2. Align Vehicle-to-Vehicle Impact Timing with Occupant Injury Curves

Why temporal alignment matters

An EDR-derived timeline tells you what the vehicle did: when the event trigger fired, when delta-v began to accumulate, and when the airbag command was issued. An occupant injury curve tells you what the occupant experienced: the head acceleration pulse that produces HIC, the chest deflection history that drives rib loading, and the timing of peak values. Each is internally consistent but they live on different clocks. The reconstruction clock is anchored to the EDR's pre-trigger buffer and its sample grid; the injury clock is anchored to whatever the test or simulation used as its time zero. Before you can say "the head reached peak acceleration 40 ms after airbag deployment," you must prove that both curves share a common origin and that neither has been shifted by filtering, resampling, or a sign convention.

The task in this case is therefore not to compute HIC or to re-derive delta-v — those belong to other steps. It is to register two time series onto one axis, identify the injury onset relative to the impact phase, and then test whether the alignment is defensible. A defensible alignment has three properties: a documented anchor event, a residual check that the anchor actually lands where the physics says it should, and a sensitivity note describing how much a plausible anchor error would move the injury-onset conclusion.

Establishing the anchor and the sign convention

Pick one anchor event that appears unambiguously in both domains. The most reliable choice is usually the airbag deployment command (or the squib firing time), because the EDR records it to the millisecond and the occupant model or sled test typically logs it as a discrete event. A second-best anchor is the delta-v onset — the first sample where the filtered acceleration exceeds the trigger threshold — but this is softer because it depends on the filter and threshold used.

Fix the sign convention before touching any data. In this case we adopt the SAE J211 / ISO 6487 convention: positive acceleration points forward (toward the direction of travel for a frontal impact), and the crash pulse is therefore negative during frontal deceleration. Chest deflection is positive toward the sternum (anterior loading). State this in the report; a sign flip between the vehicle pulse and the occupant pulse is one of the most common causes of a "the occupant loaded before the car decelerated" absurdity.

Define time zero explicitly. A practical convention: t = 0 at the airbag command, with the pre-crash phase negative and the post-crash phase positive. This makes the injury-onset question directly readable: "head peak at t = +38 ms" means 38 ms after the command.

Worked example: registering a frontal impact

The table below is a teaching example with synthetic values chosen to be physically plausible for a 50 km/h frontal barrier impact of a mid-size sedan. It is not measured data.

Read the sequence: the vehicle begins decelerating about 12 ms before the airbag command (expected — the command is issued after the algorithm confirms the crash), the head reaches peak acceleration 38 ms after the command, and chest deflection peaks 14 ms later. The head peak falling inside the airbag's inflation window and after delta-v onset is the physically expected ordering. If your registration produced a head peak before delta-v onset, the anchor or sign is wrong.

Residual check and sensitivity

After alignment, compute a residual: the difference between the time the occupant model predicts for a known event and the time the vehicle data implies. A useful residual is the airbag-contact interval — the gap between command and the first head-acceleration rise attributable to bag contact. In the example, head acceleration begins to rise at about +18 ms, giving a 18 ms command-to-contact interval, which is consistent with a typical inflator fill time. If your registration yields a negative or sub-5 ms interval, suspect a clock offset or a mislabeled anchor.

Then run a sensitivity: shift the occupant curve by ±5 ms and re-read the injury onset. If the conclusion ("head peak occurs after delta-v onset and within the bag window") survives a ±5 ms shift, the alignment is robust enough to report. If it flips, the anchor is too soft and you should switch to the squib firing time or a hard mechanical event.

Verifying temporal consistency

Three checks close the loop. First, ordering: delta-v onset must precede the airbag command, which must precede head and chest peaks. Second, interval plausibility: the command-to-head-peak interval should fall within the inflator fill plus bag-contact range for the restraint system in question; a value far outside that range signals a clock offset. Third, cross-domain residual: the time at which the occupant model's chest deflection begins to rise should coincide with the time the vehicle pulse's deceleration first exceeds the level that would load the belt — a soft check, but a useful one when the airbag anchor is unavailable.

If a check fails, the fix is almost always one of four things: a unit mismatch (ms vs. s), a filter-induced group delay (a CFC 60 filter shifts the pulse by several milliseconds relative to CFC 1000), a resampling offset (linear interpolation onto a coarser grid moves peaks), or a sign flip. Log which one you corrected; the correction is part of the reconstruction's provenance.

Deliverable: alignment record

The following record is the minimum a reviewer needs to audit the registration. Fill it once per reconstruction and attach it to the timeline figure.

With the record complete, the injury-onset statement becomes auditable: "head peak at +38 ms relative to airbag command, after delta-v onset at −12 ms, robust to ±5 ms anchor error." That sentence — not the raw curves — is what the reconstruction is defending.