Non-Metallic Pipe Stress Analysis: FRP, HDPE and PVC/CPVC in CAESAR II and AutoPIPE

Code basis: ASME B31.3-2024 Chapter VII and B31.1-2026 Mandatory Appendix N for the piping, ASME NM.1-2022 (thermoplastics) and NM.2-2022 (FRP) for the stress criteria, and the manufacturer's published data for the properties. Plastic and composite pipe breaks most of the habits a steel stress engineer relies on. It is twenty to a thousand times more flexible, it grows two to twelve times as far for the same temperature change, FRP has different properties along the pipe and around it, and thermoplastics get softer the longer a load is applied. The B31 codes tell you all of that matters and then decline to give you a stress limit. This guide is the working method: which document supplies the acceptance criteria, how each material is analysed, what to type into CAESAR II and AutoPIPE, what to ask the manufacturer for, how to support the line, and the mistakes that reviewers keep finding. It uses real vendor data and two real FRP analyses, with the project names removed.

First published 2026-11-11 · content last reviewed 2026-11-11 · page regenerated 2026-11-12 at build.

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

Why a steel model gives the wrong answer

Copy a steel model, swap the material, and the program will run and print stresses. Four things in that model are now wrong, and none of them raises a warning.

Sources: FRP values are RPS Composites A-150 / P-150 pipe (E-433 Rev Z p. 3 for moduli, Design Manual 22-03 p. 16 for the 1.4 × 10⁻⁵ expansion figure). HDPE values are Performance Pipe PP-901 (July 2021), p. 19 and Table 11 at 73 °F. Steel, PVC and CPVC values are ASSUMED typical figures for comparison only. Use your manufacturer's sheet for the product you bought.

What those four differences do to the analysis:

• Low modulus means the pipe sags between supports that a steel line would bridge easily, and that it can buckle as a column when you hold back its thermal growth. It also means thermal forces are small even though thermal movements are large, which is the opposite of the steel instinct. The RPS Design Manual puts FRP at roughly one-twentieth the stiffness of steel (p. 14).
• High expansion means the movement has to go somewhere: into offsets, into expansion joints, or into anchors that are designed for it.
• No yielding is the big one for FRP. The whole B31 idea of a higher allowable for displacement stress rests on local yielding and shakedown. A laminate doesn't shake down, it cracks. RPS says it directly: treat displacement load cases the same way as sustained loads (E-433 Rev Z §4.1).
• Time-dependent modulus means a thermoplastic line has no single stiffness. The right number depends on whether the load lasts seconds or decades.

Everyone has seen the steel-trained model of a plastic line with an EXP case passing comfortably. That result means nothing here, and the rest of this guide is about what replaces it.

Applicable codes: who actually gives you a stress limit

Start with the contract code, because it decides everything downstream, and then be honest about what it does and doesn't contain.

Read that table from the top and the practical position is this. The B31 book makes you responsible for demonstrating adequate flexibility by a method you can show is valid (¶A319.4.2), and then steps back. You close the gap one of two ways: adopt the NM.1 or NM.2 criteria, or use a recognised FRP code built into the software together with the manufacturer's allowables. Both of the FRP analyses in this guide took the second route, running BS 7159 in CAESAR II with RPS allowables, on projects whose pipe class or specification cited NM.2. Either route can be defended if the owner agrees and the design basis says so. On a B31.3 job neither is granted by the code itself, and under B31.1 the NM standards are an all-or-nothing alternative. Mixing them without saying so is the finding.

Write it down. The design basis should state the contract code, the stress criteria document, the software code setting, and the source and revision of the manufacturer data. On non-metallic jobs those are four separate decisions, and a reviewer will ask about each one.

When you don't need a formal analysis

B31.3-2024 ¶A319.4.1 and NM.2-2022 ¶2-4.4.1 both exempt a system that duplicates a successful installation, one that can be judged adequate by comparison with an analysed system, or one laid out with enough inherent flexibility or with joints and expansion devices used as the manufacturer intends. NM.2 ¶2-4.4.2 then grades the alternatives as simplified, approximate and comprehensive. Its simplified method, which may rest on table or chart data such as spans, is for systems with reasonable flexibility and less than 75 °F between the installation temperature and the exposure temperature, and it does not allow a free-floating system with no anchors. As a rule of thumb, small-bore drain and vent lines built to the manufacturer's tables sit there, and long racks, tank farms, anything tied to rotating equipment and anything in seismic service don't.

FRP: the analysis method

1. The pipe has two layers and two directions

Filament-wound FRP is orthotropic: the fibres run mostly around the pipe, so it is stiffer and stronger in the hoop direction than along the axis. A beam-element program needs six things from the manufacturer for each size: axial modulus, hoop modulus, shear modulus, the Poisson ratio that couples hoop stress to axial strain, the expansion coefficient and the density. NM.2 ¶2-4.4.2(c)(2) asks for actual orthotropic properties that represent the specified resin, wind angle and glass content, and ¶2-4.3.3 warns that the Poisson ratio varies enough that the simplified formulas used for metals aren't generally valid.

The wall is also two layers. A resin-rich liner keeps the chemical out; the structural laminate carries the load. NM.2 ¶2-4.3.5 uses nominal dimensions and counts the corrosion barrier for weight and thermal effects but not as structure.

The clean way to get that split out of a stress program is the one RPS recommends (Design Manual 22-03 p. 20; E-433 Rev Z §4.3): enter the total wall thickness as the pipe wall, enter the liner thickness as the corrosion allowance, and make sure the stress cases are evaluated on the corroded wall. The program then computes weight, stiffness and thermal force on the total wall and stress on the structural wall.

