How CFRP Materials Are Tested: A Guide to the ASTM Methods Behind the Numbers

Two data sheets, two different tensile numbers for what is supposedly the same carbon fiber fabric — every engineer who has sourced CFRP fabric or CFRP plate from more than one supplier has run into this. The fiber did not change. The test method did. A strength or modulus number by itself is not a fact about the material; it is a fact about the material plus a specific test method, specimen geometry, strain window, and pass/fail rule. Read the footnote before you read the number, or you are comparing two things that were never meant to be compared — and before you trust the rest of the sheet, read it with the same discipline described in how to read a CFRP technical data sheet.

Why do two suppliers quote different tensile numbers for the same carbon fiber fabric?

Because "tensile modulus" is not one measurement — it depends on which strain window the lab used to calculate it, and the two dominant standards use different windows. ASTM D3039 carbon fiber tension testing calculates a chord modulus from the stress-strain curve between 1000 and 3000 µε, while EN 2561 calculates a secant modulus from 10% to 50% of the failure load. On a linear-elastic material the two methods can converge; on a real carbon laminate with any curve nonlinearity, they give different numbers from the same test data (FAA DOT/FAA/AR-04/24 Table 3, p.9). SAE AMS 2980/3970 recognized the problem and modified EN 2561 to use the same 1000–3000 µε window as D3039 — so a European coupon report that cites "EN 2561" without noting the AMS modification is not directly comparable to a D3039 report, and one that does note it usually is.

The two standards also disagree on what they are allowed to test. EN 2561 is scoped to 0° unidirectional carbon plate only (its Specimen Type B); ASTM D3039 covers balanced-symmetric laminates and woven fabric as well (DOT/FAA/AR-04/24 §2.1, Table 1–3, p.4–11). A woven fabric tested under EN 2561 falls outside that method's own scope — the appropriate standard for a woven carbon product is D3039. Two further differences move the number less dramatically but still measurably: D3039 recommends a 7° tab bevel where EN 2561 mandates a 90° square-cut tab, and D3039 specifies a spherical-contact micrometer for thickness (±1%, about ±2.5 µm) where EN 2561 uses a flat-anvil micrometer (0.01 mm) — on a textured woven surface the two tools do not read the same thickness, which propagates directly into the cross-sectional area used to calculate strength.

What does a "mean" strength number actually promise you?

Less than it sounds like. ACI 440.2R requires the design ultimate strength reported on a data sheet, ffu*, to be the mean of at least 20 test specimens minus three standard deviations (the standard's own shorthand: ffu* = mean − 3σ), which corresponds to 99.87% statistical certainty that a random specimen exceeds that value (ACI PRC-440.2-23 §4.3). The elastic modulus, by contrast, is reported as the plain mean of the same 20-plus specimens — it is not knocked down the same way, because modulus scatter has a different structural consequence than strength scatter.

The three-sigma knockdown exists because roughly half of any batch of coupons tests below the simple average — that is what an average is. A data sheet that quotes a bare mean as its "ultimate strength" is quietly handing the risk in that lower half to whoever designs against the number. When a supplier's headline tensile strength looks unusually high compared to a competitor's on the same fiber, the first question is not whether the fiber is better — it is whether both numbers were knocked down the same way.

How is a lap splice in wet lay-up carbon fiber actually tested?

With a dedicated method built for exactly that joint: ASTM D7616, sometimes described informally as a lap shear test for wet lay-up FRP, measures the apparent overlap splice shear strength where one length of saturated fabric overlaps another — the joint every wet lay-up installation relies on when a single roll will not span the wrap. At least five specimens are tested, with minimum width set by layup type (25 mm for unidirectional, 38 mm for cross-ply), and required specimen length scales with splice length per a fixed table — a 100 mm splice requires a 305 mm specimen, a 200 mm splice requires 410 mm (D7616/D7616M-11(2023) Table 1). Tabs are made to D3039, tying this method back to the tension standard above.

The number that comes out, apparent shear strength per unit width (V* = Pmax/w), is only meaningful if the specimen failed in the right place. D7616 recognizes failure inside the splice — delamination, adhesive failure, cohesive failure, or fiber tear within the overlap — as valid. Failure outside the splice — net-section rupture or longitudinal splitting — does not represent the splice's shear capacity and is not a valid result for that purpose. A test report worth trusting states where each of the five specimens actually failed, not just the average load.

What does a supplier's bond-strength claim for structural epoxy actually rest on?

