Larson–Miller Parameter

Creep rupture testing has an arithmetic problem that no amount of money solves: a component is designed for 100,000 hours, which is eleven and a half years, and nobody is going to run the test. Larson and Miller's 1952 paper in the Transactions of the ASME offered the standard way around it. Their observation was that the rupture data for a given alloy at a given stress, taken over many temperatures and times, collapses onto a single curve when plotted against the combination P = T(C + log₁₀ t). Run a shorter test hotter, get the same P, and you have measured the same point on the curve.

The form is not arbitrary. If rupture is controlled by a thermally activated process, the rupture time follows an Arrhenius law, t = t₀exp(Q/RT). Take logarithms and rearrange and you get Q/(2.303R) = T(log t − log t₀), which is exactly the Larson–Miller form with C = −log t₀. So the parameter is an Arrhenius relation with the activation energy hidden inside P, and C is the logarithm of the pre-exponential factor with a sign change. That is also why C tends to land near 20 for steels: it corresponds to a t₀ of about 10⁻²⁰ hours, which is a reasonable atomic attempt frequency.

Now the two things that make this page a trap, and both are units.

The temperature is ABSOLUTE. Kelvin or Rankine, never Celsius or Fahrenheit. T multiplies the entire bracket, so a Celsius value is not a small error — at 800 °C, using 800 instead of 1073.15 makes P a quarter too small and lands you on a different part of the master curve entirely. This site pins every creep temperature's display to kelvin for exactly that reason.

The logarithm is of a time in HOURS. This is not a free choice and it is not a convention this site invented. Every published master curve, every code table and every vendor datasheet was built with the rupture time counted in hours, and C was fitted against that. Feed the same test in minutes and log₁₀t rises by 1.78, so P rises by 1.78T — about 1900 units at 1073 K — which is a very large move on a curve whose useful range is a few thousand units wide. Feed it in seconds and it is worse. The number will look entirely reasonable, which is the danger.

And now the honest limitation, which is bigger than either: THIS PAGE CANNOT GIVE YOU A RUPTURE STRESS. The Larson–Miller method has two halves. The first half is the parameter, which is arithmetic anyone can do and which this page does. The second half is the alloy's master curve — the measured relationship between applied stress and P — and that curve is not derivable from anything. It is proprietary experimental data, accumulated over decades of long-duration rupture testing on specific heats of specific alloys, and it is published in code tables (ASME Section II Part D), vendor datasheets, and the assessed collections held by NIMS in Japan and ECCC in Europe. Without your alloy's curve, P is a coordinate with no map. Any site that offers to convert a Larson–Miller parameter into a rupture stress without asking which alloy you have is inventing the answer.

C itself is a fitted constant belonging to the alloy AND the dataset. Twenty is a common default for steels and it is a default, not a law — values from about 15 to 25 are all in current use, and the right way to get one is to run rupture tests at several temperatures and stresses and choose the C that makes all the curves collapse best. Using someone else's C with your own rupture data will not collapse anything, and using a C fitted for one alloy family on another is the second most common error on this subject after the temperature.

Finally, remember what the whole method is: an extrapolation. The master curve was fitted to tests of thousands of hours and is being used to predict hundreds of thousands. That assumes no new damage mechanism appears in the interval. Long-term creep programmes have repeatedly shown that assumption to be optimistic — sigma-phase precipitation, carbide coarsening, and creep cavitation at grain boundaries all arrive on timescales that no short accelerated test can reveal, and several creep-strength-enhanced ferritic steels have had their allowable stresses revised downward once genuinely long data arrived. Treat a long extrapolation as a hypothesis, and pair it with metallographic inspection of the real component.