Engineering Stress and True Stress: the Criterion Behind the Code, With a Worked Check

Pull a steel coupon apart and a testing machine records one force and one elongation at every instant. From that single pair of readings you can compute the stress two different ways, and both are correct. Divide the force by the original cross-section it started with and you get engineering stress, σ_eng = P/A₀. Divide it by the actual cross-section at that moment, the area that has already thinned under load, and you get true stress, σ_true = P/A_i. Same force, two denominators, two numbers.

Every strength in your code, f_y the yield strength and f_u the ultimate strength, is the first one: an engineering stress. The material, meanwhile, is always feeling the second one. This article is about that gap: what it is, the one-line conversion between the two, why AISC, NBR 8800 and Eurocode all speak in engineering stress on purpose, and a worked check on a real steel tie where the CalcSteel FEM engine measures the gap to the fourth decimal. The short version is that the code chose the number it can measure, and it loses almost nothing by doing so, until you get out to necking, where the two stresses part company for good.

Here are the four definitions you can quote, all built from the same measured force P, original length L₀ and original area A₀.

• Engineering stress, σ_eng = P / A₀. The load carried by every square millimetre the bar started with.
• Engineering strain, ε_eng = ΔL / L₀. The stretch as a fraction of the original length.
• True stress, σ_true = P / A_i. The load over the instantaneous area A_i, which shrinks as the bar stretches (Poisson contraction elastically, then constant-volume flow plastically).
• True strain, ε_true = ln(L_i / L₀) = ln(1 + ε_eng). The sum of every incremental stretch dL/L, integrated over the deformation.

The distinction is invisible while the bar barely moves and grows as it deforms. In tension A_i is smaller than A₀, so true stress is always the larger of the two. Engineering stress is a bookkeeping convenience anchored to the geometry you can measure with a caliper before the test; true stress is what the surviving material is actually resisting.

The two descriptions are not independent; one line converts between them. As long as the deformation is uniform along the bar (that is, before any neck forms) and plastic flow conserves volume, A_iL_i = A₀L₀. Rearranging gives A_i = A₀ / (1 + ε_eng), and therefore:

σ_true = σ_eng (1 + ε_eng)    and    ε_true = ln(1 + ε_eng)

Read the two conditions again, because they are the fine print engineers forget. The relation holds only while the section is uniform (so it dies the instant the bar necks) and only while volume is conserved (which is a good model for plastic flow, where Poisson's ratio is effectively 0.5, but not for the elastic range, where it is about 0.3 and the bar does change volume slightly). In the elastic design range the true correction is even smaller than σ_eng(1 + ε) suggests, so using the constant-volume form there is a mildly conservative, safe simplification. The table runs the conversion across the range that matters.

Three rows are a rounding error; the fourth is a different world. That contrast is the entire story of when the distinction matters.

If true stress is what the material really carries, why does no design clause use it? Three reasons, and they compound.

1. It is what the test reports. The strengths in every steel standard come from a tension test run to a fixed procedure (ASTM A370, which invokes ASTM E8/E8M; ISO 6892-1 elsewhere). Those procedures define yield and tensile strength as force divided by the original cross-sectional area. f_y and f_u are engineering stresses by definition, not by choice; the code simply inherits the number the laboratory is required to report.

2. Design lives where the two agree. Every strength check, bending, axial, shear, the interaction equation, is a comparison at or below f_y, so at strains under about 0.18%. As the table showed, true stress is under two tenths of a percent above engineering stress there. The code gains nothing by carrying a correction smaller than the scatter in the steel itself.

3. It is the conservative direction. Because true stress exceeds engineering stress in tension, the material's real yield point sits just above the tabulated f_y. Designing to the engineering value leaves that sliver in reserve rather than spending it. A number that is directly measured, easy to tabulate, and slightly safe is exactly what a code wants.

Zoom into the part of the curve a structural engineer ever operates in, from the origin up to yield, and the engineering and true lines are visually one line. At ε = 0.000750 the conversion adds 0.075%; at first yield for MR250, ε_y = 0.00125, it adds 0.125%; even for a 345 MPa grade at ε_y = 0.001725 it adds only 0.172%. On a plotted stress axis those offsets are thinner than the ink.

