Page 273 - Separation process engineering

Use total condensers and reflux should be returned as a saturated liquid. Both columns have partial
reboilers. Feeds are saturated liquids.
Use general metric units. You can choose to either simulate the two connected columns or simulate

them one at a time. One advantage of doing one at a time is you produce a lot less results to wade
through. Since the bottoms product from the first column keeps changing, the second column does not
have the correct feed until you select the final design for the first column. Once you have the first
column finished, rerunning it every time as you optimize the second column gives no additional
information. But, you may prefer to work at simultaneously optimizing both columns. Thus, the
procedure is up to you; however, the instructions are written assuming you will do one column at a
time.

Simulate the first column. Find the minimum L/D, and the actual L/D. Then find the optimum feed stage
and the number of stages that just gives the desired separation. You may choose to do this first with
DSTWU (a short cut program similar to the Fenske-Underwood-Gilliland approach). Set N to some
arbitrary number (e.g., 40) to find the estimate of min L/D. Multiply this minimum by 1.15 and use this
L/D as the input to find estimates for N and the feed location. These numbers are then used as first
guesses for RADFRAC. With RADFRAC, first find the actual (L/D) min (set N = 100 or some other
large number, feed stage at ½ this value, and vary L/D until obtain desired separation). Then do
RADFRAC at 1.15 times (L/D) min . Even though the DSTWU estimate for (L/D) min is not accurate, the

estimates for N and feed location at L/D = 1.15 (L/D) min may be accurate. Alternatively, you can
decide to bypass the use of DSTWU and either do the Fenske-Underwood-Gilliland approach by hand
or start RADFRAC totally with guesses. The best specifications for RADFRAC are to specify the
values of L/D and D. Note: You must finish (L/D) min calculation with RADFRAC.

Find the optimum feed location for column 1 by trial and error (see the Note at the end of Lab 5).

For column 2, use the bottoms from the first column as the feed to the second column. Repeat the
procedure using DSTWU, a hand calculation with the Fenske-Underwood-Gilliland approach, a
McCabe-Thiele diagram, or guesses to find estimated values of: (L/D) min , L/D, optimum feed stage,
and total number of stages. Then do exact RADFRAC calculations to find accurate values for (L/D) min ,

L/D, optimum feed stage and N that just give desired recoveries.
Once the optimum columns have been designed, do one more RADFRAC run to determine the column
diameters at 80% of flood with a tray spacing of 2.5 feet (see section III in Lab 5). Use Fair’s method
to calculate flooding.

If you haven’t already done so, connect the columns and do a run with everything connected. This run
checks to make sure you did not inadvertently change conditions when going from the bottoms of the
first column to the feed of the second.

Report the number of stages and the optimum feed stages in each column, the reflux ratios, the
temperatures at the top and bottom of each column (in K), the heat duties in the reboilers and
condensers (kJ/s), the compositions (in mole fractions) and flow rates of the three products and of the
interconnecting stream (kmol/h), the column diameters, and any other information you consider to be
relevant. If you use DSTWU or the Fenske-Underwood-Gilliland approach or a McCabe-Thiele
diagram for the second column, compare these results with RADFRAC.