Before a single plate is cut for a tall column, somebody should drive the route. Not look at it on a map. Drive it, and note every bridge, every overhead line, every roundabout with a kerb radius that a 60 m load cannot negotiate, and the one level crossing where the police will only allow a movement between one and four in the morning.
That sounds like a logistics task. It is actually a design input, and on large columns it is frequently the design input that determines what the equipment looks like. A column is one of the few pieces of process equipment where getting it to site is a harder engineering problem than making it work.
Two ways to build something 80 m tall
The first is to build it complete, ship it as one piece, and lift it once. Everything happens under shop conditions. Long seams are welded with the shell rotated on rollers, which is faster and produces better welds than any position welding at site. Internals can be installed and checked in the horizontal, with good access. Inspection and testing are done where the instruments and the people already are. The column arrives finished and stands up in a single, carefully rehearsed lift.
The second is to build it in sections, ship those, and complete the assembly at site. This is what you do when the single piece cannot be moved or cannot be lifted. It is entirely workable and it is done constantly. It also moves a set of critical welds from a controlled shop into the open air, adds field radiography at height, and introduces alignment work that has to hold verticality over a very long object.
The choice between them is rarely about capability. It is about the road, the crane and the site. Which is why it should be settled in the first fortnight, not discovered in month five.
The crane is the real specification
Lift planning is treated as a construction activity, and it is, but it constrains the fabricator long before the construction contractor is appointed.
A crane's capacity is not a single number. It falls away with radius and it falls away with boom length, so the honest question is never "can it lift 900 tonnes" but "what can it lift at the radius the plot actually permits, with that boom configuration, on ground with that bearing capacity". Change any one of those and the answer changes.
The consequences run straight back into design. Lifting lugs and trunnions are structural items that have to be engineered into the shell, sized for the dynamic loads of the lift rather than the static weight, and positioned so that the column is stable through the rotation from horizontal to vertical. Tailing arrangements at the base need the same attention. And the shell itself has to survive being supported at two points while horizontal, which is a load case that has nothing to do with its operating duty and which occasionally governs the design of an otherwise thin walled section.
The equipment we build reaches diameters beyond 10,000 mm and heights past 100 m, with shell thickness from 4 mm up beyond 300 mm, pressures from full vacuum to 500 bar and above, temperatures from minus 196 to 1,000 degrees Celsius, and single lifts beyond 1,000 MT. Those upper figures are not there to impress. They exist because a column that has to be delivered as one piece is only possible if the shop, the transport and the lift have all been planned around the same number.
Internals are the column
The shell is a pressure envelope. The separation happens on the inside, and the internals are where a column is won or lost.
Trays and packing solve the same problem differently. Trays, whether sieve, valve or bubble cap, give you discrete contacting stages, tolerate fouling and solids better, and behave predictably across a wide range of loads. Structured and random packing offers a much lower pressure drop per stage, which is exactly what a vacuum column needs, and a higher efficiency per unit height, at the cost of being fussier about liquid distribution and less tolerant of dirt.
The component that causes the most trouble in service is the liquid distributor. Packing only performs if liquid is spread genuinely evenly across the cross section, and on a large diameter column that is a demanding piece of fabrication. A distributor that is slightly out of level, or has a few blocked orifices, produces maldistribution, and maldistribution quietly destroys separation efficiency in a way that is very hard to diagnose from outside the vessel. On a wide column, the distributor deserves as much fabrication attention as any pressure boundary.
Two practical matters follow. Tray support rings and downcomer bolting bars have to be positioned accurately relative to seams that will be welded later, so dimensional control during shell fabrication is directly a process performance issue. And manway placement has to allow a person to physically reach the trays for installation and future maintenance, which sounds obvious and is regularly compromised to suit a platform layout.
What happens at a site joint
When a column ships in sections, the circumferential welds joining them are made at site, and they deserve planning rather than assumption.
