3. Types of anomalies in crash testing

An anomaly in a crash test is any condition that compromises the validity, repeatability, or interpretability of the measured result. Not every anomaly invalidates a test, and not every invalid test is obviously anomalous. The practical skill is classification: determining where in the test chain the deviation originated, because the origin determines both the diagnostic path and the retest decision. This chapter establishes a four-domain taxonomy — equipment, setup, execution, and data acquisition — and shows how to localize a fault to one domain using observable evidence.

The four domains are not arbitrary. They follow the physical and logical order in which a crash test is constructed: the facility and instrumentation must be healthy (equipment), the specimen and its environment must be configured to the test specification (setup), the event must be triggered and proceed as prescribed (execution), and the resulting signals must be captured and conditioned correctly (data acquisition). A fault in an earlier domain frequently masquerades as a fault in a later one, which is why misclassification is the most common cause of unnecessary retests and, worse, of accepted invalid data.

Four-domain anomaly taxonomy for crash testing with typical signatures and propagation paths Diagram showing equipment, setup, execution, and data acquisition domains in test-chain order, each with representative failure modes, and arrows indicating how an upstream fault can propagate and appear as a downstream symptom. Anomaly Domains Along the Crash Test Chain Upstream faults propagate rightward and often surface as downstream symptoms 1 · Equipment Facility & instrumentation health • Propulsion / winch speed drift • Barrier face damage or shift • Load-cell zero / gain drift • Accelerometer mounting torque • Lighting / camera sync source • Track rail alignment Signature Bias across all channels or repeatable offset 2 · Setup Specimen & environment config • Dummy seating / H-point error • Belt routing / pretension • Impact angle / offset • Tire pressure / ride height • Ambient temp / humidity • Sensor channel mapping Signature Localized to one specimen or one channel group 3 · Execution Event trigger & progression • Late / early trigger pulse • Trolley pre-rotation • Airbag non-deploy / late • Secondary impact • Braking before contact • Pyro channel misfire Signature Timing shift or event sequence deviation 4 · Data Acquisition Capture & signal conditioning • Channel saturation / clipping • Aliasing / low sample rate • CFC filter mismatch • Cable break / intermittent • Time-zero misalignment • Ground loop / noise floor Signature Signal shape artifact, not physical response Common misclassification Equipment bias read as DAQ noise · Setup error read as execution fault Localization rule: Check the earliest domain whose evidence is inconsistent — do not start at the symptom. Retest trigger: Any fault that changes the physical input, event timing, or a required measurement channel.

Equipment anomalies: the facility as a source of bias

Equipment faults are the most insidious because they are usually systematic: they bias every test conducted under the same condition, so a single anomalous result may be the first visible symptom of a long-standing drift. Typical examples include propulsion speed error (the trolley arrives at the barrier at 49.2 km/h against a 50 ± 1 km/h target, still nominally in tolerance but trending), barrier face deformation or lateral shift after repeated impacts, load-cell zero offset, and accelerometer mounting torque below specification, which produces a characteristic high-frequency resonance superimposed on the true signal.

Diagnosis relies on pre-test verification records rather than the crash data itself. A daily zero-check on each load cell, a documented barrier survey after every N impacts, and a torque log for every accelerometer installation convert an ambiguous post-test question ("was the rig healthy?") into a documented fact. When an anomaly appears, compare the suspect channel against its calibration history and against sibling channels measuring the same physical quantity. If all channels on one data module show a common offset, the fault is almost certainly upstream of the specimen — equipment, not setup.

Setup anomalies: specimen and environment configuration

Setup faults are localized: they affect one specimen, one dummy, or one channel group, and they are usually traceable to a specific configuration step. Common cases include dummy H-point deviation beyond the seating procedure tolerance, incorrect belt routing or pretensioner state, impact angle or offset error (a 15° oblique test executed at 17°), tire pressure outside the specified range, and ambient temperature outside the conditioning window. A subtle but frequent setup fault is channel mapping error — the left-side thoracic accelerometer physically installed at the right-side location, or two channels swapped at the junction box. The data look physically plausible, which is exactly why mapping errors survive review.