2. Properties change with size

This catches people. The liner is a fixed thickness, so it is a large share of a small pipe's wall and a small share of a large pipe's, and the composite properties shift with diameter. An excerpt from the RPS A-150 / P-150 table shows the size of the effect:

Source: RPS Composites E-433 Rev Z (19 May 2022), p. 3, "Pipe Properties", for A-150 and P-150 pipe with a 0.11 in liner. Excerpt for illustration. Get the current sheet from the manufacturer for your product; the H-150-200 line, for example, has a 0.22 in liner and its own sheet.

The same manufacturer's design manual also quotes round-number laminate properties (1.4 × 10⁶ psi axial, 2.3 × 10⁶ psi hoop for the structural layer, p. 8). Those describe the structural laminate alone. The table above is the composite wall, liner included, and it is the one that matches the "total wall plus corrosion allowance" input method. RPS prefers the composite values for thermal load and accepts the structural modulus as a slightly conservative alternative (p. 18). What causes trouble is switching basis from one size to the next, or between the modulus and the Poisson term.

3. The allowable is not one number

There are three ways the allowable gets defined, and you need to know which one your model is using.

A manufacturer's axial allowable that varies with pressure. RPS publishes an allowable axial stress for combined pressure and bending at 0 psi and at 150 psi, built to hold a design factor of 6 on combined loading: DF = [σ_p/σ_up + σ_b/σ_ub]⁻¹, with a pressure axial strength of 20,000 psi and a non-pressure axial bending strength of 12,000 psi (E-433 Rev Z p. 3, note 3). For NPS 2 and larger the pipe allowable is 2,000 psi with no pressure and rises to between 2,300 and 2,700 psi at 150 psi, which follows from the higher strength the formula assigns to the pressure component. Fittings have their own table. The design manual's advice follows from this: enter the maximum operating pressure, not the 150 psi rating of the pipe you happened to buy (p. 20).

A strain limit, as in BS 7159. The program converts a design strain to a stress with the modulus and checks a combined stress. This is what both case files below used.

A hoop-versus-axial envelope, as in ISO 14692 and ASME NM.2. The allowable axial stress depends on the hoop stress at that point. NM.2 ¶2-4.3.4(a) describes its approach in exactly those terms.

In ASME NM.2-2022 ¶2-4.4.4(b) the checks are S_H ≤ k·S_Hmax and S_A ≤ k·S_Aallow, with:

Wind, seismic and test loads need not be taken as acting together. Notice how small the thermal credit is: ten percent, where a steel code would hand you a separate and much larger displacement range. ISO 14692 does the same job with a part factor on the envelope. The defaults the stress programs carry are 0.67 for sustained, 0.83 for sustained plus thermal and 0.89 for occasional, per Bentley's AutoPIPE training material and a Bondstrand project procedure we reviewed; we have not checked those against the current ISO text, so confirm them in your copy.

4. Load cases: no EXP

RPS's recommendation is unambiguous: run every operating load as an Operating or Sustained stress case, do not use Expansion stress cases, and run wind and seismic as Occasional cases (E-433 Rev Z §4.4). The same allowable applies to primary and secondary loads, so the case that matters is the full operating case W+T+P checked against the sustained allowable, not T alone checked against a range.

Pattern after the example in RPS E-433 Rev Z §4.4, for a single operating condition. S_h here is the allowable at the maximum sustained pressure. The source example applies a 1.05 vertical seismic factor on weight and a 1.33 occasional factor; set yours from the site seismic data and from the load duration rule you are working to (NM.2 ¶2-4.4.4(b)(2) above). The RPS pattern takes no increase on the operating case; on the NM.2 route that case is the one checked with k₁ = 1.1.

NM.2-2022 ¶2-4.4.4(c) adds two rules for the thermal input: calculate displacement stresses with the higher of the ambient and design-temperature modulus, and take expansion from the minimum installation temperature to the maximum exposure temperature, with contraction the other way round.

Add a hydrotest case, WW+HP, as an occasional case with the test factor. If the line has more than one operating condition, each one gets its own stress-checked operating case. Our load case guide covers the numbering and combination mechanics; the difference here is only which cases carry a stress check and against what.

5. Fittings: use the manufacturer's SIF and flexibility factors

The B31J factors built into the program are for steel butt-weld fittings. FRP elbows and tees are thicker than the pipe, laid up differently, and stiffer than a steel elbow of the same radius. RPS gives, per size, an elbow SIF between 1.7 and 4.0 and a flexibility factor between 1.0 and 2.5, derived by finite element analysis, plus these rules (E-433 Rev Z §3):

• Use the listed elbow thickness, which is the extrados. The intrados is 140 % to 200 % of it.
• Reduce the flexibility factor by 30 % for 45° bends.
• Divide both SIF and k by the pressure stiffening factor 1 + 2.53·(P/E_h)·(D/2T)²·(R/T)^1/3, which RPS references to BS 7159:1989 Eqn 7.7.
• Reducers: flexibility 1.0, SIF 2.5 at the large end and 1.3 at the small end.
• Tees and laterals: flexibility 1.0, pipe wall thickness, and SIFs by size band from 1.5 to 3.1 depending on run or branch and in-plane or out-of-plane.
• Flanges: SIF 1.0 and k 1.0, checked separately by equivalent pressure, P_eq = 16M/(πG³) + 4F/(πG²) + P, against 180 psi.