Usually a slant-shear test, ASTM C882, which is the standard method behind the "bonds to concrete at X MPa" line on a structural adhesive data sheet. Two half-cylinders of mortar, cut at 30° to the vertical, are glued back together with the epoxy and loaded in compression; the 30° angle puts the bond line under combined shear and compression, closer to how a real bonded joint is loaded than a pure shear or pure pull test would be, and the resulting slant-shear bond area works out to twice the cylinder's cross-section (ASTM C882/C882M-13a §4.1, Note 2). Three composite specimens are tested per resin type.

The number only means what it appears to mean if the mortar itself is strong enough. C882's own Note 1 states that for a quantitative bond-strength result, the mortar's 28-day compressive strength must exceed the expected bond strength by at least 2.5 times; below that ratio the test can only confirm a qualitative pass/fail against roughly 13 MPa (1900 psi). And the report is required to state where the specimen failed — in the epoxy, in the mortar, or at the interface (§13.1.5). A well-formulated structural epoxy should force the parent mortar to fail first, meaning the adhesive itself was never the weak link in the test.

Why does the epoxy resin itself need six different test methods — and even more footnotes?

Because tensile strength, compressive strength, flexural strength, heat deflection temperature, shear strength, and water absorption are six physically different properties of the cured resin, each governed by its own ASTM method (D638, D695, D790, D648, D732, and D570 respectively) — and every one of those six methods has at least one condition that changes the reported number without changing the resin. The clearest example is heat deflection temperature (HDT) under ASTM D648: the standard defines two fiber-stress levels, 0.455 and 1.82 MPa, and two support-span configurations, Method A at 101.6 mm and Method B at 100 mm. Switching span method alone, on the same cured material, changes the measured HDT by 1.0 to 4.5 °C (D648-18 Note 16, Table 3). A data sheet that lists "HDT 65 °C" without naming the stress level and span method has not actually specified anything comparable to a competitor's "HDT 65 °C" — and HDT is only one of two temperature figures worth checking on an epoxy sheet; the resin's glass transition temperature, Tg, is the other, and it needs its own test-condition footnote too.

Flexural testing under ASTM D790 carries a similar set of dials: Procedure A runs at a 0.01/min strain rate, Procedure B at 0.10/min (used only when Procedure A does not produce a failure within the standard's 5% strain limit); the default span-to-depth ratio is 16:1, but the standard allows — and for laminated or highly orthotropic materials, expects — ratios up to 32:1, 40:1, or 60:1 (D790-17 §4.2–4.6, §7.6). Tensile testing under D638 has its own single-point ambiguity: the standard requires the reported tensile strength to be stated as either "at yield" or "at break" — a resin with a yield point can legitimately report two different numbers depending which one is chosen (D638-22 §11.2). Water absorption under D570 has the widest spread of all: five-plus immersion conditions — 24 h, 2 h, long-term to saturation, 2 h boiling, 30 min boiling, or 50 °C for 48 h — can report entirely different percentages for the same material, and the time to reach true saturation scales with the square of specimen thickness (D570-22 Table 1: nylon-6 needs about 100 hours to saturate at 1 mm thickness but roughly 62,000 hours — about seven years — at 25 mm). And the shear number from ASTM D732 deserves its own caution: the standard itself states that punch shear strength is calculated as failure load divided by thickness times punch perimeter, and is explicitly not a thickness-independent material property — it cannot be compared to a metal lap-shear number from ASTM D1002 or to the slant-shear bond number from C882 above, because all three are measuring different things by different geometry.

Does the coupon in the lab actually predict what happens on site?

Only if the coupon itself was made correctly — specimen preparation is not a formality, it is where a surprising amount of test data goes wrong before the load is ever applied. Round-robin testing across ten industry labs found that the dominant cause of invalid composite test specimens was tab adhesive failure and dimensional or hole-position deviation, not the cutting method used to shape the coupon (NPL GPG38 §10.5). In one documented case, an entire dataset was invalidated because the tab adhesive itself failed before the composite did.

Tab material and geometry are chosen for a reason, and the reasons are counter-intuitive in places. G-10 glass/epoxy is the preferred tab material precisely because it is less stiff than the carbon laminate being tested (about 32.6 GPa versus 142 GPa for a carbon/epoxy tab) — a lower-stiffness tab produces less stress concentration at the tab tip, so a carbon-fiber tab on a carbon-fiber coupon is not the better choice it sounds like (FAA AR-02/106 §4.1, Table 5). Tension and compression testing then call for opposite tab geometry: tension specimens use a beveled tab at 10°–15° to spread the load transfer, while compression specimens use an unbeveled, square (90°) tab, because a beveled tab increases the unsupported gauge length and can trigger buckling — measured compressive strength for a 30°-beveled GFRP tab was 1255 MPa versus 1517 MPa for the same setup with a square tab (FAA AR-02/106 §5.4, Table 6). Cutting method matters too: a diamond-tipped saw with 5% oil-based coolant at roughly 3500 rpm and 700 mm/min is the industry-standard approach, while laser cutting of cured carbon laminate creates a heat-affected zone of 6–7 mm when cutting perpendicular to the fiber direction — a real change in local resin state right where the strength is supposed to be measured (NPL GPG38 §3.2.4, Table 3.4).

