Problems with COCO 3

[nrgeng]

Finding several problems with COFE 3
Running COFE 3.0.0.19 (C3) along side COFE 2.7.0.23 (C2)

1. FS4.3.1.fsd (FS) runs in C2 in 3s 432ms, but does not run when loaded into C3 unless several control loops are discarded then solves in C3 in 9m 14s 688ms. [A,C]
2. Streams do not turn green when the FS solves in C3 but do in C2. [A,B,C]
3. The log in C3 sometimes adds phrases at the end of a line because it does not provide a line feed. This produces a rather long string of information to read. [D]

Emailed images for your review that are not for public viewing.

Support Images

A. V4.3.1 Solved & Green in C2.jpg
B. V4.3.1 Solved but NOT Green.jpg
C. V4.3.1 Solved but NOT Green 2.jpg
D. V4.3.1 No Line Feed in Log.jpg

[jasper]

Is there any way you can provide me with a repeat of such behaviour? If I cannot reproduce the behaviour it is next to impossible to resolve.

An fsd would be most helpful. If the information in there is deemed important, perhaps simply change the numbers, compound slate, etc?

[nrgeng]

I understand your request for an fsd from which to work, but that will not be possible. The preparation of a shareable fsd version would just be too time consuming. I feel that spending my time debugging myself will minimize my overall troubleshooting time. I describe my technique below.

May I suggest that you keep a list of the reported problems. Then compare reports from other Forum members as they use COCO 3. My reported problems will bolster new problem reports if they compare. Hopefully the reports by other Forum members will provide an example fsd for your use. If I develop an fsd that exhibits a problem and is sharable, I will of course submit it for your use.

I the interim, I will be attempting to simplify my flowsheet by using Custom Units. My technique involves separating a section, working with it until it solves, and then integrating it into the main flowsheet. Eventually all of the subsections will solve in the main flowsheet as they did when solved independently. It is time consuming but it does work and I know of no other method to use with COCO problems. Do you have any suggestions on this subject?

On another matter: I am involved with several issues as a Windows 10 Forum member. Other members are furious with Microsoft and frustrated in having converted to Windows 10 from Win 7 or Win 8/8.1. The Forum consensus is to NOT convert to Windows 10 for ~6 months until these issues are resolved (if ever). For example, the "Startup Menu and Cortana Issue" has almost 12,000 postings. Some members are locked out from the use of their computer and use an Apple iPad to communicate with Microsoft. Not a good publicity move on Microsoft's part!

[jasper]

Cortona can be turned off (I am in Spain, it is not available here). The only issue I have with Win10 is with some very specialized debuggers; other than that I am quite happy with it. I do experience some issues with search, particularly that it does not find items that are in the start menu, even after re-indexing. Windows 7 and 8 are bothered with the same though.

Isolation of sub-problems is a good strategy in any flowsheeting application. Typically a flowsheet has three kinds of problems: 1) bugs, best solution: report and they will get resolved, if I can reproduce them, 2) unfeasible problem specifications: possible solution, put the log in verbose mode and look for debug messages on which variable has the higher residual. Check what influences that variable. 3) convergence issue. Best solution is to specify a reasonable initial guess; as with (2) look at the verbose log and the residuals. Also often used: solve with recycles not closed. Close recycles once they become reasonably estimated. And as you do, try to start with a simpler sub-problem.

Your problems are noted and I hope you will report further on them.

[jasper]

Item (3) has been resolved (CUP).

[jasper]

For item (1), what do you mean with "Does not run"? Is there an error?

[nrgeng]

Thank you for the tips on obtaining solutions. I believe that my problems with FS4.3.1.fsd are mostly, perhaps exclusively, Type 2 and Type 3, especially since the fsd solves in COCO 2. I do use Verbose mode but not to the extent that you have recommended. I will try what you have suggested regarding residuals. I also open recycles to prevent disturbances from affecting isolated sub-blocks being solved. Thanks for the great advice.

My use of "does not run" is a loose use. The fsd runs but does not complete to a solution within 30 to 40 minutes. Also, I could not stop the simulation, after 40+ minutes, short of using the computer's Off button. The red COFE Stop Button could have been connected to the moon for all the usefulness it provided in halting the simulation. What definition(s) do you use so that I may also adopt it(them) for clarity in future communication?

[jasper]

Sounds like something is unresponsive. If the simulation does not halt, I suspect a thermodynamic calculation or unit operation calculation does not terminate. Such behaviour would go hand in hand with the iteration count not increasing.

To resolve such an issue I certainly would require to reproduce it.

[nrgeng]

From our exchange here, my remaining question is if a flowsheet solves in COCO 2, why does not it solve in COCO 3? It is the same fsd file. Was the COCO 2 simulation in error? The correct answers in COCO 2 should be the correct answers in COCO 3. Right?

For example, COCO 2 ran in Windows 7 and it ran in Windows 10 without any problems. So why the difference in running the same fsd file between COCO 2 and COCO 3?

[jasper]

I cannot answer that without seeing the fsd. COCO 3 has completely new solvers, new file format, etc.