ISO 6892-2
Tensile Testing of Metals at Elevated Temperature

Written by Chris Thomas, Universal Grip Company · Reviewed by Frank Bertini · Updated October 24, 2026

ISO 6892-2 is the international standard for tensile testing of metallic materials at elevated temperature, meaning anything above the 35 °C upper limit of the room-temperature standard ISO 6892-1. It reuses the Part 1 definitions, test pieces and calculations and adds what heat changes: how closely the test piece temperature must be held and measured, how many thermocouples must touch it, how long it must soak before loading, how the extensometer gauge length is handled when the piece has expanded, which properties are determined hot and which at room temperature after fracture, and strain rates that are slower after yield than at room temperature because hot metals are more rate-sensitive. Turbine discs and blades, boiler and pressure-vessel steels, exhaust alloys, fasteners for hot service and any alloy with a design curve against temperature are tested to it.

The current edition is ISO 6892-2:2026, published in August 2026, which replaced the 2018 edition; most equipment and laboratory pages still cite 2018. This guide covers the scope and what the 2026 revision changed, the temperature tolerance table in text, the sensor and verification rules, soak time and the small-load-during-heating practice, the four ways of handling the extensometer gauge length, Methods A and B and their rates, a worked example on Inconel 718 at 650 °C with the arithmetic shown, the hot load train (why wedge grips are out and threaded or button-head pull rods are in), extensometer choices through a furnace or chamber window, a numeric comparison with ASTM E21, and the chambers, grips, extensometers and machines Universal Grip supplies.

In this guide: What ISO 6892-2 Covers and How It Relates to Part 1 · Editions: ISO 783, 2011, 2018 and the 2026 Revision · Temperature: Tolerances, Sensors and Verification · Heating, Soaking and the Small-Load Rule · Test Rates: Method A and Method B at Temperature · Test Pieces and the Hot Load Train · Step-by-Step Procedure · Worked Example: Inconel 718 at 650 °C · Common Errors in Hot Tensile Testing · ISO 6892-2 vs ASTM E21 · Chambers, Grips, Extensometers and Machines for ISO 6892-2 · FAQ

What ISO 6892-2 Covers and How It Relates to Part 1

ISO 6892-2 specifies the tensile test for metals at temperatures higher than room temperature, that is above 35 °C, with no fixed upper limit; the tolerance table runs to 1,100 °C and anything above that is by agreement. The normative references are ISO 6892-1:2019 for the test itself, ISO 7500-1 for force calibration and ISO 9513 for extensometers. Everything not restated in Part 2 is taken from Part 1: the symbols, the proportional gauge length L0 = 5.65 √S0, the specimen annexes, the fit-together elongation measurement and the rounding rules. The differences are deliberate and few:

• Dimensions are room-temperature values. L0 is marked and measured before heating, S0 is the cold cross-section, and every stress R is force divided by the cold S0. The one exception is the extensometer gauge length Le, which has its own rules (below).

• Properties normally determined hot: proof strength Rp (usually Rp0.2), upper and lower yield ReH and ReL where the material shows them, and tensile strength Rm. Determined at room temperature after fracture: percentage elongation A (fit-together, L0/3 fracture rule applies) and reduction of area Z from the cold Su.

• Not normally determined unless specified or agreed: permanent set strength Rr, yield-point extension Ae, plastic and total extension at maximum force Ag and Agt, and total extension at fracture At, because hot extensometry to fracture is difficult and the values are rate- and time-dependent.

• Added symbols: T the specified temperature, Ti the indicated surface temperature of the parallel length, and ts the soaking time in minutes, all of which go in the report.

The companion parts are ISO 6892-3 for +10 down to −196 °C and ISO 6892-4 for liquid helium; the room-temperature standard is covered in our ISO 6892-1 guide and the ASTM counterpart in the ASTM E8 guide. Our metals testing page gives the lab overview.

