A jacketed reactor is opened after fourteen months. The shell is sound. The welds are sound. But at the base of one nozzle, where the flow changes direction and the fluid sits slightly hotter than the bulk, the wall has thinned to a point where the vessel has months rather than years left in it.
Nothing about that failure is a fabrication problem. The welding was good, the code work was correct, the hydrotest passed. What went wrong happened earlier, when somebody selected a material against a nominal process description rather than against the fluid as it actually behaves at a hot spot.
Reactors are unusually unforgiving in this respect. A storage tank holds a known fluid at a known temperature. A reactor is a machine for making one substance turn into another, which means the chemistry inside it is deliberately changing, and the worst moment for the metal is often not the normal operating case at all.
Choose the metal before you draw the geometry
On most equipment, geometry comes first and material follows. On reactors that order should be reversed, because the material decision constrains almost everything downstream: wall thickness, welding approach, whether the vessel can be heat treated, how the internals are attached, and how it can be repaired in ten years.
Three broad routes are available, and they suit genuinely different problems.
Solid alloy construction, using stainless grades, duplex, or nickel alloys such as Inconel 625, Hastelloy C276 or Alloy 20, gives you a homogeneous vessel with predictable behaviour and straightforward repair. It is the most expensive route per kilogram, and on a thick walled vessel the cost becomes hard to justify when only the inner face is exposed to anything aggressive.
Clad construction answers exactly that. A corrosion resistant layer, roll bonded or explosion bonded or applied as weld overlay, is carried on a carbon steel backing that does the structural work. You buy alloy performance on the wetted surface and carbon steel economics for the strength. The trade is that clad demands more from the fabricator: the bond has to be verified, nozzle penetrations and attachment welds need careful detailing, and overlay requires strict heat input control to keep the alloy layer chemically intact.
Glass lined construction occupies a different niche altogether. For strongly acidic service the glass is chemically superior to almost any metal, and it is why glass lined reactors remain standard across large parts of fine chemicals and pharmaceuticals. The limits are mechanical rather than chemical: glass does not tolerate thermal shock or impact, repair is specialised, and a single chip in the wrong place ends the campaign.
Failures that a datasheet does not predict
General corrosion, the slow uniform loss of metal, is the easy case. You calculate a rate, add a corrosion allowance, and the vessel is sized for it. The failures that shorten reactor life are usually the localised ones.
Chloride stress corrosion cracking attacks austenitic stainless steel under tensile stress in the presence of chlorides and heat. It produces branching cracks that can pass straight through a wall with negligible weight loss, which means an inspection based on thickness measurement can find nothing at all until it finds a leak. Duplex grades resist it far better, and this single mechanism drives a large share of the duplex specified in chemical plants.
Pitting and crevice attack concentrate at exactly the places a designer creates without thinking: under a gasket, behind a support lug, in the shadow of a poorly finished weld. Design and surface finish matter here as much as alloy choice.
Then there are the temperature driven mechanisms. Sensitisation, where chromium carbides form at grain boundaries during welding and leave the adjacent zone depleted and vulnerable, is why low carbon grades such as 316L exist and why heat input control is not a shop floor detail. Hydrogen damage in high pressure hydroprocessing service is a design case of its own.
The practical consequence is that a material recommendation is only as good as the process information behind it. Tell your manufacturer the chloride content, the pH excursions during cleaning, the maximum skin temperature at the jacket, and what happens during an upset. Those four facts change alloy selection more often than the nominal process description does.
The reactor types, and what each one demands
Reactor is a broad word. The engineering underneath it varies enormously.
Batch reactors, jacketed, agitated, glass lined, vacuum or pressure rated, dominate fine chemicals and pharmaceuticals. Their defining difficulty is heat transfer through a jacket into a stirred mass, and their defining risk is a runaway exotherm, which makes jacket design and emergency cooling a safety matter rather than an efficiency one.
Continuous reactors, whether stirred tank or plug flow and tubular, trade flexibility for consistency and are sized around residence time.
Refinery reactors are their own world: hydrocrackers, hydrotreaters, hydrodesulphurisation units and fluid catalytic cracking service, where thick walls, hydrogen partial pressure and elevated temperature combine into some of the most demanding pressure equipment built anywhere.
