2. Anthropomorphic test devices versus human body models

Occupant protection engineering relies on two fundamentally different classes of surrogates to predict how the human body responds to crash loads: anthropomorphic test devices (ATDs), commonly called crash test dummies, and human body models (HBMs), which are computational representations of human anatomy built with finite element (FE) methods. Neither surrogate is universally superior. Understanding where each excels, where each breaks down, and how they are combined in modern safety development is essential for interpreting test results and making defensible design decisions.

What Each Surrogate Actually Is

An ATD is a physical, instrumented mechanical assembly. A Hybrid III 50th percentile male dummy, for example, consists of a steel and aluminium skeleton, polymer flesh components, and a network of accelerometers, load cells, and potentiometers. Its mass distribution, joint stiffness, and geometry are fixed by regulation (49 CFR Part 572 in the US, ECE R94/R95 in Europe). Because it is physical, it can be placed in a real vehicle on a real sled or crash test, and its outputs are measured directly in engineering units: head resultant acceleration in g, chest deflection in millimetres, femur axial force in kilonewtons, and the Hybrid III neck injury criterion N_ij.

An HBM is a numerical model. The THUMS (Total Human Model for Safety) family, the GHBMC models, and open-source alternatives such as the VIVA+ models discretise bones, brain, internal organs, muscles, and major vasculature into millions of elements. Material laws describe cortical bone as elastic–plastic, brain tissue as viscoelastic, and soft organs with hyperelastic formulations. An HBM is solved in an explicit FE code such as LS-DYNA or Radioss, and its outputs are field variables: strain in the corpus callosum, von Mises stress in the liver, rib cortical bone strain, and intracranial pressure gradients.

Capability Comparison

Where ATDs Fall Short

The most consequential limitation of ATDs is that they are not biofidelic in all loading directions. The Hybrid III was designed primarily for frontal impacts. Its neck responds poorly in lateral and oblique loading, its lumbar spine is stiff compared with a living occupant, and it has no representation of internal organs. Consequently, it cannot predict injuries such as liver laceration, aortic rupture, or diffuse axonal injury — injuries that dominate real-world trauma but leave no signature in the dummy's measurement channels.

A second limitation is anthropometric rigidity. A 50th percentile male dummy represents a single body size. Testing a 5th percentile female occupant requires a different physical dummy with different mass and geometry, and the range of available sizes is coarse. Obese occupants, children in booster seats, and elderly occupants with reduced bone strength are poorly represented.

A third limitation is measurement granularity. An ATD reports what its transducers can measure. If a transducer is not installed at a location, no data exists there. This makes it difficult to diagnose why an injury occurred, as opposed to simply recording that a threshold was exceeded.

Where HBMs Fall Short

HBMs are not yet validated to the same level of confidence as ATDs for regulatory decision-making. Validation is typically performed against a limited set of cadaveric impact tests, and the corridors are wide. A model that matches a chest deflection corridor may still mispredict rib fracture location because fracture is governed by local cortical bone strain, which is sensitive to element size, material parameters, and mesh quality.

HBMs are also computationally expensive. A full-body THUMS model with 2 million elements may require 10–40 hours on a 64-core node for a 150 ms frontal impact simulation. This limits the number of design iterations that can be explored. Additionally, HBMs require careful boundary condition definition — seat belt routing, airbag contact, friction coefficients, and pre-impact posture — and small errors in these inputs can produce large differences in output.

Finally, HBMs lack a standardised injury criterion. While tissue-level thresholds exist in the literature (e.g., 1.5% strain for rib cortical bone, 0.2–0.3 strain for brain axonal injury), there is no regulatory equivalent of HIC or N_ij for an HBM. This makes pass/fail decisions ambiguous.

Complementary Use in Practice

The most effective engineering workflow uses ATDs and HBMs in a deliberate sequence. ATDs provide the regulatory baseline and the physical validation anchor. HBMs provide the mechanistic insight and the population coverage that ATDs cannot deliver.

A typical complementary workflow for a frontal occupant protection programme proceeds as follows:

1. Sled test with ATD. Run a production seat and restraint system with a Hybrid III 50th male at 56 km/h delta-v. Record head acceleration, chest deflection, and femur loads. Verify that HIC < 700, chest deflection < 63 mm, and femur force < 10 kN (example targets based on FMVSS 208).
2. Build and correlate the HBM. Reconstruct the same sled environment in LS-DYNA with a GHBMC 50th male model. Apply the measured sled pulse as the boundary condition. Compare HBM-predicted chest deflection and head acceleration against the ATD measurements. Adjust belt stiffness, friction, and seat cushion properties until the HBM matches the ATD within an acceptable tolerance (e.g., ±15% on peak chest deflection).
3. Extend to unrepresented populations. Morph the validated HBM to a 5th percentile female and a 95th percentile male. Re-run the same load case. Examine rib strain and liver stress, which the ATD cannot measure.
4. Identify injury mechanisms. If the morphed HBM predicts rib cortical bone strain above 1.5% at a specific rib level, investigate whether the belt path or seat geometry concentrates load there. This is a design insight that no ATD test could provide.
5. Iterate design and re-verify with ATD. Modify the restraint geometry, then confirm that the ATD still meets regulatory targets. The HBM guides the design change; the ATD confirms compliance.

Decision Guide: Which Surrogate to Use

The choice depends on the question being asked. If the question is "Does this design pass the regulation?", the ATD is the correct tool because it is the regulatory instrument. If the question is "Why did this injury occur, and how would it change for a different occupant?", the HBM is the correct tool because it provides mechanistic, population-scalable output.

In practice, the two are not competitors. The ATD defines the compliance envelope; the HBM explores the design space within and beyond that envelope. A programme that uses only ATDs risks missing injuries that the dummy cannot detect. A programme that uses only HBMs risks producing designs that fail regulatory tests because the model's boundary conditions or material parameters were not validated against physical hardware.

Key Takeaways

• ATDs are physical, regulated, and directly measurable; they are the compliance instrument but have limited biofidelity and fixed anthropometry.
• HBMs are computational, scalable, and provide tissue-level output; they offer mechanistic insight but lack regulatory standardisation and require careful validation.
• The two are complementary: ATDs anchor physical reality and regulatory compliance, while HBMs extend coverage to unrepresented populations and reveal injury mechanisms that ATDs cannot detect.
• A robust workflow correlates the HBM to a physical ATD test before using the HBM for design exploration, ensuring that the model's boundary conditions and material parameters are grounded in measured data.