Editions: ISO 783, 2011, 2018 and the 2026 Revision

The elevated-temperature method began as ISO 783:1999. ISO 6892-2:2011 was the first edition under the 6892 number, written to match the 2009 strain-rate approach of Part 1. The 2018 second edition was a minor update that added the note on holding a small tensile load during heating and tidied the cross-references. The third edition, ISO 6892-2:2026, was published on 3 August 2026 after a 2023 to 2026 development cycle and is adopted in Europe as EN ISO 6892-2:2026; the 2018 edition is withdrawn. According to its foreword the 2026 edition revised the heating-device and temperature clause (9.3), reworded the Method A and Method B rate clauses (10.3 and 10.4), revised the test-report clause (12), reworded the informative annex on test-piece geometries and updated every reference to the 2019 edition of Part 1. The numeric requirements below follow the published text where we have it and the 2018 values where the 2026 wording was editorial; a laboratory certifying to the standard should work from its own copy of the 2026 edition.

The introduction to the standard also notes that the stress-rate control of Part 1 Method B is not offered at elevated temperature, and that the committee may delete the expanded strain-rate ranges of Part 2 Method B in a future edition, so new procedures should be written around Method A.

Temperature: Tolerances, Sensors and Verification

The heating device, which may be a resistance furnace, an environmental chamber, induction heating or a fluid bath, must hold the indicated surface temperature Ti of the parallel length within a permitted deviation from the specified temperature T, and the variation along the parallel length must not exceed the same figure:

• T up to 600 °C: Ti − T within ± 3 °C; variation along the piece not more than 3 °C.

• 600 to 800 °C: ± 4 °C; variation not more than 4 °C.

• 800 to 1,000 °C: ± 5 °C; variation not more than 5 °C.

• 1,000 to 1,100 °C: ± 6 °C; variation not more than 6 °C.

• Above 1,100 °C: by agreement between the parties, and the agreed tolerance must be reported.

Sensors. Thermocouples are placed on the parallel length: for an extensometer gauge length Le under 50 mm, one at each end of the parallel length; for Le of 50 mm or more, a third at the centre. The number may be reduced where experience with the furnace shows the variation stays inside the limits, but at least one sensor must be in contact with the test piece, with good thermal contact (tied or spot-welded, never resting loosely) and shielded from direct radiation from the heating elements. The usual types are K or N to about 1,000 °C and R or S above that.

Verification. The temperature measuring system must resolve 1 °C and be accurate to ± 0.004 × T or ± 2 °C, whichever is larger: ± 2 °C up to 500 °C, ± 2.6 °C at 650 °C, ± 3.2 °C at 800 °C and ± 4 °C at 1,000 °C. Calibration must be traceable to SI and repeated at intervals of not more than one year, with the errors recorded. Because the sensor error eats into the ± 3 to ± 6 °C tolerance, a three-zone furnace with independent control of each zone is the normal way to keep both the deviation and the along-piece variation inside the band on long pieces.

Heating, Soaking and the Small-Load Rule

• Zero the force after the hot load string is assembled and before the piece is gripped, and do not re-zero during the test.

• Heat under a small tensile load. As the piece and pull rods expand they would otherwise push the load string into compression; running the machine in force control at a few tens of newtons during heating keeps the piece straight and the thermocouples in contact. This note was added in 2018 and is standard practice.

• Soak. Once Ti is inside the tolerance band the piece is held at temperature for at least 10 minutes before loading, longer for thick pieces or furnaces with a large thermal mass, so that the core reaches the surface temperature; a short soak leaves a cooler core and reads Rp high. The temperature must not overshoot above the band during heating, because a brief excursion can temper, age or recover the microstructure. The heating time and the soaking time ts are both reported.

• Extensometer gauge length Le. Le must be at least 10 mm and centred on the parallel length. Because the piece expands, the standard gives four ways to handle it and requires the report to say which was used: Method 1, Le measured at room temperature and used as is; Method 2a, the nominal Le set at the test temperature; Method 2b, a reduced Le set at room temperature so that it reaches nominal when hot; Method 2c, the room-temperature Le corrected by the thermal expansion to the test temperature. The parts of the extensometer outside the furnace are shielded from draughts, and the device must be ISO 9513 class 1 or better for Rp and Rt.

Test Rates: Method A and Method B at Temperature

Method A is strain-rate control, as in Part 1, with a tolerance of ± 20 %. For ReH, Rp and Rt the rate is 0.00007 s⁻¹ or 0.00025 s⁻¹, under closed-loop extensometer control (A1) or estimated from the crosshead speed over the parallel length (A2, vc = Lc × ė). After the proof strength is passed the rate for Rm is raised, but to a lower value than the 0.0067 s⁻¹ of Part 1 Range 4, because the rate sensitivity of strength is larger hot; in the editions we have followed the post-yield rate is 0.0014 s⁻¹ under crosshead control. Higher rates may be used for a specific purpose, such as matching a room-temperature data set at the same rate, and must then be reported.