To put a number on the stiffening term: for an NPS 6 elbow with E_h = 1.153 × 10⁶ psi, D = 6.70 in and T = 0.350 in from the RPS elbow table, and an ASSUMED bend radius of 9 in, the factor is 1.04 at 60 psig and 1.09 at 150 psig. It is a small correction at these pressures, and it works in the direction of making the elbow slightly stiffer and the SIF slightly lower.

With no manufacturer data at all, Bentley's AutoPIPE guidance assumes a bend flexibility of 1.0 and a minimum SIF of 2.3 on bends and tees, and Hexagon's CAESAR II FRP presentation suggests 2.3 in both planes as a conservative choice. Use that to get a preliminary model running and replace it when the vendor data arrives.

6. Decide how the line handles its growth

The RPS manual frames the choice as semi-rigid or rigid (pp. 16 to 19), and it is the most useful early decision on an FRP layout.

Semi-rigid: let the pipe move and absorb the growth in offsets and loops. The required offset leg is L = (1/12)·√(3·E·I·ΔL/M) in feet, with I on the total wall and the allowable moment M = S_b·Z_s on the structural wall. RPS's example: an 8.52 in OD pipe anchored 24 ft from the elbow, 100 °F rise, α = 1.4 × 10⁻⁵, grows 0.40 in and needs a 10.85 ft leg. Add 20 % if the leg ends at a flange or tee. A loop uses the same leg length with the axial portion at half of it, and guides immediately each side.

Rigid: anchor the run and let the pipe carry the thermal load in compression. This only works because the modulus is low.

RPS's example for 6 in pipe (OD 6.48 in, ID 6.00 in, E = 1.5 × 10⁶ psi, ΔT = 50 °F): total wall area 4.70 in², thermal force F = E·A·α·ΔT = 4,940 lb, and stress on the 2.59 in² structural wall of 1,905 psi. Under five thousand pounds from a fully restrained 6 in line is the low modulus at work; the same run in NPS 6 Schedule 40 steel would carry roughly 50,000 lb (ASSUMED 29 × 10⁶ psi and 6.5 × 10⁻⁶ per °F).

The price is buckling. The manual's Euler check with the total-wall inertia (22.93 in⁴) gives a critical length of 262 in, or 21.8 ft, and concludes that every second support at about 10 ft spacing has to be a guide. E-433 §5.1 is more conservative: it uses the structural layer's modulus and inertia, E_s and I_s, and gives a rule of thumb of F_cr at least 15 % above F_th. Running the same example that way, with I_s = 13.08 in⁴ for a 6.22 in structural bore and an ASSUMED E_s of 1.36 × 10⁶ psi (the value noted on our project copy of the sheet), gives 176 in, or 14.6 ft (our calculation). Three things changed: the inertia, the modulus and the 15 % margin. Same pipe, same vendor, a third shorter guide spacing. Use the more conservative of the two unless the manufacturer tells you otherwise.

Three more points on restrained runs. E-433 §5.2 gives separate, lower allowables when the axial stress is mainly direct tension rather than bending, and a separate rule for compression. NM.2 ¶2-4.4.4(a)(2) requires restrained systems to be checked for column buckling. And internal pressure puts a restrained FRP pipe into tension, which offsets thermal compression, so NM.2 requires you to consider the hot, low-pressure case as well.

7. Pressure makes FRP grow

Internal pressure stretches low-modulus pipe measurably, and the Poisson coupling adds to it. NM.2 ¶2-4.4.4(c)(3) requires pressure elongation in the analysis and gives eq. (2-4-4) for the strain. Both programs can include it. RPS notes that depending on the program the effect may understate the pressure load on anchors, gives a hand formula for the true anchor load, and observes that the difference is usually small next to the thermal load (E-433 §5.3). Whichever way you set it, set it deliberately and say so. The two real configuration files we compared for this guide had it off; a third-party Bondstrand procedure requires it on.

FRP in CAESAR II: the input mapping

In CAESAR II, material 20 carries generic FRP defaults that you overwrite with the manufacturer's values, and the work is mapping a vendor table into a handful of fields without mixing bases. This is the mapping we use, taken from two analysed projects; field names are as they appear in versions 12 to 14, so check them against your release.

Configuration file

Keep a separate configuration file in the job folder for FRP work. These are the settings that matter:

RPS adds its own list of things to verify in the software (E-433 §4.2): that user bend flexibility factors and SIFs are actually used, that F/A axial stress and torsion are included in the code stress, that stresses are calculated on the corroded wall, that the SIF multiplier is 1 for all load cases, and that the occasional load factor is set. The sheet lists 1.2 for that last item and uses 1.33 in its own wind and seismic example, so set it from the duration rule.

Element input

The trap: one configuration, many sizes

The configuration file carries one set of FRP defaults. A model that runs from NPS 2 to NPS 10 has properties that differ by size, and anything left on the default is right for one size at most. Compare the NPS 3 and NPS 6 rows: G/E_a is 0.781 against 0.882, and the Poisson term is 0.395 against 0.403. In one of our projects the configuration held the NPS 6 values and in the other the NPS 2 to 3 values, and you could tell which just by doing the arithmetic on the vendor table. Two ways to stay out of trouble:

• Build a spreadsheet of the derived fields for every size, and type the per-size values on the first element of each size so they carry forward. The RPS manual recommends the spreadsheet (p. 20).
• Or split the model by size range, each with its own configuration. One of our reports ended up as an NPS 10 model and an NPS 4 and 6 model.