Catalytic reactors in fixed, moving, fluidised and trickle bed configurations are dominated by internals: support grids, distributors and the requirement that flow be genuinely even across the bed. A beautifully fabricated shell with a poor distributor is a poor reactor.
We also build the energy transition equipment that is growing fastest, covering hydrogenation, ammonia synthesis and sustainable fuel processes, alongside polymerisation reactors, pharmaceutical and biotech vessels including bioreactors and fermenters, and the specialty end with autoclaves and electrochemical and photochemical designs.
The jacket is usually the real constraint
Ask what limits the output of a batch reactor and the answer is rarely the volume of the vessel. It is the rate at which heat can be moved across the wall.
Most reactions of commercial interest release heat, and the reaction rate rises with temperature, which sets up the mechanism everybody in this field is trained to fear. If heat is generated faster than the jacket can remove it, the contents get hotter, the reaction accelerates, and it generates heat faster still. That is a runaway, and it is a safety event rather than a process upset.
Heat removal capacity therefore has to be designed against the worst credible case, not the normal one. It is set by three things: the temperature difference available between the contents and the cooling medium, the area of the jacket, and the heat transfer coefficient on both sides. Vessel geometry works against you as scale increases, because volume grows faster than surface area, which is why a reaction that behaves impeccably in a pilot vessel can be genuinely difficult in a production one.
Jacket design is where the engineering response sits. A conventional annular jacket is simple and cheap and gives poor velocity, therefore a poor coefficient. A half pipe coil welded in a spiral around the shell forces the cooling medium along a defined path at high velocity, gives a much better coefficient, and takes higher pressure, at higher fabrication cost. A dimple jacket offers a middle route with good coverage and low medium volume, which matters when the response time of the cooling system is important.
Internal coils add area directly, at the cost of complicating cleaning and creating surfaces that foul. Agitation is the other half of the answer, since the coefficient on the process side depends on how vigorously the contents are moved past the wall, which is the second reason mixing belongs in the design conversation early rather than late.
The practical instruction for a buyer is to state the heat of reaction and the worst case addition rate in the enquiry, not just the operating temperature. Without them a designer is sizing a jacket against a steady state that does not describe the risk.
The envelope, and the agitator nobody sizes early enough
Our reactor work runs to internal diameters beyond 8,000 mm and heights beyond 60 m, from full vacuum to 700 bar and above, across minus 196 to 1,100 degrees Celsius, with single pieces up to 1,000 MT and agitator drives beyond 2,500 kW.
That last figure deserves more attention than it usually gets. On an agitated reactor the mixing system is not an accessory bolted on at the end. The agitator decides heat transfer at the wall, it decides whether solids stay suspended, and it imposes real mechanical loads on the vessel through the shaft, the seal and the mounting. Bracket design, shaft deflection and critical speed have to be settled while the vessel is still being engineered. Reactors that vibrate in service almost always trace back to a mixing system that was specified after the shell.
Verifying the alloy actually arrived
Alloy substitution is not usually malicious. It happens because a plate got mixed in a yard, or a fitting came from a stock bin, or a repair was made with the consumable that was open. The consequence is the same either way, and on a reactor it is severe.
Positive material identification is the answer, and on alloy reactor work it should be applied to parent plate, to fittings, to weld deposits and to internals, not sampled once and signed off. Alongside it, ferrite measurement on duplex confirms the phase balance survived welding, and hardness testing confirms the heat affected zone is within limits. Where clad is used, the bond itself is examined ultrasonically.
It is a modest amount of testing against the cost of discovering in year two that one nozzle forging was the wrong grade.
Start with the chemistry
If there is a single habit that separates reactor projects that age well from those that do not, it is that the good ones settle metallurgy first and treat the drawing as a consequence.
It feels backwards to procurement, because the material decision is the expensive one and there is a natural instinct to defer it until the design is firm. But the design is not really firm until the metal is chosen, and every week that decision is postponed is a week of engineering done against an assumption. Bring the fluid, the upsets and the cleaning regime to the table on the first day. The vessel that results will be cheaper to build and considerably cheaper to own.
