Support Stiffness in Pipe Stress Models: When "Rigid" Misleads

Code basis: ASME B31.3-2024 and B31.1-2024 for the piping, AISC 360-22 for support steel, and, where the owner invokes them, PIP PNC00004 (Feb 2025), PIP PNC00001 (May 2023) and PIP RECE002 (Aug 2026). Open any pipe stress model and look at the restraints: unless someone has typed a number into the stiffness field, every one of them is rigid — CAESAR II and AutoPIPE both default an unspecified restraint to an effectively infinite translational stiffness. Real supports are nowhere near that stiff, and the gap between the default and the installation can be several orders of magnitude. For most lines, most of the time, the gap doesn't change the answer. For the lines where it does — rotating equipment, vibrating service, large hot lines on flexible structures — it can reverse the conclusion of the analysis. This guide covers how to calculate support stiffness, when the rigid assumption helps you and when it hides a real load, how to model it in CAESAR II and AutoPIPE, and a worked pump-nozzle example showing the load a rigid model reports as zero.

First published 2026-09-23 · content last reviewed 2026-09-23 · page regenerated 2026-09-23 at build.

Written and reviewed by Matthew Norris, P.E. — active P.E. licensure in Arizona, California, Kansas, Missouri, North Carolina, Texas.

What real supports actually deliver

Stiffness only matters relative to the pipe. A restraint behaves as rigid when it is much stiffer than the pipe it restrains, measured at that point and in that direction — which is why a small-bore, thin, low-temperature line gives nearly the same answer on a soft support as on a rigid one, while a large, hot, heavy-wall line does not: the pipe there is the stiff element, it generates large thermal forces, and the support is the one that should be giving.

Ranges are ASSUMED, order-of-magnitude values for illustration — every real support type spans a wide range depending on geometry and material, and there is no code table of typical support stiffness to check them against. Calculate yours with the formulas below, or get it from the structural model.

There is a practical reason the default exists: stress analysis usually runs before the structure is designed, so starting rigid is a reasonable placeholder. The problem is staying rigid after the steel is sized, without going back and checking. That is vendor software behavior in CAESAR II and AutoPIPE, not a code requirement — nothing in B31.1 or B31.3 tells the program what value to assume for an unspecified restraint, and nothing in either code endorses leaving it there once the real support exists.

When "rigid" is conservative, and when it isn't

This is the central question, and the honest answer depends on which result you're checking:

The pattern: rigid restraints over-predict loads at the restraint and under-predict loads everywhere the restraint should have shed them. Those places — equipment nozzles, adjacent supports, flanges — are usually the ones with the tightest allowables. On real projects that shows up as expansion loops added because a rigid line stop over-predicted thermal load, clamps loosened in the field to fix a thermal problem that then produce high vibration, and pump or compressor nozzle loads over the limit after start-up even though the model passed.

Rotating equipment: the case that makes this mandatory

The clearest example is a pump suction or discharge line whose first support is a shoe on a platform or rack beam near the pump. Define four reference positions for the top of the support beam under that shoe:

• Line A — the theoretical position the stress model assumes: an infinitely stiff beam, no deflection.
• Line B — the beam deflected under its own weight only.
• Line C — the beam deflected under its own weight plus the empty pipe spool, after erection and bolt-up to the pump.
• Line D — the beam deflected under its own weight, the pipe, and the operating contents.

Split the deflection into two parts. A → C happens before the flange is bolted to the pump, and field fit-up takes it out: the millwright and pipefitter align the flange to the pump, with a field weld or adjustable support to close the gap. API RP 686 2nd Ed. §4.6 "Piping Alignment Requirements" (flange parallelism, bolt-hole offset) and §4.8 "Pipe Strain Measurement" (the dial-indicator shaft-deflection check while the flange bolts are tightened) exist to catch what fit-up misses. With good fit-up, the nozzle load from A → C tends toward zero. C → D — the contents weight — happens after bolt-up, when the line fills, and nothing in the field takes it out: the beam deflects further, the pipe follows it, and because the pump nozzle doesn't move, the difference becomes load and moment on the nozzle. That is the part the stress engineer has to verify using the real beam stiffness, because a rigid model reports it as zero.

