7 replies to this topic

[hana1]

Good evening all.

I hope everyone is well, I am a newbie to the forum, a student in process engineering.

I have some confusions about a few problems that i always get marked down for in exams and courseworks, now i wish to seek help because i am afraid of asking my teacher over and over again.

After reading several notes dispatched by him i havn't got a clue how i answer it in the exam or even using hysys for that matter .
Firstly; If using a Wegstein method in a simulation (or any other method) uses a convergence block,
in what ways can i improve the convergence if it has problems? (please see diagram)

 Image047.jpg   11.27KB   106 downloads

Also whats a HYLITE compared to Hysys?..This always comes up in exams and i leave it out blank.

And my final question

If i have been a given a simple model using HYLITE, please see diagram.  Image046.jpg   10.04KB   70 downloads

If this model is fully and correctly specified . On running it everything appears to work properly and the simulator reports that the calculation has converged after 23 iterations. But on checking energy balances it is found that there is a significant in discrepancy. From validation tests, models and property data & simulator are known to be fully reliable. What is the likely cause of the problem, and what can i do to cure it?..

I would be ever so grateful if you can help me solve my problems. God bless you all.

Hana.

[joerd]

Firstly; If using a Wegstein method in a simulation (or any other method) uses a convergence block,
in what ways can i improve the convergence if it has problems? (please see diagram)

You can set a few parameters in a Wegstein block, most notably the acceleration parameter. If you set it to a lower value, the recycle iterations will be "dampened" and that should reduce swings in successive iteration steps, which sometimes improves convergence.
Also, the tolerance is a big factor. You seem to have two "nested" loops, if you set the tolerance of the inner loop (Recycle 1) somewhat higher (looser) you will usually still end up with a good answer since the outer loop (Recycle 2) will repeat until it is satisfied, which forces the Recycle 1 to be executed as well, and improve its accuracy. When the simulation has converged, you should check if the accuracy of the answer is good enough, otherwise tighten the tolerance again.

Also whats a HYLITE compared to Hysys?..This always comes up in exams and i leave it out blank.

Never heard of HYLITE. If I Google for it, it seems to be some kind of material. Strange question for an exam.

And my final question

If i have been a given a simple model using HYLITE, please see diagram.

If this model is fully and correctly specified . On running it everything appears to work properly and the simulator reports that the calculation has converged after 23 iterations. But on checking energy balances it is found that there is a significant in discrepancy. From validation tests, models and property data & simulator are known to be fully reliable. What is the likely cause of the problem, and what can i do to cure it?..

Check the tolerance on the recycle loop. It may be too loose. Also, there is a possibility that some of your components cannot exit the recycle loop. This can cause the simulator to assume convergence while the material balance is not satisfied. Next thing to check is column convergence, which may be a factor as well. Look at all the streams, and try to make material balances and energy balances around each piece of equipment, this will help you focus where the error is located.
Good luck

[hana1]

^ Thank you very much joerd May God bless you

[hana1]

Hello.

I was reading through your answers and wanted to ask about my own answers too from my research notes.

For the first question, would it be also appropiate to add that we could use mass and energy balance to detect convergence errors and do a sensitivity analysis, also repeating the calculation with a different criterion (say 10 times higher), plus to try a different starting point and examine the effect it has on the results hence if results change then their is no problem?

Also, why do we tighten the convergence criterion? would this rule apply to fast and slow convergence?

You mentioned in your answer to use an 'accelarotor parameter'; but isn't Wegstein itself an accelerator? would it be possible to use the Newton method?

Thanks in advance.

[missbk]

Hello all,
I am a newbie to process engineering industry and to this forum.
Just a problem involved Hysys so I post it here...
I seek for the book "Plantwide dynamic simulators in Chemical Processing and Control" of Sir William L. Luyben, but what it turned is just a 43 page preview of this book, from google book search engine.
There are many examples and problems using Hysys to simulate and solve inside this book. The explanations are clear and easy to get.
I wish some of you may have that book and pleased to share... My email is thaodk02@gmail.com.
Thank you very much.

[joerd]

Hello.

I was reading through your answers and wanted to ask about my own answers too from my research notes.

For the first question, would it be also appropiate to add that we could use mass and energy balance to detect convergence errors and do a sensitivity analysis, also repeating the calculation with a different criterion (say 10 times higher), plus to try a different starting point and examine the effect it has on the results hence if results change then their is no problem?

This would be appropriate, but will not solve convergence problems. It will help you identify where the problems are, and will confirm if your simulation has indeed converged.

Also, why do we tighten the convergence criterion? would this rule apply to fast and slow convergence?

I don't quite understand this question. You can sometimes get a reasonable answer by loosening the tolerance, because the solver will be satisfied more easily. If you calculate the loop again, you normally will get a better answer (because it is converging). That's what happens if you have nested loops: the inside loop gets calculated until it is converged, then the outside loop is calculated. Normally, the inside loop is happy from that point on, but if the outside loop makes large changes in the streams feeding the inside loop, the inside loop may have to re-converge. In some cases, you can get an oscillating behavior, which may be solved by adjusting tolerances and/or damping. It is often a question of trial and error.

You mentioned in your answer to use an 'accelarotor parameter'; but isn't Wegstein itself an accelerator? would it be possible to use the Newton method?
Thanks in advance.

Wegstein is an acceleration method, so you can set its parameters (q) to make bigger/smaller steps, whatever you feel is appropriate. The Newton method is the Wegstein method without acceleration.

@missbk - Please read the guidelines for posting. First, don't start a new topic inside someone else's, because we all get confused. Second, trying to get copyrighted material without properly paying for it is illegal, and not tolerated in this forum. The fact that Google Books only gives you 43 pages for free should tell you something.

[missbk]

@ joerd: thanks for your guidance, I learned the lesson. Indeed I did order one online but just can not wait 'cos it takes 1 month shipping. But anyway sorry for bothering - I think this topic implies anything involved Hysys

[hana1]

Thank you Joerd, i get it now, but can i ask that if i was an engineer and i was to test the reliabilty of my results from computer based flowsheet calculations with regards to convergence, would i need to check on my calculations to see if they have converged and to ensure that the imbalances are consistent, could i test the senitivity of the results to the convergence criterion and starting point? Are these the correct measures for an Engineer to take to ensure reliability of their work, with regards to convergences and balances?


Thanks again.

Hana.