How to improve the accuracy of component balances around distillation column???

[Zach]

Hi there,

I have been attempting to model a partial distillation column with ~45 components. The goal of this column is to separate hydrogen into the vapor, primarily propane in the liquid distillate, and everything else go to bottoms.

This should be a fairly simple process, and I had thought I was making progress, because I was mainly checking the temperature profile and the molar concentrations. However, I just realized when looking at the molar flow rates that the component mass balances are typically way off! I have seen half of the propane in the inlet vanish from the outlets of the column. In the model I am attaching, there is 5-6 times as many moles of hydrogen in the outlet streams as there is in the inlet stream. (When I saw this, I knew that I had to post on here).

Is there any way to ensure a higher level of accuracy on the component balances specifically? I don't know much about "convergence tolerance", but even lowering this value does seem to necessarily make the component balances more accurate.

If anyone has advice or a flowchart that they use when optimizing a column with this many components, I would greatly appreciate it!

PS: I noticed a bug during my troubleshooting investigation. The "Vapor Product Flow Rate" input, which is active when a partial condensor is used, reads values in [mol/s] even when [lbmol/h] is selected from the dropdown menu. After I realized this, it was easy to work around. However, this did make me a bit paranoid that all of my issues with the component balance could stem from unit conversion issues... I have not changed my simulation to SI units to test this.

[Daniel Medeiros]

Hi Zach,

Thanks for sharing the file, I'll take a look at it and get back to you shortly.

Regards
Daniel

[Zach]

Hi Daniel,

Thanks, I really appreciate your time.

I have done a bit more testing on my end, and I think I realized a few errors I was making that should help the column converge more easily. #1 - I was neglecting to adjust the pressure drop when increasing the number of column stages, and #2 - I was not paying enough attention to the vapor product flow rate (and was probably setting this value too low). By paying closer attention to these values, and using the column temperature profile as a guide, I have been able to get the column to converge at lower tolerances, and I expect that this will enable more accurate material balances across all of the individual components.

That said, I am still curious to hear your thoughts on the simulation attached above. It is surprising to me that a converged solution could be so far off in a single component, hydrogen. However, I am still fairly new to running simulations with this many components.

Best,

• Zach

[Daniel Medeiros]

Hi Zach,

For your case there is Hydrogen that makes convergence more difficult, and ideally one would use Chao-Seader or Grayson-Streed model. I spent some time yesterday trying to converge it using PR or SRK but didn't manage to do it.

The convergence tolerance is the most important parameter in order to get a reasonable mass balance. A good reference value is something around 1E-4 to 1E-5. My strategy is to start with something like 1-0.1, check the results, use them as initial estimates for a second try with 0.01-0.001 and so on... but then the first results should not have any of those 1E-10 flow rates. When you see those there is a high chance that your reboiler/condenser specs are somewhat off.

Once you get a first time convergence with Wang-Henke you can try switching to Napthali-Sandholm solver and see if it works better (=faster).

Regards,
Daniel