True stress-strain calculator
Paste engineering stress-strain data and get true stress, true strain and effective plastic strain, cut off at necking where the conversion stops being valid. Then get the hardening curve as an LS-DYNA *MAT_024 card or an Abaqus *PLASTIC table, extended past the test with a Swift or Voce fit. Free, no sign-up, your data stays in the browser.
σt = σe(1 + εe) · εt = ln(1 + εe) · Considère necking check · checked against scipy on generated tests · Updated 21 Nov 2026
LS-DYNA *MAT_024 + *DEFINE_CURVE
$ Example 1 steel (generated data) | units tonne, mm, s (stress MPa)
*MAT_PIECEWISE_LINEAR_PLASTICITY
$ MID RO E PR SIGY ETAN FAIL TDEL
1 7.8500E-9 213250.0 0.3 285.44031 0.0 0.0 0.0
$ C P LCSS LCSR VP
0.0 0.0 100 0 0.0
$ EPS1 EPS2 EPS3 EPS4 EPS5 EPS6 EPS7 EPS8
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
$ ES1 ES2 ES3 ES4 ES5 ES6 ES7 ES8
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
*DEFINE_CURVE
$ LCID SIDR SFA SFO OFFA OFFO DATTYP
100 0 1.0 1.0 0.0 0.0 0
$ effective plastic strain true stress
0.0 285.44031
5.3404685E-4 285.6038
0.0021361874 286.42309
0.0048064217 302.95242
0.0085447497 321.61713
0.013351171 340.43813
0.019225687 359.23178
0.026168296 377.01677
0.034178999 394.16773
0.043257795 410.60222
0.053404685 426.30762
0.064619669 441.30295
0.076902747 455.75665
0.090253918 469.55003
0.10467318 482.87513
0.12016054 495.72015
0.13671599 508.07774
0.15433954 519.98401
0.17303118 531.6839
0.19279091 542.9408
0.20636777 550.1664
0.23119205 562.47304
0.26333831 576.96238
0.30140573 592.42959
0.34458473 608.21823
0.39232907 623.95829
0.44423773 639.4427
0.5 654.5596
*ENDAbaqus *ELASTIC + *PLASTIC
** True stress vs plastic strain | units tonne, mm, s (stress MPa) *MATERIAL, NAME=MAT1 *ELASTIC 213250.0, 0.3 *PLASTIC 285.44031, 0.0 285.6038, 5.3404685E-4 286.42309, 0.0021361874 302.95242, 0.0048064217 321.61713, 0.0085447497 340.43813, 0.013351171 359.23178, 0.019225687 377.01677, 0.026168296 394.16773, 0.034178999 410.60222, 0.043257795 426.30762, 0.053404685 441.30295, 0.064619669 455.75665, 0.076902747 469.55003, 0.090253918 482.87513, 0.10467318 495.72015, 0.12016054 508.07774, 0.13671599 519.98401, 0.15433954 531.6839, 0.17303118 542.9408, 0.19279091 550.1664, 0.20636777 562.47304, 0.23119205 576.96238, 0.26333831 592.42959, 0.30140573 608.21823, 0.34458473 623.95829, 0.39232907 639.4427, 0.44423773 654.5596, 0.5
Converted points (true strain, true stress, plastic strain)
| # | Eng. strain | Eng. stress (MPa) | True strain | True stress (MPa) | Plastic strain |
|---|
What is true stress and true strain?
True stress is the load divided by the area the bar has at that moment; true strain adds up each small stretch divided by the length at that moment. A tensile test reports engineering values (original area and length), which understate stress as the bar thins. For loading up to the maximum load:
Both come from assuming the metal's volume stays constant while the gauge length stretches evenly, which is close to true for plastic flow:
For a reading at 10.33% engineering strain and 430.7 MPa from worked example 1 below:
The stress goes up by 10.3% and the strain goes down a little. Below 2% strain the difference is at most 2%, which is why it is ignored in elastic design and matters in plasticity, forming and crash.
Why does the conversion stop at the ultimate tensile strength?
At the maximum load a neck forms. From there the strain gathers in the neck, the gauge-length strain no longer describes it, and the stress in the neck is not purely uniaxial. So σe(1 + εe) is only valid up to the UTS. This calculator converts the points up to the highest engineering stress and leaves the rest out (it tells you how many).
The neck starts where the material's hardening can no longer make up for the shrinking area. That is the Considère condition:
For a power law σ = Kεn it gives a true uniform strain equal to n. It is a useful check on your data: in worked example 1 the fitted Swift law predicts necking at 0.1936 plastic strain and the test's engineering peak is at 0.1928. If the two disagree badly, the strain probably came from the crosshead or the peak is noise.
