4. Raw crash test data sources

Every injury-limit evaluation ultimately rests on a chain of measurements taken during a crash event that lasts 80–150 ms. Before any Excel pass/fail logic can be trusted, the analyst must understand what each sensor physically measures, how it is mounted, how its signal is conditioned and digitized, and how the resulting file is structured. This chapter builds that foundation: the three primary transducer families used in vehicle crash testing, the data acquisition (DAQ) chain that carries their signals, and the file formats you will actually open when building an automated evaluation workbook.

Accelerometers: The Primary Injury Metric Source

Accelerometers dominate occupant-safety evaluation because most regulated injury criteria are acceleration-based: Head Injury Criterion (HIC), 3 ms chest clip, and the various body-region acceleration limits in FMVSS 208, ECE R94/R95, and Euro NCAP protocols. A crash-test accelerometer is typically a piezoresistive or capacitive MEMS device with a full-scale range of ±500 g to ±2000 g for vehicle structure and ±250 g to ±500 g for occupant locations. The critical specification is not just range but frequency response: per SAE J211/1, accelerometers used for injury assessment must be filtered with a CFC (Channel Frequency Class) filter — CFC 1000 for head and rigid-body channels, CFC 180 for chest and femur channels, CFC 60 for some occupant-compartment measurements.

Mounting matters as much as the sensor. A head accelerometer is bonded to the skull of the Hybrid III or THOR dummy at the head CG; a chest accelerometer sits on the sternum or spine box; vehicle accelerometers are rigidly bolted to the sill, B-pillar, or floorpan. Any compliance in the mount introduces a resonance that corrupts the high-frequency content and inflates HIC. In the DAQ file, each accelerometer appears as a channel with a sign convention defined by the test house — typically positive X forward, positive Y to the right, positive Z upward in the vehicle coordinate system, but dummy-local axes may differ. Always confirm the sign convention from the test report before applying any limit comparison, because a reversed polarity silently inverts a pass into a fail.

Load Cells: Forces at Belt, Seat, and Structure Interfaces

Load cells measure force directly and are essential where injury criteria are force-based or where belt geometry must be characterized. Common locations include seat-belt webbing tension load cells (measuring belt force in kN), seat cushion and back-panel load cells, steering-column axial force, and pedal/steering-wheel impact forces. A typical belt load cell is a strain-gauge bridge rated to 20–30 kN with a CFC 600 or CFC 1000 filter depending on the protocol.

Load cells output a low-level millivolt signal (typically 1–3 mV/V excitation), which means the DAQ must supply stable bridge excitation and amplify before digitization. This makes load-cell channels more sensitive to excitation drift and cable resistance than accelerometers. In practice, you will see the excitation voltage and bridge factor recorded in the channel metadata; if the workbook later needs to convert raw volts to engineering units, those two parameters must be carried through. For injury evaluation, belt force is often used to check chest deflection correlation or to verify that a force limiter behaved as designed, rather than as a standalone pass/fail criterion.

Displacement Sensors: Deflection and Intrusion

Displacement measurements capture what acceleration cannot: chest deflection, steering-column rearward displacement, pedal intrusion, and structural deformation. Three technologies appear in crash labs:

• String potentiometers — a cable pulled from a spool; robust, inexpensive, used for chest deflection (via the dummy's internal potentiometer) and larger structural travel.
• LVDTs — linear variable differential transformers; high accuracy and resolution, used where small displacements (a few mm to tens of mm) must be resolved precisely.
• Laser/optical triangulation sensors — non-contact, used for intrusion and occupant-relative motion where a physical tether would alter the result.

Displacement channels are usually filtered at CFC 180 or CFC 600. Chest deflection from a Hybrid III is a direct input to the chest deflection criterion, so its zero offset and calibration slope must be verified. A common field problem is a string pot that reaches end-of-travel and clips — the trace flat-tops, and any criterion computed from it becomes artificially low. When reviewing raw data, always scan displacement traces for flat segments at the extremes.

The Data Acquisition Chain and File Formats

All three transducer families feed a DAQ system that samples at a high rate — typically 10 kHz to 20 kHz per channel for crash tests, with anti-alias filtering applied before digitization. The DAQ applies excitation, amplifies, digitizes, and time-stamps each sample against a common trigger (usually a break-beam or accelerometer threshold at bumper contact). The output is a multi-channel time-history file.

You will encounter several formats in an Excel automation workflow:

The practical rule for an Excel pass/fail workbook is to standardize on a single intermediate format — usually CSV with a documented header block — and to preserve the channel metadata (unit, CFC filter, sign convention, sample rate) alongside the numeric columns. If the metadata is lost, the workbook cannot know whether a channel is in g or m/s², and a limit expressed in g will be compared against the wrong scale.

From Raw Channels to Evaluable Signals

The figure below traces a representative signal path from transducer to the point where an Excel workbook can consume it, showing where filtering, unit conversion, and metadata capture occur.

Practical Verification Before Automation

When you receive a raw dataset, run a short sanity pass before writing any formula. Check that the time vector is monotonic and starts near the trigger; confirm the sample rate matches the header; plot each channel and look for clipping, flat-tops, or dropouts; and verify that the unit string in the header matches the magnitude of the data (a head acceleration peaking at 800 in a channel labeled "m/s²" is almost certainly in g). For a belt load cell, confirm the excitation and bridge factor are present so the raw voltage can be converted. These checks take minutes and prevent the most damaging class of error in an automated pass/fail workbook: a correctly computed criterion applied to the wrong signal.

Once the channels are verified and standardized, the workbook can safely ingest them. The next stage — computing HIC, the 3 ms clip, and deflection criteria from these time histories — depends entirely on the integrity established here, which is why raw data provenance and metadata discipline are treated as first-class engineering requirements rather than clerical overhead.