This level of process discipline — and whether a supplier's lab actually follows it — is exactly what to probe when qualifying a carbon fiber fabric manufacturer: ask not just for the strength number, but for the specimen count, the tab material, and where the failures occurred.

A coupon passed in the lab — does that guarantee the bond on the wall is good?

No, and ACI 440.2R's construction specification, ACI SPEC-440.12-22, exists specifically because a lab coupon and a field installation are not the same test. Once a wet lay-up system is installed as a bond-critical application, the installation itself has to be pull-off tested on site under ASTM D7522, and the acceptance rule requires both conditions at once: the pull-off strength must exceed 200 psi (≈1.4 MPa), and the failure must occur inside the concrete substrate itself — Mode G in ASTM D7522's failure classification — not at the bond line (ACI SPEC-440.12-22 §3.5.2.5). A high pull-off number that fails at the interface instead of in the concrete does not automatically pass; the engineer of record has to make that call. Minimum testing frequency is three pulls per 1000 ft² (≈93 m²) of installed area, with at least three per day regardless of area (§3.5.2.2).

Where witness-panel testing is specified, ACI SPEC-440.12-22 calls for one 12 in. × 12 in. panel per 5000 ft² (≈465 m²) of installed material, a minimum of two panels per day, with at least 15% of all panels sent for lab testing under ASTM D7565 (§3.5.5.2–3.5.5.4). A panel is judged deficient in strength if two or more of its five valid tensile results fall below the data sheet's rated ultimate force, or in stiffness if the average chord tensile stiffness falls below 90% of the data sheet's rated value (§3.5.5.7–3.5.5.8) — which is the direct link between the lab-derived data sheet numbers discussed above and whether a specific day's installation is accepted.

None of this is unique to carbon fiber over concrete — the same discipline applies to GFRP rebar, which is qualified by its own tensile method (ASTM D7205) using steel tube anchors grouted onto the bar ends rather than the adhesive tabs used for flat coupons — a different anchorage for a different specimen shape, with the same underlying question: is the reported strength coming from the fiber, or from how the specimen was held? The common thread across every method in this article is the same: know which standard produced the number, know which option inside that standard was selected, and only then decide whether two numbers are actually the same measurement.

FAQ

Is ASTM D3039 or EN 2561 the correct standard for testing carbon fiber fabric?

ASTM D3039, for woven or balanced-symmetric fabric layups. EN 2561 is scoped only to 0° unidirectional carbon plate (its Specimen Type B) and does not cover fabric. A tensile report for carbon fiber fabric citing EN 2561 without stating why is outside that method's own defined scope.

Why does ACI 440.2R require at least 20 test specimens?

Because the design strength, ffu*, is calculated as the mean minus three standard deviations of the sample — a statistic that only converges to a reliable 99.87% guarantee with an adequately large sample. Fewer specimens produce a less reliable estimate of both the mean and the standard deviation feeding into that calculation (ACI PRC-440.2-23 §4.3).

What is a lap shear test in the context of CFRP strengthening?

In wet lay-up CFRP, it commonly refers to ASTM D7616, which measures the apparent shear strength of an overlap splice where two lengths of saturated fabric join. The result is only valid if the specimen fails inside the splice itself — delamination, adhesive, or fiber-tear failure — rather than in the net section outside it.

Can a high pull-off strength on site still fail acceptance?

Yes. ACI SPEC-440.12-22 requires both a pull-off strength above 200 psi and a failure mode inside the concrete substrate (Mode G per ASTM D7522). A result above 200 psi that instead fails at the bond line does not automatically pass and is subject to the engineer of record's judgment.

Why do two structural epoxy data sheets show different HDT values for what looks like the same resin?

Because ASTM D648 defines two fiber-stress levels (0.455 and 1.82 MPa) and two support-span methods (101.6 mm and 100 mm), and switching the span method alone can shift the measured HDT by 1.0 to 4.5 °C on the same cured material. An HDT number without both variables stated is not a complete specification.

Does testing a resin's shear strength with ASTM D732 give the same number as a bond test like ASTM C882?

No. D732 punch shear strength is calculated from load divided by thickness times punch perimeter and the standard itself states it is not an independent material property. ASTM C882 measures adhesive-to-concrete bond strength under combined shear and compression at a 30° slant plane. The two numbers describe different physical tests and cannot be substituted for one another.