Reading the output

For the FRP codes the program reports a code stress and a ratio at each node, with the allowable varying from node to node because it depends on pressure and on whether the node is pipe, bend or tee. Note which kind of component governs. Read the hoop and axial components as well as the ratio, and check that the fitting factors echoed in the output are the vendor's.

FRP in AutoPIPE

AutoPIPE comes at it from the other side: it ships FRP component and material libraries and implements ISO 14692 with the failure envelope and part factors built in, so less of the mapping is by hand. What follows is from Bentley's own non-metallic training workbook and presentation (the presentation is copyright 2014; the workbook is undated), so treat the dialog names as a guide and check your release.

1. General Model Options: set the piping code to ISO 14692 or BS 7159. The default component and material libraries follow the code.
2. Pipe Properties: actual ID and wall thickness, fluid specific gravity, and an FRP material from the library. Cold properties come from the library and can be edited in the dialog; hot properties follow the operating temperature.
3. Poisson ratio: enter the larger of the two ratios from the manufacturer's sheet; the program derives the other from the modulus ratio.
4. Envelope inputs: the long-term axial strength with and without pressure, the hoop strength, the partial factors A1 (temperature), A2 (chemical resistance) and A3 (cyclic service), and the part factor f₂ by load category. All editable.
5. Bends and tees: give them their own pipe identifiers with the real, thicker wall, then use Joint Type and User SIF to set the joint type, the qualified stress and the biaxial ratio. A tee that hasn't been qualified by test carries a pressure stress multiplier that the training material puts between 1.8 and 3.0.
6. Load sets: create the default static set, tick pressure extension, and add a hydrotest set.
7. Code compliance report: each node lists hoop and axial stress, SIFs, the allowable and the ratio, with the governing ratio flagged.

For FRP on a B31.3 job without an FRP code, the training material describes a simpler route: a user-entered allowable equal to a design stress of one-tenth of the minimum tensile strength. In B31.3-2024 that is the ¶A302.3.2(b) basis for laminated (contact-molded) RTR, valid to 180 °F. Filament-wound pipe falls under ¶A302.3.2(c), a hydrostatic design basis times a service factor, so check which one your pipe is. For non-standard materials, expansion is entered as inches per 100 ft, so 3.0 × 10⁻⁵ per °F over a 30 °F rise is 3.0 × 10⁻⁵ × 30 × 1,200 = 1.08 in per 100 ft.

Which program is easier? Our view: AutoPIPE gets to a first ISO 14692 result with less typing because the envelope and libraries are already there, and CAESAR II makes you enter the material yourself, which is more work and more exposure to input error but keeps every assumption in front of you. In either program the library or default values are not your manufacturer's, and the job isn't done until they have been replaced.

Reference results: two FRP analyses

Numbers help you know whether your own output is in the right neighbourhood. These are from two issued analyses, with client, site and document numbers removed.

Case file A: plant drains and clarifier piping

Vinyl ester filament-wound FRP with a 0.11 in liner, NPS 2 to NPS 10, 120 °F, design pressure 5 to 15 psig on the clarifier lines and 60 psig on the de-watering lines (90 psig test), slurry up to 1.34 specific gravity, supported at grade and in trenches. BS 7159 in CAESAR II with RPS allowables. The project support specification limited deflection between supports to 0.1 in.

Two things to notice. Hydrotest governed one model, so it can't be left as an afterthought. And the transfer-area ratio rose from 86 % to 96 % between revisions even though the code stress fell, from 3,387 psi to 2,163 psi, because the allowable fell further, from 3,943 psi to 2,260 psi. That revision corrected the wall, the corrosion allowance, the modulus, the Poisson term and the expansion coefficient. Correct inputs gave a less comfortable answer.

Case file B: vapor headers off a bank of FRP tanks, high-seismic site

NPS 3 to NPS 6 FRP vapor lines connecting eight FRP storage tanks, 131 °F design temperature, 3.7 psig design pressure, in a region where seismic load governs. BS 7159 in CAESAR II. Forty-seven load cases: weight, three operating conditions, wind from four directions, seismic in six directions, seismic anchor movement of the tanks, and a relief valve reaction. Rubber expansion joints at the tank nozzles.

The nozzle line is the one worth studying. An FRP tank nozzle is rated for a few hundred pounds, so a pipe that passes its stress check by a mile can still tear the nozzle off the shell. The layout near the tanks, with a flexible joint at each nozzle and restraints close behind it, is what keeps those loads low. The check in that report covered the operating cases; run it for the occasional cases as well. Get the allowable nozzle loads from the tank fabricator in writing before you start, and give the fabricator your seismic displacements at the nozzle in return.

HDPE: the analysis method

1. There is no single modulus

Polyethylene is viscoelastic. Load it quickly and it is fairly stiff; hold the load and it keeps deforming. The design values are published as an apparent modulus by load duration and temperature.

Apparent elastic modulus, 1,000 psi, PE 4710. Typical values from Performance Pipe PP-901 (July 2021), Table 11, which credits the PPI Handbook of Polyethylene Pipe, 2nd Ed. Excerpt; the full table runs from −20 °F to 140 °F and seven durations.

Part of the choice is a code requirement. ASME NM.1-2022 ¶2-4.2(b)(1) requires the modulus used for thermal expansion analysis of HDPE to be consistent with the duration of the thermal load, and ¶2-4.2(b)(2) has seismic loads use the half-hour modulus from NM.3.3 increased by 25 %. The rest of the table below is our practice, built around those two rules. It means at least two models, or two property sets in one model:

2. Worked example: what the modulus choice does

NPS 4 DR 11 PE 4710, restrained between anchors, 50 °F temperature drop. OD 4.500 in, wall 0.409 in, so the wall area is 5.26 in². With α = 8 × 10⁻⁵ per °F (PP-901) the restrained stress is σ = E·α·ΔT and the anchor load is F = σ·A.

