Executing a Column Successfully

Craig Spears

May 04, 2011

It’s no secret that distillation columns tend to be more difficult to solve successfully in process simulation than most other units. There are many reasons for this, but the following are some hints to help you get a green tower:

  • Check all Inlet Streams – First and foremost check each of your column feed streams. If these are not at the expected conditions, then the column is more likely to fail.
  • Check Specifications – Confirm that the specifications do not imply an impossible separation.
  • Delete Specification Tolerances – If a tolerance is set, delete the value. In general, setting the specification tolerance is discouraged, as the program will automatically choose an optimal tolerance.
  • Try Easier Specifications – The ratio specifications automatically created by ProMax are natural variables for the solver. These specifications will be easiest for ProMax to solve, and may help obtain an initial column profile, allowing you to then enable a more difficult specification (such as component compositions, TVP, or temperatures).
  • Confirm Vapor and Liquid on All Stages – Vapor and liquid must be present on each stage of the distillation column. If the column does not have a condenser or reboiler, confirm that a feed containing some liquid or vapor, respectively, exists on the top or bottom stage.
  • Change Enthalpy Model – This option is found in the column Convergence tab; choices are Boston-Britt or Composition-Dependent. The default Boston-Britt should be used when there are no convergence problems.
    • Composition-Dependent is a unique enthalpy model and is useful for unstable or difficult-to-converge columns (e.g. ionic columns, especially amine strippers, or columns with extremely wide or narrow boiling point differences between the components).
    • The Composition-Dependent enthalpy model can be prohibitively slow for a large number of components. If the column does not converge and oscillates within the first 50 iterations, stop execution and check the validity of input data, especially specifications.
  • Change Inner Loop Model – This option is also found in the column convergence tab; choices are Boston-Sullivan or Boston-Sullivan Non-ideal. Boston-Sullivan is the default and should be used when there are no convergence problems.
    • Boston-Sullivan Non-ideal is useful for unstable or difficult-to-converge columns (e.g. highly-loaded amine absorbers, absorbers where the amine becomes the limiting reagent and strippers operating near minimum reflux).
    • The Non-ideal model should be used in combination with the Composition-Dependent enthalpy model. This model can be prohibitively slow for a large number of components.
  • Add Initial Estimates – Add initial estimates, especially for more "difficult" specifications such as component flows or draw temperatures. Reflux ratio and boil-up ratio estimates are often helpful, as are flow rate estimates for side draws.
  • Change Column Type – The polar liquid column type can aid convergence in columns which have a non-polar vapor phase in contact with a polar liquid phase (e.g. methanol-protected cold process).
  • Use K Damping – Try changing the K damping parameter in the column convergence tab.
    • Although some damping is performed internally in the column calculations, specifying a K damping parameter can be beneficial. The value for the K damping should be non-negative integers up to about 10 (using too high of a value can lead to erroneous convergence).
    • Increasing the K damping can be useful if the outer loop error oscillates or when receiving the error “For staged column X the inner loop calculation failed due to an invalid stage temperature calculation”.
  • Enable/Disable Boston-Sullivan Kb – Try using Boston-Sullivan Kb found in the convergence tab; or, if Boston-Sullivan Kb is enabled, clear the checkbox to disable. This parameter can aid convergence for very wide or very narrow boiling mixtures.
  • Increase Maximum Iterations – If the column appears to be converging, but exceeds the maximum iterations, this value can be increased in the solver grouping of the convergence tab.
  • Monitor Specifications and Variables – Monitor the specifications and variables in the tables on the specifications tab, and compare the calculated value to the specified target value. For example, if the reflux ratio is set to 1, the column fails, and the calculated value is 0.1, there is likely not enough liquid flow at the top of the column.
  • Modify Variable Estimates and Bounds – The column provides an estimate for variables if an estimate is not provided by the user. The Target values for these variables can be changed on either the specifications tab or convergence tab, variables grouping. Lower and upper bounds can also be specified.
  • Add a Recycle to the Pump-around Loop – If the column has a pump-around loop and fails to converge due to errors in the pump-around blocks or the loop itself, or if the column calculations are exceedingly slow, consider adding a recycle block to the pump-around loop. When the recycle is added to the loop, the pump-around effectively becomes a "draw". The draw rate is still specified in the column Specifications tab, but no pump-around estimates are required for exchangers or pumps in the loop.

If you have attempted these steps and a column still refuses to successfully converge, please send it in to our support engineers. They will be happy to take a look at it.