Elect_Aspect_Course_2_1_2 - nasa/gunns GitHub Wiki

GUNNS is a ‘steady-state’ solver, which means that it assumes that the state of the system is steady and unchanging over the duration of each time step, and does not model any different transient states that the real-world system would exhibit in between the time steps. This applies to any time-dependent internal effects (like capacitance or inductance) as well as discrete external events (switches opening and closing). Higher-fidelity models than GUNNS will usually have a dynamic solver that solves the governing PDE of the system (Maxwell’s Equations for electrical circuits or Navier-Stokes for fluid, etc.), and calculates the transient state of the system at any point in time. This is illustrated below. This is an example of what the output voltage of a switch would look like after it is closed. In the real world, there is an inrush of current fed by an upstream supply that will oscillate until it converges on the regulated voltage (dynamic), whereas in GUNNS, the model simply jumps to the regulated voltage (steady-state).

TODO make this smaller

Real-world electrical systems are full of transient effects like this. These are typically very high frequency and damp out on the order of micro-seconds, and aren’t seen at the relatively large time steps (tenths of seconds) that we typically run GUNNS networks at. GUNNS networks of any realistic circuit are too computationally slow to update fast enough to model such real-world electrical transients, and even if you could update GUNNS that fast, it still wouldn’t model those effects, because they are dynamic.

TODO more talking points:

• why is gunns steady-state vs. dynamic: linear vs. PDE and rectangular (Euler) integration of time effects
• why this matters more for elect vs. fluid: fluid has lower freq. transients
• we still model transients, however these are discrete/logic models, and limited to time step