2. Airbag and seatbelt system fundamentals

Reduced-order models of occupant restraint systems are only as trustworthy as the physical understanding they encode. Before any surrogate model, lumped-parameter network, or response-surface calibration is attempted, the analyst must be able to describe what the airbag and seatbelt actually do during a crash, which quantities govern that behavior, and which metrics define success. This chapter establishes that foundation: the deployment physics of a pyrotechnic airbag, the load-path mechanics of a three-point belt with pretensioner and load limiter, and the performance metrics that later chapters will treat as model outputs.

Deployment physics: from inflator output to cushion pressure

An airbag is a gas-handling device with a very short duty cycle. A typical driver frontal inflator delivers on the order of 100–200 g of gas-generant combustion products plus stored compressed gas, producing a mass flow rate that peaks within roughly 5–15 ms of the crash pulse and decays over 40–80 ms. The cushion is not a rigid pressure vessel; it vents through sew holes and fabric porosity, and its volume grows as it unfolds. The governing relation is a mass and energy balance on the cushion control volume:

d(m_c·u)/dt = ṁ_in·h_in − ṁ_out·h_out − p·dV/dt + Q̇_wall
p = m_c·R·T / V(t)      (ideal-gas approximation for early-phase sizing)

For reduced-order work the practical consequence is that cushion pressure is a coupled function of inflator mass flow, vent area, and cushion volume history. Volume history is itself driven by the occupant: the cushion inflates freely until it contacts the occupant, after which occupant penetration compresses the gas and raises pressure. This is why a deployment-only simulation (no occupant) and an occupant-interaction simulation give very different pressure traces, and why any ROM must declare which regime it represents.

Two timing quantities dominate deployment physics. The time to fire is set by the sensing and arming logic, typically 10–25 ms after the crash pulse exceeds threshold. The time to cushion — the moment the cushion reaches the occupant — must precede occupant forward excursion into the deployment zone. A useful design check is the "5-inch–30 ms" rule of thumb: the occupant should not be closer than about 5 inches (127 mm) to the module at 30 ms after fire, otherwise the cushion strikes the occupant while still accelerating outward. Out-of-position (OOP) occupants violate this rule, which is why suppression and dual-stage inflators exist.

Restraint mechanics: belt load path, pretensioner, and load limiter

The three-point belt restrains through a load path: lap belt across the pelvis, shoulder belt diagonally across the chest, both anchored at the retractor, buckle, and anchor points. The belt force is limited by two devices working in opposition. The pretensioner removes slack — typically 40–80 mm of webbing — within 10–20 ms, coupling the occupant to the vehicle earlier and reducing excursion. The load limiter then caps shoulder-belt force, commonly in the 3–6 kN range for frontal systems, by allowing controlled webbing payout through a torsion bar or tearing stitch. The pretensioner raises force early; the load limiter caps it later. Their interaction defines the chest deflection versus time history.

The occupant's motion is governed by a simple but instructive balance. Treating the torso as a lumped mass m restrained by belt stiffness k and damping c, with the vehicle decelerating at a(t):

m·ẍ_rel + c·ẋ_rel + k·x_rel = −m·a(t) + F_limit·sign(ẋ_rel)
x_rel(0) = x_rel_dot(0) = 0

where x_rel is occupant motion relative to the vehicle. The load limiter appears as a force saturation: once belt force reaches F_limit, the effective stiffness drops and the occupant "rides down" the crash pulse at roughly constant force. This is the central trade: a stiffer belt reduces excursion and contact with interior surfaces but raises chest acceleration and deflection; a softer belt does the reverse. The pretensioner buys excursion margin so the load limiter can be softened without the occupant striking the steering wheel or knee bolster.

Key performance metrics

Restraint performance is judged against regulated and consumer metrics, and these are exactly the quantities a ROM must reproduce. The most important are summarized below; values are typical frontal-crash targets, not universal limits.

HIC₁₅ is worth a closer look because it is the metric most sensitive to airbag tuning. It is defined as

HIC = max over t1,t2 of [ (1/(t2−t1)) ∫ a(t) dt ]^2.5 · (t2−t1),  t2−t1 ≤ 15 ms

with a in g. Because of the 2.5 exponent, HIC is dominated by short, high-amplitude acceleration spikes. A cushion that is too stiff at contact produces exactly such a spike; a cushion that is too soft lets the head travel further and may produce a later, broader pulse. The ROM must therefore capture the shape of the head acceleration trace, not just its peak — a point that recurs when we discuss model validation.

How the subsystems interact: a worked example

Consider a 50th-percentile male dummy, frontal 50 km/h sled, dual-stage inflator, retractor pretensioner, and 4 kN load limiter. Suppose the analyst runs a baseline ROM and observes HIC₁₅ = 820 with chest deflection 44 mm. The head spike occurs at 62 ms, shortly after cushion contact. Two levers are available: soften the inflator second stage (reduce ṁ_in after 30 ms) or raise the load limiter to 5 kN. Softening the inflator lowers the contact pressure and typically drops HIC by 80–150 points but increases head excursion by 10–20 mm; raising the load limiter reduces excursion but raises chest deflection by roughly 4–6 mm per kN. The correct move depends on which metric is closer to its limit — here HIC is the binding constraint, so inflator softening is preferred, provided excursion margin remains. This is precisely the kind of trade a well-built ROM lets you explore in minutes rather than days.

What this means for reduced-order modeling

Three modeling implications follow directly from the physics above. First, the airbag and belt cannot be modeled as independent force generators; the cushion pressure depends on occupant penetration, and the belt force depends on occupant motion, so the ROM must close the loop between them. Second, the metrics are shape-sensitive, not just peak-sensitive — HIC's 2.5 exponent and the 3 ms chest clip both reward accurate pulse timing, so a ROM that matches peaks but shifts events by 5 ms will fail validation. Third, the design levers (inflator staging, vent area, pretensioner stroke, load-limiter force) map onto a small number of ROM parameters, which is what makes surrogate-based optimization tractable in later chapters.

With these fundamentals in place — inflator mass flow, cushion volume–pressure coupling, belt load path with pretensioner and load limiter, and the metric definitions — the next step is to translate them into a parameterized reduced-order architecture whose states and outputs correspond one-to-one with the quantities discussed here.