A quasi-1D performance model built from the governing equations up — and the validation pass that turned up a discrepancy worth chasing.
Designing engine hardware like DEIMOS means predicting performance before metal is cut — chamber pressure decay, characteristic velocity, thrust coefficient, and specific impulse all need an estimate that geometry decisions can be checked against. Off-the-shelf rocket performance tools assume steady deflagrative combustion; an RDE's chamber pressure is anything but steady, so I built a quasi-1D model specific to the detonation cycle instead of adapting a tool that wasn't meant for it.
The model couples two pieces of prior work rather than inventing the physics from nothing: Shepherd & Kasahara's pressure decay relation governs how chamber pressure evolves behind the detonation front, and Stechmann's temporal framework provides the time-resolved structure needed to track that decay through a full cycle rather than just at a single instant.
A model is only useful if it's been checked against something that actually fired. I validated the toolbox's outputs against published RDE hotfire data, comparing predicted c*, Cf, and Isp against measured values across the available test points rather than relying on a single favorable comparison.
That validation pass surfaced a consistent pattern rather than scattered noise: the model overpredicts specific impulse by roughly 23–45% relative to measured hotfire values, and the size of the gap moves with operating condition instead of sitting at a fixed offset. A fixed-offset error would point to a calibration constant; a condition-dependent one points to a missing physical effect.
Candidates under investigation include unmodeled heat loss through the chamber wall, incomplete combustion efficiency assumptions inherited from the pressure-decay relation, and nozzle expansion losses not captured by the current cycle-averaged Cf treatment.
I'm treating the overprediction as the most useful output of the validation pass, not a flaw to patch over quietly — isolating which term in the cycle-averaged integration is responsible will tell us something real about RDE performance, not just about this model. That investigation is ongoing alongside continued use of the toolbox to inform geometry decisions on DEIMOS.