Material Stress: Why the Same 226.60 MPa Passes in One Steel and Fails in Another, with a Worked Check

Here is a beam with nothing interesting about it. An IPE 360, simply supported, 8.0 m of span, carrying 10.0 kN/m of permanent load and 7.5 kN/m of variable load. Composite floors, mezzanines and office bays are full of it.

Run it. The design combination is 1.35 × 10.0 + 1.5 × 7.5 = 24.75 kN/m, the CalcSteel engine returns a peak bending moment of 198.0 kN·m, and against the elastic section modulus of the IPE 360, Wel = 873.77 cm³, that is a bending stress of 226.60 MPa.

Now answer one question: does the beam pass?

You cannot. Not because the number is uncertain, but because the question is only half asked. In MR250, the ordinary Brazilian structural steel, the screening limit is fy/1.10 = 227.27 MPa and the utilisation is 0.997, a pass. In ASTM A36, which has the exact same yield strength of 250 MPa, the limit is 0.90 × 250 = 225.0 MPa and the utilisation is 1.007, a fail. Same steel. Same beam. Same stress. Opposite verdicts, separated by 2.27 MPa of pure convention.

And in 1.4301 stainless, which most people would call the premium material, the utilisation is 1.187 and the beam is not close. Meanwhile in S460, the strongest steel in the sweep, the bending utilisation drops to 0.493 and the beam still fails, for a reason that has nothing to do with strength.

That is what this article is about. “Material stress” fuses two questions that behave in completely different ways, and every one of the numbers above was produced by the same engine that runs inside CalcSteel, with the material changed and nothing else.

Search “material stress” and you get two different topics wearing one label. One is the stress state: the internal force intensity at a point, in MPa, the thing that σ = N/A and σ = M/W compute. The other is the material limit: the intensity at which that particular metal stops behaving the way the model assumed.

They belong to different owners. The stress state is owned by the load and the geometry. The limit is owned by the material and by the code that judges it. Keeping them apart is not pedantry, it is the difference between two sentences that sound alike and are not:

• “This steel has a stress of 226 MPa.” Meaningless. Steel does not have a stress. A point in a loaded member does.
• “This member is at 226 MPa, and this steel is allowed 225.” A complete engineering statement, and the reason the beam above fails.

There is a real question hiding inside the confusion, though, and it is worth asking properly: does the stress itself depend on the material? The honest answer is that it depends on whether the structure is determinate. In a determinate member, no. In a redundant one, yes, and by a lot. The engine settles both cases below, and the second one is where the modelling errors live.

One model, defined once, solved many times. Only the material identifier on the bars changes between runs.

Before any of it is trusted, the engine has to reproduce the textbook. A calibration run on an IPE 400 at 8.0 m under 30.0 kN/m returned M = 240.0 kN·m against wL²/8 = 240.0, V = 120.0 kN against wL/2 = 120.0, and a midspan deflection of 35.99968 mm against 5wL⁴/384EI = 35.99968 mm. Agreement to five decimal places, so what follows is the model talking, not the solver.

One note on the section properties, because it matters for reproducing these numbers with a steel table in hand. The profile registry builds the shape from the rolled geometry without the root fillets, which gives Wel = 873.77 cm³ where the published table for an IPE 360 says 903.6 cm³, about 3 % higher. Demand and capacity here are computed on the same basis, so every utilisation is consistent, and slightly conservative.

The check itself is the elastic screening the CalcSteel beam calculator runs: σ = M/Wel against the code's design stress. A full verification of a compact section would use the plastic moment and would be more generous. That is deliberate. The point of this article is a comparison between materials, and the elastic route compares them on identical terms.

The beam was solved ten times. Structural steels from four traditions, one austenitic stainless and one aluminium alloy, each one carrying the engine's own catalogue values for E, fy, fu and density. The bending moment came back identical in all ten runs: 198.0 kN·m, with a spread between the largest and the smallest of 0.000 kN·m.

That is not a coincidence and it is not a rounding artefact. The beam is statically determinate, so equilibrium alone fixes the internal forces: M = wL²/8 contains the load and the span and nothing else. E, fy, fu and density are simply absent from the equation. So σ = M/Wel = 226.60 MPa in all ten cases, including the aluminium one.

This is the single most useful thing a student can take from a stress chapter, and the least often stated plainly: in a determinate structure the material does not appear in the demand at all. Choosing a stronger steel does not reduce the stress you have to resist. It raises the stress you are allowed to reach.

