2. Crash test data sources and instrumentation

Restraint performance cannot be evaluated from video alone. The physical evidence that determines whether a belt, airbag, or pretensioner did its job comes from a synchronized instrumentation chain: accelerometers that capture the vehicle and occupant kinematics, load cells that measure the forces transmitted through belts and seats, and a data acquisition system (DAS) that samples, filters, and stores all channels on a common time base. This chapter explains what each sensor actually measures, how the measurement chain is configured, and how the raw signals become the injury metrics used in restraint dashboards.

Accelerometers: capturing the crash pulse and occupant response

Accelerometers convert acceleration into a voltage or digital count proportional to the applied acceleration. In a full-scale frontal sled or vehicle crash test, the most important channel is the vehicle or sled acceleration pulse, measured by a uniaxial accelerometer rigidly mounted to the vehicle structure or sled frame, typically in the longitudinal (X) direction. This pulse defines the crash severity and is the input to every downstream calculation. Occupant-mounted accelerometers — head, chest, and pelvis — capture the body's response and are the basis for head injury criteria and chest deflection correlation.

Two conventions must be fixed before any data is interpreted. First, the sign convention: in this course, positive X points forward in the vehicle direction of travel, and a frontal impact produces a deceleration pulse that is negative in the vehicle frame. Second, the mounting orientation: an accelerometer's sensitive axis must be aligned with the measurement direction, and the channel metadata must record that orientation. A head accelerometer mounted with its sensitive axis reversed will produce a mirrored signal that looks plausible but yields a wrong HIC. Always verify orientation against the test setup sheet before processing.

Accelerometer signals are sampled at high rates — commonly 10 kHz or more for head and chest channels — because the pulse contains high-frequency content that matters for injury metrics. The raw signal is filtered according to the relevant standard before use: head acceleration for HIC is typically filtered with a CFC 1000 (channel frequency class) low-pass filter, while chest and vehicle channels often use CFC 180 or CFC 60 depending on the analysis. The filter class is not cosmetic; it changes the peak values and therefore the reported metrics.

Load cells: measuring restraint forces

Load cells measure force directly and are the primary evidence for how much load a restraint system applied. In restraint testing, the most common configurations are:

  • Belt load cells — in-line transducers spliced into the shoulder and lap belt webbing, measuring tension in the belt. They reveal pretensioner firing, force limiting, and the timing of belt engagement.
  • Seat and anchor load cells — mounted at seat tracks or belt anchor points to capture the reaction forces transmitted into the structure.
  • Steering column and knee bolster load cells — measuring forces from the driver's interaction with the interior.

Belt load cells are calibrated in tension, and their output is typically reported in kilonewtons. A shoulder belt load cell on a force-limited retractor might show a rapid rise to a plateau around 4–6 kN during the ride-down phase; the plateau value and its duration are exactly what the restraint engineer needs to judge whether the force limiter is behaving as designed. Load cell channels are usually sampled at 10 kHz as well, and are filtered with CFC 600 or CFC 1000 depending on the channel's role.

From acceleration to HIC: the head injury calculation

The Head Injury Criterion (HIC) is computed from the resultant head acceleration. The standard definition is:

HIC = max over t1,t2 of [ (t2 - t1) * ( (1/(t2-t1)) * ∫ a(t) dt )^2.5 ]

where a(t) is the resultant head acceleration in g,
      filtered per CFC 1000,
      and (t2 - t1) is constrained to 15 ms or less.

In practice, the calculation is done on the filtered resultant acceleration vector, and the window search is performed over all valid 15 ms intervals. A HIC of 700 is a common regulatory threshold for the 15 ms window in frontal occupant protection evaluations; values above that indicate elevated head injury risk. The key practical point is that HIC is extremely sensitive to the filter class and to the resultant construction — computing HIC from a single-axis channel instead of the resultant, or from unfiltered data, produces numbers that are not comparable to regulatory results.