On projects that invoke PIP RECE002 (Aug 2026), that verification is not optional. §4.1.1.5 lists the loads that shall be considered when calculating piping loads, including restraint reactions (item (g)), other attached piping that may exert significant loading (item (h)), and adjacent equipment attached to the same header (item (i)). §4.3.1.5 requires anchor deflection at adjacent restraints to be included for vital and sensitive machinery, and §4.3.1.4 requires the adjacent supports to be adjustable without needing re-adjustment between ambient and operating. A model with rigid restraints next to the nozzle cannot demonstrate compliance with any of those.

Rule for rotating equipment: replace the rigid default with calculated stiffness at every restraint within the nozzle's zone of influence — in practice, at least the first two supports on each nozzle line, and any restraint within about 10 pipe diameters. RECE002 §4.3.1.3 already prohibits anchors and restraints inside 10 D unless otherwise specified.

B31.3-2024 gives the general permission behind all of this at ¶319.2.1(b), "Restraint Flexibility": when a restraint is not treated as rigid, the engineer may take its flexibility into account when computing the displacement stress range and the reactions. The code leaves the decision to the engineer; the owner criteria, where invoked, turn it into a requirement for vital and sensitive machinery.

How to calculate support stiffness

The definition

Support stiffness in a given direction is the force needed to move the pipe attachment point by one inch in that direction:

k = F / δ (lb/in)

Equivalently, apply a 1 lbf load at the pipe and compute the deflection; the stiffness is the inverse of that deflection.

Components combine in series

A load path from pipe to ground passes through several elements — the clamp or shoe, the beam, the column, the connection — and their flexibilities add:

1 / k_support = Σ 1 / k_i

The softest element controls: a clamp at 10⁶ lb/in on a beam at 10⁴ lb/in gives a combined support stiffness of about 9,900 lb/in — improving the clamp does almost nothing. For elevated pipe on racks and posts, the structure usually governs, not the clamp.

Beam-theory formulas for the common cases

E is the modulus of the support steel at its own temperature, usually ambient — don't use the pipe's hot modulus here. I is the section's moment of inertia about the bending axis in the direction of the pipe load: a vertical load bends a rack beam about its strong axis, but an axial pipe load on a line stop pushes the beam laterally, about its weak axis, and often puts it in torsion too. A support is typically stiff vertically and much softer laterally and axially. Calculate stiffness per direction — don't copy the vertical number into the guide and line-stop directions.

Getting it from the structural model

Once the civil or structural engineer has a model, ask for the deflection under a unit load at each pipe-support location, in each restrained direction — or better, the stiffness matrix at those nodes. CAESAR II can take the steel in directly through its structural steel modeller, with the pipe connected to structural nodes through CNODEs, so the structure's flexibility and coupling — loading one beam deflects the next — come in automatically.

Worked example — pump suction shoe on a platform beam

Situation. An NPS 8 Schedule 40 carbon-steel line comes off a pump suction nozzle. The first support is a shoe 72 in from the nozzle, on a W8×31 platform beam. When the line fills, the contents add about 1,500 lbf at the shoe location. Every input is marked ASSUMED where it isn't looked up.

Step 1 — beam stiffness at the shoe

k = 48·E·I / L³:

• Case 1 (L = 240 in): k = 48 × 29×10⁶ × 110 / 240³ = 153,120,000,000 / 13,824,000 = 11,080 lb/in
• Case 2 (L = 120 in): k = 153,120,000,000 / 1,728,000 = 88,600 lb/in

In series with the 1 × 10⁶ lb/in shoe (1/k_total = 1/k_beam + 1/10⁶): Case 1 → 1/11,080 + 1/1,000,000 = 9.128×10⁻⁵ → 10,960 lb/in; Case 2 → 1/88,600 + 1/1,000,000 = 1.229×10⁻⁵ → 81,400 lb/in. The beam governs, as expected.

