3. Pretensioner Gas Generator Pressure–Time Tuning for Stroke Consistency

Why Stroke Consistency Is a Gas-Output Problem

In a pyrotechnic pretensioner, the stroke delivered to the webbing is not set directly — it is the integrated result of the gas generator's pressure–time history acting on the piston/retractor mechanism. Two design levers dominate that history: the propellant grain geometry (which controls the burn-rate profile and therefore the shape of the pressure rise and decay) and the orifice size (which controls how quickly combustion gas is metered into the working volume). Because propellant burn rate is strongly temperature-dependent, a grain/orifice combination that produces a clean stroke at +23 °C can under-deliver at −35 °C (slow burn, low early pressure) and over-deliver at +85 °C (fast burn, high peak pressure). The calibration task is to find a geometry–orifice pair whose integrated pressure impulse stays within a stroke tolerance band across the full −35 °C to +85 °C range, verified against measured pressure–time curves.

Defining the Target and the Acceptance Window

Before tuning, fix the acceptance criteria so the search has a stopping condition. A representative specification for this exercise (example values, to be replaced by the program's real requirements):

The stroke tolerance is the binding constraint: it must hold at every temperature, not just at the nominal one. This is why a single-temperature tune is insufficient and the pressure–time curve must be captured at multiple soak temperatures.

How Grain Geometry and Orifice Shape the Pressure–Time Curve

The propellant burn follows a pressure-dependent rate law, approximately ṁ_b = A_b · a · Pn, where A_b is the burning surface area, a is the temperature-sensitive burn coefficient, and n is the pressure exponent. The gas generator pressure then follows a mass balance between generation and outflow through the orifice:

dP/dt = (R·T_g / V) · (ṁ_b − ṁ_out) − (P/V)·dV/dt
ṁ_out = C_d · A_orifice · P / √(R·T_g)   (choked-flow approximation)

Three geometry families give distinctly different pressure–time signatures, and the choice drives temperature robustness:

• Progressive grain (burning area increases, e.g., internal-burning star or slotted): pressure rises through the burn — good for maintaining late stroke but risks high peak at +85 °C.
• Neutral grain (constant burning area, e.g., tubular with inhibited ends): flat pressure plateau — the most temperature-stable impulse, usually the best starting point for stroke consistency.
• Regressive grain (burning area decreases, e.g., single-perforated with exposed ends): pressure peaks early then decays — fast initial stroke but weak tail, sensitive to cold under-delivery.

The orifice acts as the metering element: a smaller orifice raises chamber pressure and lengthens the pressure tail (more impulse, but higher peak); a larger orifice vents faster, lowering peak but shortening the useful impulse. The tuning problem is therefore a two-variable search — grain geometry (which sets the shape) and orifice area (which sets the level and duration) — evaluated against the integrated impulse that maps to stroke.

Working the Tuning Search

The figure above shows the failure mode this case targets: a neutral-grain baseline tuned at +23 °C delivers 181 mm, but the cold curve's slow burn drops the impulse to 168 mm (−13 mm) while the hot curve's fast burn pushes it to 192 mm (+11 mm) — both outside the ±8 mm band. The tuning objective is to compress that temperature spread. Two adjustments act on different parts of the curve:

1. Reduce temperature sensitivity via grain geometry. Shift from a mildly progressive grain toward a neutral (constant-area) grain so the burn-area profile does not amplify the temperature-driven burn-rate change. This flattens the spread between the cold and hot curves rather than shifting the whole family.
2. Re-center the level via orifice area. After the shape is stabilized, trim the orifice to bring the nominal impulse to the center of the band. A slightly larger orifice lowers the hot peak (protecting the 42 MPa ceiling) at the cost of some cold impulse, so the orifice is adjusted after the geometry, not before.

A practical search sequence: hold orifice fixed, sweep grain geometry across the three families at the three soak temperatures, and select the family with the smallest cold-to-hot impulse spread. Then sweep orifice area in small increments (e.g., 0.1 mm diameter steps) to center the nominal stroke. Because each fire is destructive, the sweep is done on a design-of-experiments basis with a small number of builds per point, and the pressure–time curve is the primary observable — stroke is inferred from integrated impulse and confirmed on a reduced confirmation set.

Reading the Measured Curves

For each soak temperature, capture chamber pressure versus time and check three features against the acceptance table:

• Peak pressure — must stay under 42 MPa at +85 °C. If the hot peak exceeds the ceiling, enlarge the orifice or reduce grain web thickness.
• Impulse (area under P–t) — the proxy for stroke. Compute the integral to t₉₀ and compare the cold/hot integrals to the nominal. A cold integral more than ~4% below nominal signals under-delivery; a hot integral more than ~4% above signals over-delivery.
• t₉₀ — time to reach 90% of stroke. If the cold curve's t₉₀ exceeds 18 ms, the burn is too slow for the crash-pulse window and the grain needs a faster-burning formulation or a thinner web.

Typical failure signatures and their corrective direction:

Confirmation and Sign-Off

Once the geometry–orifice pair holds the predicted stroke within ±8 mm across all three soak temperatures and the hot peak stays under 42 MPa, run a reduced confirmation set (multiple builds per temperature) to verify repeatability, since pyrotechnic output has build-to-build scatter. Record the final grain geometry, orifice diameter, and the measured P–t curves at each temperature as the calibration evidence. The pass condition is a stroke band that stays inside tolerance at every temperature with margin, not a single nominal point that happens to land on target.