Method B keeps the expanded strain-rate ranges inherited from ISO 783: a slow band of the order of 0.001 to 0.005 per minute for ReH and Rp, and a faster band of the order of 0.02 to 0.2 per minute for Rm, with A and Z measured as at room temperature. It exists for specifications written to the old method and is flagged for possible deletion. There is no stress-rate method at elevated temperature. The method, the rate or rates, the Le method, T, Ti, ts and the heating time are all mandatory in the report, and a result without them cannot be compared with another laboratory's.

Rate matters more hot than cold. Creep and recovery operate during the test, so Rp0.2 of a nickel superalloy at 650 °C can move by several percent between 0.00007 and 0.00025 s⁻¹, and the stress-strain curve of a steel at 500 °C can show serrated flow whose amplitude depends on rate. The ± 20 % band is tight by design, and a machine that cannot hold it under extensometer control at 0.00007 s⁻¹ should run A2 with an allowance for the compliance of the long hot load string.

Test Pieces and the Hot Load Train

The test pieces are those of ISO 6892-1 Annexes B and D, proportional with k = 5.65 or non-proportional, but the geometry is constrained by the furnace: wedge grips do not fit inside a furnace, the jaws would creep at temperature, and bite marks initiate fracture in hot alloys. The informative annex therefore shows three furnace-compatible geometries, and almost every hot tensile laboratory uses one of them:

• Round with threaded ends: M10 to M20 threads on 5 to 14 mm diameter pieces, screwed into threaded pull rods with at least one diameter of engagement. The default for bar, forgings, castings and superalloys; self-aligning when the threads are concentric with the gauge section, and the only geometry that cannot slip.

• Round with shoulders (button-head): the piece is held in split collets inside a shoulder grip; used where threads cannot be machined (hard or brittle alloys, small diameters) and in automated test cells. Alignment depends on the collet seats.

• Flat with locking-pin holes: sheet pieces with a hole in each tab, loaded through pins in slotted pull rods; the hole-to-edge distance must be large enough that the tab does not tear out at temperature.

Pull rods and adaptors. The load train inside the furnace is made of a material that keeps its strength and does not creep at the test temperature: stainless steel to about 600 °C, Nimonic, Inconel or similar nickel superalloys to about 1,000 °C, and cobalt superalloys or ceramics above that. The rods pass through the furnace end plates to water-cooled adaptors that protect the load cell and the machine's standard couplings; spherical seats or alignment adapters outside the furnace take out bending, which is verified at room temperature with a strain-gauged piece. The furnace bore must clear the pull-rod diameter and the extensometer rods, and the thermocouple leads need a path that does not short against the heating elements.

Step-by-Step Procedure

• 1. Prepare the piece to ISO 6892-1 Annex B or D with threaded, shouldered or pinned ends; measure S0 and mark L0 at room temperature; record d0 or b0 and a0.

• 2. Assemble the hot load train, zero the force, fit the piece, attach two (Le < 50 mm) or three (Le ≥ 50 mm) thermocouples on the parallel length, fit the extensometer and set Le by the chosen method (1, 2a, 2b or 2c).

• 3. Close the furnace and heat under a small tensile load in force control, without overshooting the tolerance band.

• 4. Soak for at least 10 minutes after all thermocouples are inside the band; record the heating time and ts.

• 5. Load under Method A: 0.00025 s⁻¹ (or 0.00007 s⁻¹) under extensometer control to beyond Rp0.2, then the post-yield rate under crosshead control to fracture, removing a contact extensometer if it cannot survive fracture.

• 6. Record Ti throughout; a reading outside the band during the determination of Rp or Rm invalidates the test.

• 7. Cool and remove the halves, fit them together, measure Lu to 0.25 mm and the minimum section for Su at room temperature.

• 8. Calculate Rp0.2, Rm, A and Z with the Part 1 formulas and rounding; check the L0/3 fracture rule for A.

• 9. Report the standard and edition, T, Ti, ts, heating time, method A or B with rates, extensometer class and Le method, test-piece geometry and k, and the fracture position.