The practical control is a setup verification gate before the trigger is armed: a signed checklist covering dummy positioning measurements, belt routing photographs, impact geometry (angle, offset, height), environmental readings, and a channel-to-location cross-check performed by a second person. Setup anomalies that pass this gate are rare; setup anomalies that bypass it are common.

Execution anomalies: the event did not proceed as prescribed

Execution faults occur during the event itself and are identified primarily from timing and sequence evidence. Examples include a trigger pulse arriving late relative to the intended contact time, trolley pre-rotation before impact, airbag non-deployment or deployment later than the specified time window, unintended secondary impacts, and braking applied before contact. These faults change the physical input to the specimen, which means the measured response no longer corresponds to the test specification — a retest is almost always required, regardless of how clean the signals look.

Execution anomalies are best detected with independent timing references: a contact-tape or laser trigger at the barrier face, high-speed video at 1,000 fps or higher, and a trigger-monitor channel recorded alongside the crash data. If the video shows contact at t = 0 but the accelerometer time-zero is 8 ms later, the discrepancy is an execution or time-alignment fault, not a specimen response. Always resolve timing questions from the independent reference first, then reconcile the data channels to it.

Data acquisition anomalies: correct event, corrupted record

DAQ faults leave the physical event intact but corrupt its representation. The classic signatures are channel saturation (a clipped peak on a chest acceleration trace where the true peak exceeded the ±500 g range), aliasing from an inadequate sample rate, incorrect CFC filter class applied during post-processing (CFC 1000 applied to a channel specified for CFC 60), cable breaks producing a flat-line or step discontinuity, ground-loop noise raising the noise floor, and time-zero misalignment between channels. These are distinguishable from physical anomalies because they violate signal plausibility: a step discontinuity, a perfectly flat segment, or a peak that terminates in a horizontal line is not a physical response.

A saturated channel is not a valid measurement. If a required channel clipped, the test cannot be reported as compliant even if every other channel is clean — the missing peak is precisely the value the test exists to measure. Do not extrapolate or "reconstruct" a clipped peak; schedule a retest with an adjusted range or an appropriately ranged sensor.

Localizing the fault: a worked example

Consider a frontal sled test where the driver chest resultant acceleration shows an unexpected 12 ms plateau at 62 g, while the pulse from the sled carriage accelerometer is nominal. The naive reading is "specimen response anomaly." The correct procedure is to walk the domains in order:

  1. Equipment: check the chest accelerometer's calibration date and mounting torque log. Both current — equipment is unlikely.
  2. Setup: verify the dummy seating record and channel map. The chest channel is mapped correctly, but the seatback angle was recorded at 27° against a 25° ± 1° specification.
  3. Execution: the sled pulse and trigger timing are nominal; no pre-rotation or secondary impact is visible in video.
  4. DAQ: the plateau is smooth and repeatable across the three chest channels, not a step or clip — consistent with a physical cause, not an acquisition artifact.

The evidence localizes the fault to setup: the seatback angle error changed the chest-to-belt geometry and produced a real but non-representative loading pattern. The correct disposition is a retest with the seatback reset to specification, not a data-processing correction. This example illustrates the central discipline of this chapter: the domain in which the anomaly appears is frequently not the domain in which it originated, and the retest decision must follow the origin.

Domain Primary evidence source Typical signature Retest implication
Equipment Calibration and verification logs Common bias across channels or tests Recalibrate; retest affected tests
Setup Pre-test checklist, photos, measurements Localized to one specimen or channel group Correct configuration; retest
Execution Independent timing reference, high-speed video Timing shift or sequence deviation Almost always retest
Data acquisition Signal plausibility, range and filter settings Clip, step, flat-line, aliasing Retest if a required channel is lost
Record the anomaly domain explicitly in the test report, together with the evidence that localized it. A retest justified by "the data looked wrong" is indefensible in a regulatory review; a retest justified by "seatback angle recorded at 27° against a 25° ± 1° specification, confirmed by the setup photograph and the seating record" is not.
← Previous 📖 Contents Next →