3. Sensor selection for crash testing

Crash test instrumentation is a chain of transducers, signal conditioning, cabling, and recorders, and the chain is only as good as its weakest transducer. Selecting the right sensor for each measurement channel is therefore one of the highest-leverage engineering decisions in the whole data acquisition workflow. A sensor that is too slow, too fragile, or mounted in the wrong place will produce data that looks plausible but is physically wrong — and in a destructive test there is no second chance to re-measure. This section reviews the six transducer families most commonly used in vehicle crash testing, the physical quantity each one actually measures, and the criteria that drive selection.

3.1 The Six Transducer Families and What They Really Measure

Accelerometers dominate crash instrumentation because occupant injury criteria (HIC, chest deflection rate, head resultant acceleration) are defined in terms of acceleration. Piezoresistive (PR) accelerometers are the workhorse for crash pulses: they are DC-coupled, so they capture the sustained deceleration of a barrier impact, and they tolerate the high shock levels (500 g to 2000 g) typical of vehicle body and dummy locations. Capacitive MEMS accelerometers are common in lower-g environments and in angular rate packages. Piezoelectric accelerometers have excellent high-frequency response but are AC-coupled, so they cannot measure a DC or slowly varying crash pulse and are generally avoided for primary crash channels.

Load cells measure force directly through a strain-gauged elastic element. In crash testing they appear in two main roles: instrumented barrier load cells (which measure the force distribution across a rigid wall) and load cells embedded in fixtures such as seat belt anchorages, steering columns, or pedal assemblies. A load cell's rated capacity must exceed the expected peak force with margin, but not so much that the output at the forces of interest falls into the noise floor.

Displacement sensors capture the geometric deformation that acceleration alone cannot describe — steering column intrusion, pedal travel, seat track displacement, and dummy chest deflection. String potentiometers (string pots) are the classic choice for large, slow-to-medium displacements; LVDTs and eddy-current sensors serve shorter ranges with higher bandwidth. Displacement is often the most direct evidence of occupant compartment intrusion.

Strain gauges measure local surface strain on structural members. Bonded foil gauges in quarter-, half-, or full-bridge configurations are used on B-pillars, door beams, and suspension components to validate CAE models and to detect yielding. Strain gauges are small, inexpensive, and sensitive, but they require careful surface preparation and temperature compensation.

Pressure sensors are used for airbag inflator output, fuel tank and brake line pressure, and tire pressure during rollover. Piezoresistive pressure transducers with high burst pressure ratings are typical; the key selection issue is surviving the pressure spike without saturating or rupturing.

Angular rate sensors (gyroscopes) measure rotational velocity and are essential for rollover and occupant kinematics studies. MEMS gyros are now standard, often packaged with accelerometers in a six-degree-of-freedom inertial measurement unit (IMU).

3.2 Selection Criteria

Choosing a sensor is a multi-criteria decision, and the criteria interact. The most important are:

• Measurand and range: What physical quantity, and what is the expected peak? Select a full-scale range that places the peak at roughly 50–80% of full scale — high enough to avoid clipping, low enough to preserve resolution.
• Bandwidth: The sensor and its signal conditioning must pass the frequencies present in the event. A rule of thumb is to choose a sensor with a resonant frequency at least five times the highest frequency of interest. Crash pulses contain energy up to several kHz, so accelerometers with 10–20 kHz resonance are common.
• Environment: Temperature, humidity, electromagnetic interference, and mechanical shock all constrain the choice. Engine-bay and exhaust locations demand high-temperature sensors; underbody locations demand water and debris protection.
• Mounting: The mounting method is part of the sensor. A stud-mounted accelerometer has a much higher usable frequency than a magnetically or adhesively mounted one. Mounting stiffness and mass loading can shift the effective response.
• Mass loading: A heavy sensor on a light structure changes the structure's dynamics. Small, lightweight sensors are preferred on thin panels and dummy components.
• Signal conditioning and cabling: PR accelerometers need excitation and amplification; piezoelectric sensors need charge amplifiers. Cable routing must survive the crash and not introduce triboelectric noise.
• Calibration and traceability: Every channel should be traceable to a calibration certificate, and the calibration should bracket the expected range.

3.3 A Worked Selection Example

Consider instrumenting a frontal 56 km/h rigid-barrier test on a passenger car. The body-mounted accelerometer at the B-pillar is expected to see a peak of about 60 g with significant energy up to 1 kHz. A piezoresistive accelerometer with a ±200 g range and a resonant frequency above 10 kHz satisfies both the range and bandwidth criteria, and its DC response captures the full crash pulse. The dummy head accelerometer, by contrast, may see peaks near 300 g; a ±500 g PR accelerometer with a similar resonance is appropriate. For the steering column, a string pot with a 150 mm range and a 100 Hz bandwidth is sufficient because column intrusion is a low-frequency, large-displacement event. For the seat belt load cell, a 20 kN capacity with a 5 kHz bandwidth covers the expected belt force peak of roughly 8–10 kN.

The table below summarizes the mapping from measurement need to sensor family and the dominant selection criterion.

3.4 Matching the Sensor to the Data Channel

Selection does not end with the transducer. Each sensor must be matched to a data channel with appropriate anti-alias filtering and sample rate. As a practical rule, sample at least ten times the highest frequency of interest; for a 1 kHz crash signal, a 10–20 kHz sample rate per channel is typical. The anti-alias filter corner should be set below half the sample rate to prevent aliasing of high-frequency content back into the passband.

The figure below shows the decision logic for selecting a sensor family based on the physical quantity to be measured and the key constraint that dominates the choice.

3.5 Common Selection Mistakes

Three mistakes recur in crash instrumentation. First, over-ranging: choosing a 5000 g accelerometer for a 50 g measurement because it is the only one on the shelf. The signal then occupies a tiny fraction of the ADC range and the effective resolution collapses. Second, ignoring mounting effects: a sensor with a 20 kHz resonant frequency mounted on a flexible bracket may have an effective resonance of 2 kHz, which is inside the measurement band and corrupts the data. Third, neglecting cable and connector survival: a perfectly selected sensor with a cable that fails at 20 g of cable whip produces a dropout exactly when the data matters most. Each of these is a selection error, not a test-day error, and each is preventable with a short written justification for every channel before the test.

In practice, sensor selection is documented in an instrumentation plan that lists, for each channel, the measurand, the sensor model and serial number, the full-scale range, the resonant frequency, the mounting method, the signal conditioner, the sample rate, and the calibration reference. This plan becomes the reference against which the recorded data is validated after the test, and it is the first document reviewed when a channel looks suspicious.