3. Sources of variability in crash tests

Crash test repeatability is never limited by a single factor. When two nominally identical tests produce different occupant responses, the difference is the sum of many small contributions acting simultaneously. To manage repeatability, engineers must decompose total variability into its physical sources and estimate how much each source contributes. This section examines five categories: the test article, the facility, the instrumentation, the environment, and the operator.

A useful mental model is a variance budget. If the total standard deviation of a response metric such as Head Injury Criterion (HIC) or chest deflection is σ_total, then under the assumption of independent contributions:

σ²_total = σ²_article + σ²_facility + σ²_instrumentation
         + σ²_environment + σ²_operator

This decomposition is not exact — sources interact — but it forces disciplined thinking about where repeatability is lost and where investment pays off.

Test Article Variability

The test article is often the dominant contributor. Even vehicles from the same production line differ in body-in-white weld quality, material thickness, adhesive application, and trim mass. For a frontal offset test, a ±2% variation in front-rail gauge can shift longitudinal crush distance by several millimetres, which propagates into occupant compartment deceleration.

For Anthropomorphic Test Devices (ATDs), variability arises from sensor calibration drift, joint friction, and previous test history. A Hybrid III dummy that has been subjected to several high-severity tests may exhibit changed lumbar spine stiffness. Best practice is to track each ATD's calibration certificate and cumulative test count, and to re-certify when response falls outside the manufacturer's tolerance band.

Practical controls include:

• Documenting vehicle build date, VIN, and any pre-test modifications.
• Weighing and recording ATD segment masses before each test.
• Performing a low-severity pendulum calibration check on the ATD within 30 days of the test.

Facility Variability

The facility includes the tow track, barrier, propulsion system, and sled. Key parameters are impact speed, impact angle, and barrier alignment. A nominal 64 km/h offset test with ±0.5 km/h speed tolerance and ±0.5° yaw tolerance is typical, but even these small windows produce measurable response differences.

Barrier face flatness and foundation stiffness matter. A deformable barrier with uneven honeycomb density will load the vehicle asymmetrically. Sled buck fixtures introduce their own compliance; if the buck is not rigidly bolted, energy is absorbed by the fixture rather than the test article.

Facility variability is controlled through:

• Daily speed calibration using a light gate or radar.
• Barrier alignment checks with a laser tracker before each test series.
• Torque verification on all fixture bolts to a specified value.

Instrumentation Variability

Instrumentation contributes through sensor accuracy, mounting, signal conditioning, and data processing. Accelerometers have transverse sensitivity, temperature drift, and cable noise. Load cells in the barrier face are sensitive to preload and bolt torque. High-speed cameras introduce timing jitter if not genlocked.

Consider a chest accelerometer with a ±5% sensitivity tolerance. If the true peak is 60 g, the reported value could range from 57 g to 63 g purely from sensor tolerance. When multiple sensors are combined into a single injury metric, these errors can accumulate or partially cancel.

Mitigation strategies include:

• Using end-to-end calibration with a known reference accelerometer.
• Applying the SAE J211 channel filter class consistently (e.g., CFC 180 for chest acceleration).
• Documenting cable routing and strain relief to avoid triboelectric noise.

Environmental Variability

Ambient conditions affect both materials and measurements. Temperature influences foam stiffness in the barrier and airbag inflator output. Humidity affects adhesive cure and sensor insulation resistance. Air density and wind can alter aerodynamic drag on the vehicle, though this is minor at crash speeds.

A common specification is 20 ± 5 °C and 50 ± 20% relative humidity. When a test is conducted outside this band, the result should be flagged. For example, a barrier foam tested at 5 °C may be 15–20% stiffer than at 20 °C, increasing peak force.

Operator Variability

Operator contributions are often underestimated. They include ATD positioning, seat belt routing, marker placement, camera setup, and data review decisions. A belt routed 10 mm higher on the shoulder can change chest deflection by several millimetres. A seat track position set one notch differently changes occupant kinematics.

Operator variability is reduced through standardized work instructions, photographic records of setup, and independent verification by a second technician. A useful practice is a pre-test checklist signed by both the setup technician and the test engineer.

Relative Contribution Example

The table below shows a hypothetical variance budget for a frontal offset test, based on engineering judgment for illustration. It is not measured data.

In this example, the test article and operator together account for nearly 70% of the variance. This suggests that improving vehicle build consistency and standardizing setup procedures would yield the greatest repeatability gains.

Practical Guidance for Reducing Variability

When planning a repeatability study, allocate at least three repeat tests per condition. Compute the standard deviation for each key metric and compare it to the tolerance required by the program. If the observed σ exceeds the target, use the variance budget to prioritize corrective actions.

A simple decision rule: if operator and facility together contribute more than 40% of variance, focus on work instructions, fixture torque checks, and speed calibration before changing vehicle design. If test article variance dominates, the issue is likely in the build process or ATD condition.

Finally, document every test with a setup sheet that captures the five source categories. This record becomes the foundation for any future uncertainty analysis and allows engineers to trace a repeatability problem back to its origin.