3. Characterization test methods

An airbag inflator is a small, sealed gas generator that must deliver a precisely shaped pressure pulse into a cushion within tens of milliseconds. Because the device is consumed during deployment, its performance cannot be measured directly in the vehicle. Instead, characterization relies on a controlled surrogate environment: a fixed-volume tank test in which the inflator discharges into a closed pressure vessel and the resulting pressure history is recorded. From that single pressure–time curve, engineers reconstruct the mass flow rate, total delivered mass, gas temperature, and effective thermodynamic properties that later feed occupant-restraint simulations. This chapter explains how those tests are configured, how the pressure signal is acquired and conditioned, and how the raw trace is converted into the physical quantities a CAE model actually needs.

Tank Testing: The Controlled Surrogate Environment

The standard configuration is a rigid steel or aluminum vessel of known internal volume, typically 20 L, 60 L, or 100 L depending on inflator class. The inflator is mounted in a boss on the tank wall so that its discharge axis points into the free volume, avoiding direct impingement on the pressure transducer. A piezoelectric or piezoresistive pressure sensor is flush-mounted in the opposite wall, and the tank is instrumented with a thermocouple or resistance temperature detector (RTD) to record the initial wall and gas temperature. The vessel is sealed and, for most tests, evacuated or filled with dry nitrogen at a defined initial pressure and temperature.

Three parameters dominate test validity. First, tank volume must be large enough that the peak pressure stays within the linear range of the transducer but small enough that the pressure rise is well above noise; a 60 L tank with a driver-side inflator typically peaks between 200 and 400 kPa absolute. Second, initial temperature must be recorded because the ideal-gas relation used in post-processing is temperature-sensitive; a 10 K error in initial temperature propagates directly into the derived gas mass. Third, sensor bandwidth must exceed the frequency content of the event. A typical inflator pressure rise reaches 90 % of peak in 10–30 ms, so the useful signal energy lies below roughly 1 kHz, but combustion instability and diaphragm rupture produce sharp transients that demand at least 20 kHz sampling to avoid aliasing.

Pressure–Time Curve Acquisition

The pressure trace is the primary measurand, and its quality determines everything downstream. The sensor output passes through a charge amplifier or bridge conditioner, then an anti-aliasing low-pass filter set to roughly one-third of the sampling rate. The digitizer must be triggered synchronously with the firing command so that t = 0 corresponds to the instant current is applied to the initiator. A pre-trigger window of 5–10 ms is recorded to establish the baseline pressure and to verify that no drift or offset is present before the event.

After acquisition, the raw trace is conditioned in three steps. Baseline correction subtracts the mean pre-trigger pressure so that the curve starts at zero gauge pressure. Offset and drift removal addresses slow thermal drift in the transducer, which can otherwise appear as a false pressure rise over the 200 ms record. Filtering applies a zero-phase low-pass filter, typically a fourth-order Butterworth at 2–5 kHz, to suppress ringing without shifting the curve in time. The result is a smooth, monotonic pressure history P(t) that rises from zero to a peak and then decays as the gas cools against the tank walls.

Deriving Mass Flow Rate and Effective Gas Properties

The pressure trace alone does not directly give mass flow. The conversion rests on the ideal-gas law applied to the tank volume, with the crucial assumption that the gas inside the tank is well mixed and at a uniform temperature. At any instant, the mass of gas in the tank is

m_gas(t) = (P(t) · V) / (R_eff · T(t))

where V is the known tank volume, R_eff is the effective specific gas constant of the combustion products, and T(t) is the instantaneous gas temperature. The mass flow rate is then the time derivative:

ṁ(t) = d/dt [ (P(t) · V) / (R_eff · T(t)) ]

The difficulty is that T(t) is not measured directly. A thermocouple in the tank responds far too slowly to track a 30 ms event. Instead, T(t) is obtained from an energy balance: the enthalpy entering with the gas is balanced against the internal energy increase of the tank contents and the heat lost to the walls. For a first-order estimate, the process is treated as adiabatic over the short event duration, giving

T(t) = T_initial · (P(t) / P_initial)^((γ_eff − 1) / γ_eff)

with γ_eff the effective ratio of specific heats. This is where the term effective gas properties becomes important. Real inflator gas is not a single species; it is a mixture of N₂, CO₂, H₂O, and trace species, with temperature-dependent heat capacities. Rather than model every species, characterization collapses the mixture into two effective constants, R_eff and γ_eff, chosen so that the reconstructed mass flow matches the known total propellant mass and the measured peak pressure. A typical fitting procedure is:

1. Assume an initial guess for γ_eff (e.g., 1.25 for a typical pyrotechnic mixture).
2. Compute T(t) from the adiabatic relation.
3. Compute m_gas(t) and integrate ṁ(t) over the event to get total delivered mass.
4. Compare with the propellant mass loaded into the inflator (known from the bill of materials).
5. Adjust γ_eff and R_eff iteratively until the mass balance closes within 2–3 %.

The resulting ṁ(t) curve typically shows a sharp initial spike as the initiator and booster charge burn, followed by a broader plateau as the main propellant grain deflagrates, and a decay as the gas cools and the pressure driving force drops. The integral of ṁ(t) over the event is the total gas mass, which for a driver-side inflator is on the order of 50–120 g depending on the design.

From Curve to Model Input

The final step is to package the derived quantities for simulation. A typical inflator model in a restraint CAE code takes the form of a mass flow rate table ṁ(t) plus a temperature table T(t), or alternatively a pressure-dependent discharge coefficient C_d·A_eff and the effective gas constants. The tank test provides all of these. Because the test is repeatable and the tank volume is fixed, the same inflator can be characterized at multiple temperatures (−35 °C, 23 °C, 85 °C) to capture the temperature sensitivity of the propellant burn rate, and the resulting family of curves becomes the input deck for the occupant simulation.

In practice, the quality of the final restraint model is bounded by the quality of the tank test. A clean baseline, a correctly triggered record, a properly filtered curve, and a mass balance that closes are the four conditions that separate a usable characterization from a curve that merely looks plausible. When those conditions are met, the derived mass flow rate and effective gas properties carry the physical fidelity needed for downstream simulation.