Worked Example: Inconel 718 at 650 °C

A solution-treated and aged Inconel 718 bar is tested at 650 °C on a threaded-end round. d0 = 6.00 mm, so S0 = π × 6² ÷ 4 = 28.27 mm². L0 = 5.65 × √28.27 = 5.65 × 5.317 = 30.04 mm, marked as 30 mm; Lc = 36 mm (at least L0 + d0/2 = 33 mm); Le = 25 mm, a high-temperature rod extensometer set by Method 2a at the test temperature. Because Le is under 50 mm, two Type K thermocouples are tied at the ends of the parallel length. The tolerance band for 650 °C is ± 4 °C and the sensor system must be accurate to ± 0.004 × 650 = ± 2.6 °C. Heating takes 25 minutes under a 50 N load, soak ts = 15 minutes, Ti stays between 648 and 652 °C.

• Rate to Rp0.2: Method A1 at 0.00025 s⁻¹ on Le = 25 mm, an extension rate of 25 × 0.00025 = 0.00625 mm/s, or 0.375 mm/min.

• Rate after yield: 0.0014 s⁻¹ over Lc = 36 mm, a crosshead speed of 36 × 0.0014 = 0.0504 mm/s, or 3.0 mm/min.

• Rp0.2: force at 0.2 % plastic offset 27.2 kN → 27,200 ÷ 28.27 = 962.2 → 962 MPa.

• Rm: maximum force 31.4 kN → 31,400 ÷ 28.27 = 1,110.7 → 1,111 MPa.

• A (room temperature): Lu = 35.4 mm → (35.4 − 30) ÷ 30 × 100 = 18.0 %; fracture 12 mm from the nearer mark, more than L0/3 = 10 mm, so valid.

• Z (room temperature): du = 4.95 mm → Su = π × 4.95² ÷ 4 = 19.24 mm² → (28.27 − 19.24) ÷ 28.27 × 100 = 31.9 → 32 %.

• Capacity: 31.4 kN at temperature, but the same bar at room temperature would reach about 1,400 MPa × 28.27 mm² = 40 kN; size the frame and the pull rods for the cold strength of the largest piece, here a 100 kN machine with a 50 kN load cell in its class 1 range.

The report then reads: ISO 6892-2:2026, Method A1, 0.00025 s⁻¹ to Rp0.2 then 0.0014 s⁻¹, T = 650 °C, Ti 648 to 652 °C, ts = 15 min, heating 25 min, Le = 25 mm by Method 2a, class 1 extensometer, round d0 6 mm threaded, k = 5.65, Rp0.2 962 MPa, Rm 1,111 MPa, A 18.0 %, Z 32 %.

Common Errors in Hot Tensile Testing

• Citing the wrong edition. ISO 6892-2:2018 is withdrawn; test reports and procedures should reference the 2026 edition.

• Thermocouples not touching the piece. A sensor reading furnace air can be 10 to 30 °C from the piece surface, far outside the ± 3 to ± 6 °C band; at least one sensor must be in contact and all must be shielded from the elements.

• Short soak. Loading as soon as the surface sensor reads T leaves the core cooler and reads Rp high by a few percent on thick pieces; 10 minutes is the minimum, not the target.

• Overshoot during heating. A 20 °C excursion above an ageing temperature changes the precipitate structure of a superalloy or tempers a steel; a controller tuned for the cold furnace will overshoot a hot one.

• Unreported Le method. The strain at Rp0.2 differs by the thermal-expansion factor (about 1 % at 650 °C for a nickel alloy) between Method 1 and Methods 2a to 2c; it is small on Rp0.2 but must be declared.

• Wedge grips or bite-marked ends inside the furnace, which creep, slip at 0.00007 s⁻¹ or start the fracture at the jaw; use threaded or shouldered pieces.

• Compressive preload from expansion when the load string is not held in force control during heating, which buckles thin flats and lifts thermocouples off the surface.

• Post-yield rate copied from Part 1. 0.0067 s⁻¹ at temperature raises Rm of a rate-sensitive alloy by several percent against the Part 2 rate; the method and rates must be reported so the data can be compared.

• Pull-rod material at its limit, stretching or creeping during the test and taking the strain rate out of band under A2 control; superalloy rods above about 600 °C.