The deflection is a different story, and the difference is instructive. It carries E explicitly, δ = 5wL⁴/384EI, so it moved in every run where E moved. Under the serviceability combination the same beam came back at 28.26 mm with the European steels at 210 GPa, 29.67 mm with the American and Brazilian steels and the stainless at 200 GPa, and 84.77 mm in aluminium at 70 GPa. The ratio against the steel case is 3.0000, which is 210/70 to four decimals, exactly as the closed form demands.

Look at the first two numbers again, because there is something odd in them. S235 and A36 are, physically, almost the same metal. Their real Young's moduli differ by well under one per cent. But EN 1993-1-1 declares E = 210 GPa and both AISC 360 and NBR 8800 declare E = 200 GPa, and that declaration is what a solver uses. The identical beam therefore deflects 5.000 % more in the American model than in the European one, and no laboratory would find the difference. Part of what we call a material property is a decision written in a committee.

Here is the whole comparison in one table. The demand column is constant because the previous section proved it has to be. Everything that varies is on the capacity side.

Three things in that table are worth more than the rest of the article.

The stainless fails while the mild steel passes. Austenitic 1.4301 costs several times what S235 costs and everybody calls it a better material, which it is, for corrosion. Structurally its 0.2 % proof stress is 210 MPa, below plain S235, and EN 1993-1-4 divides it by 1.10 rather than 1.00. “Better material” is not a structural category. It is only ever better at something specific.

The aluminium passes on strength. With fo = 260 MPa the 6082 T6 profile is allowed 236.36 MPa and lands at 0.959, better than S235. Anyone stopping the check there would ship it. The next section is what happens to that beam.

Two identical yield strengths, opposite verdicts. MR250 and A36 both have fy = 250 MPa, and the codes that judge them disagree by 1.0 %. That deserves its own section.

MR250 and ASTM A36 are, for this purpose, the same steel: fy = 250 MPa, fu = 400 MPa, E = 200 GPa. The two codes then treat the number differently.

• NBR 8800 divides. The design strength is fy/γa1 with γa1 = 1.10, so 250/1.10 = 227.27 MPa, and 226.60/227.27 = 0.997.
• AISC 360 LRFD multiplies. The design strength is φb·fy with φb = 0.90, so 0.90 × 250 = 225.00 MPa, and 226.60/225.00 = 1.007.

The gap is 2.27 MPa, or 1.01 %, and it is the entire difference between a beam you detail and a beam you resize. It is not a disagreement about steel. 1/1.10 = 0.909 and 0.90 are two roundings of the same reliability target, calibrated against different load factor sets by different committees in different decades. Neither is more correct.

Which is exactly why the CalcSteel beam calculator prints both utilisations side by side on every solve instead of picking one. When a member lands in the 0.99 to 1.01 band, the governing code is not a detail of the report, it is the answer.

The practical rule that follows is short. A utilisation is only a number next to the code that produced it. Copying 0.997 out of a Brazilian job into an American one is not a units conversion, and the beam does not know it moved.

Return to the beam that fails. The instinct of every engineer under time pressure is to upgrade the steel, so let us do exactly that and measure what it buys.

S235 to S460 is 1.9574 times the yield strength, the largest jump the European catalogue offers in ordinary construction. The strength utilisation falls from 0.964 to 0.493, cut almost in half, and it feels like a solved problem.

The deflection under the serviceability combination is 28.26 mm in S235. In S460 it is 28.26 mm. Not approximately: the engine returned a spread of exactly 0.000 mm across S235, S275, S355 and S460, because all four have the same E = 210 GPa. Against the L/300 limit of 26.67 mm, the utilisation is 1.0597 before the upgrade and 1.0597 after it.

The beam fails identically in the strongest steel available. Not marginally better, not slightly worse. Identically, to the last digit the solver carries.

The reason is one line long. Deflection depends on E and I, and the grade changes neither. Yield strength is a chemistry and processing outcome; Young's modulus is set by the interatomic bond stiffness of iron and is essentially the same in every carbon steel ever rolled. You buy strength with grade. You buy stiffness with shape.

Which is why so many real floor beams are governed by serviceability. The moment you are in that regime, the grade column of the catalogue stops being useful, and the money you spend there buys nothing at all.

If the material cannot fix a stiffness problem, the section has to. Same span, same loads, same S235, only the profile changes.

The IPE 400 costs 16.1 % more steel than the IPE 360 and takes the deflection utilisation from 1.060 to 0.750. Problem closed, in the cheapest possible currency.

The HEA 300 is the instructive failure. It weighs 54.6 % more than the IPE 360, 33 % more than the IPE 400, and still ends at 0.948, worse than the lighter IPE. All that extra metal went into flange width, which serves stability and column work, not into depth, which is what a second moment of area is made of. Ix scales with the cube of the depth. Nothing else in the catalogue comes close to that leverage.