Step 2 — how the contents weight splits between the shoe and the nozzle

Idealize the pipe between the nozzle (treated as fixed) and the shoe as a cantilever. Its tip stiffness is k_pipe = 3·E·I/L³ = 3 × 29×10⁶ × 72.5 / 72³ = 6,307,500,000 / 373,248 = 16,900 lb/in. The contents weight at the shoe divides in proportion to stiffness — the shoe takes the series-combined support stiffness from Step 1, not the beam alone, since the shoe and the beam already act together: shoe carries k_s / (k_s + k_pipe), and the pipe carries the rest back to the nozzle.

Check: Case 2 fraction = 81,400/(81,400+16,900) = 0.828 → shoe 1,242 lbf, nozzle shear 258 lbf, moment 258 × 72 in = 18,576 in-lb ≈ 1,548 ft-lb. Case 1 fraction = 10,960/(10,960+16,900) = 0.393 → shoe 590 lbf, nozzle shear 910 lbf, moment 910 × 72 in = 65,520 in-lb ≈ 5,460 ft-lb. This is a deliberately simple two-spring model: it ignores nozzle flexibility, continuity beyond the shoe and thermal load. Its purpose is to show the size of the effect, not to replace the stress model.

The finding. The rigid model reports zero contents-weight load at the nozzle. The same layout with the shoe at midspan of a 20 ft platform beam puts on the order of 5,500 ft-lb on the nozzle from the contents alone, before any thermal load. For an NPS 8 suction nozzle that is likely a large fraction of the allowable, and under PIP RECE002 vital/sensitive machinery — limited to 50 % of the industry allowables at layout, §4.2.2.3 — it could be the entire budget.

The fix is usually cheap if found at design: move the support to near a column, or add a post directly under the shoe to grade; stiffen the beam (a deeper section, or a knee brace); or use an adjustable support (a screw-jack stanchion) so the fitter can take out A → C and the model can then be run with the real stiffness for C → D.

The dynamic side of the same example

Mechanical natural frequency scales with the square root of stiffness, √(k/m). The same beam-and-shoe idealization that softens the static nozzle load also softens the dynamic picture: replacing the rigid default with the Case 1 series-combined stiffness (about 11,000 lb/in versus 10¹²) drops the effective stiffness feeding the span's first mode by several orders of magnitude. Because frequency scales as the square root of stiffness rather than linearly, even a comparatively modest reduction — an order of magnitude, far short of what this example shows — is enough to pull the predicted natural frequency down substantially. A span whose rigid-model frequency clears a pump's running speed by a comfortable margin can have little or none left once the real support stiffness is used, because the rigid model always reports the highest frequency the system could have.

Vibration service: supports have to be stiff and hold the pipe down

For lines with dynamic loads — reciprocating machinery pulsation, slug flow, water hammer, flow-induced turbulence — the support design goal changes. You are no longer only controlling static stress and deflection; you are keeping the mechanical natural frequency away from the excitation and limiting vibration amplitude.