Check: 130,000 × 8 × 10⁻⁵ × 50 = 520 psi; 520 × 5.257 = 2,734 lb, shown rounded to 10 lb. Area = π(4.500² − 3.682²)/4 = 5.26 in².

A factor of 4.5 on the anchor load from one input. PP-901 makes the practical point that heat moves slowly through a polyethylene wall, so temperature changes are not quick, and that the average temperature is often used when picking the modulus. Our reading of that: a cold snap overnight is a ten-hour load. A seasonal swing is a thousand-hour load. Neither is short-term, and neither is fifty years.

3. Allowables

The material rating starts with the hydrostatic design basis. For PE 4710 the HDB at 73 °F is 1,600 psi, and the hydrostatic design stress for water is 1,000 psi, which is the HDB times a design factor of 0.63. PPI TR-3 is the policy behind those listings. That is a pressure rating basis, and NM.1 then builds the piping checks on allowables S and S_A from NM.3.3:

• Sustained longitudinal stress, eq. (2-3-8), against S at the design temperature, with the SIF from Mandatory Appendix IV applied to the sustained moment.
• Occasional, eq. (2-3-9), with a factor that depends on how long the load acts.
• Cyclic displacement range, eq. (2-3-11): i·M_c/Z against S_A, which is a fatigue allowable at the design cycle count (¶2-2.3.3(b)).
• Non-repeated anchor movement, eq. (2-3-12), checked separately.

We don't hold NM.3.3, so we quote no allowable from it. Read S and S_A for your resin, temperature and cycle count from your own copy. If the project isn't working to NM.1, one owner's HDPE stress guideline that we reviewed, built on the PPI Handbook, sets the allowable at HDS times an environmental factor times a temperature factor and checks sustained, thermal and occasional stress against it, with no 1.33 increase on occasional loads.

A public worked example is EPRI report 1021094 (December 2010), a research report written around nuclear Code Case N-755, which analysed a 4 in PE 4710 line in CAESAR II and cross-checked it in a general finite element program. At 100 °F and 150 psi it reported sustained stress of 366 psi against 670 psi and thermal stress of 901 psi against a 1,100 psi fatigue-based limit, with long-term deflection of 0.02 in against a 1 in limit on supports at 4 ft. It used the code case's tensile modulus (110,000 psi at ambient, 100,000 psi at 100 °F) for the code stresses and the long-term apparent modulus for deflection, the same kind of split as above.

4. Thermal growth

Performance Pipe's rule of thumb is 1/10/100: an inch of movement for every 10 °F for every 100 ft. Unrestrained HDPE on the surface copes with that by snaking sideways, and PP-901 recommends laying it with deliberate slack so that it snakes where you want it to. The handbook also says expansion joints are in general not recommended for HDPE, especially in pressure service, and that any joint used has to be made for HDPE so that it activates at very low force. The usual tools are offsets, loops, slack, and anchors sized for the restrained load. Check Poisson pull-back at mechanical joints and transitions too: pressurising the pipe shortens it, with a Poisson ratio of 0.45 for long-term and 0.35 for short-term loading (PP-901).

5. Fittings

NM.1 Mandatory Appendix IV, Table IV-1, gives stress intensification and flexibility factors for HDPE straight pipe, butt-fusion joints, molded and mitered elbows, tees, saddle and electrofusion branches, reducers and flange adapters. Use those, typed in as user values, in place of the program's steel factors. Fabricated mitered elbows are often made from heavier-wall pipe than the run (Code Case N-755, as quoted in the EPRI report, asks for one dimension ratio lower), so give them their own wall thickness in the model.

6. In the software

In CAESAR II the EPRI example shows the pattern, there under a nuclear code setting: a user-defined material with its own modulus, Poisson ratio and density, user-entered allowables, and user SIFs at bends and tees. Add separate runs for the short-term and long-term modulus. AutoPIPE's training material describes an HDPE option based on ASME Code Case N-755, a nuclear case, which may or may not suit a process job. We have not confirmed what either current release implements for NM.1, so check what yours offers before you build a workaround, and say in the report which equations were actually applied.

PVC and CPVC

Solvent-cemented PVC and CPVC fall under the same thermoplastic rules as HDPE: B31.3 Chapter VII or B31.1 Appendix N for the piping, NM.1 for the stress criteria. The engineering problem is different, though, because the material is stiff and brittle where polyethylene is soft and tough.

• It doesn't snake. At roughly 0.4 × 10⁶ psi (ASSUMED typical), PVC is three times stiffer than short-term HDPE, and it grows about 3.6 in per 100 ft per 100 °F. That growth turns into bending at changes of direction, so offsets and loops have to be sized. Our expansion loop calculator has a thermoplastic mode for the guided-cantilever check.
• It loses strength fast with temperature. One manufacturer's span chart (Harvel) for Schedule 40 PVC drops from about 7½ ft at 60 °F to about 4 ft at 140 °F for 4 in pipe carrying water. We read those two values from a scanned chart, so confirm them against your manufacturer's current table before use. CPVC holds its properties to a higher temperature, which is the reason it exists.
• Joints need attention. Bentley's guidance assumes a minimum SIF of 2.3 on bends and tees where there is no data. Get the manufacturer's position on solvent-cement and threaded joints.
• It must not carry gas. B31.3-2024 ¶A323.4.2(a)(3) prohibits PVC and CPVC for compressed air or other compressed gases.
• Anchors. Design the anchor pattern so that each run has a defined way to move, following B31.3-2024 ¶A321.5.2. The one-anchor-per-straight-run rule of ¶A321.5.3 is written for brittle piping such as glass; applying it to PVC and CPVC is a conservative practice, not a stated requirement.