How do you get effective plastic strain for LS-DYNA or Abaqus?
Subtract the elastic strain from the total true strain. The material card gets true stress against this plastic strain, starting from zero at the initial yield stress:
- LS-DYNA *MAT_PIECEWISE_LINEAR_PLASTICITY (MAT_024): put the curve in a *DEFINE_CURVE (abscissa = effective plastic strain, ordinate = true stress) and its ID in LCSS. With LCSS set, the EPS1-8/ES1-8 points and ETAN are ignored, and so is SIGY unless VP = 1 (then a nonzero SIGY adds a strain-rate term), so the cards here use VP = 0. The manual asks for the first point at zero plastic strain, at the initial yield stress.
- Abaqus *PLASTIC: rows of true stress, true plastic strain; the first row must have zero plastic strain, and it defines the initial yield stress.
- Units: E and the stresses in the same unit as the rest of the model. In kg-mm-ms, stress is in GPa (1 kg/(mm·ms²) = 1 GPa): 250 MPa goes in as 0.25.
Where does "initial yield" go? Real curves bend gradually, so the calculator takes the true stress at a chosen plastic strain (0.2% by default, as in Rp0.2), puts it at zero plastic strain, and follows the measured points beyond it. The small plastic strain before that point is ignored. For most steels and aluminium that costs little. For materials with no straight elastic part, like cast iron, it can throw away most of the curve: pick a smaller offset (see the challenge results below).
Worked example 1: a steel test with a known answer
To test a method you need data whose answer is known. We generated a tensile test from a Swift hardening law, σ = 750(0.006 + εp)0.2 MPa, with E = 210.0 GPa, then converted it to engineering stress and strain and rounded the readings to 0.00001 strain and 0.1 MPa, like a lab file. After the peak we added a made-up drop in load, standing in for necking. It is the steel sample in the calculator.
Modulus. The least-squares fit over 10-40% of the peak stress gives E = 213.2 GPa, 1.5% above the true 210.0 GPa: with only 3 readings in the elastic band, the 0.00001 strain rounding shows. A real extensometer file has more elastic points.
Cut at necking. The highest engineering stress is 446.6 MPa at 0.2157 strain; the 6 readings after it are dropped. Converting that peak reading by hand:
The exact law gives 542.9 MPa at that plastic strain, so the conversion recovers it to within the rounding: across all 33 hardening points the largest difference is 0.11 MPa.
Initial yield. The true stress at 0.2% plastic strain is 285.4 MPa (exact 285.5 MPa). Fits. Swift: K = 750.2 MPa, ε0 = 0.0060, n = 0.2001 (the generating values were 750, 0.006, 0.2), RMS error 0.03 MPa. Voce: A = 294.2, Q = 260.5 MPa, b = 13.08, RMS error 5.5 MPa.
The LS-DYNA card for this example in tonne-mm-s units (20 points from the data, 8 from the Swift extension to 50% plastic strain):
$ Example 1 steel (generated data) | units tonne, mm, s (stress MPa)
*MAT_PIECEWISE_LINEAR_PLASTICITY
$ MID RO E PR SIGY ETAN FAIL TDEL
1 7.8500E-9 213250.0 0.3 285.44031 0.0 0.0 0.0
$ C P LCSS LCSR VP
0.0 0.0 100 0 0.0
$ EPS1 EPS2 EPS3 EPS4 EPS5 EPS6 EPS7 EPS8
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
$ ES1 ES2 ES3 ES4 ES5 ES6 ES7 ES8
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
*DEFINE_CURVE
$ LCID SIDR SFA SFO OFFA OFFO DATTYP
100 0 1.0 1.0 0.0 0.0 0
$ effective plastic strain true stress
0.0 285.44031
5.3404685E-4 285.6038
0.0021361874 286.42309
0.0048064217 302.95242
0.0085447497 321.61713
0.013351171 340.43813
0.019225687 359.23178
0.026168296 377.01677
0.034178999 394.16773
0.043257795 410.60222
0.053404685 426.30762
0.064619669 441.30295
0.076902747 455.75665
0.090253918 469.55003
0.10467318 482.87513
0.12016054 495.72015
0.13671599 508.07774
0.15433954 519.98401
0.17303118 531.6839
0.19279091 542.9408
0.20636777 550.1664
0.23119205 562.47304
0.26333831 576.96238
0.30140573 592.42959
0.34458473 608.21823
0.39232907 623.95829
0.44423773 639.4427
0.5 654.5596
*ENDHow do you extend the curve past necking?
