NGL debutanizer with a bottoms recycle

The flowsheet a real CAPE-OPEN interchange deck describes, built on the three specifications that deck states and this product can now solve. An NGL cut (16% propane / 13% isobutane / 26% n-butane / 45% n-pentane, 40 mol/s at 15.5 bar) mixes with a bottoms recycle, is preheated to exactly 10 mol% vapour, and is fractionated in a 22-stage column with a 1.40 -> 1.45 MPa pressure profile; 15% of the bottoms returns to the feed and the loop is torn and converged.

The preheater is specified by the VAPOUR FRACTION it must leave behind (outlet_vapor_fraction = 0.10), not by a duty or an outlet temperature — which is how a column preheater is always specified, because the feed-stage flash is what the number is for. The solver roots the duty for it: 0.404 MW, landing the feed at 373.6 K. The column carries a real PRESSURE PROFILE rather than one pressure: trays resist vapour flow, so the condenser sits at 14.00 bar and the reboiler at 14.50 bar and each stage boils at its own value. The cut is set the way the deck sets it, by the absolute distillate rate: 20.833 mol/s (75 kmol/h).

Result: an LPG distillate at 345.3 K carrying 30.7% propane, 23.7% isobutane, 43.6% n-butane and 2.0% n-pentane, against a 411.9 K bottoms of 91.8% n-pentane with 6.9% n-butane. Reboiler 1.021 MW, condenser 1.003 MW. The recycle settles at 3.382 mol/s — a fixed point of the loop's own balance, specified nowhere.

The finding worth carrying, because it is a trap this deck sets. The only composition the deck states anywhere sits on its recycle stream, and that stream is 96% n-pentane — only 11.6 mol% propane + butanes. Read it as a guide to the FEED and the deck's own 75 kmol/h distillate rate stops being achievable: a clean overhead would need a feed of at least 179 mol/s, so at the 43.4 mol/s this flowsheet runs, 76% of the distillate would have to be n-pentane by MATERIAL BALANCE alone — whatever the reflux, with no solver involved. Underwood agrees from the other side, putting minimum reflux at 8.8 for a 95%/98% key recovery on that composition against 1.465 on the NGL feed shipped here. That stream is the BOTTOMS, not the feed, and on the real feed the deck's reflux of 2.0 is 1.37x minimum: a sensible design point, so the deck is self-consistent after all. The same lesson as a distillate ratio set equal to a feed's light fraction — the cut specification has to fit inside what the feed carries.

Bounds, stated rather than implied. The fresh feed's rate and composition are ours — the deck states neither, and the recycle estimate it does state is for a scale it never gives. The feed is mixed ON-STAGE (the whole feed joins its stage's balance); the deck declares Above-Stage, RadFrac's default, which differs by about 0.02 K on a feed this lightly vaporised. The reboiler is a kettle. The recycle returns at the bottoms pressure, so the mixed feed sits at 14.50 bar and the preheater's 0.5 bar drop lands it at the column's top pressure, 0.24 bar below its own feed stage — a real plant closes that with a feed pump the deck does not show. The tear is seeded with the converged circulation, which cuts the solve from 109 s to 32 s; a rough seed converges to the same fixed point, slower.