For the stress model, the method is the HDPE method with different numbers: a user material with the manufacturer's modulus at temperature, the manufacturer's expansion coefficient, allowables on the hydrostatic design stress basis, user SIFs, and the full operating case checked against the sustained allowable. Because the modulus is higher than HDPE's, restrained thermal loads are higher, and valves and flanged equipment connections deserve a careful look.

Supports, spans and standard details

Most non-metallic piping failures we have seen attributed to "stress" are support failures: a point load, a clamp that crushed the wall, a valve hanging on the pipe, or a span set from the steel table. The codes agree with each other here. NM.2-2022 ¶2-5.1.1(b) says point loads and narrow areas of contact shall be avoided and calls for padding where damage could occur. NM.1-2022 ¶2-4.1.1 wants vertical clamps to spread load over a pipe diameter of length and forbids tightening clamps to compress the pipe. B31.3-2024 ¶A321.5.2(d) and NM.1-2022 ¶2-4.1.4(d) both refuse to count clamp friction as an anchor unless the manufacturer says so.

Spans

Sources: RPS Composites Design Manual 22-03 p. 15 and E-433 Rev Z p. 4, both for P-150 and A-150 pipe; derate by 0.90 for 180 °F service. HDPE rows are our calculation.

Look at the first four rows. These are two documents from the same manufacturer for the same pipe, and the E-433 spans run 14 % to 35 % longer than the design manual's, because they are built on different criteria. Neither is wrong. The point is to cite which one you used, and to notice that both assume water or 1.3 specific gravity and no concentrated loads. Then check the owner's specification: case file A's support specification limited deflection between supports to 0.1 in, five times tighter than either vendor table, and the spacing had to meet it.

HDPE span, worked. Performance Pipe recommends a long-term deflection limit of 1 in and the simple-beam form L = [384·E·I·y / (5·(W_p + W_f))]^1/4 (PP-901 pp. 23 to 24). For NPS 4 DR 11: I = π(4.500⁴ − 3.682⁴)/64 = 11.11 in⁴; pipe weight 0.182 lb/in at 0.0347 lb/in³ (0.960 g/cm³, PP-901 Table 3); water 0.384 lb/in at 0.0361 lb/in³. With the 50-year modulus at 73 °F of 29,000 psi, L = [384 × 29,000 × 11.11 × 1.0 / (5 × 0.567)]^1/4 = 81 in, or 6.8 ft. At 140 °F, with 12,500 psi, it is 66 in, or 5.5 ft. Above some temperature and below some size, continuous support in a tray or channel is cheaper than the hangers.

Rules that hold for all three materials

• Saddles and pads. Cradle the pipe over a wide arc, with an elastomer or bonded wear pad between pipe and steel. RPS's catalog saddle and clamp supports include cushioning, and its notes ask for the supporting steel to bear on at least half of the support base length; its design manual calls for a 1/8 in rubber lining in all clamps, hangers and anchors.
• Guides. The RPS manual replaces every fourth hanger with a guide to control sideways oscillation, and more often where a restrained run needs them against buckling. Standard guide and line-stop gaps in the RPS catalog are 1/8 in. Model the gap you are going to build.
• Anchors. Anchor each straight leg once, and only once, and take the load through thrust collars, bonded on FRP and fused on polyethylene, as the manufacturer details them.
• Valves. Independently support valves and equipment that would otherwise load the pipe, and evaluate actuator weight and overhang (NM.2-2022 ¶2-5.1.1(c)). A lined knife gate valve can outweigh several spans of the pipe it sits in.
• Fittings. Keep supports off fittings and leave room for the joint overlay on FRP.
• Flanges. Full-face gaskets, flat-face mating flanges or a spacer, torque to the manufacturer's table (RPS installation manual), and never pull a misaligned FRP flange into place with the bolts.
• Risers. Carry the weight on a collar, not on clamp friction.
• Friction. Use a coefficient that matches the actual sliding pair. On case file A the reviewer asked for the 0.15 in the model to be raised to the 0.30 they took as typical for steel sliding on concrete.

Rubber expansion joints in FRP and plastic systems

Rubber joints show up on FRP jobs at tank nozzles and pumps, to absorb differential movement and to isolate vibration. They are soft, which is what you want next to a nozzle rated for a few hundred pounds, and they come with a pressure thrust that the model has to carry somewhere.

Source: Proco Products Style 240 data sheets for the two sizes, and Proco's control unit table for Series 240. Movements are non-concurrent. Thrust at 60 psig is our calculation, thrust factor times pressure. Excerpt for illustration; use the sheet for the joint you are buying.

