A 30 m Stainless Run Grows 20 mm in Summer: Setting Joint Spacing

Scope declaration. This article covers sol-air temperature effects on member temperature, free expansion, two restraint models (fully fixed and elastically restrained by posts), sealant movement-capacity back-calculation and sliding joint travel. It does not cover concrete shrinkage and creep, heat of hydration, welding distortion, seismic or wind load combinations, or long-term wear and seizure of sliding surfaces. Worked examples use 304 (06Cr19Ni10) polished round tube 60 x 2 mm, posts 1100 mm high at 1200 mm centres, and Foshan climate data. For dark-coated or flat-plate sections the temperature range must be multiplied by 1.9 throughout; the polished-tube figures cannot be applied directly. Every value below can be reproduced from the equations given.

A 30 metre stainless run grows by roughly 20 mm in summer. The statement is correct, but three numbers behind it are usually guessed: the member temperature range, the joint spacing and the joint detail. This article calculates all three, and points out one common misjudgement along the way: for an ordinary balustrade, thermal stress is almost never the governing check. Movement is.

1. First fix delta T: air range is not member range

Most calculations simply take the local annual air temperature range as the member range. That is the first trap. Metalwork outdoors is heated above air temperature by the sun during the day and cooled below it at night by radiation to the sky. The engineering description is the sol-air temperature:

• alpha_s: solar absorptance. Polished or 2B stainless 0.35 (handbooks give about 0.30 for polished metal, 0.25 to 0.40 for light surfaces); dark fluorocarbon or powder coating 0.85 (0.80 to 0.95 for dark surfaces)
• I: solar irradiance, 950 W/m2 at summer noon on a horizontal surface
• h_o: exterior surface heat transfer coefficient, 18 W/(m2 K) (ASHRAE at 15 mph converts to about 23; still air about 12 to 15)
• phi: geometry factor = projected sunlit width divided by dissipating perimeter, the term most often missed
• delta T_night: night sky radiation, about 2 to 4 K below air temperature; taken as 3 K

The geometry factor is the key to this check. A round tube receives sun over a projected width D but dissipates over the full circumference pi D, so phi = 1/pi = 0.318. A flat coping or cover plate is heated on its top face while the underside barely dissipates, so phi = 1.0, which is 3.14 times the tube.

Foshan values and resulting member ranges

Air temperatures for Foshan: summer extreme maximum 37 degrees Celsius (recorded extremes about 38 to 39 degrees Celsius), winter extreme minimum 3 degrees Celsius. The annual air range is therefore 34 K, not the 40 K so often written into method statements. The member range is a different story:

The conclusion is hard: same city, same 30 metres, polished tube 43 K versus dark flat plate 82 K, a factor of 1.9. Movement is governed by surface colour and section shape, not by climate. Applying a 40 degree rule of thumb to a dark coated coping understates the movement by roughly half.

Reconciliation with the related article on this site

The 13 October 2026 companion article on facade grating uses the same heat balance with slightly different inputs: air 35 degrees Celsius, west-facing afternoon irradiance I about 700 W/m2, h = 11.4 + 5.7v (still air v = 0.5 gives h = 14.25), dark PVDF alpha_s = 0.75, giving a surface of 71.8 degrees Celsius. This article takes air 37 degrees Celsius, horizontal noon irradiance 950 W/m2, h = 18 in a light breeze and dark coating alpha_s = 0.85, giving 81.9 degrees Celsius.

The gap of about 10 K is fully explained: different irradiance and heat transfer assumptions, plus the geometry factor introduced here. A grating blade behaves roughly as a flat plate with phi about 1, whereas a round tube is only 0.318, so the same finish runs far cooler on a tube. The two articles do not conflict; design should take the worst case, meaning still air, leeward recesses and cavities. The shared anchor is alpha(304) = 17.3 x 10-6/K, which is one reason this article adopts that value, since some sources quote 16.5 and would understate movement by 4.6 percent.

2. Free expansion: delta L = alpha * L * delta T

Alpha also has a units-of-record problem. GB/T 20878-2007 gives 304 (06Cr19Ni10) a coefficient of 17.2 x 10-6/K over 0 to 100 degrees C and mill datasheets for 20 to 100 degrees C typically quote 17.3 x 10-6/K, whereas Annex C of EN 1991-1-5 gives stainless steel 16 x 10-6/degree C, which is where the 16.0 to 16.5 figures come from. This article uses 17.3 x 10-6/K, and an external check of independent material tables and expansion calculators found the same 17.3 x 10-6/K value for 304. The 16.5 figure gives results 4.6 percent lower; projects drawn to the Eurocode system should use 16 x 10-6 and recalculate throughout.

