Search All of the Math Forum:

Views expressed in these public forums are not endorsed by NCTM or The Math Forum.

Replies: 8   Last Post: Sep 22, 2000 8:09 AM

 Messages: [ Previous | Next ]
 Dwarf Posts: 22 Registered: 12/7/04
Posted: Sep 20, 2000 2:28 PM

On Wed, 20 Sep 2000, Jacek Kierzenka wrote:

> Let me clarify the matter:
>
> Simulink algebraic loop solver parameters (tolerances, number of
> iterations, etc.)
>
> The 'relative and absolute tolerances under Simulation|Parameters' apply to
> the ODE solvers, only. This may or may not have impact on solving algebraic
> loops, but only indirectly.
>
> As for solving algebraic loops: Simulink simply tries to solve them as
> accurately as possible.
>
> Hope this helps,
> Jacek

It answers the question, but it raises another, and I'm starting to think
that I'm approaching the problem wrong (or rather, that I could be
approaching th eproblem differently), so I'll pose it here.

I'm going to be developing a model of a pipe network. We want to use
Simulink because we eventally need to overlay a control system on the
model, and with Stateflow it seems like it would work admirably.

I'm having a problem writing the block that will represent a pipe section.

Currently, each pipe-block takes an inlet pressure, an outlet pressure,
and a flow rate. It operates on these data, and calculates a new flow
rate, and (if appropriate for the pipe location in the network) a new
inlet and/or outlet pressure. The new outlet pressure and new flow are
fed to the next pipe downstream, but the new inlet pressure must be fed
back to the upstream pipe. This forms an algebraic loop.

My hope was that Simulink could solve this loop at each timestep and
balance the flow. It's not happening. I get random errors that it's
having trouble solving or that it's encountered a singularity. That, and
the pipe flows don't match (as much as a 2-5% change in flow within a pipe
section - conservation of mass is really an optional constraint, after all
:)

I'm willing to try just about any solution (if someone could point me to a
third party package for solving piping networks in Simulink, I'd be
greatly in their debt)

Thanks!

+----------------------------+ Want to know the secret to riding a bike
| | from Fairbanks to Anchorage?
| Greg Anderson | 1. Push down with left foot.
| Rogue Squadron - Rogue 31 | 2. Push down with right foot.
| dwarf@wam.umd.edu | 3. Repeat.
| timbwolf@glue.umd.edu | 4. Stop when you reach Anchorage.
| AIM - Aranthon |
| www.wam.umd.edu/~dwarf | I know it seems pretty simple, but it
| | worked for me. AAVR 1, 8/20-8/26, 2000
+----------------------------+ Want to help me wipe out AIDS? Ask me how!

Date Subject Author
9/18/00 Dwarf
9/18/00 Wayne S. Hill
9/19/00 Dwarf
9/19/00 Wayne S. Hill
9/20/00 Jacek Kierzenka
9/20/00 Dwarf
9/20/00 Wayne S. Hill
9/21/00 Wolfgang Hammer
9/22/00 Kelvin Hales