3. THOR dummy architecture

The THOR (Test Device for Human Occupant Restraint) dummy represents a fundamental rethinking of the frontal impact anthropomorphic test device. Where Hybrid III evolved incrementally from the 1970s Part 572 design, THOR was developed from the outset around a modular architecture, a more human-like kinematic chain, and an instrumentation philosophy that treats the dummy as a measurement platform rather than a pass/fail indicator. Understanding this architecture is essential before interpreting any THOR injury metric, because the physical construction directly determines what the sensors can and cannot measure.

Modular architecture: assemblies, not a monolithic body

THOR is organized into discrete, replaceable modules: head assembly, neck, upper and lower torso, pelvis, and four limb assemblies. Each module carries its own sensor set and its own structural interface, and modules can be removed and replaced without disturbing adjacent instrumentation. This is a deliberate departure from Hybrid III, where the head, neck, and upper torso are mechanically and electrically intertwined to the point that a neck change often requires partial upper-torso disassembly.

The practical consequence is calibration and serviceability. A THOR head can be certified on its own drop tower, a neck can be certified on its own pendulum, and the certified module can then be installed on a torso that has been separately certified. The interface between modules is a defined mechanical datum with a defined bolt pattern and a defined connector pinout, so the assembled dummy's geometry is repeatable. In a test lab, this means a damaged lower leg can be swapped between runs without re-certifying the entire dummy, and a channel failure can be traced to a single module rather than to a harness running the length of the body.

The modularity also enables configuration variants. The same torso and pelvis can accept different arm and leg builds, and the neck can be exchanged for a different stiffness variant to match a specific occupant population or a specific research question. This is why THOR is often described as a family of builds rather than a single part number.

Improved biofidelity: where the human likeness actually lives

Biofidelity improvements in THOR are concentrated in the regions that dominate frontal restraint loading: the head–neck junction, the clavicle and shoulder, the rib cage, and the pelvis–femur interface.

The neck uses a flexible column with a distributed stiffness profile and a geometry that produces a more human-like flexion–extension moment–angle response, particularly in the early phase of head rotation. Hybrid III's neck is a segmented rubber-and-metal column whose response is comparatively stiff and whose hysteresis differs from human data. In practice, the THOR neck changes the timing and magnitude of head rotation in a frontal pulse, which propagates into head injury criteria and into the interaction between the head and a deploying airbag.

The shoulder and clavicle assembly is articulated. THOR's clavicle can deflect and rotate under belt loading, allowing the shoulder to retract and the belt to migrate in a way that Hybrid III's rigid upper torso cannot reproduce. This matters directly for belt-restraint tests: the belt-to-shoulder interaction determines chest deflection magnitude and the load path into the spine, and a rigid shoulder tends to over-constrain the belt and under-predict excursion.

The rib cage uses a different rib geometry and a different damping treatment, and it is instrumented for multi-point deflection rather than a single chest-potentiometer reading. THOR typically carries chest deflection measurements at multiple locations plus chest acceleration, so the deflection field across the sternum and ribs can be reconstructed rather than reduced to one number. The pelvis and femur interfaces are also revised to better represent hip loading and femur kinematics under knee-bolster and belt loading.

Key departures from Hybrid III

The most consequential differences are not cosmetic. They change what a test can conclude.

