2. Crash test fundamentals

Before any AI-assisted simulation can be trusted, the physical crash test it is calibrated against must itself be well understood. A crash test is not a single event but a measurement system: a controlled impact, a set of transducers, a data-acquisition chain, and a post-processing pipeline that converts raw voltages into the metrics regulators and engineers use to judge occupant protection. This chapter establishes that foundation, because every HBM (Human Body Model) validation exercise ultimately compares simulated kinematics and injury criteria against these measured references.

Test Types and What Each One Is For

Crash tests are categorized by impact direction, barrier configuration, and purpose. The most common regulatory and consumer-rating configurations include:

  • Full frontal rigid barrier (e.g., 100% overlap at approximately 56 km/h into a fixed rigid wall): exercises the full front structure, restraint systems, and airbag timing symmetrically.
  • Offset deformable barrier (ODB) (e.g., 40% overlap at 64 km/h into a honeycomb deformable barrier): concentrates loading on one side, exposing structural asymmetry and intrusion.
  • Side impact (moving deformable barrier striking the driver side at roughly 50 km/h): evaluates door intrusion, side airbags, and lateral occupant loading.
  • Pole impact: a narrow rigid pole concentrates energy into a small area, testing side structure and curtain airbags.
  • Rear impact: primarily a whiplash and seat/head-restraint assessment.
  • Pedestrian and cyclist protection tests: subsystem impacts (head, legform) against the vehicle front end.

Each configuration defines a test matrix: impact speed, barrier mass and stiffness, overlap percentage, dummy type and seating position, and pass/fail criteria. The speed and barrier stiffness are not arbitrary — they set the total kinetic energy and the deceleration pulse shape that the vehicle and occupants experience.

Instrumentation: The Measurement Chain

The measurement chain begins at the transducer and ends at a validated time-history. Key instrument categories:

  • Accelerometers on the vehicle structure (B-pillar, sill, tunnel, engine cradle) and on the dummy (head, chest, pelvis). They measure acceleration in g, later integrated to velocity and displacement.
  • Load cells in barriers, seat belts (belt force), and steering columns to capture force time-histories.
  • Potentiometers and string pots for displacement, such as steering-column collapse or pedal travel.
  • High-speed cameras (1,000–10,000 fps) for kinematic tracking and event verification.
  • Dummy internal sensors — chest deflection potentiometers, neck load cells, and abdomen pressure sensors — that feed injury criteria directly.

Transducer selection is governed by range, sensitivity, and frequency response. A head accelerometer might be rated to 2,000 g with a resonant frequency well above the signal bandwidth of interest (typically filtered to CFC 1000 per SAE J211). Mounting rigidity matters: a loosely mounted accelerometer will resonate and corrupt the signal.

Data Acquisition and Signal Conditioning

Signals pass through amplifiers and anti-alias filters before digitization. The standard reference for crash-test signal processing is SAE J211, which defines channel classes (CFC 60, 180, 600, 1000) and the corresponding low-pass filter characteristics. A typical acquisition setup samples at 10–20 kHz per channel, well above the filter corner, to avoid aliasing.

Two practical steps dominate data quality:

  1. Zeroing and offset correction: each channel is zeroed immediately before impact, and any DC offset is removed so integration does not drift.
  2. Filtering and integration: acceleration is filtered to the appropriate CFC class, then integrated to velocity and displacement. Integration drift is a classic failure mode — a small residual offset produces a large displacement error over the impact duration.
Always verify that integrated velocity returns to a physically plausible value (e.g., the vehicle's post-impact velocity) before trusting displacement-derived metrics. A mismatch usually indicates an offset or filter error, not a real phenomenon.

Key Safety Metrics

Raw time-histories are converted into injury criteria that map to regulatory limits. The most important are:

MetricDefinitionTypical Limit (illustrative)
HIC (Head Injury Criterion)Integral of resultant head acceleration over a 15 ms windowHIC₁₅ ≤ 700 (frontal)
Chest deflectionPeak sternum-to-spine compression≤ 50–60 mm depending on dummy
Chest acceleration (3 ms clip)Peak chest acceleration sustained 3 ms≤ 60 g
Femur axial forceCompressive load along the femur≤ 10 kN
Neck injury (Nij)Combined axial force and bending moment≤ 1.0

These criteria are computed from filtered channels, so the filter class directly affects the reported value. HIC, for example, is conventionally computed from CFC 1000 head acceleration. Reporting a HIC from an unfiltered or differently filtered channel is a common source of disagreement between labs.

From Measurement to HBM Validation

For AI-assisted simulation, the crash test provides the ground truth: measured occupant kinematics (from high-speed video and sensor fusion), restraint forces, and injury criteria. An HBM is validated by reproducing the same impact pulse and comparing simulated head/chest/pelvis trajectories and criteria against these measurements. The quality of that validation is bounded by the quality of the test data — which is why instrumentation, filtering, and metric definitions must be documented and consistent.

Crash Test Measurement Chain → HBM Validation Impact Event Barrier + vehicle speed / overlap Transducers Accelerometers, load cells potentiometers, cameras Signal Conditioning Amplify, anti-alias zero / offset Data Acquisition 10–20 kHz sampling SAE J211 channels Filtering CFC 60 / 180 / 600 / 1000 integrate → v, d Injury Metrics HIC, chest deflection Nij, femur force HBM Validation Compare kinematics and criteria vs. test Iterate: refine HBM parameters, re-check against measured channels Key checkpoints • Zero channels before impact; verify integrated velocity is physically plausible. • Report metrics with the filter class used (e.g., HIC from CFC 1000).

Practical Data-Quality Checks

Before a test dataset is released for HBM validation, engineers run a standard set of checks. A reproducible example workflow looks like this:

1. Load raw channels (accel, load, displacement) at native sample rate.
2. Apply zero-offset correction using a 50 ms pre-trigger window.
3. Filter each channel to its assigned CFC class per SAE J211.
4. Integrate head/chest acceleration to velocity and displacement.
5. Verify: post-impact velocity ≈ expected residual; displacement returns near zero.
6. Compute HIC, chest deflection, Nij, femur force from filtered channels.
7. Cross-check camera-derived kinematics against sensor-derived trajectories.

Typical failure symptoms and their causes: a displacement that drifts steadily upward usually indicates a residual DC offset; a HIC value that changes dramatically when the filter class is switched indicates the signal contains high-frequency content near the filter corner; a camera trajectory that diverges from the accelerometer-derived trajectory often points to a mounting or synchronization error. Each of these must be resolved before the data can serve as a validation reference.

Do not treat a single test as ground truth for HBM validation. Repeatability varies with dummy positioning, barrier alignment, and sensor mounting. Use multiple tests and report the spread of key metrics so the simulation is compared against a distribution, not a single number.

With test types, instrumentation, acquisition, and metrics established, the next step in the course is to understand how these measured references are structured into validation datasets and how HBM outputs are compared against them quantitatively.

← Previous 📖 Contents Next →