7 replies to this topic

[bhautesh.mistry@gmail.com]

Hello Engineers!

 

I am simulating a column using Aspen HYSYS v8.4 where sometime I do not get convergence of it even when equilibrium and Heat/spec errors are under the specified values. Please look at the screenshots attached here.

 

Does anyone have faced same problem? If so then kindly suggest any possible way to come out of this problem.

 

Thanking you in advance.

 

Regards,

 

Bhautesh

Attached Files

•  Column solver.png   50.35KB   31 downloads
•  Column unconverged.png   56.02KB   18 downloads

[Bobby Strain]

First, change the specifications to make the column converge.

 

Bobby

[bhautesh.mistry@gmail.com]

Dear Mr. Strain,

 

Thank you for your response.

 

I tried that option. The problem is sometime column converges right after I change specifications and sometime the solver just run with the said allowable spec tolerances as shown in the attached image.

 

My task is to use different top and bottom stage temperature as specificaitons and do the analysis.

 

I hope to get some more valuable suggestion from Aspen HYSYS users.

 

Thank you.

 

Regards,

 

Bhautesh

[Bobby Strain]

You need to choose some different parameters for the specification. One thing you must learn about simulators. If something doesn't work, try something different. Don't keep beating a dead horse.

 

 

Bobby

[bhautesh.mistry@gmail.com]

Dear Mr. Strain,

 

I know that (and I did) to converge any unit process or operation using simulation tools, different approaches need to be considered and implemented. The purpose of this post is to get the knowledge from the people who faced the exact said problem.

 

Thank you for your kind reply.

 

Best regards,

 

Bhautesh

[DanW]

I think that what Bobby is trying to say is that first you converge the column with different specifications (slightly modified) and then try again with your desired specs using the previous converged solution as the starting point.

[bhautesh.mistry@gmail.com]

I think that what Bobby is trying to say is that first you converge the column with different specifications (slightly modified) and then try again with your desired specs using the previous converged solution as the starting point.

Dear DanW,

 

I already did that. I am curious to know why this problem had happened. Just a value addition to my simulation knowledge.

 

Best regards,

 

Bhautesh

[Luke1939]

Bhautesh,

 

Please excuse my interruption into this dialog. I am a "Johny come lately" here and am a recent member of CheResources.com. I know nothing about Aspen HYSYS other than what I just quickly observed browsing in the Internet. 

 

Having worked with some dynamic simulation programs (ACSL, in particular) in the past going back to the mid '60s and having written many of my own routines in Basic, Fortran and for the past 35 years in Mathematica and having seen major improvements in simulation confidence as computers have improved. While working in Mathematica, I am privileged to be able to dictate the computational accuracy whether it be 20 digits, 200, or 2000 digits. I am not limited to machine accuracy issues  

 

For your issue, I am making an assumption that Aspen HYSYS is using a finite differences algorithm. This is common. Finite difference solutions sometimes have a personality problem in that it will not allow computational conversion because:

1. the step size is too large

2. the forcing function is too large

3. the desired computational accuracy is too small

4. because of limits in the computer computational accuracy. If the error is lost in the weeds of computing accuracy, you will absolutely never get convergence.

5. something is grossly in error with the input data. I have had the dubiously personal distinction of having created my own problems. On many occasions I have verified the expression of GIGO, Garbage-In-Garbage-Out. One must be particularly mindful that the input is accurate and that the units are compatible. Another insight gained was that I was calculating garbage with great accuracy. Unfortunately, one can be deceived in to believing that accuracy is guaranteed just because it was calculation was done on a computer.

 

There is another possible issue that I have run into and that is that the computational algorithm may be written with first order approximations to the fundamental differential equations. Your parameters may require a higher order approximation of the differential equations. I have definitely run into that. I have learned to write all my routines with 2nd or 3rd order approximations.

 

If you were to study and understand the mathematical dynamics of finite differences, you would find that there are mathematical tests such as the Von Neuman stability method to determine the maximum allowable step size for a given forcing function that will allow for efficient conversion.

 

I seriously recommend that the computational algorithms used in the Aspen HYSYS software be fully understood otherwise, you are just blindly using the software package with total ignorance and having no understanding of how the results are computed. That is scary.

 

​I suggest that you generate a closed-form solution to a simplified text book problem and compare your answer to the software package output if only to generate confidence with that particular software package.

 

Please excuse my professorial response, I hate to see somebody flounder around in circles in frustration.

 

Good Luck,

 

Luke1939