Two sentences to keep, then, and they are the practical core of the whole topic: a strength problem is a material problem, and a stiffness problem is a geometry problem. Diagnose which one you have before you spend money, because the two remedies do not substitute for each other in either direction.

Run it yourself below. Change fy and watch only the strength utilisation move. Change the profile and watch both move. The calculator prints the NBR 8800 and AISC 360 screenings side by side, which is the 2.27 MPa argument from three sections ago, live.

Everything so far rested on the beam being statically determinate. Remove that and the conclusion inverts.

Take the same IPE 360 in S355 and hang its midspan from a 25 mm round bar that rises 12 m to a fixed point overhead, the way a floor beam is picked up by a roof truss in an industrial building. That is one degree of redundancy. Equilibrium no longer determines the split between the two load paths: compatibility does, and compatibility is written in stiffnesses.

Swapping the hanger from steel to aluminium, and touching nothing else, moves the peak moment in the beam from 57.847 kN·m to 96.942 kN·m. That is 1.6758 times more bending, and the beam's bending stress goes from 66.2 MPa to 110.9 MPa. The load did not change. The geometry did not change. Only what the hanger is made of changed, and the stress in a member 12 m away moved by 68 %.

Each engine result was checked against the compatibility closed form R = δ₀/(L³/48EI + Lh/EhAh), where δ₀ = 39.965 mm is the free midspan deflection of the unpropped beam. Steel: 90.962 kN from the engine against 90.962 kN by hand. Aluminium: 59.456 against 59.456.

Now the control that stops the wrong lesson being learned. Solve it again with the beam and the hanger both in aluminium: the hanger force comes back at 90.9621 kN, and the all-steel case was 90.9621 kN. A difference of 0.0000 kN.

So a soft material does not attract less load. A relatively soft one does. E cancels out of a linear redundant structure whenever it is uniform, and only the ratio between the parts survives. That is why mixing materials in one load path is a decision and not a detail: steel ties on aluminium frames, stainless hangers on carbon frames, timber on steel. The stiffer partner takes the load, and it takes it in proportion to a ratio nobody wrote in the specification.

Stay with the propped beam and delete the loads. Every load case is empty. Now raise the temperature of the hanger by 30 K, which is a sunny afternoon on an exposed member.

The bar wants to grow by α·ΔT·L. The structure will not let it grow freely, and the part of the growth that is prevented becomes force. Applied to the engine as the equivalent thermal load, N₀ = E·A·α·ΔT, and checked against N = -N₀·kbeam/(kbeam + khanger):

Three things here are worth slowing down for.

First, the stress exists with no load in the model. This is as literal as “material stress” ever gets: the number was produced by a material property, α, and by nothing else.

Second, the fully restrained stress E·α·ΔT is a trap. Steel reads 75.6 MPa and aluminium only 48.3 MPa, because aluminium's larger α is more than cancelled by its smaller E. From that column you would conclude the aluminium hanger is the safer choice.

Third, the real answer reverses it. The aluminium bar ends up at 25.09 MPa against the steel bar's 20.03 MPa, 1.2528 times more, because a soft bar in series with a beam of finite stiffness keeps a bigger share of its own thermal load: the restraint ratio kbeam/(kbeam + khanger) is 0.52 for aluminium and 0.26 for steel. Full restraint is a bound, not an estimate, and in a real structure the fraction that survives is decided by the stiffnesses around the member.

The engineering habit that follows is old and reliable. Thermal force lives on redundancy, so if you do not want it, do not build the redundancy: a slotted hole, a sliding bearing or an expansion joint turns the whole table into zeros.

Every check so far compared a stress to fy. The limit that a material actually imposes is a package, and fy is only the entry that governs on a warm day with a static load.

fu, at the sections where a net area governs. Yielding on the gross section is ductile and self-limiting, so it is judged against fy with a mild factor. Rupture through a bolt line is neither, so it is judged against fu with a harsher one, EN 1993-1-1 §6.2.3 pairing A·fy/γM0 with 0.9·Anet·fu/γM2 at γM2 = 1.25. The ratio fu/fy is itself a material property: 1.380 in S355 and 2.476 in 1.4301, which is why stainless connections and carbon connections do not fail in the same order.