The elastic Poisson caveat makes the real gap smaller still. Below yield the bar has not started constant-volume plastic flow, so its area shrinks with Poisson's ratio 0.3, not 0.5. Carrying that correctly, the true stress at first yield is about 250.19 MPa rather than the 250.31 MPa the constant-volume formula returns, a 0.075% gap instead of 0.125%. Either way the message is the same: in the design range, engineering stress is true stress for every practical purpose, and the tidy σ_eng(1 + ε) form is the safe one to quote because it slightly overstates the gap. This is the deepest sense of the criterion behind the code: the code did not approximate anything, it worked in the regime where the approximation is exact to three decimals.

The distinction stops being academic exactly where uniform deformation stops: the neck. Up to the ultimate strength f_u the whole gauge length stretches together, and σ_eng(1 + ε) still holds. For a mild steel that reaches f_u = 400 MPa at a uniform strain of about 16%, true stress at that point is 400 x 1.16 = 464 MPa, already 16% above the engineering value, at a true strain of ln(1.16) = 0.148.

Past f_u the two curves do opposite things, and it is worth being precise about why. Deformation localizes into a neck; the load the bar can carry starts to drop. Engineering stress, still dividing that falling load by the fixed original area A₀, falls with it, sliding down to fracture at maybe 300 MPa. True stress divides the same falling load by the real area, which is collapsing faster than the load, so it keeps rising all the way to fracture, well past 600 MPa for a ductile structural steel. The engineering curve's famous downhill run to fracture is not the material getting weaker; it is an artefact of dividing by an area that is no longer there. The material never softens. That single insight, that the drop is a bookkeeping shadow of a shrinking cross-section, is why anyone studying fracture, forming or crash uses true stress and never engineering stress.

Time to measure the gap rather than assert it. Take the most honest possible test object for a stress this pure: a single-angle L 100x100x10 acting as a tension tie, a bracing member or hanger 4 m long in structural steel. An axial member has one stress, uniform over the section, exactly the coupon geometry the whole engineering-versus-true distinction was born on.

We model it in the CalcSteel FEM engine, the same solver behind the calculators, and read the results. Under a service tension of N = 285 kN the engine returns the axial force in the member as 285.0 kN (equilibrium, as it must) and reports the section area as A = 19.0 cm2. It also returns the elongation of the loaded end: 3.000 mm. That last number is the check on the engine itself, because closed-form N L / (A E) = 285 x 400 / (19.0 x 20000) = 0.300 cm = 3.000 mm, matching to the third decimal. The stiffness is right, so the stress and strain we pull from it are trustworthy.

Now turn the analysis into the two stresses. Engineering stress is the axial force over the original area: σ_eng = N / A₀ = 285 kN / 19.0 cm2 = 15.0 kN/cm2 = 150.0 MPa. The strain the fibre is feeling is that stress over the modulus: ε = σ / E = 150.0 / 200000 = 0.000750, which is the same 0.075% the engine's 3.000 mm elongation over the 4000 mm length gives directly (3.000 / 4000 = 0.000750). Two independent routes, one strain.

True stress applies the conversion: σ_true = σ_eng(1 + ε) = 150.0 x 1.000750 = 150.11 MPa. The material is carrying 150.11 MPa; your analysis, your code check and your calculator all say 150.0 MPa. The gap is 0.11 MPa, or 0.075%, on a tie loaded to 60% of MR250's yield. This is not a coincidence of this member; it is the general fact that in the elastic range ε is a fraction of a percent, so (1 + ε) is one to three decimals. The utilisation your software prints, σ/f_y = 150/250 = 0.60, would read 0.601 in true stress. Nobody sharpens a design on the third decimal of the utilisation, which is precisely why the code does not carry it.

You can reproduce the whole exercise. In the calculator below, set up a member and a load and read the stress it reports. Divide that stress by 200,000 MPa and you have the strain ε. Multiply the stress by (1 + ε) and you have the true stress. You will watch the correction stay in the third or fourth decimal for anything short of yield, and you will see with your own numbers why the code keeps the engineering value.

It is the same engine that produced the worked tie above, free, with no login for the maths.

Raise the load until the tie first yields and watch the gap at the one point the code actually cares about. First yield in MR250 is σ = f_y = 250 MPa, which needs N = 25 kN/cm2 x 19.0 cm2 = 475 kN. Re-run the same member at N = 475 kN and the engine returns σ = 250.0 MPa, an elongation of 5.000 mm (so ε = 5.000 / 4000 = 0.001250, exactly the yield strain of MR250), and the axial force back as 475.0 kN. Everything closes.