• Resting supports, rod hangers and springs can't hold down a vibrating pipe. Friction on a shoe is rarely enough to resist dynamic force, and a hanger has no stiffness against upward or lateral motion. PIP PNC00001 §4.2.10 requires lines subject to pulsation to be restrained perpendicular to their axis with clamp-type or hold-down supports. Reciprocating-machinery piping practice steers away from resting supports for pulsating service; API 618 (reciprocating compressors) discusses support and foundation stiffness for pulsating systems qualitatively — this guide does not cite a specific clause, because that clause has not been confirmed against your own copy of the edition in use, and neither a clause number nor a numeric minimum-stiffness value should be quoted without checking it there.
• Hold-down clamps, and the structure under them, set the frequency. A clamp is only as stiff as what it's bolted to — a well-made clamp on a flexible post can be softer than a plain shoe on a sleeper. The Energy Institute's Guidelines for the Avoidance of Vibration Induced Fatigue in Process Pipework (2nd Ed., 2008) is the standard reference for vibration screening and support recommendations, and PIP PNC00004 points to it by name.
• Supports near reciprocating machinery are spaced by frequency, not only by stress and deflection. Spans that satisfy the static criteria are often far too long for pulsating service. Spacing then comes from the required mechanical natural frequency, which in turn depends on the support stiffness you model.
• Damped clamps. Clamps lined with a visco-elastic damping material trade some stiffness for a lot of damping. Model them with the vendor's reduced stiffness — don't use the plain steel-clamp value.

The circular trap: the natural frequency the model predicts depends on the support stiffness assumed. Assume rigid, get a comfortable frequency margin, build it on flexible steel, and the margin disappears. For any dynamic evaluation, use calculated support stiffness — never the default.

Modelling it: CAESAR II and AutoPIPE

1. Stiffness on the restraint

Every restraint type (+Y, Y, X, Z, guides, line stops, anchors) has a stiffness field. Enter the calculated k in lb/in for that direction — leaving it blank gives the vendor's rigid default.

Restraint: Y      Stif: 1.1E4 lb/in     Mu: 0.3
Restraint: Z      Stif: 3.0E3 lb/in     (lateral — beam weak axis + shoe)

2. Connect to the structure with a CNODE

Model the support steel explicitly — as beam elements, or through the structural modeller — and link the pipe node to the structural node with a restraint that carries a CNODE. The structure's flexibility and coupling then come in without a hand calculation. It's the better approach where several lines share one beam.

3. Bilinear restraints for preloaded clamps

A bolted hold-down clamp holds the pipe by friction until the axial force exceeds the clamp's grip. Model that in the axial direction with a bilinear restraint: K1 is the initial stiffness up to the yield force F_y, F_y is the clamp preload times the total friction coefficient, and K2 is the post-slip stiffness, near zero. This beats adding the preload as an external force, which creates spurious stresses and deflections where the stiffness is low. Get the preload and friction coefficient for the actual clamp from the clamp vendor — a clamp lined on both faces has two sliding surfaces, so the friction coefficients add.

4. Gaps and friction

Model design gaps as built. Run with and without friction and take the governing result — a PIP PNC00004 §3.2.3 requirement where invoked, and good practice everywhere. Friction also needs a stiffness of its own (the "friction stiffness" in CAESAR II); the default is usually fine, but watch for convergence problems at low normal loads.

5. Springs and rod hangers are already "stiffness"

A spring hanger is a support with modelled stiffness: the spring rate. Include rod length and swing for rod hangers — PNC00004 §3.2.14.2 models rod length and limits rotation to 5°, and PNC00001 §4.3.3.3 limits swing to 4°. Use the free Spring Hanger Variability Calculator to check the load change your spring rate implies.

6. Vessel and tower nozzles

A vessel shell nozzle is not an anchor. WRC Bulletin 297 offers only a narrow flexibility dataset — two figures, limited to thin shells with simply supported boundary conditions — not a general set of axial and rotational nozzle stiffnesses; the bulletin itself notes flexibility is very sensitive to the assumed boundary conditions, especially for large penetrations. For anything outside that narrow range, FEA-derived nozzle flexibilities (axial, and in-plane and out-of-plane rotation) are the general route, modelled at the nozzle node. On a tall column, include the column's own thermal growth and its flexibility to the foundation, modelled as pipe elements.

7. Buried and soil-supported pipe

Soil is stiff overall but weak locally. It takes a long run of buried pipe — often several hundred feet — before the soil acts as an effective anchor. Model soil springs over that length rather than using a rigid anchor at the ground entry.

