2. SID-IIs Small Female Pelvis Response Baseline

Purpose and Scope of the Baseline

A SID-IIs small female pelvis baseline is not a pass/fail test. It is a documented reference response that later injury-diagnosis work is compared against. The deliverable of this exercise is a small, auditable dataset: the raw and filtered acceleration and force channels recorded from the pelvis of a SID-IIs dummy during a standardized lateral impact, together with the exact channel scaling, filter class, and coordinate convention used to produce them. If those three metadata items are not fixed before the test, the numbers are not reusable, and any later injury assessment built on them is unsound.

The SID-IIs pelvis instrumentation of interest here is the triaxial pelvis accelerometer cluster and the iliac force load cells. The accelerometers are typically mounted on the pelvis structure near the H-point region; the load cells sit in the iliac wing load paths. Because the dummy is a small female anthropometry, the pelvis is narrower and the sensor mounting geometry differs from the WorldSID 50th male, so baselines cannot be borrowed across dummy types.

Fixing the Test and Channel Definition Before Impact

The baseline is only meaningful if the impact condition is standardized and recorded. A practical standardized lateral impact for this purpose is a sled or impactor configuration with a defined impact velocity, a defined impactor face geometry, and a defined contact height relative to the dummy pelvis. Record the following before the run:

• Impact velocity and its measurement method (e.g., sled velocity from a flag or light gate), with tolerance.
• Impactor or barrier face geometry and the vertical position of the contact center relative to the dummy H-point.
• Dummy serial number, build configuration, and suit/clothing state.
• Seat or fixture surface, friction condition, and restraint state (belted or unbelted).
• Ambient temperature and soak time, since dummy foam and pelvis compliance are temperature-sensitive.

Channel scaling must be written down per channel, not assumed. For each accelerometer axis, the scale factor converts the recorded voltage or bridge output to engineering units (g or m/s²), and the sign convention must be stated explicitly. A defensible convention for a lateral impact is: +Y toward the impacted side, +X forward, +Z up, with the origin at the dummy H-point. State this in the log and keep it identical in every derived table and figure.

Filter Class and Why It Changes the Number

Pelvis acceleration in a lateral impact contains a high-frequency structural ringing component that is not physiologically or injury-relevant. The standard practice is to apply a low-pass filter and report the filter class alongside the peak. For dummy pelvis acceleration, a CFC 180 or CFC 1000 class filter is common depending on the downstream injury criterion; the choice must be fixed and documented. The same raw channel filtered at CFC 180 and at CFC 1000 will give different peaks, sometimes by a meaningful margin, so a baseline without a stated filter class is ambiguous.

Force channels from the iliac load cells are typically filtered at a lower class than acceleration, and the two must not be conflated. Record the filter class per channel group in the metadata table, and apply the filter in the analysis pipeline rather than relying on the data acquisition default, so the result is reproducible.

Worked Baseline Extraction

The following is a compact, self-contained extraction and reporting script. It assumes the raw channels are already exported as a CSV with a time column and one column per channel, in volts, and that the per-channel scale factors and offsets are known from the calibration sheet. The numbers used here are illustrative example values, not measured results.

import numpy as np
import pandas as pd
from scipy.signal import butter, filtfilt

# --- Channel definition (example values; replace with calibration sheet) ---
CHANNELS = {
    "pelvis_ax": {"scale": 0.0500, "offset": 0.0, "unit": "g", "filter": "CFC180"},
    "pelvis_ay": {"scale": 0.0500, "offset": 0.0, "unit": "g", "filter": "CFC180"},
    "pelvis_az": {"scale": 0.0500, "offset": 0.0, "unit": "g", "filter": "CFC180"},
    "iliac_fy": {"scale": 220.0,  "offset": 0.0, "unit": "N", "filter": "CFC60"},
}

# CFC class -> approximate -3 dB cutoff (Hz) per SAE J211 convention
CFC_CUTOFF = {"CFC60": 100.0, "CFC180": 300.0, "CFC600": 1000.0, "CFC1000": 1650.0}

def lowpass(x, fs, fc, order=4):
    b, a = butter(order, fc / (fs / 2.0), btype="low")
    return filtfilt(b, a, x)

def load_and_scale(csv_path, fs):
    df = pd.read_csv(csv_path)
    t = df["time"].to_numpy()
    out = {"time": t}
    for name, meta in CHANNELS.items():
        raw = df[name].to_numpy()
        eng = raw * meta["scale"] + meta["offset"]
        fc = CFC_CUTOFF[meta["filter"]]
        out[name] = lowpass(eng, fs, fc)
    return out

def baseline_summary(data):
    rows = []
    for name, meta in CHANNELS.items():
        x = data[name]
        peak = float(np.max(np.abs(x)))
        t_peak = float(data["time"][int(np.argmax(np.abs(x)))])
        rows.append({
            "channel": name, "unit": meta["unit"],
            "filter": meta["filter"],
            "peak_abs": round(peak, 3),
            "t_peak_ms": round(t_peak * 1000.0, 2),
        })
    return pd.DataFrame(rows)

if __name__ == "__main__":
    fs = 20000.0  # Hz, example acquisition rate
    data = load_and_scale("sid2s_pelvis_run01.csv", fs)
    summary = baseline_summary(data)
    print(summary.to_string(index=False))
    summary.to_csv("sid2s_pelvis_baseline_summary.csv", index=False)

Run this against the exported run file and inspect the summary table. The expected observable is a peak absolute value and a time-of-peak for each channel, with the filter class printed alongside. If a channel shows a peak at t = 0 or at the very end of the record, suspect a scaling or offset error or a truncated record rather than a real response. If the two lateral axes disagree in sign for a symmetric impact, recheck the sign convention before trusting the baseline.

Reading the Baseline Response

The figure below shows the structure of the baseline record and the relationships that matter for later diagnosis: the raw channel, the filter stage, the scaled engineering-unit output, and the summary quantities that get carried forward. It is a signal-chain view, not a test procedure.

Verification and Failure Symptoms

Before accepting the baseline, verify three things. First, the scaled channels should be near zero before impact and return toward zero after the event; a persistent offset indicates a scaling or offset error. Second, the time of peak on the lateral pelvis axis should fall within the impact window, not at the record boundary. Third, the iliac force peak should be physically plausible for the impact severity; an implausible magnitude usually means a wrong load-cell scale factor or a unit mismatch between N and kN.

Once the summary table and metadata are consistent, the baseline is complete and can be filed as the reference for subsequent SID-IIs pelvis injury-diagnosis work. The value of this exercise is not the peak number itself but the fact that the number is reproducible: same impact condition, same sign convention, same filter class, same scaling.