1-805-484-2992
Quotes - Engineering - SalesDistillation Condensers
Total and Partial Condensing on Distillation and Fractionation Columns
On a distillation column the condenser is part of the separation, not a utility bolted to the end of it. Total or partial, at pressure or under vacuum, we size it against the real condensing curve and the real turndown range.
A distillation column separates because vapor rising through it contacts liquid running down it. That descending liquid is reflux, and reflux comes from the condenser. Change the condensing duty and you change the reflux ratio, which changes the separation, which changes product purity. This is why a condenser on a column cannot be specified in isolation from the column's operating envelope.
Total condensers. The entire overhead vapor is condensed, usually into a reflux drum, from which part returns to the column as reflux and part leaves as distillate. The drum is the reason this is the most controllable arrangement: it holds inventory, so reflux flow can be measured and manipulated directly, and the stored liquid absorbs a swing in cooling water, boil-up or inert load before the trays ever see it.
The liquid running down a distillation column that the rising vapor contacts, and it comes from the condenser. Change the condensing duty and you change the reflux ratio, which changes the separation, which changes product purity.
Partial condensers. Only part of the overhead is condensed. Liquid returns as reflux and an uncondensed vapor product leaves the top. This effectively buys an additional separation stage and suits systems with a light component you want to take as vapor. The cost is that the condenser is now coupled directly into the column's mass balance: its duty sets both reflux and vapor product rate simultaneously, so the two cannot be adjusted independently in the way a total condenser and drum allow.
The condensing curve is the whole game. A binary mixture with close-boiling components behaves almost isothermally. A wide-boiling multicomponent overhead does not: it begins condensing at its dewpoint and finishes at its bubble point, and between those the local temperature difference and the local coefficient both change substantially. Sizing on a mean temperature difference derived from terminal temperatures will mislead you, usually toward an undersized unit. Rating the curve in increments is the only reliable approach, and it is standard practice for us on any multicomponent overhead.
Inerts and light ends. Even a modest noncondensable fraction accumulates in the cold end of a condenser and blankets surface. In a column context this is worse than a simple loss of duty, because the resulting change in condensing rate propagates into reflux and shows up as a product specification excursion. A properly located vent and a design that accounts for the gas load prevents an intermittent, hard-to-diagnose quality problem.
Vacuum columns. Where the column runs under vacuum to keep bottoms temperature below a thermal degradation limit, allowable overhead pressure drop becomes the binding design constraint. Pressure lost in the condenser raises column pressure, which raises bottoms temperature, which is precisely what the vacuum was for.
Turndown. Columns run at reduced rates for all sorts of reasons. A condenser generously sized for peak duty can behave awkwardly at low load, where excess surface, condensate drainage and gas accumulation interact and reflux control becomes lively. We rate across the range rather than at the design point alone.
Related pages: overhead condensers, reflux condensers, dephlegmators, and fractionation condensers.
Questions from the column side
Pick a total condenser unless there is a light component you specifically want to take out as vapor, and then the partial condenser earns its place by doing part of the separation. The classic case is a column with a small amount of a very light component, say methane or nitrogen in an LPG stream, that would otherwise accumulate in the drum and have to be vented anyway; a partial condenser takes it off as a vapor product and is credited with roughly one theoretical stage in the column design.
The price is coupling: the condenser duty now fixes both the reflux rate and the vapor product rate, and the operator cannot move one without the other. A total condenser with a drum holding 5 to 10 minutes of reflux flow gives the operator both handles independently and a buffer against upsets, which is why it is the default on any column with a tight purity target.
Because a wide-boiling overhead delivers most of its heat at a temperature very different from the arithmetic the terminal temperatures imply, and the LMTD has no way of knowing where along the curve the duty sits. Take an overhead with a dewpoint of 200 F and a bubble point of 120 F against cooling water running 85 F to 105 F: the terminal LMTD works out to about 60 F.
If the light key makes up most of the stream, the bulk of the latent heat is released below 150 F, where the local temperature difference has already dropped to 45 F or less, and the true weighted mean is closer to 45 F then to 60 F. That is a 30 percent shortfall in area before fouling is considered. Rating the curve in ten or so increments catches it; a single-point calculation never will, and the column pays in reflux it cannot make.
The problem usually arrives on a schedule, which is the clue: purity drifts off during the hour after a feed tank switch, or every time the vacuum column's air in-leakage climbs, and comes back when someone opens the vent. Inerts enter dissolved in the feed, with nitrogen from a blanketed feed tank, or as air on a vacuum tower, and each carries a different signature.
On a pressure column the drum pressure controller will slowly open the vent as gas accumulates, which looks like a leak somewhere. On a vacuum column the overhead pressure rises instead, the bottoms temperature follows it and the heavy product degrades. In both cases the condensing rate has changed without the duty changing, so the reflux the column receives is not the reflux the operator thinks has been set. A continuous vent sized for the real inert rate, with a small purge of product vapor, is cheaper than the excursions.
It decides which pressure control scheme is available to you, and pressure control is the loop the whole column hangs on. With a total condenser and a drum you have three workable options: throttle the coolant, which is slow and fouls the water side; a hot vapor bypass around the condenser, which is fast and common; or a flooded condenser where a valve in the condensate line backs liquid up over the tubes to vary the active surface. The flooded scheme in particular needs the condenser mounted above the drum with a few feet of elevation to work with.
A partial condenser gives you only the vent and the coolant, and a dephlegmator only the coolant, so the pressure loop has a single, slower handle. Deciding this at the data sheet stage, rather than after the exchanger has been bought with the wrong nozzle elevation, is where most of the trouble is avoided.
Keeping the vapor-side pressure drop inside a budget that is usually 2 to 5 torr for the whole overhead system, and designing the condenser around that figure before anything else is considered. A worked case: a column holding 50 torr at the top to keep a heat-sensitive bottoms product below 300 F has a bottoms pressure of perhaps 60 torr once the tray drop is counted. A condenser and overhead line that lose 10 torr push the top to 60 torr and the bottom to 70, and the bottoms boiling temperature rises by something like 15 F, which is the whole protection margin gone.
The hardware answer is shell-side condensation in a horizontal unit with oversized nozzles and a short overhead line, or a condenser mounted directly at the column top so there is no line at all. Trading a slightly higher coolant flow for a larger shell is almost always the right trade here.
As far down as the column is expected to run, which for most plants means a reliable 40 to 50 percent of design and sometimes lower during startup and product changeovers. The number should come from operations, not from the design basis, because the condenser behaves differently at the low end than at the high. Below about half load the excess surface subcools the condensate, condensate drainage slows, inerts settle into the idle part of the bundle and the pressure control loop starts to cycle.
A condenser that holds pressure smoothly at 50 percent and acceptably at 30 percent has to be rated at those points, with the vent and the control scheme checked at each. If the required range runs below about 30 percent, two shells in parallel with one taken out of service at low rates often gives better control than one large unit trying to cover the whole span.
