4. Yield criteria calibration

Yield criteria define the boundary between elastic and plastic response in a constitutive model. Calibrating them means extracting the constants that make the criterion match the material's measured yield behavior under the stress states the component will actually experience. This chapter covers three widely used criteria — von Mises for isotropic metals, Hill48 for orthotropic sheet metals, and Drucker-Prager for pressure-sensitive materials such as polymers, soils, and concrete — and shows how to derive their parameters from uniaxial, biaxial, and triaxial test data.

Von Mises: A Single Parameter from Uniaxial Yield

The von Mises criterion assumes yielding occurs when the second deviatoric stress invariant J₂ reaches a critical value. In terms of the equivalent stress:

σ_eq = √(½[(σ₁−σ₂)² + (σ₂−σ₃)² + (σ₃−σ₁)²])

Yielding initiates when σ_eq = σ_y. The only calibration parameter is the uniaxial yield stress σ_y, obtained directly from a tensile specimen's 0.2% offset yield point. For a typical annealed 304 stainless steel, a uniaxial test might give σ_y = 310 MPa. Because von Mises is pressure-insensitive and isotropic, this single value automatically predicts yield under pure shear (τ_y = σ_y/√3 ≈ 179 MPa) and under balanced biaxial tension (σ_biaxial = σ_y).

Verification is straightforward: run a biaxial cruciform test or a thin-walled tube under combined axial-tension/internal-pressure loading. If the measured biaxial yield deviates from 310 MPa by more than about 5%, the material is not well described by von Mises, and an anisotropic or pressure-sensitive criterion should be considered.

Hill48: Six Parameters from Directional Yield Stresses

Hill48 extends von Mises to orthotropic materials, typically rolled sheet. Its quadratic form in the material coordinate system (1 = rolling direction, 2 = transverse direction, 3 = through-thickness) is:

2f = F(σ₂−σ₃)² + G(σ₃−σ₁)² + H(σ₁−σ₂)² + 2Lτ₂₃² + 2Mτ₃₁² + 2Nτ₁₂²

The six coefficients F, G, H, L, M, N are calibrated from directional yield stresses. Using the associated flow rule and assuming the through-thickness stress σ₃ ≈ 0 for sheet, the in-plane parameters follow from three uniaxial yield stresses and one biaxial yield stress:

For a 6016-T4 aluminum sheet, suppose uniaxial tests give σ_0° = 125 MPa (rolling), σ_90° = 118 MPa (transverse), and a biaxial cruciform test gives σ_b = 130 MPa. The through-thickness yield σ₃₃ is often approximated as the average of σ₁₁ and σ₂₂ when direct measurement is impractical. Substituting:

σ11 = 125 MPa, σ22 = 118 MPa, σ33 ≈ 121.5 MPa
F = 0.5 * (1/118² + 1/121.5² − 1/125²) ≈ 0.5 * (7.18e-5 + 6.77e-5 − 6.40e-5) ≈ 3.78e-5
G = 0.5 * (1/121.5² + 1/125² − 1/118²) ≈ 0.5 * (6.77e-5 + 6.40e-5 − 7.18e-5) ≈ 3.00e-5
H = 0.5 * (1/125² + 1/118² − 1/121.5²) ≈ 0.5 * (6.40e-5 + 7.18e-5 − 6.77e-5) ≈ 3.41e-5
N from shear test: τ12 = 72 MPa → N = 1/(2 * 72²) ≈ 9.65e-5

These coefficients are then normalized so that the reference yield stress (usually σ_0°) matches the input. The resulting yield locus is an ellipse elongated along the direction of higher yield stress. A common validation is to compare the predicted equibiaxial yield stress against the measured value; if the error exceeds 5–8%, the quadratic Hill48 form may be insufficient, and a non-quadratic criterion such as Hill90 or Yld2000-2d should be evaluated.

Drucker-Prager: Two Parameters from Triaxial Data

Drucker-Prager introduces pressure sensitivity through a linear combination of the first stress invariant I₁ and the second deviatoric invariant J₂:

√(J₂) = k − α I₁

where α controls pressure sensitivity and k is the shear strength at zero confining pressure. These two parameters are calibrated from at least two triaxial tests at different confining pressures. For a polymer such as polycarbonate, suppose triaxial compression tests yield:

Fitting a straight line to √J₂ versus I₁ gives the slope −α and intercept k. Using the first and third data points:

α = −(68.1 − 35.8) / (72.7 − 20.7) = −32.3 / 52.0 ≈ −0.621
k = 35.8 + (−0.621)(20.7) ≈ 35.8 − 12.9 ≈ 22.9 MPa

Note the sign convention: α is negative here because yield strength increases with confining pressure. In many FEA codes the Drucker-Prager parameters are entered as the friction angle β and cohesion d, related to α and k by:

tan β = −3α  →  β ≈ 61.8°
d = k √(1 + 3α²)  →  d ≈ 22.9 * √(1 + 3*0.386) ≈ 22.9 * 1.48 ≈ 33.9 MPa

These values are then entered into the material card. A validation run should reproduce the three triaxial yield points within 5%; if the fit is poor, a parabolic or exponent-form Drucker-Prager model may be needed.

Calibration Workflow and Verification

The three criteria share a common workflow: acquire yield data under distinct stress states, fit the criterion's parameters, and verify against an independent stress state not used in the fit. The SVG below summarizes the data-to-parameter mapping and the verification loop.

Common Pitfalls and Troubleshooting

• Using only uniaxial data for Hill48: A single tensile test cannot resolve orthotropy. At minimum, three directional yield stresses plus one biaxial point are needed. Without biaxial data, the equibiaxial yield prediction is unconstrained and often wrong by 10–20%.
• Ignoring the sign of α in Drucker-Prager: Most FEA codes expect α positive for materials that weaken under tension (e.g., concrete) and negative for materials that strengthen under compression (e.g., polymers, soils). Check the code's sign convention before entering values.
• Confusing yield stress with flow stress: Calibration must use the initial yield point, not the ultimate tensile strength. Using UTS inflates σ_y and shifts the entire yield locus outward.
• Neglecting r-value consistency in Hill48: If the calibrated Hill48 parameters predict r-values that differ significantly from measured ones, the flow rule assumption (associated plasticity) may not hold, and a non-associated model should be considered.

In practice, the calibration sequence is: (1) collect yield data across the stress states relevant to the application; (2) select the simplest criterion that captures the observed behavior; (3) fit parameters using least-squares or closed-form expressions; (4) verify against held-out data; (5) if verification fails, either add more test data or upgrade to a more flexible criterion. The three criteria covered here form a natural progression: von Mises for isotropic pressure-insensitive metals, Hill48 for orthotropic sheet, and Drucker-Prager for pressure-sensitive materials. Selecting the right one is as important as fitting its parameters accurately.