I remember this one from quite a while back, a model was said to be very slow and not very reliable in it's convergence. The model was supposed to be part of a data management system on a plant, but because of the encountered problems the engineers decided not to use it on-site.
Upon inspection the model didn't look particularly large or complex, so there wasn't any obvious reason for it to slow or unstable. There was a pump around external to the column that needed a recycle operation to make things converge. That is usually slowing things down, but not to the slowness of this case. Closer inspection revealed that the flow rate of the column draw was being calculated by a spreadsheet. The recycle operation that was converging the pump around loop was continuously perturbed by the changing draw rate of the column. Analysis of the spreadsheet showed that the intention of this setup was to change the draw rate such that the draw was taking all the liquid from the column. This might have been an ingenious way of using the spreadsheet and the recycle to achieve a goal if there hadn't been a more simple and faster way.
In the column you can specify the flow rate of the liquid stream leaving a tray, so in this case the most efficient solution was to de-activate the specification of the draw rate and to create an active specification of the liquid flowing down and to set the value of the specification to zero. This removes the need for the spreadsheet calculation and provides the recycle on the pump around stream with a much stabler feed. The column also becomes more robust as otherwise the calculated liquid draw rate from the column sometimes is impossible to achieve during the iterations of the recycle operation.
This change in the model along with a couple of other changes reduced the solution time of the model from more than 5 minutes to less than 30 seconds.
So, it is of course ok to be inventive, but do not forget to check out first if there isn't a built-in features that will take care of things in a more direct manner.
Sunday, 30 December 2007
The HYSYS column solver may be able to do more than you think
Posted by
Kwizim
at
14:39
Labels: Column, Steady State
Subscribe to:
Post Comments (Atom)
Contents By Category
Dynamic
- Did You Know Series: Manipulating flows in dynamics
- Dynamic Modelling of Flare Networks
- The HYSYS Process Data Tables - The Ugly Duckling
- Modelling Valves in HYSYS Steady State and Dynamics
- Modelling process piping in HYSYS Dynamics
- Use the HYSYS Carry-over capability to unclutter your PFD
- How NOT to specify equipment elevation in dynamic modelling
Thermodynamics
- Did you know series: When using Lee-Kesler Enthalpies with the PR EOS, you get better enthalpies AND ASME steam enthalpies for water as a bonus
- Using Electrolytes in HYSYS
- Using Electrolytes in HYSYS - Some things are good to know
- Modelling TEG Dehydration accurately
- Dealing with thermodynamics: the difference between petroleum industry and chemical industry
- Matching the oil density from PVT analysis with HYSYS results
Utility
Steady State
- Did you know series: You can set a component ratio using the balance block
- The HYSYS Process Data Tables - The Ugly Duckling
- Modelling Valves in HYSYS Steady State and Dynamics
- What does "Compressible Flow" mean? Isn't any gas compressible?
- Using Electrolytes in HYSYS - Some things are good to know
- My column initially converges, but with a different feed it fails to converge
- Causes of inconsistencies can be elusive
- Make sure you have an understanding of the complete process when modelling
- Use the HYSYS Carry-over capability to unclutter your PFD
- Don't blame the simulator, at least not right away
- Matching the oil density from PVT analysis with HYSYS results
- Depressuring studies and the HYSYS depressuring utility
- The HYSYS column solver may be able to do more than you think
No comments:
Post a Comment