Data acquisition: synchronization, sampling, and channel integrity

The DAS is the backbone that makes the sensors usable. Its job is to sample every channel at its required rate, timestamp all samples on a common clock, apply anti-alias filtering, and store the data with complete metadata. In a typical sled or vehicle test, the DAS is triggered by a hardware event — the sled release or a breakaway switch — and records a pre-trigger buffer so the moments before impact are captured.

Three configuration details determine whether the data is trustworthy. First, sample rate versus channel bandwidth: the sample rate must be at least twice the highest frequency of interest, and in practice much higher for head channels. Second, anti-alias filtering must be applied before sampling, not after, or high-frequency content will fold back into the signal. Third, channel metadata — sensor ID, location, orientation, sensitivity, filter class, and units — must be recorded with the data. A channel with a missing orientation or an unrecorded sensitivity is effectively unusable for injury analysis.

Before every test, run a shunt calibration or a known-input check on each load cell and accelerometer channel, and verify the trigger and pre-trigger buffer. A channel that reads zero throughout the test is usually a wiring or excitation fault, not a genuine zero-force result.

Putting the chain together: a worked example

Consider a frontal sled test at 50 km/h with a belted Hybrid III 50th percentile male. The instrumentation set includes a sled frame accelerometer (X, CFC 60), a head accelerometer triad (CFC 1000), a chest accelerometer (CFC 180), a shoulder belt load cell, and a lap belt load cell. The DAS samples all channels at 10 kHz with a 200 ms pre-trigger buffer.

After the test, the analyst first checks the sled pulse: it should show a deceleration ramp reaching roughly 30–40 g within the first 30–50 ms, consistent with the target crash severity. Next, the belt load cells are examined: the shoulder belt should show a rapid rise as the pretensioner fires, followed by a force-limited plateau. If the shoulder belt load cell shows no rise at all, the likely causes are a disconnected transducer, a failed pretensioner, or a belt routing error — each of which is diagnosed by checking the channel's pre-test calibration and the physical belt path. Finally, the head resultant acceleration is filtered at CFC 1000 and HIC is computed over the 15 ms window; the analyst compares the result against the threshold and against the chest deflection channel to judge overall restraint performance.

Restraint Test Instrumentation Chain Sensors Sled / Vehicle Accel X-axis, CFC 60 Head Accel Triad CFC 1000 Chest Accel CFC 180 Shoulder Belt Load Cell Tension, kN Lap Belt Load Cell Tension, kN Data Acquisition System Anti-alias Filter before sampling Sample @ 10 kHz common clock Trigger + Pre-buffer 200 ms pre-trigger Channel Metadata ID, axis, sensitivity, CFC Stored Raw Data Injury Metrics HIC (15 ms) resultant head accel threshold ≈ 700 Belt Force Plateau force limiter behavior ≈ 4–6 kN example Chest Response accel / deflection correlation Restraint Verdict pass / investigate per channel evidence Sign convention: +X forward; frontal impact produces a negative vehicle deceleration pulse. All channels share one time base.

Common failure modes and how to diagnose them

Instrumentation problems usually announce themselves as implausible signals rather than as outright errors. A head accelerometer channel that is flat during the impact but shows noise elsewhere suggests a loose connector or a failed excitation supply. A belt load cell that rises before the trigger event suggests a pre-load or a wiring short. A sled pulse that looks too smooth or too low in peak may indicate a filter class mismatch or a saturated channel. The diagnostic habit that resolves most of these is to compare each channel against its expected physical behavior and against a redundant channel — for example, cross-checking the shoulder belt load cell against the chest acceleration timing. When the two disagree, the instrumentation chain, not the restraint, is usually the first suspect.

Do not compute HIC from a single-axis head channel or from unfiltered data. The resultant vector and the CFC 1000 filter are part of the definition; omitting either produces a number that cannot be compared with regulatory thresholds or with other tests.

The instrumentation chain described here is the foundation for everything that follows in restraint analytics. Once the sled pulse, belt forces, and head and chest responses are captured on a synchronized, properly filtered time base, those channels can be reduced to the injury metrics and time-series features that populate a crash response dashboard — and the quality of that dashboard is bounded by the quality of the measurement chain behind it.

← Previous 📖 Contents Next →