4. Distinguishing Tension-Extension from Compression-Flexion Nij Modes
Why the Nij Quadrant Is a Classification Problem, Not a Single Number
The Normalized Injury Criterion (Nij) is often reported as a single peak value, but the criterion is actually defined over four independent loading quadrants. Each quadrant combines one axial force direction with one bending moment direction, and each has its own critical intercept pair. A neck loading event that produces a peak Nij of 0.9 in tension-extension is a fundamentally different injury mechanism from one that produces 0.9 in compression-flexion, even though the scalar magnitude is identical. Root-cause analysis therefore cannot stop at "Nij exceeded the threshold"; it must first establish which quadrant the peak belongs to, because the quadrant determines the tissue path, the restraint interaction, and the corrective action.
The four quadrants are conventionally labeled:
- NTE — Neck Tension-Extension: axial tension combined with extension (rearward head rotation relative to the torso).
- NTF — Neck Tension-Flexion: axial tension combined with flexion.
- NCE — Neck Compression-Extension: axial compression combined with extension.
- NCF — Neck Compression-Flexion: axial compression combined with flexion (the classic frontal-impact, airbag/restraint-driven mode).
Each quadrant is normalized by its own critical intercept values, so the same raw force and moment can map to very different normalized values depending on sign convention and intercept selection. Misclassification usually arises from three sources: an inconsistent sign convention between the load cell and the criterion definition, a peak-picking window that captures the axial peak and the moment peak at different instants, or a coordinate transform that flips the moment sign when the head rotates through neutral.
Sign Convention and Quadrant Assignment
Before any classification is possible, the analysis must fix a single, documented convention. A workable and common convention for a six-axis upper-neck load cell is:
- Axial force Fz: positive in tension (neck lengthening), negative in compression.
- Bending moment My: positive in extension (head rotating rearward relative to torso), negative in flexion.
With this convention the quadrant is read directly from the signs of the two channels at the instant of interest. The critical step is that the moment must be evaluated in the same head-relative-to-torso frame as the force, and both must be taken at the same time sample. If the axial force peaks at 12 ms and the moment peaks at 19 ms, the "peak Nij" computed from the two independent maxima is a synthetic value that may not correspond to any real loading state — and its quadrant label can be wrong.
Worked Classification Example
Consider a frontal sled test with a belted 50th-percentile male ATD. The upper-neck load cell records the following (illustrative example data, not measured results):
| Time (ms) | Fz (N) | My (N·m) | Fz/Fz_crit | My/My_crit | Nij sum |
|---|---|---|---|---|---|
| 40 | −180 | −12 | −0.09 | −0.15 | −0.24 |
| 55 | −620 | −38 | −0.31 | −0.48 | −0.79 |
| 62 | −740 | −44 | −0.37 | −0.55 | −0.92 |
| 70 | −510 | −30 | −0.26 | −0.38 | −0.64 |
Using critical intercepts of Fz_crit = 2000 N and My_crit = 80 N·m (illustrative values for this example only), the peak normalized sum is −0.92 at 62 ms. Both channels are negative: compression and flexion. The event classifies as NCF, with a peak Nij magnitude of 0.92 occurring at 62 ms. Note that the peak of the sum (62 ms) coincides with neither the axial-force peak nor the moment peak taken in isolation if those occurred at different samples — which is exactly why sample-wise evaluation matters.
Now suppose the same raw data were processed with an inverted moment sign (a common result of applying a coordinate transform without re-checking the convention). The My column would become positive, the sum would read +0.18 at 62 ms, and the event would be mislabeled as NCE. The peak magnitude would also collapse, hiding a near-threshold loading. This single sign error changes the root-cause conclusion from "restraint-driven compression-flexion loading" to "extension loading inconsistent with the frontal pulse" — a conclusion that would send the investigation toward the wrong subsystem entirely.
Reading the Quadrant Map
The SVG below plots the four quadrants in the Fz–My plane with the example trajectory overlaid. The trajectory stays in the lower-left (compression-flexion) quadrant throughout the event, which is the visual signature of an NCF classification. A trajectory that crosses an axis indicates a sign change in one channel and a quadrant transition — a state change that must be flagged, because the peak Nij may occur on either side of the crossing and the two sides represent different mechanisms.
How Misclassification Changes the Root-Cause Conclusion
The quadrant label is not cosmetic; it selects the causal chain that the analyst will investigate. The table below summarizes how the same peak magnitude leads to different conclusions depending on the assigned quadrant.
| Assigned quadrant | Implied mechanism | Typical root-cause direction | Consequence of a wrong label |
|---|---|---|---|
| NCF | Compression + flexion, head driven forward/down | Airbag over-pressure, belt geometry pulling torso forward, head-to-restraint contact | If mislabeled NCE, investigation wrongly targets rearward head motion and ignores the frontal restraint path |
| NCE | Compression + extension, head driven rearward | Seatback recline, head restraint too low/forward, rearward pulse | If mislabeled NCF, seatback and head-restraint geometry are exonerated incorrectly |
| NTE | Tension + extension, neck stretched rearward | Rear impact, head restraint gap, whiplash-type loading | If mislabeled NTF, the rear-impact signature is lost and the pulse direction is misread |
| NTF | Tension + flexion, neck stretched forward | Frontal pulse with head lag, belt-only restraint without airbag | If mislabeled NTE, the frontal restraint deficiency is hidden |
The practical rule is: classify first, then interpret. A defensible classification requires (1) a documented sign convention, (2) sample-wise evaluation of the normalized sum, (3) confirmation that the peak sample lies unambiguously inside one quadrant, and (4) a check for axis crossings that would indicate a quadrant transition during the event. When the peak sample sits close to an axis (normalized component near zero), the classification is fragile and should be reported with the raw channel values and the time of peak so a reviewer can verify the quadrant assignment independently.
Verification Checklist for a Classification
- Confirm the load cell sign convention matches the criterion definition; document it in the analysis log.
- Compute Fz/Fz_crit and My/My_crit at every sample, then sum sample-wise; do not combine independent channel maxima.
- Locate the peak of the sum and record its time, raw Fz, raw My, and both normalized components.
- Read the quadrant from the signs of the two normalized components at that sample.
- Scan the trajectory for axis crossings; if any occur, confirm the peak is on the reported side and note the transition time.
- Cross-check against the crash pulse direction and restraint configuration; a quadrant inconsistent with the pulse is a red flag for a sign or transform error.
Completing these six checks produces a classification that is reproducible and auditable, and it makes the downstream root-cause conclusion defensible: the quadrant is established from the data, not assumed from the test setup.