3. Calibration and traceability of crash test instrumentation

Crash test data is only as credible as the measurement chain that produced it. A barrier impact lasting 100–150 ms generates peak accelerations above 500 g, structural crush forces in the hundreds of kilonewtons, and rapid pressure transients inside airbags and fuel systems. If a load cell drifts by 2 % or an accelerometer's sensitivity is stale, the resulting occupant injury criteria (HIC, chest deflection, femur load) shift accordingly — and an audit finding is almost guaranteed. This section explains how to keep load cells, accelerometers, pressure sensors, and the data acquisition system (DAS) calibrated and traceable to national standards, and how to prove it during an audit.

The Measurement Chain and Where Traceability Lives

Traceability is an unbroken chain of comparisons, each with stated uncertainty, linking a field sensor back to a national metrology institute (NMI) such as NIST, PTB, or NMIJ. In a crash lab the chain has four links:

• Sensor — accelerometer, load cell, pressure transducer, or displacement/angular-rate device.
• Signal conditioning — bridge excitation, amplifier gain, anti-alias filter, and the shunt or inline calibration resistor.
• DAS — the on-board or off-board digitizer that samples, filters, and stores channels.
• Reference standards — the lab's working standards (dead weights, reference accelerometer, pressure deadweight tester, precision voltage source) that are themselves calibrated by an accredited laboratory.

Every link must have a current certificate, and the certificates must connect. A sensor calibrated by a lab whose own reference is not traceable breaks the chain, no matter how good the sensor is.

Calibration Types and Their Intervals

Three calibration activities are routinely confused. Keeping them distinct is the core of audit readiness.

Primary calibration is not optional and cannot be replaced by an in-house check. End-to-end calibration, by contrast, is the engineer's daily tool: it injects a known physical or electrical stimulus and verifies that the recorded engineering-unit value matches.

Load Cells: Shunt, Deadweight, and In-Situ Checks

Crash load cells (seat belt, femur, steering column, barrier) are strain-gauge bridges. Their primary calibration uses deadweights or a reference load cell in a compression/tension machine, producing a sensitivity in mV/V per kN. In the lab, the practical check is the shunt calibration: a precision resistor is switched across one bridge arm, producing a known simulated strain.

Shunt resistor: R_sh = 100 kΩ across bridge arm
Bridge resistance: R_b = 350 Ω
Excitation: V_ex = 5.000 V
Gauge factor: GF = 2.00

Simulated strain: ε = R_b / (4 · R_sh · GF)
                  = 350 / (4 × 100000 × 2.00)
                  = 4.375 × 10⁻⁴ (437.5 µε)

Expected output: V_out = ε · GF · V_ex / 4
                      = 4.375e-4 × 2.00 × 5.000 / 4
                      = 1.094 mV

The DAS should read a value within the tolerance stated on the calibration certificate (commonly ±1 % of the shunt-equivalent load). If it does not, suspect the cabling, connector, or conditioning gain before blaming the sensor. Record the shunt result, the resistor value, and the ambient temperature on the test log.

Accelerometers: Sensitivity, Frequency Response, and Mounting

Piezoresistive accelerometers dominate crash testing because they respond to DC (needed for velocity integration). Primary calibration establishes sensitivity (mV/g), frequency response to at least 1–2 kHz, and transverse sensitivity. In the lab, the two practical checks are:

1. Static/DC check — invert the accelerometer in the Earth's gravity field. A ±1 g step should appear. This catches cable and offset faults immediately.
2. Shunt or back-to-back comparison — compare against a reference accelerometer on a shaker or drop tower.

Mounting matters as much as calibration. A sensor mounted with a soft adhesive or a loose bracket will show a resonance well below its calibrated range, producing ringing that corrupts HIC. Use rigid stud or cyanoacrylate mounting, torque to the manufacturer's specification, and verify with a tap test or coherence check where available.

Pressure Sensors: Static and Dynamic Behaviour

Airbag, fuel-tank, and brake pressure sensors are calibrated against a deadweight tester or a reference pressure standard. Because many crash pressure events are dynamic, a static calibration alone is insufficient — check the sensor's dynamic response and its line-volume effects. A common audit finding is a pressure sensor calibrated statically but used to measure a fast transient without documented dynamic characterization.

Data Acquisition Systems: The Scaling Chain

The DAS converts conditioned analog signals to engineering units. Its calibration has two parts: the analog front-end (gain and offset verified with a precision voltage source) and the digital scaling (the sensitivity and zero entered in the channel definition). Both must be documented. A typical channel definition looks like:

Channel:  ACC_CH07
Sensor:   Endevco 7264C-2K, S/N 12345
Sensitivity: 0.512 mV/g   (cert. 2024-03-11, due 2025-03-11)
Zero offset: +0.003 g
Excitation: 5.000 V
Anti-alias filter: 1650 Hz, 4-pole Butterworth
Sample rate: 20 kHz
Scaling: EU = (V_measured - V_zero) / (Sensitivity × Gain)

Before each test, run a zero check (all channels at rest) and a calibration pulse (shunt or known voltage). After the test, repeat the zero check. A post-test zero shift larger than the lab's tolerance (often 2–5 % of full scale) invalidates the channel and must be flagged in the report.

Building the Audit File

Audit readiness is a documentation discipline. For each test campaign, assemble a file containing: current primary calibration certificates for every sensor; the lab's working-standard certificates; pre- and post-test verification records; channel definition sheets; and any non-conformance notes. The file should let an auditor trace any recorded channel to a national standard without asking a single question.

When these controls are in place, calibration stops being a paperwork exercise and becomes what it should be: the evidence that the numbers in the report actually mean what they claim.