A practical workflow

1. Preliminary (rigid) run. Size the layout and support types, and get initial loads for the structural engineer. Mark the model "rigid supports — preliminary."
2. Screen for sensitivity. Flag restraints where stiffness is likely to matter: within the zone of influence of rotating equipment or sensitive nozzles; line stops and anchors on large hot lines; supports on cantilevers, long beams, platforms or tall posts; every support on a vibration-sensitive line.
3. Request stiffness from the structural engineer for the flagged points, per direction, once the steel is sized — or calculate it with the formulas above for simple cases.
4. Update and re-run. Report both the rigid and the real-stiffness runs for the flagged lines, and take the governing result for each check: the rigid run for restraint loads, the real-stiffness run for nozzle loads, displacements and frequencies.
5. Close the loop with the structural engineer. Send the updated reactions — stiffer supports attract load, so the steel may need re-checking. Use the free AISC 360 Steel Member Check and Trapeze / Strongback Calculator for quick checks.
6. Document it. In the calc basis, state which restraints use calculated stiffness, the source of each value, and which run governs each check. "All supports rigid" is a legitimate preliminary basis; it is a reviewer finding in a final rotating-equipment calc.

The checklist below is the version of that workflow a reviewer can check against, item by item:

• Calc basis states the stiffness assumption — rigid (preliminary) or calculated (final) — and for which restraints.
• Rotating-equipment lines: calculated stiffness at supports in the nozzle's zone of influence; RECE002 §4.3.1.5 anchor deflection included where invoked.
• Stiffness calculated per direction, including lateral and axial (weak-axis, torsion).
• Series combination with clamp or shoe stiffness shown.
• Vibration-service lines: natural frequencies computed with calculated stiffness, never the default.
• Friction run both ways; gaps as built.
• Updated reactions returned to the structural engineer after the stiffness update.
• Field fit-up requirement for the A → C deflection stated on the isometric or construction notes (adjustable supports, API RP 686 pipe-strain check).

Tools for this workflow

None of the calculators below solve a system flexibility model — that is what CAESAR II, AutoPIPE and their peers exist to do. What they cover is the arithmetic this guide walks through: beam and steel checks for the support itself, and the downstream check the stiffness feeds.

None of the site's paid tools calculates support stiffness itself today — the beam-theory and series-combination arithmetic in this guide, or a request to the structural engineer for the stiffness matrix, is the current path. Once you have the number, the free spacing and steel calculators size the support, and the nozzle load check carries the resulting load into the equipment allowable.

FAQ

Is modelling supports as rigid conservative?

For the restraint loads and the expansion stress at the restraint, modelling a support as rigid is usually conservative: the pipe cannot use the give the real support would have provided, so the model over-predicts the load at that point. For equipment nozzle loads, displacements at the support, and mechanical natural frequencies, the answer flips to no. A rigid model reports zero contents-weight load at a nozzle whose real support deflects, reports zero displacement at a support that actually moves under load, and reports a higher natural frequency than the field will measure because it has removed every source of flexibility below the pipe. Those are exactly the results a rigid model under-predicts, and they tend to be the ones with the tightest allowables: an equipment manufacturer's nozzle table, a clearance to adjacent steel, a margin to a pump's running speed. Whether the rigid assumption is conservative depends entirely on which result you are checking, not on the assumption itself — which is why a final calculation on a machinery line needs calculated stiffness at the restraints near the nozzle, not a blanket rigid model.

What stiffness should I use if I don't have the structural design yet?

