Last verified 19 August 2026. Every candidate below has a property page with sourced values; nothing appears here that is not sourced there.
What decides it
Oxidation limit before strength. A heat-resisting grade is chosen first for the temperature at which its oxide layer stays adherent. Above that, the scale spalls, fresh metal oxidises, and the section thins in cycles. The maximum service temperature in the table is a continuous-service figure from the mill datasheet and it is not a design allowable; cyclic service is more demanding than continuous service at the same peak.
Strength at temperature, not at room temperature. A room temperature proof strength tells you almost nothing here, which is why the two candidates below are shown at 600 °C and why the four grades in the second table cannot be placed against them. A grade that is strong cold and weak hot is a common and expensive mistake in ducting that has to hold its own shape.
Thermal expansion, which is why ferritics dominate the automotive end. Austenitic grades expand roughly half again as much as ferritic ones for the same temperature rise. In a restrained assembly that expansion becomes strain every time the system heats and cools, and thermal fatigue, not oxidation, is what cracks an exhaust. This is the reason a stabilised ferritic beats a stronger austenitic below red heat — and our data does not contain the expansion coefficients that would let this page demonstrate it.
What the gas is doing. Sulphur, chlorides from waste streams, and condensate below the acid dew point in a flue are different problems from clean hot air, and they change the answer completely. Nothing on this page addresses them.
Candidates
| Grade | Class | Maximum service temperature in air | Proof strength Rp0.2 at 600 °C | Thermal conductivity |
|---|---|---|---|---|
| 310S | heat-resistant austenitic | 1050 °C · E3 | 82 MPa · E3 | 15 W/(m·K) · E3 |
| 253 MA | heat-resistant austenitic | 1150 °C · E3 | 140 MPa · E3 | 15 W/(m·K) · E3 |
Every figure carries its evidence level, and on this site that level is E3 throughout — real values from resolvable supplier documents, none of them verified against the standard they cite. What the levels mean, and what our data does not cover. These are screening figures. None of them is a design allowable.
Commonly considered here, and not ranked above
These grades come up in this application and are missing from the table on purpose. We hold a page for each, but not the property this application turns on, and ranking a material on a property we have not sourced is the one thing this site will not do.
| Grade | Why it comes up | What we would need to rank it |
|---|---|---|
| 441 | Stabilised ferritic, the mainstream automotive exhaust material below red heat, chosen for low thermal expansion and thermal fatigue resistance rather than strength | Any elevated temperature row at all. We hold room temperature minima only |
| 444 | Molybdenum-bearing stabilised ferritic used where condensate is acidic as well as hot | Elevated temperature strength and an oxidation limit |
| 321 | Titanium-stabilised austenitic, the standard answer where the duty cycles through the sensitisation range | Elevated temperature strength and a maximum service temperature |
| 347 | Niobium-stabilised austenitic, chosen over 321 where the section is welded and stays hot | Elevated temperature strength and a maximum service temperature |
What usually goes wrong
The failure mode that actually appears: thermal fatigue at restrained joints, and scale spalling that thins the section a cycle at a time rather than a year at a time.
The design is checked against a peak temperature and the system fails from cycling.
Thermal fatigue at restrained joints. A duct fixed at both ends grows when hot and cannot, so the strain goes into the material at the stiffest point — a flange, a bracket, a weld toe. The crack appears there after some number of cycles that no property table on this page predicts. The fix is almost always in the design of the restraint rather than in the grade.
Scale spalling, which is loss by the cycle. A protective oxide that survives at temperature can crack and shed on cooling, exposing bare metal that oxidises again on the next heat. A grade rated for continuous service at a given temperature can be consumed far faster in a duty that heats and cools daily, and the datasheet figure gives no warning of it.
Sensitisation on the way down. Non-stabilised austenitics passing through roughly the 500–800 °C range precipitate chromium carbides at grain boundaries, and the weld heat-affected zone loses corrosion resistance where it is most needed — which is the entire reason 321 and 347 exist and appear in the second table. Where the service condenses anything acidic, that matters more than the peak temperature.
The engineering judgement on this page carries no source
The property values are sourced and the reasoning around them is not. Which failure mode dominates, which grade is forgiving to fabricate, which screening number misleads — those are domain knowledge, and this page states them without a citation because we have not found one worth citing. They are the author’s judgement, held to the same standard as everything else here: if one of them is wrong, it is wrong in public and we will say so in the changelog.
Treat the table as evidence and the prose as an opinion from someone who has read the datasheets.
What this page cannot tell you
- Four of the six grades on this page are not ranked, because we hold no elevated temperature data for them. That is a gap in our data, not a judgement about the materials — and it is the honest state of the page.
- Thermal expansion coefficients, which this page argues are decisive, are not in our data layer for any grade.
- Creep and creep rupture. Above roughly 550 °C — a rule of thumb we have not sourced, not a threshold from a standard — time-dependent deformation governs design, and we hold none of it. It is a named gap on both candidate pages.
- Cyclic oxidation data. The maximum service temperatures quoted are continuous-service figures. Cyclic duty is more severe and we have no figures for it.
- Sulphidation, chloride attack and acid dew point corrosion in real flue gas, which routinely decide this selection.
- Whether either candidate suits your duty. Two grades with a temperature limit each is not a selection; it is the part of one we can evidence.
Take this further
If you are choosing for a real part, the useful output is not a grade — it is a record of why you chose it that survives the review in eighteen months. Build a decision record, or start from the selection process if the requirements are not settled yet.
Generated from data/applications/high-temperature-exhaust.yaml. Regenerate rather than editing this file by hand.