ISO 6892-2 vs ASTM E21

ASTM E21 is the corresponding American method and the two are similar but not equivalent; a laboratory with a furnace and hot load train can run both, with these differences:

• Threshold: ISO above 35 °C; ASTM above 38 °C (100 °F).

• Temperature tolerance: ISO ± 3 °C to 600 °C, ± 4 to 800 °C, ± 5 to 1,000 °C, ± 6 to 1,100 °C, with the same limit on variation along the piece; ASTM ± 3 °C (± 5 °F) up to 980 °C and ± 6 °C (± 10 °F) above.

• Sensors: ISO two thermocouples for Le under 50 mm, three for 50 mm and over, at least one in contact, accuracy ± 0.004 × T or ± 2 °C, calibrated within a year; ASTM requires sensors on the gauge section with calibration and checks per E220.

• Yield rate: ISO 0.00007 or 0.00025 s⁻¹ ± 20 %; ASTM 0.005 ± 0.002 min⁻¹, which is 0.000083 s⁻¹ ± 40 %.

• Post-yield rate: ISO 0.0014 s⁻¹ under crosshead control; ASTM 0.05 ± 0.01 min⁻¹ (0.00083 s⁻¹), typically set as 0.05 times the reduced parallel length per minute.

• Extensometer: ISO 9513 class 1 with the four Le methods; ASTM E83 class B-2 and a modulus check within 10 % of nominal as evidence of alignment and temperature.

• Alignment: ISO through the Part 1 references; ASTM bending not more than 10 % of axial strain, verified per E1012 at room temperature.

• Elongation and reduction of area: both at room temperature after fracture; ISO on 5.65 √S0 or 50/80 mm gauge lengths to 0.25 mm, ASTM on 4 d0 or 2 in to 0.01 in (0.3 mm).

As with ISO 6892-1 and ASTM E8, the rates are the biggest source of difference between the two sets of hot data; a report that states the standard, method and rate lets the two be compared.

Chambers, Grips, Extensometers and Machines for ISO 6892-2

Heating. For tests to about 300 °C an environmental chamber around the standard load string is the practical choice: it keeps the room-temperature grips (threaded or shoulder grips in the chamber, with extension rods through the ports) and allows a video extensometer to read through the window. Above that a tube furnace on its own stand, with pull rods through the end plates and water-cooled adaptors, holds the ± 3 to ± 6 °C bands to 1,100 °C. Our chamber buyer's guide and chambers for universal testing machines posts cover the choice between the two.

• Grips and load train: bolt and threaded grips and threaded pull rods for the M10 to M20 threaded-end round, the first choice for every hot test; shoulder grips with split collets for button-head pieces; pinned clevises for flat pieces with locking-pin holes. Rods and grips in the hot zone are made in stainless steel, Inconel or Nimonic according to the temperature, with water-cooled couplings and adapters to the machine's standard connection.

• Extensometers: high-temperature contact extensometers with ceramic or alumina rods that reach into the furnace and keep the transducer outside, ISO 9513 class 1, Le 10 to 50 mm; or a video extensometer reading through a chamber or furnace window, which has no rods to seat on the piece, no mass to disturb the thermocouples, handles Method 2a or 2c gauge-length setting directly and measures to fracture. A clip-on extensometer is suitable only in a chamber at modest temperature within its rating.

• Thermocouple handling: fine-wire Type K or N couples tied to the parallel length with ceramic-insulated wire, radiation shields, and a three-zone furnace controller that lets each zone be trimmed to equalise the along-piece variation.

• Machines: the 100 kN benchtop machine with a furnace stand covers rounds to 10 mm in most alloys; the 200 kN floor-standing machine gives the daylight for a long furnace plus cooled rods and covers 14 mm rounds and sheet of the strongest alloys. Both run closed-loop strain control from the extensometer channel for Method A1 at 0.00007 s⁻¹ and force control for the heating phase.

A hot tensile system is specified from four numbers: the maximum temperature (which sets the furnace and rod material), the largest test piece at its room-temperature strength (which sets the frame and cell), the gauge length and whether strain to fracture is wanted (which sets the extensometer), and the longest piece (which sets the furnace length and the number of zones). Give us those and we will quote a complete, aligned load train.

ISO 6892-2 FAQ

What is the current edition of ISO 6892-2?

