Pipe Rack Load Path — and Why Friction Governs

The bent that wind did not size

A two-tier rack in a refinery unit. Twenty lines, the biggest a 24″ hot header. The structural engineer has the wind speed, the seismic coefficient, the pipe weights and the plot layout. The longitudinal bracing in the bay next to the pumps keeps coming back overstressed — by about 50 %. Nobody has touched the wind data.

Guess first: which load is failing that brace?

Wind? It is the load everyone names. Seismic? It is the load everyone fears. The answer is neither. It is friction — dozens of pipes sliding a few millimetres across their shoes as the unit warms up, each one dragging a small force into the steel, all of them dragging it the same way at the same time.

The misconception, stated plainly: "Friction is a small secondary effect — we add a bit of longitudinal load and move on."

Let it stand for a second, because it feels right. μ is a number smaller than one. Each pipe weighs a few hundred kilos per metre. How bad can it be?

It is bad because friction is proportional to weight, and a pipe rack is the heaviest thing on the plot per square metre of structure. Wind is proportional to projected area, and a bundle of round pipes shields itself so effectively that its wind area is a fraction of its silhouette. One load scales with mass, the other with air. On a loaded rack, mass wins — usually by three to five times, in the longitudinal direction where the bracing lives.

The load path: shoe → beam → column → footing → soil

Every load on a rack has to reach the ground through a chain of members, and the chain is what you are really designing:

pipe → shoe → slide plate → tier beam → longitudinal strut / brace
     → column → base plate & anchor bolts → pedestal → footing → soil

Three things about that chain matter more than the arithmetic:

Gravity goes straight down the shortest path — shoe, beam, column, footing. Short, stiff, easy.
Transverse load (wind, seismic, guide reactions) is resisted by the bent itself — the moment frame formed by the two columns and the tier beams. That is why rack columns are usually oriented with their strong axis transverse.
Longitudinal load has nowhere local to go. A tier beam is pinned-ish about that axis; the friction from one shoe walks along the rack until it reaches a braced bay (or an anchor bay), collecting the friction from every bent on the way. Ten bents feeding one braced bay means ten bents' worth of friction in one diagonal.

That accumulation is the whole story, and it is the part a per-bent hand check misses.

Watch it happen: ▶ open the interactive: civil rack friction 3d — a bent with ten lines on shoes, sliding as the unit heats. Each shoe grows its own friction arrow; the arrows sum into the columns and again into the footings. Toggle the PTFE slide plates and every arrow in the chain shrinks to a third, right down to the base reaction.

Symbol key — every symbol on this sheet

μ (mu) — friction coefficient: sliding force ÷ normal force at the shoe · dimensionless
W — operating weight carried by one bent = Σ(pipe + contents + insulation) × bent spacing · kN
S — bent spacing: centre-to-centre of transverse frames · m
F_fr — friction force = μ·k·W, acting along the rack at shoe level · kN
k — simultaneity factor: fraction of lines sliding the same way at the same instant · —
q — design wind pressure at rack height · kPa (kN/m²)
C_f — force (shape) coefficient: 0.7 for round pipes, ~1.6 for open steel sections · —
A_eff — effective (shielded) wind area per bent · m²
F_w — transverse wind force on one bent = q·ΣC_f·A · kN
θ (theta) — inclination of the bracing diagonal from horizontal · degrees
P_br — brace axial force = accumulated shear ÷ cos θ · kN
D, L, T, W, E — code load symbols: dead, live, thermal (friction sits here), wind, seismic

Two different μ's, and why the letter is overloaded

This μ is the Coulomb friction coefficient — Amontons' laws, 1699: friction force is proportional to normal force and independent of contact area. It is not the ductility ratio μ used in blast and seismic work, and it is not Poisson's ratio ν. In rack work you will also meet μ written as f or CoF in vendor slide-plate catalogues. One warning that matters: Coulomb's "independent of area" holds for metal-on-metal, but PTFE is not Coulombic. Its μ falls as bearing pressure rises — typically 0.06–0.08 at 15–20 MPa and 0.12–0.15 at 2–3 MPa. So the light lines you were least worried about get the worst PTFE coefficient. Size the plate area for the load, not the pipe.

