Default FEA visualization for metal structures plots a single scalar at every node: the von Mises equivalent stress. Code-compliance checks under Eurocode 3, the ASME BPVC, and DNV recommended practices all reduce to comparing this scalar against the design yield strength, divided by a material partial factor. The criterion itself goes back to Tadeusz Huber’s 1904 paper, was independently rederived by Richard von Mises in 1913, and received its physical interpretation through distortion energy from Heinrich Hencky in 1924.
What follows examines what the criterion actually computes and where its assumptions stop working.
From Huber to Hencky: Origins of the Criterion
The earliest publication belongs to Tadeusz Huber. His 1904 paper in Czasopismo Techniczne (Lemberg, Vol. 22, p. 181) formulated yield onset through distortion strain energy. The text was in Polish and received almost no attention outside Poland.
Nine years later, Richard von Mises arrived at the same criterion independently. In ”Mechanik der festen Körper im plastisch-deformablen Zustand”, submitted to the Göttingen Academy of Sciences on November 1, 1913, von Mises justified the choice explicitly. Instead of the Tresca-Mohr hexagon, he proposed a circumscribed circle for “much simpler analytical treatment.”
Heinrich Hencky added the physical interpretation through deviatoric strain energy in 1924 (Z. Angew. Math. Mech., vol. 4, 323-334). Taylor and Quinney’s 1931 experiments on copper, aluminum, and mild steel showed that the Mises criterion captures the onset of plastic flow more accurately than Tresca. Modern literature sometimes calls the chain of ideas the Maxwell-Huber-Hencky-von Mises theory.
The Deviatoric Stress Tensor and Its Second Invariant
At any point in a continuum, the stress tensor σ_ij decomposes into hydrostatic and deviatoric parts:
σ_ij = (1/3) δ_ij σ_kk + σ’_ij.
The first term captures uniform pressure. The second carries every shape distortion. The Mises criterion uses only the deviatoric part. Compactly:
σ_VM = √(3/2 · σ’_ij σ’_ij) = √(3 J₂),
where J₂ = (1/2) σ’_ij σ’_ij is the second invariant of the deviatoric tensor. In principal stresses, this gives the classical form:
σ_VM = √[(1/2) · ((σ₁ − σ₂)² + (σ₂ − σ₃)² + (σ₃ − σ₁)²)].
In arbitrary Cartesian components:
σ_VM = √[(1/2) · ((σ_xx − σ_yy)² + (σ_yy − σ_zz)² + (σ_zz − σ_xx)²) + 3 · (τ_xy² + τ_yz² + τ_zx²)].
The factor of 3/2 is not arbitrary. Total elastic strain energy splits into hydrostatic and deviatoric parts with no cross terms, thanks to their orthogonality, and the scaling is chosen so that σ_VM equals the applied stress in uniaxial tension. A direct consequence: in pure shear, σ_VM = √3 · τ, so the shear yield strength is τ_y = σ_y / √3 ≈ 0.577 σ_y.
Geometry of the Yield Surface
In principal stress space (σ₁, σ₂, σ₃), the Mises yield surface is an infinite circular cylinder whose axis coincides with the hydrostatic diagonal σ₁ = σ₂ = σ₃. In Haigh-Westergaard coordinates, the cylinder radius equals √(2/3) · σ_y.
The geometry encodes a physical fact: pure all-around compression or tension does not yield metals. Percy Bridgman’s experiments at Harvard in the 1940s showed metals remaining elastic under enormous hydrostatic pressures. Moving the loading point along the cylinder axis never brings it closer to yield.
In plane stress (σ₃ = 0), the cylinder cuts an ellipse rotated 45° from the σ₁ and σ₂ axes. Points inside the ellipse remain elastic, the ellipse itself marks the yield onset, and points outside are physically unreachable within elastic analysis.
Comparison with Alternative Criteria
The Tresca criterion limits the maximum shear stress: τ_max = σ_y / 2. Its surface in principal stress space is a hexagonal prism inscribed in the Mises cylinder. Pure shear yields at τ_y = 0.5 σ_y, roughly 15% more conservative than the Mises value of 0.577 σ_y. The smaller elastic region produces heavier cross-sections at the same load, so Tresca appears where a regulator requires an extra margin.
