Creep Test Analysis — Creep Stages, Steady-State Rate and Burgers Model Fit
Paste the time and strain columns off your creep frame’s logger and get the creep curve, the strain rate against log time with primary, secondary and tertiary creep marked, the steady-state creep rate, and Burgers, power-law and logarithmic fits side by side with their standard errors — stage by stage for a stepped test. Nothing is uploaded.
What a creep test measures
A creep test holds a specimen at constant stress and records how it keeps deforming. Rock does not stop at the strain it reached on loading: it carries on, quickly at first and then slowly, and under a high enough stress it eventually speeds up again and fails. Those three phases are primary (transient) creep, where the rate falls; secondary (steady-state) creep, where it is roughly constant; and tertiary creep, where it accelerates towards rupture.
The numbers a report needs come out of those phases: the instantaneous strain on loading, the steady-state creep rate, how long primary creep lasted, and whether and when tertiary creep began. Behind them sit the constants of a rheological model, fitted to the curve, which is what a numerical model of a tunnel or a salt cavern takes as input.
Data
Creep tests are a Pro tool. The preview below is fully live — every option works and you can export the figure. Pro unlocks your data.
See pricingThis is the sample dataset; every row of it can be read and copied. The plot beside it is live — change any option and watch it redraw.
| # | Specimen(opt) | Stage(opt) | Time | Axial strain | Axial stress(opt) | Confining pressure(opt) | Lateral strain(opt) | Temperature°C(opt) | |
|---|---|---|---|---|---|---|---|---|---|
| 1 | |||||||||
| 2 | |||||||||
| 3 | |||||||||
| 4 |
0 rows of data
Loading the plotting engine…
Enter data on the left to plot it
SVG is vector — resize it in Illustrator or Inkscape without losing quality. PNG 4× is roughly 600 dpi at figure width.
Free downloads carry a small “made with strataplot.com” credit along the bottom edge. The plot on screen is never watermarked. Remove it
Options
This plot has no adjustable options yet.
Your data is processed entirely in this browser tab. Nothing is uploaded to a server, so confidential client data stays confidential.
How the stages are found
By the strain rate, not by eye on the strain curve. Secondary creep is the stretch where the strain rate is at its minimum and roughly constant (Brantut et al., 2013), so the stages are read off a plot of strain rate against log time, with the strain smoothed before it is differentiated, because a derivative of raw logger data is mostly noise. This page does that: at a hundred log-spaced times it takes the least-squares slope of the readings within a window a fixed number of log cycles wide (0.3 by default), and draws the result in its own panel.
The minimum of that curve marks secondary creep. Readings whose rate is within 20 % of the minimum form the secondary window, bracketed on the figure, and the steady-state rate is the least-squares slope of strain on time across that window — not the minimum of a noisy derivative. If the rate then leaves the band and keeps climbing, the page marks tertiary creep from that point. If the rate is still falling when the stage ends, there is no secondary creep in the record, and the page says the rate it gives is an upper bound. Both the band and the window are options, and you can type your own secondary window: the table marks it as yours.
The three models
The Burgers model — a Maxwell spring and dashpot in series with a Kelvin-Voigt unit — gives ε = σ/E_M + σt/η_M + (σ/E_K)(1 − e^(−E_K t/η_K)): an instantaneous strain, a delayed elastic strain that levels off, and a viscous flow that does not. Its four constants are the ones numerical codes ask for, so it is drawn by default whenever the stress column is filled in. The power law ε = ε₀ + A·tⁿ (Andrade’s form with a free exponent) and the logarithmic (Lomnitz) law ε = ε₀ + A·ln(1 + αt) describe transient creep with fewer constants and no claim about steady flow.
All three are fitted to every stage and compared in the table by r², RMSE and Akaike’s information criterion, and every constant carries a linearised standard error. A constant whose standard error is larger than itself is flagged: it means the record does not determine it — usually a stage too short to see the Kelvin unit finish, or too short to separate steady flow from the tail of primary creep.
Stepped (multi-stage) tests
Many laboratories raise the load in steps on one specimen rather than testing one specimen per stress. The page splits such a record into stages from a stage column, or, without one, wherever the stress column steps by more than 2 %. Each stage is then analysed on its own: time is measured from its load step, and strain and stress are taken as increments over the stage before, so a Burgers fit to stage three is a fit to the stress added at stage three. That is the stage-by-stage reading of a stepped test, and the figure footer says which reading is used. With several stages, the report section adds isochronous strains — total strain at 1, 10, 100 and 1000 time units after each step.