How to model one, and what to check:

1. Stiffness. Enter the axial, lateral and angular spring rates. They are published at zero pressure and room temperature, and compression and extension differ, so run the governing direction or bracket it.
2. Thrust. Apply thrust factor times pressure as opposing forces on each side of the joint unless control rods carry it. Twelve-inch joints at modest pressure put several tons on the first anchor or, worse, on the nozzle.
3. Anchoring. Proco's rule is that rubber joints go between two fixed anchor points, and where that can't be done control units are required. On FRP flanges, RPS advises avoiding tie rods where possible because they concentrate load on the flange, and using a steel backing ring where they are needed. Proco's Series 260R, its range for plastic and FRP systems, calls for an extra set of retaining rings behind the mating flange when control units are fitted to PVC, CPVC or FRP flanges, to stiffen them.
4. Concurrent movement. The rated movements are for one direction at a time. Sum the actual movement over rated movement in each direction; the total has to be 1.0 or less. Take the movements from the operating and occasional cases, including seismic anchor movement.
5. No pipe weight on the joint. Support the pipe each side.
6. Vacuum and temperature. The vacuum rating is at neutral length, and Proco says not to install the joint extended in vacuum service. Check the elastomer limit against the service.

Our expansion joint workspace runs the concurrent-movement rating check for rubber joints from the values you enter off the vendor sheet.

Common mistakes and oversights

Most of these come from review comments on real FRP reports, including our own. The rest are named in the manufacturer and code text.

1. Running an EXP case and checking it against a displacement range. There is no such allowable for FRP, and for thermoplastics the range check is a fatigue check against S_A at a cycle count. Stress-check the full operating case.
2. Mixing structural and composite properties. A laminate modulus on one size and a composite one on the next, or a laminate modulus with a composite Poisson term, gives a model nobody can check. On case file A the first issue used 1.4 × 10⁶ psi from the laminate property block where the composite value for the size was about 1.2 × 10⁶ psi.
3. Wrong corrosion allowance. The liner goes in the corrosion field at its real thickness. Case file A started with 0.0325 in where the liner was 0.11 in, which credited most of the liner as structure.
4. One property set for every size. See the configuration trap above. Check G/E_a, the Poisson term and the expansion coefficient for each size in the model.
5. Leaving the steel SIFs and flexibility factors on. Confirm in the output that the vendor's values were used at every bend and tee.
6. Using the pipe's pressure rating as the analysis pressure, or reading the allowable at the wrong pressure. The FRP allowable is pressure-dependent.
7. Treating hydrotest as an afterthought. On low-pressure FRP it can govern. It governed a model in case file A.
8. Applying 1.33 to every occasional load. NM.2 gives 1.20 for loads lasting up to 8 h and 1.33 only for loads up to 1 h and for testing. B31.3-2024 ¶A302.3.4 gives 1.33 for RTR and nothing for thermoplastics.
9. Zero-weight rigids. The program warns and carries on. Lined valves, actuators and steel backing rings are heavy relative to the pipe; get the weights from the manufacturer.
10. Default reducer and branch geometry. The program assumes a reducer angle and reports unusual branch angles as notes. Read the warning file.
11. Ignoring nozzle allowables on FRP tanks and equipment. They are an order of magnitude below steel nozzle allowables and they govern layout.
12. Leaving out the expansion joint thrust, or counting on control rods that aren't on the purchase order.
13. Clamp friction as an anchor. The codes don't allow it unless the manufacturer specifically recommends that support as an anchor, and the pipe creeps out from under the clamp.
14. Too many anchors. Extra restraint on a low-modulus line doesn't add much load, but it adds buckling risk on FRP and over-constrains thermoplastics. Anchor each leg once.
15. Steel spans, or the vendor span with the wrong fluid. Slurry at 1.3 specific gravity and insulation both shorten the span, and so does the owner's deflection limit.
16. One modulus for HDPE. A long-term modulus understates anchor and seismic loads; a short-term modulus understates sag.
17. Using this year's sheet for last year's pipe. Manufacturer data is revised. The RPS sheet we quote is Rev Z. Record the revision in the report and attach the sheet.
18. Isometric details. Line numbers reused on two sizes, a joint called out on the wrong size, a liner thickness on the drawing that differs from the vendor sheet. We found all three while preparing this guide. Reconcile the stress isometrics against the vendor data before issue.

A reviewer's checklist

The generic stress checklist most companies use has no line for any of this. Add these.

• Design basis names the contract code, the stress criteria document, the software code setting, and the manufacturer document and revision.
• Manufacturer data sheet attached, for the product line and liner thickness actually purchased.
• FRP: OD, total wall and liner entered as diameter, wall and corrosion; stress cases on the corroded wall.
• FRP: axial modulus, hoop modulus, G/E_a, Poisson term, expansion and density checked for each size in the model.
• FRP: vendor SIF and flexibility factors at bends, tees and reducers confirmed in the output; pressure stiffening basis stated.
• No EXP stress case used for code compliance. Operating cases stress-checked against the sustained allowable.
• Occasional factor matches the load duration; wind, seismic and test not combined with each other.
• Hydrotest case present, with test weight and pressure.
• Pressure elongation setting stated.
• Restrained runs: thermal force, guide spacing against buckling, and the compressive allowable checked.
• Thermoplastics: modulus per load case stated, with duration and temperature; S and S_A source stated; SIFs from NM.1 Table IV-1 or the manufacturer.
• Spans checked against the manufacturer's table for the real fluid and temperature and against the owner's deflection limit.
• Every rigid has a weight. Valve and actuator weights from the manufacturer.
• Tank and equipment nozzle loads compared with allowables supplied in writing.
• Expansion joints: spring rates, thrust, anchors or control units, and the concurrent movement sum.
• Friction coefficients match the sliding surfaces; guide gaps as built; lift-off reviewed.
• Support details on the isometric: saddle and pad, guide gap, collar-type anchors, independent valve supports.