A tensile test gives hardening data only to the uniform strain (19.5% true strain here), but crash, forming and fracture models reach 50-100% plastic strain in the elements that matter. The usual answer is to fit a hardening law to the measured part and extend it:
Both fit the steel data, but they part ways outside it. At 50% plastic strain the Swift fit gives 655 MPa and the Voce fit 554 MPa, 18% apart, from the same test. Here the data came from a Swift law, so Swift is right; for a real metal, nobody knows from the tensile test alone. The Voce fit is visibly worse inside the data too (RMS 5.5 vs 0.03 MPa), which is a hint, not proof.
In practice: steels are often extended with Swift or a Swift-Voce blend, aluminium with Voce; crash teams check the choice against a notched or shear test, or run an FE model of the tensile specimen and adjust the extension until the simulated force-displacement curve matches the test past the peak. If you give no extension, the two codes behave differently past your last point: LS-DYNA extrapolates material curves from their end slope, while Abaqus holds the last stress constant (perfectly plastic). A noisy last segment can therefore send an LS-DYNA curve up, flat or even down, so end the curve on a clean point or extend it on purpose.
Worked example 2: aluminium with a saturating curve
The aluminium sample is generated from a Voce law, σ = 270 + 60(1 − e−25εp) MPa, with E = 70.0 GPa entered (not fitted). Its engineering peak is 296.8 MPa (316.2 MPa true) at 6.34% true strain. The calculator finds an initial yield of 272.9 MPa (exact 272.9) and fits Voce A = 270.0, Q = 60.0 MPa, b = 25.02: the generating law back. Considère on that fit puts necking at 0.0619 plastic strain; the peak reading is at 0.0588, the nearest reading before it.
Here a Swift fit to the same points gives 385 MPa at 50% plastic strain against the true 330 MPa: 17% too high. The law you extend with matters as much as the data.
What 135 student reports on building MAT_024 from raw data show
Skill-Lync's LS-DYNA course has a challenge: digitise curve 2 of a published plot of true stress-strain curves for graphite cast iron (stress in ksi, strain in %), use E = 20.9 × 106 psi, build a MAT_024 card in kg-mm-ms and check it with a dog-bone tensile model. We read the 135 public reports on Skill-Lync (as of 21 Nov 2026). What stood out:
- "Yield" is a definition, not a number, for cast iron. 36 reports state the yield stress they used: from 131 to 346 MPa for the same curve. 20 sit between 143 and 152 MPa, where plastic strain first turns positive; others used a 0.2% offset. Both are defensible, and they differ by a factor of two: one report's own digitised table reaches 0.2% plastic strain at 334.7 MPa, starting from 150.8 MPa, and the whole curve ends at 0.00372 plastic strain. A 0.2% offset start would drop more than half of the plastic curve. That is why this calculator lets you set the offset, and warns when it is a big share of the data.
- Converting true data again. 35 reports call the plot true stress-strain data, in the aim most of them copy from the challenge, yet at least 11 reports describe it as engineering data and run it through σ(1 + ε) and ln(1 + ε). On this curve (under 1% strain) that changes stress by less than 1%. On worked example 1's steel the same mistake adds 19.5% to the stress at the UTS. Set "Data is: already true" for curves that are already true.
- The modulus. 4 reports type E as 20 × 106 psi instead of 20.9 × 106; 2 of them then compute with 137.9 GPa instead of 144.1 GPa (4.3% low), which shifts every plastic strain. 1 psi = 6,894.757 Pa, so 20.9 × 106 psi = 144.1 GPa = 144.1 in kg-mm-ms.
- Formula slips. 3 reports write the plastic strain as (true strain − true stress)/E, which mixes units. The correct form subtracts σt/E from εt.
Paste your digitised points into the calculator with "Stress in: ksi", "Strain in: %", "Data is: already true" and E = 144.1 GPa to check your curve and card.
What this calculator does not do
- Strain rate and temperature. One curve, one rate. For rate-dependent crash models MAT_024 takes a table of curves (LCSS pointing to a *DEFINE_TABLE) or Cowper-Symonds C and P.
- Post-necking calibration. It extends with a fitted law; it does not run an inverse FE fit of the specimen.
- Anisotropy, damage and failure. Sheet anisotropy (r-values), failure strain and damage are separate material inputs.
- Crosshead strain. If the fitted E is far below the handbook value, the strain includes machine and grip stretch, and every plastic strain is wrong. The calculator warns, it cannot correct it.
Turning test data into MAT_024 and checking it with a specimen model is a week of Skill-Lync's LS-DYNA course, and a standard interview question for CAE roles.
- LS-DYNA for Structural Mechanics/FEA (material cards from raw data, hyperelastic models, contacts)
- Basics of FEA: Explicit and Implicit
- Post Graduate Program in CAE (48 weeks)
Frequently asked questions
How do you convert engineering stress to true stress?