Temperature. Above roughly 400 °C, fy stops being fy. EN 1993-1-2 reduces it by ky,θ: 0.78 at 500 °C, 0.47 at 600 °C, 0.23 at 700 °C. Under the accidental combination this beam carries g + 0.3q = 12.25 kN/m, so M = 98.0 kN·m and σ = 112.16 MPa, 49.5 % of its cold demand. The critical temperature is where ky,θ·fy drops to that:

• S235 needs ky,θ = 0.477, reached at 597.7 °C.
• S355 needs ky,θ = 0.316, reached at 664.2 °C.
• S460 needs ky,θ = 0.244, reached at 694.2 °C.

Note what just happened. The grade upgrade that bought exactly nothing against deflection buys almost 100 °C here, and 100 °C of a standard fire curve is real minutes and real millimetres of intumescent paint. Whether a material property helps depends entirely on which limit state you are in.

Cycles. Under fatigue, fy drops out of the check completely. A welded detail in S460 and the same detail in S235 share a detail category and an S-N curve; the stress range decides, and the strongest steel on the site carries no advantage whatsoever. It is the clearest case of a material limit that is not a material strength.

There is an assumption buried under every number in this article: that the material is linear up to a yield plateau, so E is a constant and fy is a corner. Carbon steel earns that assumption. Austenitic stainless does not.

A 1.4301 tensile curve rounds over from the very beginning and never shows a plateau. There is no yield point to read, so the codes define one: the 0.2 % proof stress, the stress that leaves 0.002 of permanent strain behind. For 1.4301 that is 210 MPa, and it is a convention with a measurement attached, not a physical corner.

The consequence lands on stiffness, not on strength. EN 1993-1-4 models the curve as ε = σ/E + 0.002(σ/f₀.₂)ⁿ with n ≈ 6 for austenitics, so the secant modulus falls as the stress rises. At this beam's serviceability stress of 160.22 MPa, which is 0.763 of the proof stress, the total strain is 1.1957 × 10⁻³ against an elastic 8.011 × 10⁻⁴. The secant modulus is 134,005 MPa, 67.0 % of the nominal 200 GPa.

Which means the deflection of a stainless beam is not the 29.67 mm a linear solver reports. It is roughly 44.28 mm, 49.2 % more, and this is precisely why EN 1993-1-4 asks for a secant modulus at serviceability instead of the nominal E. A stainless member designed by a linear run and checked against a deflection limit can be wrong by half.

Aluminium alloys behave the same way, for the same reason, and are handled the same way in EN 1999-1-1. The general lesson is worth more than either case: the elastic constants you feed a solver are a model of the material, and models have a validity range. Ask where the range ends before you trust the third decimal.

One more place the material property behaves nothing like the intuition. A HEA 200 column, buckling about the weak axis, checked with EN 1993-1-1 §6.3.1 on curve b, in S235 and in S460.

At 1 m the stubby column is a squash problem and S460 delivers 1.897 times the capacity, almost the full 1.9574 ratio of the yield strengths. At 12 m it delivers 1.041, so 96 % more yield strength has bought 4 % more column.

The reason is that a slender column is not limited by strength at all. The Euler load of this member at 6.0 m is 768.1 kN and it contains E, I and the length: fy is nowhere in it. The design curve only bends the elastic answer down for imperfections and residual stress. As the member gets more slender the answer converges on Euler, and Euler does not care what grade is stamped on the flange.

So the same catalogue upgrade is a bargain in a short column, a waste in a long one, and precisely nothing in a deflection-governed beam. The material is not good or bad on its own. It is good or bad relative to the limit state that is actually governing, which is a property of the structure, not of the metal.

Everything above collapses into a short procedure.

• Compute the stress without thinking about the material. In a determinate member it cannot depend on the material, and that is a check on your model, not an approximation.
• Then ask which limit governs. Strength responds to grade. Stiffness responds only to E and I. Stability responds mostly to E. Fatigue ignores grade completely. Fire responds to grade strongly.
• Read the utilisation together with the code that produced it. The same fy = 250 MPa is 227.27 MPa in NBR 8800 and 225.00 MPa in AISC 360.
• Look for mixed materials in one load path. That is the only place where changing a material rewrites the stress, and this beam's moment moved 1.6758 times when the hanger changed.
• Check whether your elastic constants are still valid. Stainless at 0.763 of proof stress is already at 67.0 % of its nominal modulus, and a linear deflection is then 49.2 % optimistic.

The beam that opened the article is a good place to finish. It was at 226.60 MPa, which passed at 0.997 in Brazil and failed at 1.007 in the United States, was comfortable at 0.493 in S460, and failed all three of those times anyway because it deflected 28.26 mm against a 26.67 mm limit. Its actual answer was never a material. It was an IPE 400.