I think there is a real opportunity out there for an automated bank system that would allow an accurate progression through the banks at optimal flows. This could save a lot of energy and let the diver walk away with an acceptable over without consuming overly compressed gas. It wouldn't be much more than a collection of pressure switches, controlled valves, a PLC/HMI and a little software. A fun weekend project!
disclaimer: not a fill tech, but amature mad scientist. mods, if this goes too deep, feel free to break it off in its own thread. if i am incorrect in my understanding from research, please smack me down gently.
been mulling on this since you posted it. its one of those problems that seems simple, but as you look closer, gets very complex very quickly. using closed loop feedback for a single variable is pretty straightforward, but with filling you're juggling a lot more of em, while they are all interacting with each other, while trying to stay within regulation:
environment: if you dont want to get sued out of house and home, so you have to play by the rules. that means not knowingly exceeding industry limits, for me, this means CGA's fill limit of 70deg f @ working pressure. so no working backwards from tank air mass and calculating that for this fill's situation 3150psi will settle to 3000psi at 70f. the obvious cop out is a manual fill pressure adjustable by the operator... but that opens the liability window for willful misconduct of the operator.
flow rate: 300~600psi/min seems to be the common 'best practice', and i believe 600psi/min is the absolute max allowed by CGA. ive gotten 600psi/min before at a local quarry fill right off the compressor; 3200psi on the dock, 2800psi after splash. this is also diminishing returns, how close to 'perfect' do you want to be? with electronic control, what is the best psi/min vs psi curve? classic minmax problem: smack it with 600psi/min through 2000psi, then slow it down through the final pressure, or start with a slow 300psi/min or less, and then hammer it home? i think the former, but only experimentation would bear proper results.
working gas temp: what state is our fill gas in in the first place? hot off the compressor, or ambient stabilized banks? lets not even get into ppO2 combustion issues when dealing with banking nitrox.
feed pressure: the whole point of using cascades, larger usable psi range per bank tank. this is relatively straightforward for a single fill whip, just cycle through the cascade in order of increasing pressure. if you want to handle multiple whips, then you're looking at a much more complicated setup: cascade is routed to a pair of manifolds, one low, one high pressure, and the whips are switched from off to low pressure manifold to high psi manifold. oh, and pressure and temperature are related, so you have to figure out what feed pressure yields the best end result. more experimentation!
tank gas temp management: not directly measurable, but can be modeled. 3000psi line pressure does not mean much unless we have a good idea of how close to 'optimal' (again, working fill pressure @ 70f) the situation in the filled tank is. this depends on everything mentioned thus far.
compressor duty: how big is the compressor vs the cascade banks vs the fill rate? oh, and since we're talking intelligent control of cascade banks, do you use 1 compressor, or split it into two units
then you've got all the little production details. do you just mill and tap out a distro block and stick relays on it? or do you go full press and integrate your valving into the overall design? the former is $$$$$ for parts (air handling 'noids are pricey!) but may be best for a single whip design; the latter, $$$ for production complexity and time requirements, but allows much more flexibility in a multi whip design. for the simple single whip 3 cascade design, you're looking at least 5 'noid controlled valves (comp in, whip, bank [123]). for the dual manifold multi whip you've got two valves per bank, two valves per whip, and one or two valves for compressor input. thats ten (10!!) valves just for a three bank two whip setup.
while we're at it, might as well include the compressor and prime mover into the design. single compressor with binary load/unload is tried and true, but there may be some benefit to having a variable speed prime mover and/or breaking the compression cycle into two units. a two compressore design would also be very tightly coupled with the cascade controller and feed manifold pressure ranges. there are better layouts for multistage compressor that would possibly lower noise and vibration (boxer layout comes to mind). being more operator friendly on maintenance would be nice as well... seems nobody designs for the wrench monkey anymore.
oh, and we can't forget getting the unit certified! thats more $$$.
all this adds up to a very very high cost of entry, and subsequently a large sticker for any produced units. not really sure how bit the tank-fill market is, but i cant really see much room for a profitable super premium automated compress / bank / fill system. i think it'd be pretty awesome to design, implement and product such an item, but don't have 6 figures worth of capital laying around to throw at it and see if it sticks.
TL;DR: engineering is hard. making a profit is harder.