Fit up is the first difficulty. Two large shell sections have to be brought into alignment within tolerance, which means rolling tolerance and out of roundness from the shop are being reconciled outdoors, on a laydown area, with cranes holding pieces steady. The remedy is dimensional control in the shop, including trial assembly of adjacent sections before dispatch, so that any mismatch is discovered where it can still be corrected cheaply.
The weld itself is a full penetration joint in a pressure boundary, made in position rather than rotated, requiring qualified site welders, weather protection and radiography at whatever height the joint sits. Where material or thickness demands post weld heat treatment, it has to be done locally with electrical elements and insulation, which is slower and harder to control than a furnace cycle.
Verticality then has to be held over the full height once the sections are stacked, because a column out of plumb loads its skirt and anchor bolts in a way the design did not anticipate, and can disturb internal liquid distribution.
Tall equipment carries wind and earthquake
A short vessel is designed for pressure. A tall column is designed for pressure and then checked, often more onerously, against wind and seismic loading.
A slender structure a hundred metres tall is a cantilever, and the loads it sees in a storm can govern shell thickness in the lower courses and dominate the design of the skirt and anchor bolts entirely. Vortex shedding is a genuine consideration on slender columns and is one of the reasons helical strakes appear on some stacks and towers. The base is where those loads are resolved, and skirt design, base ring detail and anchor bolt arrangement carry more engineering than their appearance suggests.
This is also the reason platform and ladder layout should be discussed early. Attachments welded to a pressure shell are local stress raisers, and clips added late, after the stress work is complete, are a familiar source of rework.
A column is not one material
It is tempting to specify a metallurgy for a column the way you would for a vessel: one grade, applied throughout. On a tall column that is frequently both over specified and under specified at the same time, because conditions inside a column change substantially from the bottom to the top.
The base sits in hot, heavier material at the highest temperature in the equipment. Halfway up, conditions are milder. The overhead is where vapour condenses, and condensation is where corrosion often becomes severe, because the first liquid to form can be far more aggressive than the bulk process. In crude and many refinery services this is the classic problem: chlorides carried overhead hydrolyse and form acid at the point where water first condenses, attacking a zone perhaps a few metres tall while the rest of the shell is essentially untouched.
The engineering answer is to treat the column as a set of zones. Carbon steel where carbon steel is adequate. A corrosion resistant alloy, or a clad section, only in the band where the mechanism is active. Alloy trays or packing in the sections that see the aggressive condition, with cheaper material elsewhere. Sometimes a corrosion allowance in one zone and a lining in another.
Done well, this is one of the largest cost levers on a big column, because alloy in a 300 mm thick shell is expensive and applying it only where it earns its place can change the total materially. Done badly, by defaulting to a single grade for simplicity, you either overspend across the whole height or you install a shell that will need a section replaced in eight years.
The information that makes zoning possible is a corrosion profile along the column: temperature, composition and phase at intervals up the height, including what happens at the dew point. That is process engineering rather than mechanical engineering, and it needs to be supplied to the fabricator rather than assumed by them.
Where a column runs at very low temperature, as in air separation and liquefied natural gas service, the same logic applies in reverse and the whole envelope moves to low temperature materials with impact tested properties throughout.
The range of towers we build
Distillation and fractionation columns are the largest population, but the family is wider. Absorption and stripping columns handle mass transfer duties where separation is achieved against a solvent or a gas rather than by boiling. Reaction columns combine reaction and separation in a single shell. The specialty end covers quench and wash towers, fluid catalytic cracking main fractionators and sulphur recovery service, where the process conditions are severe and the metallurgy earns close attention. Cryogenic columns for air separation and liquefied natural gas sit at the opposite temperature extreme and demand low temperature materials throughout.
Plan backwards from the crane
The columns that go up without drama are the ones where the sequence was worked out in reverse. Start with the lift that has to happen on a particular week. Work back to the transport that gets the piece to the laydown area. Work back from that to the sections the shop will produce, and only then decide how the equipment is designed and split.
Done in that order, a large column is a demanding but predictable project. Done in the usual order, with process design first and transport considered when the drawings are nearly complete, it becomes a series of expensive compromises made under time pressure. The steel is the straightforward part. Everything around it is the project.
