Assumptions and Limits
A pressure number is only as defensible as the assumptions behind it. This page lists all of them, says which way each one moves the answer, and marks the point where the model stops being the right instrument.
The Six Modelling Assumptions
| Assumption | What it means | Effect on the result | When to worry |
|---|---|---|---|
| Quasi-static equilibrium | The gas reaches thermodynamic equilibrium within each instant, so an equilibrium equation of state can describe a transient event | Pressure waves and local transients are not resolved | Almost never for valve selection. Pressure equalises at the speed of sound, far faster than the event |
| Ideal gas | The vented mixture behaves ideally, with no intermolecular forces and negligible molecular volume | Below roughly 1 MPa the error is small. Above it, real gas behaviour makes the true pressure higher than modelled | Only for unusually high pressure systems. Typical pack peaks are tens of kPa above ambient |
| Lumped, uniform volume | One pressure and one temperature everywhere inside the enclosure, with no spatial variation | Local pressure near the venting cell is underestimated | When you care about local structural loading or the valve sits far from the vent path |
| Constant composition | A single effective gas constant applies for the whole event | Real composition shifts as venting progresses, from electrolyte vapour to permanent gases | When your venting data was measured on a very different composition to your case |
| Prescribed gas temperature | Temperature is an input, either constant or a supplied curve, not something the model solves | The answer is only as good as the temperature you assume | Always worth a sensitivity check. See below |
| Rigid enclosure | Void volume does not change with pressure | Conservative. A real lid bulges, which increases volume and relieves pressure | Rarely a problem. It errs on the safe side, but a very compliant enclosure makes the model pessimistic |
The last two deserve expanding, because they are the two most likely to mislead.
Temperature Is an Input, Not a Result
The model does not solve an energy balance. It does not know how hot the vented gas is, how much heat it loses to the enclosure walls, or how the cold air already inside the pack mixes with it. You tell it the gas temperature, and it uses that.
This matters because temperature enters the source term directly: the same
number of moles at a higher temperature produces more pressure. Running with
Assumed Constant at ambient is a common starting point, and it will
understate the peak if the vented gas is genuinely hot.
The practical response is to bracket it. Run the case at ambient temperature and again at a plausible elevated gas temperature, and see how much the peak moves. If the answer barely changes, the assumption is not carrying your result. If it changes a lot, you need real temperature data before you trust the number. See Temperature Modes.
Rigid Enclosure Is the Conservative Choice
An enclosure under internal pressure deflects. That deflection adds volume, and added volume slows the pressure rise. Holding the volume fixed therefore predicts a higher peak than reality, which is the direction you want to err in for a safety component.
It stops being harmless when the enclosure is compliant enough that the added volume is a meaningful fraction of the void volume. Then the model is not just conservative, it is pessimistic enough to reject valves that would have been adequate.
Scaling Assumptions
Two inputs scale a single measured curve, and both scalings are linear.
Number of Venting Cells multiplies the venting curve. This assumes every cell follows the same curve, starting at the same moment. Real propagation is sequential, with delays between cells, which spreads the same total gas over a longer time and produces a lower peak. Simultaneous venting is therefore the conservative reading, and it is the right one when you are demonstrating that an enclosure survives a stated number of cells.
If you need the sequential case instead, build a combined venting curve externally, with the delays included, and load it as your own data.
Number of Valves multiplies the flow curve. This assumes identical valves seeing identical conditions.
What the Model Does Not Include at All
- Combustion, ignition, and explosion. The gas is treated as inert.
- Particle ejection and clogging. Real venting carries solids and liquid droplets. These can restrict a flow path, and the effect is not represented.
- Membrane behaviour under real ejecta. A breathing membrane may block vaporised electrolyte instead of passing it, so the real relief path can underperform its clean flow curve.
- Enclosure deformation and failure mechanics. The tool reports a pressure. Whether your structure survives that pressure is a structural question.
- Backflow and oxygen ingress. Flow through the valve is treated as outward only.
- Anything downstream of the valve. Dispersion, accumulation, and ignition outside the pack are out of scope.
- Whether thermal runaway occurs at all. The venting curve is an input.
Where the Numbers Are Least Certain
In practice, the model itself is rarely the dominant source of error. Input data usually is, in this order:
- The venting curve. Cell venting measurements vary widely with trigger method, state of charge, and instrumentation. A curve measured under a different protocol can shift the peak substantially.
- The gas temperature. See above.
- The valve flow curves, particularly whether they were measured over the pressure range your case actually reaches.
- The void volume, which is often estimated rather than measured.
The model contributes less scatter than any of these. Improve the data before questioning the equation.
When to Stop Simulating
This tool is for narrowing a field of candidate valves quickly and cheaply, and for understanding which variables your design is sensitive to. It is not a substitute for a validated test.
Move to physical testing when any of the following is true:
- the margin between predicted peak and enclosure limit is small
- the answer swings significantly across plausible input ranges
- you are close to a final design freeze
- someone must certify the result
Treat a comfortable pass as a reason to proceed with confidence, and a marginal pass as a reason to test.