Logger files, units and checks
One row is one reading, in the order the logger wrote it. Time, strain and stress go in as recorded — hours, minutes, seconds or days; per cent, microstrain or dimensionless; MPa, kPa or psi — and the results come out in the same units, with moduli a thousand times larger (GPa for MPa). Files of up to 100 000 readings are fitted in full; only the drawing is thinned. A reading that goes back in time is reported and skipped rather than sorted, a stage whose stress wanders by more than ±1 % is flagged, and so is a temperature range wider than the drift threshold you set.
Common questions
- How do I calculate the steady-state creep rate?
- Find the stretch of the test where the strain rate is at its minimum and roughly constant, and take the slope of strain against time across it. This page finds that stretch from the strain rate against log time, brackets it on the figure, and reports the least-squares slope of every reading inside it. You can widen the band or type your own window.
- How do I get Burgers model parameters from creep test data?
- Fit ε = σ/E_M + σt/η_M + (σ/E_K)(1 − e^(−E_K t/η_K)) to the strain–time record at the stress applied. This page does it by least squares for every stage and prints E_M, E_K, η_K and η_M with standard errors. If the record shows no steady creep at all, η_M is reported as unbounded rather than as a large number the data cannot support.
- Is there a standard for creep testing of rock?
- ASTM D7070, the old standard for creep of rock core under constant stress and temperature, was withdrawn in 2025 with no replacement. This page follows the open literature instead: the stages from the minimum strain rate as in the review by Brantut et al. (2013), and the Burgers model as fitted by Wang & Li (2018). Every source the page cites is free to read online.
- Is this a Pro tool?
- Yes. The page, the sample tests, every option and the figure export work for everybody. A licence lets you replace the sample with your own readings.
- Does my data leave the browser?
- No. The fits run in this tab; there is no upload and no account.
Methodology
A methods paragraph to put beside the exported figure. It describes the tool’s default settings, so edit a clause if you changed one. References are given as the tool’s source code records them.
Creep records were analysed in StrataPlot stage by stage, time and strain measured from each load step, and in triaxial tests against the stress difference σ1 − σ3. Primary, secondary and tertiary creep were identified from the strain rate against log time, secondary creep being the stretch at the minimum rate (Brantut et al., 2013), the strain smoothed by a moving least-squares slope over a 0.3-decade window. The Burgers model (Wang & Li, 2018), the power law (Andrade, 1910) and the logarithmic law (Weiss & Amitrano, 2023) were fitted by least squares with linearised standard errors, the Burgers constants also given in shear-modulus form, and the Norton–Bailey law (Norton, 1929; Bailey, 1935) was fitted across all stages in log space.
References
- Andrade, E.N. da C. (1910). On the viscous flow in metals, and allied phenomena. Proceedings of the Royal Society of London A 84(567), 1–12. doi:10.1098/rspa.1910.0050. Free to read.
- Bailey, R.W. (1935). The utilization of creep test data in engineering design. Proceedings of the Institution of Mechanical Engineers 131, 131–349. doi:10.1243/PIME_PROC_1935_131_012_02.
- Brantut, N., Heap, M.J., Meredith, P.G. & Baud, P. (2013). Time-dependent cracking and brittle creep in crustal rocks: A review. Journal of Structural Geology 52, 17–43. doi:10.1016/j.jsg.2013.03.007. Open copy: hal.science/hal-00834718.
- Norton, F.H. (1929). The Creep of Steel at High Temperatures. McGraw-Hill, New York. Free scan: archive.org/details/creepofsteelathi00nort.
- StrataPlot. (2026). StrataPlot: Creep Test Analysis [Computer software]. https://www.strataplot.com/creep-test-analysis
- Wang, R. & Li, L. (2018). Burgers creep model used for describing and predicting the creep behaviour of a rock under uniaxial and triaxial compression test conditions. GeoEdmonton 2018, 71st Canadian Geotechnical Conference, Edmonton. Open: members.cgs.ca/conferences/GeoEdmonton/papers/geo2018Paper572.pdf.
- Weiss, J. & Amitrano, D. (2023). Logarithmic vs Andrade's transient creep: the role of elastic stress redistribution. arXiv:2206.06682.
How to cite this tool
If a figure made here goes into a report, thesis or paper, cite the tool so readers know how it was drawn. Add the date you accessed it if your style guide asks for one.
StrataPlot. (2026). StrataPlot: Creep Test Analysis [Computer software]. https://www.strataplot.com/creep-test-analysis
@misc{strataplot_creep,
author = {{StrataPlot}},
title = {StrataPlot: Creep Test Analysis},
year = {2026},
howpublished = {\url{https://www.strataplot.com/creep-test-analysis}},
note = {Computer software}
}