Got a few ideas bouncing around now - 1) scale up your 90/10 test by 100 then you can get it in a tall column and use ultrasonics to measure the level - but I've got a feeling that the height you need to get acceptable accurancy will mean a very narrow column and a it may distort the ultrasonic return, then again, perhaps not, so long as it gives a definite level, then you can calibrate it and figure out (automagically) the height of the least dense column. 2) Specific Gravity could be used (known volume of methanol + known volume of process fluid, compared against SG of pure methanol) but you'd need to know what SG is acceptable compared to the SG of Methanol, but that could be fine tuned on the fly.
How difficult would it be to set up a trial?
If you know that glycerin will form after XX minutes, then you just sit and wait for a bit until the timer expires, then drain until a float level indicates no more glycerin in the tank (you set the float to be heavier than the oil, but lighter than the glycerin and a simple push switch for it to land on indicates that the glycerin is drained - there's something like this in the fuel filter on landcruisers to indicate there's water in the fuel filter) once the glycerin is removed, you've changed the SG of your process fluid, and if you've hit the right level, you move it out to your post-wash, or you dose up more methanol and wait a bit longer.
The glyc is produced throughout the reaction process, with the greatest amount being produced in the first 5 - 10 minutes. So the first process runs for say 90 minutes then 45 mins to allow the glycerin to drop out of the fuel - so sounds like this may work well!
Might need a bit of trickery to filter out unwanted vibrations and sudden changes, but the real question is what sort of variance in specific gravity do you see between batches? If it swings about a bit, then it may not be a good indicator.
I would imagine that the SG of biodiesel from similar feedstocks will remain reasonably constant at a given temperature?? But I could soon do a few tests and graph out any variation at say 40 C then graph differences at maybe 5 deg. Would this help us to calibrate?
Managed to find something a bit more interesting - direct measurement of water content in oil:
http://www.meas-spec.com/product/fluid-property-sensor-HTM2500B3C4OIL.aspx
So you can probably measure direct water content, does make me wonder how it reacts to methanol though....
The sensor could be used firstly to ensure that the feedstock is dry enough, before dosing. After this any water created in the reaction is irrelevant in that you can't do anything about it. Then during drying all of the methanol has been flush out by the water down to very, very low ppms. Water content of finished fuel should be <500ppm or half an ml in a litre. There is also an offline tester called a sandy brae http://www.sandybrae.com/watertest.htm . It's made in the U.S. and $250 plus p&p but it simply uses a pressure gauge to measure the increased pressure from gas when the fuel is reacted with Calcium Hydride - could that principle be used anywhere?
Valves start getting expensive - a 1mm port valve (typical air actuator) is less than £20 per valve (around £16 on RS) but the actual pneumatic valve is around £200 each (solenoid valves are not suitable for the sort of pressure/flow I see going on in the appleseed reactors) and if you need half a dozen of them, then you're looking at the thick end of £1k before you've even looked at electronics to drive it all (and at this point I'd say get a dedicated PLC).
I'm wondering if the torque from an RC servo can be scaled up to operate a ball valve (the valves don't need to be fast acting do they) and if that can work, then there's a cheaper solution there and that can be driven with something cheap