Control of coupled tanks

I don't see why you don't treat the two tanks like one big tank when
the level is above the equalizing line and as two separate tanks
below.  Tank level control using pumps on the outfeed has been covered
here in the last year.  There are other here like Fred Thomasson that
seem to have practical experience.   I can help with the theortical or
model based control.

The previous threads have shown that level control is usually very
simple.  A simle P gain will work most of the time.  The only reason
why you would have problem is if the pumps can keep up with the peak
inflow.

Peter Nachtwey

Are you measuring the outflow on the tanks and the level in
each tank?  From that, you can deduce the inflow.

Going further, if you look at the Cv of the control valves,
and measure the flow rates, then you will know the head and
hence the depth in the tank.

We need to know a bit about the exit flows.  What are they
being controlled to optimize?

Michael

On Mon, 28 Jan 2008 12:37:42 +0900, Bruce Varley wrote:


How wild is "wild"?  Peter's suggestion of simple proportional control
(at least if it's proportional control on the outflow valve position or
pump rate) should work well if you can use a high enough gain to keep
things within limits while still keeping the system stable.  With simple
proportional control I don't even think you need to worry about the
equalizer -- just operate each outflow from the level of it's tank.

If you can't get away with a simple proportional controller then you'll
have to accept using something like a PI controller -- only if you have a
integrators in there does the cross-flow become an issue, and then only
if both tanks are above the equalizer.  Were I doing this I'd consider
implementing PI controllers that stayed independent whenever the
equalizer was not in play, but slaved the two integrators together
whenever the level was above the equalizer in both tanks.

Come to think of it, you may be able to keep independent PI control on
both tanks, unless the equalizing line is huge: as long as the level in
each tank is at least somewhat dependent on the inflow, a PI controller
should work.  Your apparent plant gains will vary depending on whether
the level is below or above the lip of the equalizer tube, so you'll have
to tune it up to be well behaved in both circumstances, but it ought to
work.

This sounds like something that should be considered, and modeled, and
simulated well before you start implementing your control rules...

--
Tim Wescott
Control systems and communications consulting
http://www.wescottdesign.com

Need to learn how to apply control theory in your embedded system?
"Applied Control Theory for Embedded Systems" by Tim Wescott
Elsevier/Newnes, http://www.wescottdesign.com/actfes/actfes.html

Thanks for the comments guys, I appreciate your input. This was a bit of a
fishing post, unfortunately I can't post more details, of which there are
some. Basically what we need is an inner control shell that will enable us
to present the system as stable linear models across the whole operating
range, for an overarching controller that extends beyond the tanks, and
which needs to have access to the discharge handles.

Bruce,

I'm  not convinced there is an especial problem here: since each tank
has its own outflow controller, and the inflows are wild, then any flow
(in or out) between tanks through the equalising line is simply a
variation of the net inflow as seen by each individual tank.

Industry standard methods of calculating the level-controller tuning
settings on the basis of expected flow excursions and vessel dimensions
should be applicable.

Having said that, it is very easy in a thought experiment to miss some
unexpected behaviour, and since this arrangement would be trivially
simple to model and simulate in something like Simulink, I'd probably
do just that to demonstrate to myself (and prove to the client) that
there is no untoward behaviour.

Best Regards,
Kelvin B. Hales
Kelvin Hales Associates Limited
Consulting Process Control Engineers
Web: www.khace.com