The Rankine criterion (maximum principal stress) checks σ_I ≥ σ_c and applies to brittle materials such as cast iron, ceramics, and glass. It captures none of the shear physics behind metal plasticity.
For pressure-dependent media, the Mises form extends into the Drucker-Prager cone (Drucker and Prager, Quarterly of Applied Mathematics, 1952): √J₂ = A + B · I₁. Mises emerges as the special case B = 0. Anisotropic materials, including rolled steel with strong texture and unidirectional carbon-fiber composites, call for the Hill generalization. Concrete, soils, and rocks use Mohr-Coulomb with its hexagonal pyramid surface.
Von Mises Stress in Design Codes
Eurocode 3 EN 1993-1-1 formalizes the criterion in Section 6.2.1. Formula 6.1 writes von Mises stress in normalized form:
(σ_x,Ed / (f_y / γ_M0))² + (σ_z,Ed / (f_y / γ_M0))² − σ_x,Ed · σ_z,Ed / (f_y / γ_M0)² + 3 · (τ_Ed / (f_y / γ_M0))² ≤ 1.
The recommended partial factor is γ_M0 = 1.00. The code itself notes that this check is conservative: it excludes partial plastic stress redistribution that elastic design otherwise allows. Engineers apply Formula 6.1 only when interaction through N_Rd, M_Rd, and V_Rd resistances is not available.
EN 1993-1-5 Section 10 (Reduced Stress Method) uses the same criterion to assess the interaction of σ_x, σ_z, and τ in plate buckling. The factor α_ult,k is the smallest load multiplier at which the equivalent stress in the panel reaches f_y.
ASME BPVC Section VIII Division 2, Part 5 takes a different route. The stress tensor is first linearized along a Stress Classification Line, the equivalent stress is then computed from the linearized tensor, and the result is compared to the limits in Figure 5.1 (P_m ≤ S, P_L ≤ 1.5 S, P_m + P_b + Q ≤ 3 S_m). Reading the peak FEA value directly and comparing it to ASME allowables is methodologically incorrect.
DNV-RP-C208 (Edition 2019-09, Amended 2022-10) states it briefly: the Mises yield function is considered suitable for most capacity analyses of steel structures. For nonlinear FEA of offshore structures, it serves as the de facto standard.
Numerical Pitfalls
In a finite-element solve, the equivalent stress is evaluated at Gauss integration points, extrapolated to nodes, and averaged across adjacent elements. Discontinuities in the averaged nodal values at element boundaries flag inadequate mesh density.
Singularities at sharp re-entrant corners and at concentrated load points remain a separate problem. The theoretical stress at such a point is infinite, so refining the mesh does not converge it to a finite value. Sound practice requires either a realistic fillet on the geometry or a hot-spot stress extrapolation procedure for fatigue checks.
The standard mesh convergence threshold is a change of less than 3% in the peak stress under the next refinement. ASME VIII-2 checks require at least 2-3 elements through the thickness, otherwise the bending contribution to the linearized tensor is captured incorrectly.
The σ_VM value is always non-negative. A single number does not distinguish tension from compression, so interpreting a loading state still requires the principal stresses or a hydrostatic-signed version of the equivalent stress.
What the Criterion Does Not Predict
Mises assumes material isotropy and independence from the first stress invariant. Rolled steel with pronounced texture, unidirectionally reinforced composites, and α-titanium are better handled by the Hill criterion. Concrete, soils, and porous polymers require pressure-dependent forms: Drucker-Prager, Mohr-Coulomb, and cap models.
The criterion describes only the onset of plastic flow. It does not address structural stability (separate procedures in EN 1993-1-1 Sections 6.3 and EN 1993-1-5), fatigue (S-N curves, hot spot, fracture mechanics), creep, or the failure of welded and bolted connections. In modern structural verification workflows, σ_VM is one of the mandatory checks, not a substitute for the others.
The criterion is often called an empirical coincidence. Statistics of dislocation motion in a randomly oriented polycrystalline medium happen to align with the distortion-energy condition, yet there is no rigorous physical law behind it. Even so, a century of testing on copper, steel, aluminum, and various alloys keeps it the default choice for ductile metals under moderate conditions.