Thirty metres of 304 at 43 K grows 22.3 mm, so the 20 mm in the title holds, and this is the comparatively mild polished-tube case.

3. Ranking three materials: aluminium largest, carbon steel smallest

Compared at 30 m and 43 K:

The correct ranking is 6063 aluminium 30.3, then 304 stainless 22.3, then carbon steel 15.5. Stainless grows 1.44 times as much as carbon steel and 0.74 times as much as aluminium, so it sits in the middle. The trade belief that stainless moves least is wrong; that position belongs to carbon steel. Mixed-material projects are where this bites: a 304 handrail on steel posts differs by 6.8 mm over 30 m, and that difference has to be absorbed entirely at the joints.

4. Fully restrained stress: only valid when both ends are truly fixed

For 304, E = 193 GPa. For 6063-T5, E = 70 GPa, alpha = 23.5 x 10-6/K and the design strength is 90 MPa. Note also that 205 MPa for 304 is the 0.2 percent proof strength, not a design strength already reduced by a partial factor; the two should not be used interchangeably.

One common writing error deserves correction: describing 143.6 MPa as having exceeded the design strength, or as causing rupture, is not supportable; it is 70 percent of the proof strength. Likewise at delta T = 50 K the stress is 166.9 MPa, or 81.4 percent, still below yield. What genuinely does exceed yield is a dark flat plate baking at 80 degrees Celsius: at delta T = 82 K a fully restrained 304 member reaches 273.8 MPa, 1.34 times the proof strength.

5. The real case: posts are not rigid, so the stress is a few MPa

The formula above assumes both ends are fixed to something immovable. Balustrade posts are cantilevers and are flexible. Take a post of 60 x 2 mm tube at 1100 mm:

Its elastic limit thrust at the top is F = Rp0.2 * W / h = 205 * 5114 / 1100 = 953 N, reached at a top deflection of 14.3 mm. Treating the line of posts as a distributed spring against a rail axial stiffness EA = 70.3 MN gives a characteristic length of:

Because a segment is far shorter than that characteristic length, the springs are soft in relative terms and the peak stress at the anchorage is:

For a 15 metre segment the thermal stress in the rail is only 3.1 MPa, one forty-sixth of the fully restrained 143.6 MPa. The mechanics are self-consistent: the axial force at the anchorage is 1142.5 N and the sum of all post spring reactions matches it exactly, a relative difference of 0.000 percent.

The same conclusion comes from a simpler route. Locking the whole segment up would need 143.6 * 364.4 = 52.3 kN of axial force, while one post can contribute at most 953 N, so twelve posts over 15 m have a ceiling of 11.4 kN, and reaching even that would require every post to lean 14.3 mm when the actual free movement is only 5.6 mm. The posts cannot lock the rail.

So what actually fails on site?

If the rail cannot pull itself apart, something else gives. Where the rail is welded or bolted tight to the posts, the posts are dragged along and the end post moves furthest:

Every stress stays within capacity, but the movement is visible: a 30 m segment leans its end post 11.2 mm, or 0.58 degrees, which the eye picks up immediately. The site complaints about opened end joints, mismatched splices and cracked abutments all trace back to movement with nowhere to go, not to the tube running out of strength.

6. Why a sealed joint is not an expansion joint

The most common mistake is to leave a 6 to 10 mm gap, fill it with weatherproof sealant and call it a movement joint. Back-calculating against the movement capability classes of GB/T 22083 shows why this fails. Total relative movement at a joint is M = alpha * s * delta T; measured from the mid-temperature installation point it is M/2 each way, so the required width is:

Even with the tightest 12 m spacing and the best class 25 sealant, the joint has to be 17.9 mm wide. On a dark flat plate at delta T = 82 K the same case needs 34.0 mm, beyond class 25 entirely. No handrail or coping is built with joints that wide. Sealant keeps water out; it does not accommodate movement. Movement must be taken by a joint that can actually slide.