Run the preliminary layout rigid, as usual, but do not stop there for the lines that matter. Flag the rotating-equipment and vibration-service lines and the restraints on cantilevers, long beams, platforms or tall posts, then re-run those with a realistic lower-bound stiffness — something on the order of 10,000 lb/in vertical on typical platform steel is a reasonable bracket before the structural design exists — and compare the two results. If the nozzle loads, displacements or natural frequency change meaningfully between the rigid and bracketed runs, you have identified which supports the structural engineer needs to design stiff, and you can raise it before the steel is sized rather than after. If nothing changes, the line is insensitive to support stiffness and the rigid preliminary result stands. Once real steel sizes exist, replace the bracketed value with a calculated one from the beam-theory formulas or the structural model, and re-run the flagged lines a final time before the calculation is issued for review.

Why does support stiffness matter for pump nozzle loads if the pump flange is aligned in the field?

Field alignment only removes the deflection that exists at the moment the flange is bolted up: the beam's own weight plus the empty pipe spool, taken up by the millwright and pipefitter aligning the flange to the pump with a field weld or an adjustable support. Deflection from the operating contents weight, and from thermal growth, happens after that bolt-up is complete, and nothing in the field removes it — the beam deflects further as the line fills, the pipe follows the beam, and because the pump nozzle itself does not move, the difference becomes load and moment carried into the nozzle. A rigid stress model reports that load as zero in both stages, because it has no deflection to lose. The stress engineer has to capture the second stage — contents weight and thermal load after bolt-up — with the real calculated support stiffness, because that is the part field alignment cannot fix and the part the equipment manufacturer's nozzle allowable actually has to absorb.

Does the clamp or the structure matter more?

For elevated pipe on racks, platforms or posts, the structure underneath almost always governs, because support flexibilities combine in series and the softest element controls the total. A clamp rated at a million pounds per inch sitting on a beam that deflects at ten thousand pounds per inch gives a combined stiffness barely different from the beam alone — roughly 9,900 lb/in in that example — so making the clamp stiffer buys almost nothing. For a shoe or clamp on a concrete sleeper or pier at grade, the relationship reverses: the foundation is comparatively very stiff, and the clamp or shoe hardware itself, along with its bearing and any elastomeric pad, can become the softer and therefore governing element. The practical rule follows directly from the series formula: identify the softest link in the load path from pipe to ground before spending engineering effort stiffening anything else, because stiffening a link that is not the softest one changes the calculated support stiffness by only a small fraction.

Does B31.3 require me to model support stiffness?

B31.3-2024 does not hand you a stiffness value, but it does address the concept directly: ¶319.2.1(b), "Restraint Flexibility," permits the engineer to account for a restraint's flexibility in the displacement stress range and the reactions whenever that restraint is not treated as rigid. That permissive wording puts the decision on the engineer, not a default value in the software. Owner practices close the gap further and are often the binding requirement on a project: PIP RECE002 requires anchor deflection at adjacent restraints to be included for vital and sensitive machinery, and PIP PNC00004 requires a modelled rod-hanger length with rotation checked and limited. Put together, a final calculation on a rotating-equipment nozzle line that leaves every restraint at the software's rigid default, with no stated basis for that choice, is difficult to defend in review even though no single B31.3 paragraph mandates a numeric stiffness — the code gives you the permission to model it properly and the owner criteria, where invoked, gives you the requirement.

How much does support stiffness change a mechanical natural frequency?

Mechanical natural frequency scales with the square root of stiffness, so it does not take an enormous change in support stiffness to move a frequency by a meaningful amount — dropping the effective stiffness in a support's load path by roughly an order of magnitude, which is a realistic gap between a rigid model and a real beam-and-clamp support, can lower the predicted first mode by tens of percent. A span designed to keep its natural frequency comfortably clear of a pump's running speed or blade-pass frequency on a rigid model can end up with little or no margin once the real, softer support stiffness is included, because the rigid model always reports the highest frequency the system could possibly have. This is the core reason vibration-service and pulsating lines are never screened on default stiffness: the frequency margin the rigid run shows is the best case, not the expected case, and the gap between them is exactly the support flexibility a calculated stiffness captures.