  • Instrumentation density and channel count. Hybrid III carries a relatively small set of channels: head accelerations, chest accelerations, chest deflection, upper and lower neck loads, lumbar loads, and femur loads. THOR adds multi-point chest deflection, additional spinal load measurement, improved head angular rate sensing, and in many builds a more extensive lower-extremity load set. The result is a much higher channel count per test, which raises data-acquisition, filtering, and channel-integrity requirements.
  • Kinematic chain and posture. THOR's articulated shoulder and revised pelvis produce a different seated posture and a different response to belt geometry. A restraint system tuned on Hybrid III will not necessarily produce the same occupant excursion or the same chest deflection on THOR, and the difference is largest in oblique and far-side frontal loading.
  • Injury criteria. Because THOR measures a deflection field and additional load paths, its injury assessment uses criteria that Hybrid III cannot support, including rib-based deflection metrics and criteria derived from the additional spinal and shoulder channels. Hybrid III criteria such as HIC and single-point chest deflection remain computable, but they do not exploit the added measurement capability.
  • Certification and handling. Modular certification means the lab's calibration schedule is organized per module, and the dummy's mass properties and joint stiffnesses must be verified per build. Handling procedures, lifting points, and sensor zeroing differ from Hybrid III practice.
THOR Modular Build and Measurement Chain THOR — modular assemblies Head assembly tri-axial accel + angular rate Neck flexible column, upper/lower loads Upper torso articulated clavicle / shoulder multi-point rib deflection Lower torso / spine spinal load channels Pelvis hip / femur interface Four limb assemblies femur / tibia load cells mechanical datum + connector Data acquisition high channel count per-module harness channel integrity check filtering per channel Injury assessment deflection-field metrics spinal / shoulder loads HIC still computable but not sufficient Hybrid III — reference build Head + neck + upper torso mechanically intertwined Rigid upper torso fixed shoulder geometry Single chest potentiometer one deflection value Lower channel count simpler DAQ, fewer checks Whole-dummy certification neck change disturbs torso Restraint tuned on HIII ≠ same response on THOR departure departure Each THOR module is certified independently; the assembled build defines the measurement set and the applicable injury criteria.

What this means in practice: a module swap and its verification

Consider a lab preparing a frontal sled test with a belt and driver airbag, where the previous run produced an out-of-range lower-neck load channel. The workflow illustrates how the architecture is used.

First, isolate the fault at module level. The neck module is removed at its mechanical datum, and the connector is inspected for pin damage and continuity. Because the neck carries its own load cells, the suspect channel can be exercised on the bench with a known shunt or a calibrated load, independent of the torso harness. If the channel responds correctly on the bench, the fault is in the harness or the torso-side connector, not the transducer.

Second, replace and re-certify only the affected module. A spare neck is installed, and the neck's own pendulum certification is checked against its certificate. The assembled dummy's head–neck geometry is then verified against the build drawing using the specified datum references, and the head is zeroed in the seated reference posture.

Third, verify the assembled build before the run. The checks that matter are: all channels present and responding to a tap test; zero offsets within the build's tolerance; joint stiffness at the shoulder and pelvis within the build specification; and total mass and center of gravity within the build tolerance. A common failure symptom at this stage is a channel that reads plausibly but is scaled incorrectly because the module was installed with a different bridge configuration than the DAQ expected — the tap test will show a response, but the magnitude will be wrong, and only a shunt calibration exposes it.

Fourth, interpret results with the correct criteria set. After the run, chest deflection must be examined as a field across the measurement locations, not as a single peak. A localized deflection concentration with a modest single-point peak is a different injury signal than a broad deflection with the same peak, and Hybrid III's single-channel output cannot distinguish them. Similarly, the shoulder and clavicle channels should be reviewed for load path anomalies that would indicate belt migration or submarining, which the rigid Hybrid III shoulder tends to mask.

Do not carry Hybrid III acceptance thresholds directly onto a THOR build. The different kinematic chain, different chest construction, and additional channels mean that a response which passes on Hybrid III may fail on THOR, and vice versa. Criteria must be taken from the THOR-specific assessment method associated with the build in use.
When comparing THOR and Hybrid III results for the same restraint design, hold the pulse, belt geometry, and seat position constant, and report the channel list alongside the metrics. A metric without its channel definition and filter class is not reproducible.

The architectural lesson is that THOR's modularity is not merely a maintenance convenience. It defines the certification boundary, the harness topology, and the set of measurements available for injury assessment. Treating the dummy as a set of independently certified modules — each with its own datum, connector, and calibration record — is what makes the higher channel count and the richer criteria set usable in a production test environment.

← Previous 📖 Contents Next →