Everything a rack carries — and which ones actually move

Load	Where it comes from	Character
Dead (D)	structure, empty pipe, valves, insulation, fireproofing	always on, vertical
Operating contents	product in the lines; often classed with live	always on, vertical
Hydrotest	water in one line (or one defined group) at a time	temporary, vertical, local
Live (L)	platform/walkway 2.5–5 kPa, cable trays, future trays	vertical
Thermal friction (T)	μ·W as lines slide on their shoes	longitudinal, everywhere
Anchor & guide loads	pipe stress model output at specific nodes	localised, large
Wind (W)	q·C_f·A on shielded pipe bundle + open steel	transverse or longitudinal
Seismic (E)	mass × spectral acceleration (pipes are most of the mass)	either direction
Future growth	typically a 20 % spare-weight and spare-space allowance	vertical
Erection	crane picks, temporary stability, partially built frames	transient

Two entries in that table are movement loads, not weight loads: friction and anchor/guide loads. They exist only because the pipe is going somewhere. If nothing moves, they are zero. That single fact explains most of the confusion around them — and it is why an ambient utility rack genuinely does not need the same longitudinal bracing as a hot process rack of the same weight.

Friction is a reaction, not an action — and that caps it

μ·W is the most friction can deliver: the force required to break the shoe loose and keep it sliding. It is not a force the pipe wants to apply. If the thermal push behind the pipe is smaller than μ·W, the pipe simply does not move and the shoe carries whatever the push actually is — less than μ·W, and the line stays locked (which then loads the equipment nozzle instead, because the expansion has to go somewhere). Two practical consequences:

Friction never exceeds μ·W, but it can be much less on short runs and small lines.
A line that never slides is not friction-free — it is a hidden anchor. The "friction is small" mistake and the "this line is too short to matter" mistake are the same mistake seen from opposite ends.

On long hot headers the thermal push is enormous (tens to hundreds of kN), so μ·W is reached comfortably and the full value is the right design load.

Friction vs anchor loads — and what slide plates buy you

These two get lumped together as "pipe loads on the rack". They could not be more different:

	Friction	Anchor load
Where	every shoe, every bent	one node, by design
Size	1–10 kN per shoe	50–300 kN, sometimes more
Direction	along the rack, either way	usually along the rack, one way
Effect on design	sizes struts, bracing, columns everywhere	sizes one bent — the anchor bay
Reduce it by	slide plates, fewer sliding lines	relocating the anchor, adding a loop

The design response is different too. An anchor load gets a dedicated anchor bay: heavier columns, a deeper footing, sometimes a concrete shear wall. Friction gets slide plates — PTFE on a stainless backing plate, or graphite above ~260 °C where PTFE degrades. Dropping μ from 0.30 to 0.10 takes two-thirds off every longitudinal force in the rack. That is often the difference between a 2L 90×90 diagonal and a heavy built-up brace across twenty bays, and it is almost always cheaper than the steel — which is why the "should we use slide plates?" conversation belongs in the layout phase, not after the frame analysis fails.

Try the trade-off yourself: ▶ open the interactive: civil rack friction calc — set the line mix, the bent spacing and μ, and watch the longitudinal friction, the transverse wind and the brace utilisation move together. The single most instructive readout is "μ at which friction = wind": on a loaded rack it lands around 0.08–0.10, which means even PTFE does not make friction go away — it only makes it survivable.

Worked example — one refinery bent, start to finish

Two tiers, bent spacing S = 6 m, rack height 12 m, columns 0.25 m wide. Lines: twelve 6″ (55 kg/m operating), six 12″ (160 kg/m), two 24″ (440 kg/m).