7. Sliding joints: slotted hole length is calculated, not guessed

Every sliding joint has the same one-way travel requirement:

One requirement rarely drawn: set the initial offset for the installation temperature. The table assumes the bolt sits at the centre of the slot, which means installing at mid-range temperature. Installing on a 35 degree day needs a pre-offset towards the contraction side of alpha * s * (T_install - T_mid) / 2, otherwise the slot runs out of travel the following winter. Skipping this is the direct cause of railings installed in summer splitting at the splices the next cold season.

8. The 15 to 20 metre rule of thumb is too generous

No mandatory spacing exists for metal balustrades, so 15 to 20 metres is a rule of thumb. The closest hard number comes from Table 8.1.1 of GB 50010-2010, the concrete structures code:

More important is note 4 to that table: local movement joint spacing for exposed structures such as cast-in-situ canopies and overhangs should not exceed 12 m. That is the code's explicit figure for exposed, directly sunlit elements, and it is the closest match to the case here. Concrete provisions do not transfer directly to metalwork, but for an exposed sunlit member controlling movement joints at 12 m is far more defensible than 15 to 20 m, and it costs almost nothing: at 12 m the slot is only 8 x 20.

How many joints does 30 m need?

An easy place to slip: the number of internal joints is the number of segments minus one, not total length divided by spacing.

• At 12 m target spacing: 30 / 12 = 2.5, round up to 3 segments of 10 m, so 2 internal joints
• At 15 m target spacing: 30 / 15 = 2 segments of 15 m, so 1 internal joint
• At 20 m target spacing: 30 / 20 = 1.5, round up to 2 segments of 15 m, so 1 internal joint

Do not forget the two ends. Wherever the run meets a wall, column or adjoining building, that side also needs a clear gap sized for the half-span movement: with 10 m segments the end travels about 3.7 mm.

Common mistakes and their costs

1. Mistake: Using the air temperature range as the member range. Cost: Foshan air range is only 34 K while a dark flat plate sees 82 K, an underestimate of 2.4 times. Correct approach: Use the sol-air relation with separate absorptance and geometry factors, 0.318 for tube and 1.0 for flat plate.
2. Mistake: Assuming stainless moves least. Cost: Over 30 m, 304 moves 22.3 mm against 15.5 mm for carbon steel, leaving 6.8 mm of differential with nowhere to go. Correct approach: Rank them aluminium 30.3, 304 22.3, carbon steel 15.5.
3. Mistake: Judging a balustrade with sigma = E * alpha * delta T. Cost: 143.6 MPa gets reported as an overstress when the real figure under post restraint is 3.1 MPa, 46 times lower. Correct approach: Reserve that equation for ends genuinely fixed to rigid structure; otherwise design for movement.
4. Mistake: Treating a sealant-filled gap as a movement joint. Cost: A 12 m segment needs a 17.9 mm joint even with class 25 sealant, which cannot be built. Correct approach: Sealant seals; slotted holes absorb movement.
5. Mistake: Centring the bolt in the slot on a summer installation. Cost: Not enough one-way travel left, and the splice pulls open in winter. Correct approach: Pre-offset by alpha * s * (T_install - T_mid) / 2.

In one line: fix the member temperature range first (dark flat plate 82 K against polished tube 43 K, a factor of 1.9), then the growth (22.3 mm over 30 m), then stop checking stress (it is only 3.1 MPa) and start detailing movement: 12 m segments with 8 x 20 slotted holes, plus an installation-temperature offset.

FAQ

How much does 30 metres of 304 stainless grow in Foshan?

For a polished round tube with a member range of 43 K, delta L = 17.3e-6 * 30000 * 43 = 22.3 mm. For a dark-coated flat plate the range is 82 K and the growth reaches 42.6 mm, nearly double.

Does fully restrained 304 exceed yield under thermal loading?

sigma = E * alpha * delta T = 193000 * 17.3e-6 * 43 = 143.6 MPa, which is 70 percent of the 205 MPa 0.2 percent proof strength, so no. At the dark flat plate range of 82 K the figure is 273.8 MPa, 134 percent of proof strength, and the restraint must be released.

What joint spacing should an outdoor stainless balustrade use?

Control it at 12 m, following note 4 to Table 8.1.1 of GB 50010-2010, which limits local joint spacing for exposed structures to no more than 12 m. Pair it with M8 slots at 8 x 20 for polished tube or 8 x 30 for dark flat plate, and pre-offset for the installation temperature.