Tools for this workflow

None of the calculators below replaces the system model. They cover the hand checks this guide walks through, before the model exists and when you are reviewing someone else's.

Everything here is engineering guidance, not a substitute for the code, the manufacturer's current data or the judgement of the engineer of record, who verifies the inputs and owns the result.

FAQ

Does ASME B31.3 require a stress analysis for FRP or plastic piping?

B31.3-2024 Chapter VII requires the piping to have adequate flexibility, and ¶A319.4.2 puts the burden on the designer to demonstrate it by a simplified, approximate or comprehensive analysis using a method that is valid for the case, unless one of the ¶A319.4.1 exemptions applies. Those exemptions cover a system that duplicates a successful installation, one that can be judged adequate by comparison with a previously analysed system, and one laid out with conservative inherent flexibility or with joints and expansion devices used as the manufacturer intends. What the code does not do is tell you how to run the analysis or what stress to accept: ¶A319.1.2(a) states that no specific stress-limiting criteria or methods of stress analysis are presented. So the answer is yes for anything that falls outside the exemptions, and the criteria have to come from somewhere else, normally ASME NM.1 for thermoplastics or NM.2 for FRP, or a recognised FRP code such as BS 7159 or ISO 14692 together with the manufacturer's allowables. Record which one you used in the design basis.

Why shouldn't I use an expansion stress case for FRP?

The higher allowable that steel codes give to displacement stress depends on the material yielding locally and shaking down to elastic behaviour over the first few cycles. A glass-reinforced laminate does not yield. Loaded past its limit it cracks the resin and starts to weep, and it does that just as readily under a thermal load as under weight. For that reason the manufacturer we quote in this guide recommends treating displacement load cases the same way as sustained loads and not using expansion stress cases at all, and ASME NM.2-2022 allows only a ten percent increase, k1 of 1.1 against 1.0, when thermal displacement and settlement are included with the sustained loads. In the stress program that means the full operating case, weight plus temperature plus pressure, is given a sustained stress type and checked against the sustained allowable. The algebraic difference between operating and sustained is still useful for understanding where the thermal load goes, but it carries no code check.

What data do I need from the FRP manufacturer before I start?

For each pipe size: outside diameter, structural and total wall thickness, liner thickness, axial modulus, hoop modulus, shear modulus, the Poisson ratio for axial strain caused by hoop stress, the expansion coefficient, density, and the allowable axial stress with the pressure it applies at. For fittings: the outside diameter and thickness of elbows and tees, stress intensification factors and flexibility factors, and whether pressure stiffening should be applied. For supports: the recommended spans with their basis, guide spacing guidance and standard support details. Ask also for the rules on occasional load increases, hydrotest, compressive and direct tensile loading, and flange loading, and for the revision and date of the document. The better manufacturers publish all of this in a stress analysis data sheet for each product line. If the data comes as round-number laminate properties in a brochure instead, ask for the composite wall properties by size, because the liner makes them differ from the laminate values and from each other.

Which modulus do I use for HDPE in a stress model?

More than one. Polyethylene's apparent modulus falls with load duration and with temperature, by more than a factor of four between a short-term load and a fifty-year load at room temperature. ASME NM.1-2022 ¶2-4.2(b) requires the modulus for thermal expansion analysis to be consistent with how long the thermal load is applied, and has seismic loads use the half-hour modulus increased by 25 percent. In practice, check sag and long-term deflection with the long-term modulus at the service temperature, check thermal stress and anchor loads with a modulus matched to the length of the temperature change, and check wind, seismic, surge and relief loads with the short-term modulus. Support loads should be taken from both ends of that range. A single run with a long-term modulus will understate restraint and seismic loads, and a single run with a short-term modulus will understate deflection. State the value, duration and temperature used for each load case in the report.

Can I use the same pipe support spans for FRP or plastic pipe as for steel?

No. The modulus is a small fraction of steel's, so deflection governs long before stress does, and the spans come out much shorter. For water-filled NPS 4 pipe at moderate temperature, one FRP manufacturer's tables give roughly 9 to 10 ft for a single span, and a deflection-based calculation for DR 11 polyethylene gives under 7 ft at room temperature and less as it warms up. Use the manufacturer's span table for the actual product, and read its basis: the temperature, the fluid specific gravity, the deflection limit and whether it assumes single or continuous spans. Two tables from the same manufacturer can differ by up to a third because they use different criteria. Then derate for hotter service, heavier fluid, insulation and concentrated loads such as valves, which should be supported independently, and check the owner's specification, which may set a tighter deflection limit than the manufacturer does. Support type matters as much as spacing: wide padded saddles, no point contact, and no clamps tightened onto the wall.

Is CAESAR II or AutoPIPE better for non-metallic piping?

Both will do the job and both need the same care. AutoPIPE ships FRP material and component libraries and implements ISO 14692 with the failure envelope and part factors built in, so a first result arrives sooner. CAESAR II offers BS 7159, UKOOA and ISO 14692 but starts from a blank FRP material, so you map the manufacturer's table into the configuration file and element fields yourself, which takes longer and leaves more room for a wrong basis, but keeps every assumption visible. For thermoplastics, neither program removes the need to choose the modulus for each load case and to enter the fitting factors and allowables by hand, and we have not confirmed what the current releases implement for NM.1, so check yours. The quality of the answer depends far more on the manufacturer data, the load case setup and the support details than on the program. Whichever you use, replace every library or default property with the value for the product being installed and say in the report which code equations the program applied.