Weight per metre of rack  = 12(55) + 6(160) + 2(440) = 2 500 kg/m
Operating weight per bent W = 2 500 × 6 × 9.81/1000  = 147.2 kN
Friction  F_fr = μ·k·W = 0.30 × 1.00 × 147.2         =  44.1 kN   (longitudinal)
Spec floor     = 0.10 × W                            =  14.7 kN

Now the wind, using the common shielding shortcut (largest pipe in the tier plus 10 % of the rest):

Σ pipe heights (incl. insulation) = 12(0.19) + 6(0.36) + 2(0.71) = 5.86 m
Shielded height h_eff = 0.71 + 0.10(5.86 − 0.71)                 = 1.23 m
A_pipes = 1.23 × 6 = 7.35 m²      A_steel = 2 × 12 × 0.25 = 6.0 m²
F_w = 0.9 [0.7(7.35) + 1.6(6.0)]                                 = 13.3 kN

44.1 kN of friction against 13.3 kN of wind — 3.3×. And friction does not stop at the bent:

Braced bay serving 5 bents:  ΣF = 5 × 44.1 = 220.7 kN
Diagonal at θ = 45°:         P_br = 220.7 / cos45° = 312.2 kN
Light diagonal (capacity 200 kN):   utilisation = 156 %   ✗
Same rack with PTFE (μ = 0.10):     P_br = 104 kN → 52 %  ✓

The brace did not fail because the wind data was wrong. It failed because 220 kN of pipe friction arrived from five bents away, and nobody added it up.

Common pitfalls

Applying friction to one bent and stopping there. It accumulates to the braced bay. Check the strut, the brace and the base shear at the anchor/braced bay, not just the local frame.
Forgetting friction acts both ways. Heat-up drags one direction, cool-down the other. Both signs must be run; bracing that works in tension-only fails if you only checked one.
Combining full friction with full seismic (or refusing to combine them) without reading the project criteria. Both practices exist; the criteria decide, not the analyst.
Taking credit for PTFE that will not be installed — or specifying PTFE on lines whose bearing pressure is so low that μ is really 0.15.
Using empty pipe weight for friction. Friction follows the operating weight, contents included; but use the empty weight when you are checking uplift and stability.
Ignoring the 10 % floor. Many structural criteria require a minimum longitudinal load of 10 % of the pipe weight regardless of μ and simultaneity — a cheap insurance policy against exactly the argument that "not all lines slide together".
Assuming a hot line always slides. If it cannot break friction loose it becomes a virtual anchor and the load reappears at the equipment nozzle instead.

Outcome

The rack load path is shoe → beam → column → footing → soil; transverse load is taken by the bent frame, longitudinal load walks along the rack to a braced or anchor bay and accumulates.
Friction F_fr = μ·k·W is a movement load, distributed over every shoe, and on a loaded rack it typically exceeds transverse wind by 3–5× in the longitudinal direction.
The crossover coefficient — the μ at which friction equals wind — is around 0.08–0.10 on a full rack, i.e. below PTFE. Slide plates reduce friction; they never remove it.
Anchor loads are localised and large (one bent, heavier everything); friction is small and everywhere (every member, all the way to the braced bay). Different loads, different fixes.
Interactive: ▶ open the interactive: civil rack friction calc — line mix, μ, wind, brace utilisation.
3D: ▶ open the interactive: civil rack friction 3d — the arrows building at each shoe and summing into the footing, with a PTFE before/after toggle.

Open items

A worked load-combination table against a real project structural design criteria (PIP STC01015 or client equivalent) — the friction/wind/seismic simultaneity rules vary more than they should.
Transverse friction at guides, and the guide reaction itself, deserve their own topic.
Extend the explorer with a second tier having its own μ (PTFE on the hot tier only), which is the optimisation most projects actually make.

Know why, not just what.

One sheet from EPiC Circle

Interactive engineering lessons across every EPC discipline — calculators, 3D models and misconception-first teaching. Free to read; sign in to track progress, take the quizzes and ask questions.

Open in EPiC Circle →

Educational teaching material — simplified and illustrative. Not for engineering design use.
© EPiC Circle · provided by PipeAxis