ISO 6892-2:2026, published 3 August 2026 and adopted as EN ISO 6892-2:2026; it replaces ISO 6892-2:2018, which is withdrawn. The revision reworked the temperature clause, the Method A and B rate wording, the test report and the references to ISO 6892-1:2019.

What temperature range does ISO 6892-2 cover?

Any temperature above 35 °C, the upper limit of the ISO 6892-1 room-temperature band. The tolerance table runs to 1,100 °C; above that the tolerance is agreed between the parties and reported.

What are the ISO 6892-2 temperature tolerances?

Indicated surface temperature within ± 3 °C of the specified temperature up to 600 °C, ± 4 °C from 600 to 800 °C, ± 5 °C from 800 to 1,000 °C and ± 6 °C from 1,000 to 1,100 °C, with the variation along the parallel length limited to the same figure.

How many thermocouples does ISO 6892-2 require?

One at each end of the parallel length when the extensometer gauge length is under 50 mm, plus a third at the centre when it is 50 mm or more. The number may be reduced with documented furnace experience, but at least one sensor must be in contact with the test piece and all must be shielded from direct radiation.

How accurate must the temperature measurement be?

Resolution of 1 °C and accuracy of ± 0.004 × T or ± 2 °C, whichever is greater: ± 2 °C to 500 °C, ± 2.6 °C at 650 °C, ± 3.2 °C at 800 °C, ± 4 °C at 1,000 °C. Calibration must be traceable and repeated at intervals of not more than one year.

How long is the soak time in ISO 6892-2?

At least 10 minutes after the test piece is inside the tolerance band, longer for thick pieces so that the core reaches the surface temperature. The temperature must not overshoot the band during heating, and both the heating time and the soaking time ts are reported.

What strain rates does ISO 6892-2 use?

Method A: 0.00007 s⁻¹ or 0.00025 s⁻¹ ± 20 % for ReH, Rp and Rt under extensometer (A1) or crosshead (A2) control, then a lower post-yield rate than at room temperature, 0.0014 s⁻¹ under crosshead control, for Rm. Method B keeps slower expanded ranges for older specifications. There is no stress-rate method at temperature.

Which properties are measured hot and which at room temperature?

Rp0.2, ReH, ReL and Rm are determined at the test temperature; percentage elongation A and reduction of area Z are measured at room temperature after fracture by fitting the halves together. Ag, Agt, At, Ae and Rr are not normally determined hot unless specified.

What is the extensometer requirement in ISO 6892-2?

ISO 9513 class 1 or better for proof strength, gauge length at least 10 mm centred on the parallel length, with the parts outside the furnace shielded from draughts. The gauge length may be handled by Method 1 (room-temperature Le), 2a (nominal Le set hot), 2b (reduced Le set cold) or 2c (cold Le corrected for expansion), and the method used must be reported.

Why are wedge grips not used for ISO 6892-2?

They do not fit inside a furnace, their jaws creep and lose clamping force at temperature, and the bite marks start fractures in hot alloys. Test pieces are made with threaded ends, shoulders for split collets or pin holes, and loaded through superalloy pull rods with water-cooled couplings outside the furnace.

What is the difference between ISO 6892-2 and ASTM E21?

ISO applies above 35 °C with ± 3 to ± 6 °C tolerances by band and rates of 0.00007 or 0.00025 s⁻¹ ± 20 % then 0.0014 s⁻¹; ASTM E21 applies above 38 °C with ± 3 °C to 980 °C and ± 6 °C above, rates of 0.005 ± 0.002 min⁻¹ to yield then 0.05 ± 0.01 min⁻¹, E83 class B-2 extensometers and a 10 % bending limit per E1012. Both measure elongation and reduction of area at room temperature.

What machine capacity is needed for hot tensile testing?

Size for the room-temperature strength of the largest piece, not the hot strength: a 6 mm Inconel 718 round reaches about 31 kN at 650 °C but 40 kN cold, so a 100 kN machine with a 50 kN cell; a 10 mm round of the same alloy needs 110 kN cold, a 200 kN floor-standing machine. The pull rods must carry the same force at temperature without creeping.

Need a temperature chamber, hot pull rods, threaded or shoulder grips or a video extensometer for ISO 6892-2? Contact Universal Grip with your temperature range, specimen geometry and expected strength and we will quote the right setup.