Multiply the engineering stress by one plus the engineering strain: σt = σe(1 + εe). True strain is the natural log of one plus the engineering strain: εt = ln(1 + εe). Both assume constant volume and uniform stretching, so they hold only up to the maximum load (the ultimate tensile strength), where necking starts.
What is the formula for true strain?
True strain is εt = ln(L/L0) = ln(1 + εe), the sum of every small stretch divided by the length at that moment. In terms of area, from constant volume, εt = ln(A0/A). At 10% engineering strain the true strain is 9.53%; at 50% it is 40.5%. Below about 2% the two are almost the same.
Why is true stress higher than engineering stress?
Engineering stress divides the load by the original area, but a bar in tension gets thinner. True stress divides by the real, smaller area, so it is always higher in tension. Past the ultimate tensile strength the engineering curve falls because the neck's area shrinks faster than the metal hardens, while true stress keeps rising.
Can you convert engineering stress to true stress after necking?
Not with σe(1 + εe). After the maximum load, strain concentrates in the neck, so the gauge-length strain no longer describes it and the stress is no longer uniaxial. Use the curve only up to the UTS, then extend it with a hardening law (Swift, Voce or Hollomon) or calibrate it against the test with an FE model of the specimen.
What is effective plastic strain?
It is the permanent part of the strain: total true strain minus the elastic part, εp = εt − σt/E. In a uniaxial test it equals the plastic strain along the bar. LS-DYNA's MAT_024 and Abaqus's *PLASTIC both want hardening as true stress against effective plastic strain, starting at zero plastic strain at the initial yield stress.
What does LCSS mean in MAT_024?
LCSS is the ID of a load curve (*DEFINE_CURVE) giving effective stress against effective plastic strain for *MAT_PIECEWISE_LINEAR_PLASTICITY. When it is set, LS-DYNA uses that curve for hardening and ignores EPS1-EPS8, ES1-ES8 and ETAN (and SIGY, unless VP = 1). The first point should be at zero plastic strain, at the initial yield stress.
How do I convert engineering stress to true stress in Excel?
With engineering strain in A and engineering stress in B: true stress =B2*(1+A2), true strain =LN(1+A2), plastic strain =D2-C2/E with E in the same stress unit. Fill down only to the row with the highest stress (=MAX(B:B)). If strain is in percent, divide it by 100 first.
Related on Skill-Lync
- 0.2% offset yield strength calculator (Rp0.2 and E from raw test data)
- Hyperelastic curve fit calculator (rubber: Mooney-Rivlin, Ogden, MAT_077)
- Consistent units calculator (kg-mm-ms, tonne-mm-s and others)
- LS-DYNA project ideas
- Explicit time step calculator (TSSFAC, DT2MS and mass scaling)
- LS-DYNA courses compared
- All free engineering tools
Sources
- G. E. Dieter, Mechanical Metallurgy, SI metric edition, McGraw-Hill, 1988: true stress and true strain, constant volume, instability in tension (Considère).
- A. Considère, Mémoire sur l'emploi du fer et de l'acier dans les constructions, Annales des Ponts et Chaussées, 1885.
- H. W. Swift, Plastic instability under plane stress, J. Mech. Phys. Solids 1 (1952) 1-18; E. Voce, The relationship between stress and strain for homogeneous deformation, J. Inst. Metals 74 (1948) 537-562.
- LS-DYNA Keyword User's Manual R14, Volume II: *MAT_PIECEWISE_LINEAR_PLASTICITY (LCSS, SIGY, ETAN, Remark 1a; "The first point must be zero corresponding to the initial yield stress"); Volume I: *DEFINE_CURVE remarks ("In the constitutive models, extrapolation is employed if the values on the abscissa go off scale").
- Abaqus documentation, Defining plasticity ("you must use true stress and true strain"; the nominal-to-true relations "are valid only prior to necking"; first value at zero plastic strain; constant response outside the data) and Classical metal plasticity.
- D. Roylance, Stress-strain curves, MIT 3.11 (constant volume before necking; a power-law material necks at true strain n); DoITPoMS, University of Cambridge, Necking (Considère).
- NIST Special Publication 811, Appendix B.9: 1 psi = 6.894 757 × 103 Pa.
- Worked examples: generated by Skill-Lync with scripts/verify/true_stress.py (numpy/scipy least squares from many starting points, Considère by root finding) and recomputed by this page's calculator code at every build. Example data is generated, not from a real specimen. Challenge figures: Skill-Lync's reading of 135 public student reports, 21 Nov 2026.