Solve Dynamic Problems

[rhsrxs]

Benjamin Horwitz wrote an article in CEP about using a steady state simulator to solve dynamic problems (Avoid Nausea When Solving Dynamic Problems). He referenced using Chemcad which had the ability to set the Max Recycle iterations to 1. I noticed the lowest COCO can be set is 2. Can COCO be used in this way?

[jasper]

It is not intended to be used in this way; it is intended to be a steady state flowsheeter.

You can of course try. Note that the minimum iteration count is 2 as the initial guess counts as the first iteration.

[colancto]

I remember using such an approach with ProSim Plus a long time ago, like 15 years ago. You may create a sense of dynamic by moving from one loop iteration to another but it is easy to make mistakes especially since the time constants of each block have to be carefully evaluated. There is no time step you may easily set and you have to relate loop iteration to time step, making sure the time step and the time constants of each model are consistent. Even if dynamic simulation is not yet a very current practice, it has made good progress with such tools as INDISS Plus, Dynsim, Aspen Dynamics, gPROMS and so on. CAPE-OPEN has been looking into standardizing what is a dynamic Unit Operation and this has been prototyped in INDISS Plus. We need however to progress some more in that direction.

[rhsrxs]

I have implemented the salt dilution problem in a COCO flowsheet. I have also worked on the dynamic problems from the CEP article "The Case of the Collapsing Can". I will upload when I document the flowsheets. I have tried uploading an attachment but I get an error - "The extension fsd is not allowed". How do I upload flowsheets?

[colancto]

here it goes

[jasper]

That seems to work now.

[rhsrxs]

Attached is the salt dilution example.

[HenkF]

Regarding dynamics with a steady state simulator : at Billiton/BRA in the past (late 80's) we developed a dynamic simulator, which actually was a steady state simulator, without convergence aids, just a plain direct substitution of recycle streams, slow but sure. Timing was mimicked by limiting recycle stream flows/flow fractions. You also need residence times / volumes for important units to be set. And definitely a direct substitution convergence mechanism, no newton / broyden. Later this 'steady dynamics' approach has been succesfuly used with Aspen Plus simulations as well. I guess current Chemcad approach is based on something like this as well?
Henk

[colancto]

Good discussion. For sure you may get direct substitution convergence scheme to mimick dynamics. I consider that then an Euler explicit scheme is used. Not very fast for sure, not very accurate either and not always stable. But you get answers at a low license cost. Somebody well aware of the pitfalls of that approach will find it quite adequate. Back in the 90s I developed a dynamic simulation course in a chemical engineering school that indeed made use of this approach: was giving a good feeling of how dynamics worked, was easy enough to implement for students and easy to get unstable when necessary.

[jasper]

Here I have to point out that COFE may not be suitable for performing this type of iteration. The first iteration for sure will be direct substitution if the Wegstein scheme is chosen (unless Wegstein is not possible, for example because controllers are part of the recycle) as the first 2 or 3 iterations of the WegStein scheme are in fact direct substitutions, but the initialization of recycles also performs a Calculate on zero of more of the unit operations in a recycle.

The example that was posted however was not set up using a single iteration. It was set up using an entire solution and a manual direct substitution. This of course works, potentially (not sure whether there are any pitfalls here). The specified iteration count in this case of course does not matter either.

In case no recycles are present (which would be the case here if the controller loop is broken) there are no iterations. Solution is direct, from front to end. Such a case would of course also work.

The picture gets more complex in case there are multiple recycles that are not coupled. In this case COFE will solve recycle by recycle after analysing the connectivity.