs, detailed skeletal structures, and active or passive musculature. Their credibility argument is fundamentally different: there is no physical hardware to correlate against, so validation relies on cadaveric and volunteer test data, and the model's biofidelity must be established per body region and per loading mode. An HBM validated for frontal thoracic loading may have limited credibility for lateral pelvic loading. This regional and directional dependence is a defining feature of HBM credibility assessment and must be carried into any corridor definition.

The practical consequence is that the two families require different validation evidence. ATD models are typically validated against certification-level and component-level hardware tests with tight tolerances; HBMs are validated against a patchwork of post-mortem human subject (PMHS) tests, volunteer data, and organ-level experiments, with wider scatter and more condition-specific applicability. A credibility corridor built for a Hybrid III model cannot be transferred to an HBM without re-establishing the evidence base.

Restraint system models and their coupling to the occupant

The restraint system model comprises the seat belt (webbing, retractor, pretensioner, load limiter, buckle, anchor points), the airbag (module, inflator, folded cushion, venting), the seat structure, and the contact interfaces between these and the occupant. Each subsystem has its own modeling fidelity requirements. Belt webbing is typically modeled with membrane or shell elements and a defined force–strain characteristic; the retractor and load limiter are represented by discrete elements with force–displacement curves; the pretensioner is a time-dependent pull-in. The airbag is the most credibility-sensitive restraint component: its deployment kinematics depend on the inflator mass flow curve, the folding pattern, the cushion permeability, and the venting behavior, and small changes in any of these can shift occupant loading by a meaningful margin.

The critical point is that restraint models are not validated in isolation from the occupant. A belt model validated by pulling on a rigid fixture tells you about the belt, not about the belt–occupant interaction. The interface between the belt and the occupant's chest, the airbag and the occupant's head, and the seat and the occupant's pelvis are where the coupled response is determined. These interfaces are credibility-critical because errors on either side compound rather than cancel.

The simulation chain: sled, body-in-white, and full vehicle

Occupant CAE is executed across a hierarchy of environments. At the lowest level, the sled test isolates the occupant and restraint system on a rigid buck subjected to a prescribed acceleration pulse. The sled model's credibility depends on the accuracy of the pulse, the fidelity of the seat and restraint models, and the occupant model itself. Because the boundary conditions are prescribed rather than computed, the sled is the cleanest environment for validating the occupant–restraint couple, and it is where most component-level corridors are established.

At the next level, the occupant and restraint models are embedded in a body-in-white (BIW) or trimmed-body model subjected to a full-vehicle crash pulse. Here the boundary conditions are no longer prescribed: the pulse at the seat is an output of the structural crash simulation, and it carries its own uncertainty. The seat mounting stiffness, the floor deformation, and the steering column collapse all become part of the loading path. The occupant response is now sensitive to structural model fidelity in a way that the sled environment does not capture.

At the top level, the full-vehicle model includes the complete structure, closures, powertrain, and interior, with the occupant and restraint models embedded at their physical locations. This is the environment in which regulatory and consumer-rating assessments are ultimately made, but it is also the environment with the largest accumulated uncertainty. A corridor that holds at the sled level may not hold at the full-vehicle level, not because the occupant model changed, but because the loading path changed.

Identifying credibility-critical interfaces in practice

The interfaces highlighted in the diagram are the points where credibility is most easily lost. Consider a concrete example: a frontal 50 km/h sled model with a Hybrid III 50th percentile occupant, a three-point belt with a 4 kN load limiter, and a driver airbag. The occupant model has been validated against certification tests; the belt model has been validated against a rigid pull test; the airbag model has been validated against a static deployment test. Each component has evidence. Yet the coupled sled response may still miss the physical test, because the belt–chest contact stiffness was calibrated on a rigid fixture and the airbag–head contact timing depends on the interaction between the deploying cushion and the occupant's forward excursion — neither of which is captured by the isolated component validations.

A practical way to expose these interface sensitivities is to run a controlled perturbation study before defining the corridor. Vary the belt–occupant friction coefficient within a plausible range (for example, 0.2 to 0.4), vary the airbag vent area by ±10%, and vary the seat mounting stiffness by ±15%. Record the resulting spread in the occupant injury metrics — head resultant acceleration, chest deflection, and pelvis acceleration. The interfaces that produce the largest metric spread are the ones that most need dedicated validation evidence and the tightest corridor control. This perturbation study is itself a credibility activity: it identifies where the model is sensitive and therefore where the validation evidence must be strongest.

From landscape to corridor

The model landscape described here is the prerequisite for corridor definition. A corridor is only meaningful when it is attached to a specific model configuration at a specific level of the simulation chain, with the interfaces that dominate its response explicitly identified. The next step in the credibility workflow is to translate this landscape into quantitative acceptance bands — but that translation is only defensible if the analyst can state, for each interface, what evidence exists and what remains an assumption. The landscape map is therefore not a preliminary formality; it is the structural basis on which every subsequent corridor decision rests.