Every line is dimensioned cold and works hot

A piping drawing carries a length, a schedule and a grade, all true at 20 degrees Celsius. At operating temperature only the length changes, and it changes more than most buyers expect. Thirty metres of 304 pipe heated from 20 to 120 degrees grows 51.6 mm. That is a layout problem, not a tolerance one, and if the layout has no answer the pipe finds one: a bellows crushes, a weld cracks at the toe, or a pump flange is dragged off its base.

The coefficient is published, and the spread is wider than buyers assume

The 304 data sheet gives a mean coefficient of 17.2 × 10^-6 per kelvin, a modulus of 193 GPa and a conductivity of 16.2 W per metre kelvin. The austenitic family barely moves from there: 317LMN sheet is quoted at 17.5 over 0 to 100 C, and 316 and 316L sit in the same band. 310S is the useful exception, 14.4 at 0 to 100 C rising to 17.5 at 649 C, and its sheet claims resistance to warping. 330 is tabulated at 14.4, and the 904L page says plainly that its coefficient is extensively higher than non-alloyed steel, worth remembering before 904L is chosen for a long hot run.

Duplex changes the picture. The 2205 plate page publishes 7.5 to 8.1 × 10^-6 per degree Fahrenheit across its elevated rows, roughly 13.5 to 14.6 per kelvin against 17.2 for 304. 2507 lists a low expansion rate among its selling points. 825 is quoted at 7.8 × 10^-6 per degree Fahrenheit, about 14 per kelvin, and 625 runs from 13.1 at room temperature to 15.8 at 871 C. The low-alloy steels sit lowest: P91 is tabulated at 11.3 × 10^-6 per kelvin at 200 C, two thirds of the austenitic figure.

Run the number before you run the layout

Movement equals coefficient times length times temperature change. On the same 30 metre run and 100 kelvin rise, 304 gives 51.6 mm, 317LMN 52.5, 2205 about 42, P91 33.9. Loop leg length scales with the square root of that figure, so the duplex line's smaller movement shortens the loop by about a tenth rather than a third: a change of alloy is not a substitute for a layout. Spool lengths compound it: API 5L line pipe ships in fixed 5.8, 6, 11.8 and 12 metre lengths, so a 12 metre spool of 304 grows 20.6 mm on its own, and a duplex pipe at 12,000 mm runs the same arithmetic with a smaller coefficient.

Restraint has a price, and the pipe is not the one paying it

If both ends of that run are anchored the steel cannot grow, so it is loaded instead: 193,000 MPa times 17.2 × 10^-6 times 100 gives about 332 MPa in 304, well past the yield of annealed austenitic sheet. It does not politely hold that load: the line buckles, moves the anchor, or pulls the nozzle off the vessel.

Duplex behaves differently. At 190 GPa and 14 × 10^-6, the same 100 kelvin rise produces about 266 MPa, and 2205 pipe is listed at a 550 MPa yield, so the material stays elastic and the load moves into the anchor instead. A DN200 schedule 40 pipe carries near 5,400 square millimetres of metal, and 266 MPa across it is roughly 1.4 meganewtons, some 145 tonnes of thrust on a structure detailed for wind and snow. A loop is cheaper than an anchor every time. P91 behaves like the austenitic case rather than the duplex one: 218 GPa times 11.3 × 10^-6 times 100 is 246 MPa against a 205 MPa yield. The P22 header below it is the same 205 MPa material with a higher coefficient, so it does not get an easier answer.

What actually absorbs the movement

Three devices do the work and they are not interchangeable. The first is geometry: an offset or loop built from the same pipe and sized as a guided cantilever. The useful property of that formula is that the leg grows with the square root of the movement, so quadrupling the expansion only doubles the leg, and a lower coefficient buys a modest shortening rather than an escape from the layout.

The second is a bellows or an expansion joint. 625 is listed for bellows and expansion joints by name, from cryogenic to 2000 F, because a thin convolution in a 17.2 coefficient austenitic is a fatigue item rather than a pressure item. The third is cold pull, where part of the calculated movement is built in against the operating direction at erection; half is the usual figure in general practice, trading erection tolerance for a lower stress range.

Anchors and guides are drawn last and matter first. An anchor defines where the movement goes; a guide keeps the pipe on the line so the loop bends instead of bowing sideways. General practice puts the first guide close to the anchor and spaces the rest by pipe size, and each becomes a structural attachment on the rack. The welded H-section and the steel structure under it expand too, at about 12 × 10^-6 per kelvin, in the same direction at the same time.

Where the catalogue changes the answer

Long lengths remove field welds and concentrate movement. A106 is tested length by length and shipped in random lengths, while the alloy headers reach 16,000 mm and the P91 page states that maximum explicitly, one reason a hot header has fewer joints to absorb growth than a carbon utility line. Temperature ceilings then decide whether the alloy is available at all: 2507 is limited to service below 316 C, and the 2205 window of minus 50 to 600 F comes with conditions on welded structures. 347H and 321 exist for the top of that range, stabilised against sensitisation that comes from holding at temperature rather than from cycling. S32205 pipe and 310S pipe cover the two ends of the same argument, one for chloride duty with a lower coefficient and one for heat with a 2000 F ceiling.

The tank at the end of the run

The first nozzle is where a calculation becomes a mechanical problem. A tank shell in 316L sheet is stiff and does not want to be pushed, so the run has to arrive with an offset, a guide and a support close enough that the nozzle sees a bending moment rather than shear. The sequence matters more than the detail: nozzle, first guide, anchor, then the loop. Get it wrong and the tank becomes the expansion joint.

What to send with the enquiry

Five items decide the layout: design and operating temperature, the anchor points, the allowable nozzle loads, whether the line is insulated, and the order of hydrotest against heat-up. A line that passes a cold hydrotest can still fail on the first hot cycle, because the test proved the welds and said nothing about the movement. Add the steam-out temperature if the line sees one, since a 140 C steam-out is a second thermal case and often the governing one.