Convert: σ_true = 250.0 x 1.00125 = 250.31 MPa, and ε_true = ln(1.00125) = 0.0012492. At the yield point the code writes f_y = 250 MPa; the material's true yield stress is 250.31 MPa; the difference is 0.31 MPa, three tenths of one percent, and in the safe direction. If the criterion behind the code is really a strain reaching a threshold, then the stress that names that threshold is the same number, 250 MPa, whether you measure it against the original area or the real one. The two definitions only diverge once you leave the elastic world entirely, and structural design never does.

None of this means true stress is a curiosity. It is indispensable, just not in a member strength check. Reach for it, and never engineering stress, whenever the steel is deliberately taken into large plastic strain:

• Cold forming. Press-braking a plate or roll-forming a cold-formed section drives strains of 10% to 30% at the bend. The forming force, the springback and the residual stress all depend on the flow (true) stress at those strains; the engineering curve, already falling there, would mislead you badly.
• Finite-element modelling of ductile failure. Any nonlinear FE run that follows steel past yield, connection ductility, seismic fuses, blast, crash, must be fed a true stress versus true (or plastic) strain curve. Feeding a solver the engineering curve tells it the material softens after f_u, which it does not, and the analysis diverges or lies.
• Fracture and low-cycle fatigue. Reduction of area, true fracture strain and the strain-life relations that govern seismic detailing are all written in true measures, because that is where the material's real state lives.

There is one place the two worlds meet inside the code, and it is worth naming. The minimum f_u/f_y ratio (roughly 1.1 to 1.25) and the minimum elongation that every structural grade must pass are engineering-stress requirements whose whole purpose is to guarantee a long plastic runway, the very region where true and engineering stress diverge. Plastic hinges, moment redistribution and seismic energy dissipation all spend that runway. So the code checks strength in engineering stress and buys ductility, measured the same way, to protect a plastic reserve it will describe in true stress the day it has to model it failing.

Thinking f_u is the highest stress the steel ever reaches. f_u is the peak of the engineering curve. The true stress at fracture is far higher, often 1.5 to 2 times f_u, because the real area at the neck is a fraction of the original.

Reading the post-ultimate drop as the material weakening. The engineering curve falls after f_u only because it divides a shrinking load by a fixed original area. The material strain-hardens right up to fracture; true stress never decreases.

Applying σ_true = σ_eng(1 + ε) after necking. The formula assumes uniform deformation and dies at the neck. Past that point the state is triaxial and localized, and true stress needs a direct area measurement (or a Bridgman correction), not the tidy conversion.

Feeding an engineering stress-strain curve to a nonlinear solver. A very common and dangerous error. Material models in FE expect true stress versus true (or plastic) strain; the engineering curve will make the model soften artificially past f_u.

Is the difference ever significant in structural design? Not in a normal strength check: at and below yield it is under 0.2%. It becomes significant only in large-strain problems, forming, ductile fracture, crash, seismic detailing.

Which one does my calculator report? Engineering stress, always, because it computes force over the modelled (original) section. That is the right number to compare against a code f_y or f_u, which are engineering stresses too.

Does the same apply in compression? The sign flips: in compression the area grows, so true stress is slightly below engineering stress. At design strains it is again negligible, and compression is governed by buckling and stability well before the distinction could matter.

Engineering stress is the number the code keeps because it is the number the test reports and the number design can use without a correction.

• σ_eng = P/A₀, σ_true = P/A_i; before necking, σ_true = σ_eng(1 + ε) and ε_true = ln(1 + ε).
• f_y and f_u are engineering stresses by definition, because ASTM E8 and ISO 6892 report force over the original area.
• In the design range the gap is a rounding error: 0.075% at a service tie, 0.125% at first yield, and in the safe direction.
• Worked L 100x100x10 tie (A = 19.0 cm2, N = 285 kN): the engine gives a 3.000 mm stretch (= NL/AE), so σ = 150.0 MPa and σ_true = 150.11 MPa.
• At N = 475 kN the engine returns σ = 250.0 MPa = f_y; the two stresses split for real only at necking, where engineering stress falls to fracture while true stress climbs past 460 MPa.