Faculty of Materials Science and Technology
in Trnava

Material Data in DEFORM

Lecture 3  ·  Modelling and Simulation of Processes I – Forming (MSPO2_6I)

Ing. Martin Necpal, PhD.

Winter semester 2026/2027
Materiálovotechnologická fakulta  ·  STU Bratislava

DEFORM simulation of a piston forging punch

Piston forging – punch and workpiece in DEFORM (© SFTC)

Today's agenda (120 min, no break)

TimeBlockContent
0 – 101 · Why it mattersGarbage in, garbage out — one part, two flow curves
10 – 202 · Material windowLayout, library, Regular vs Mixture, units
20 – 303 · Elastic dataE, ν, α(T), hyperelastic materials
30 – 604 · Flow stressTesting, tables, power law, Johnson–Cook, Zerilli–Armstrong, sinh-Arrhenius, choosing a model
60 – 805 · Yield & hardeningvon Mises, Hill 48, R-values, Barlat; isotropic, kinematic, Yoshida–Uemori
80 – 906 · CreepCreep curve, DEFORM creep models, relaxation
90 – 1057 · Thermal datak, heat capacity, emissivity, density, interface HTC
105 – 1158 · Beyond mechanicsGrain (JMAK), phases, hardness, diffusion, EM, fracture
115 – 1209 · Wrap-upChecklist per process, quiz

Structure follows DEFORM V14 manual, Chapter 10 “Material Data”

Why it mattersGarbage in, garbage out

Same part, same mesh, same friction — two flow curves

  • Upsetting Ø20 × 30 mm → 15 mm, μ = 0.1
  • A: library curve “C45, cold”
  • B: own test of the “same” grade (different heat, annealed)
  • End force: 540 kN vs 427 kN — Δ ≈ 27 %
  • Mesh refinement changes the force by a few % — the material data by tens of %
Material data drive: force & energy · temperature · die fill · damage · springback · grain size · hardness
0 5 10 15 0 200 400 600 800 Stroke [mm] Upsetting force [kN] Flow curve A: σ = 860 ε^0.17 MPa Flow curve B: σ = 700 ε^0.25 MPa Δ ≈ 27 %

Slab-method estimate with σ = Kεⁿ (computed for this slide)

Material windowLayout

The DEFORM Material window

DEFORM material window, Elastic tab, material C10100

Material window, Elastic tab — copper C10100 (DEFORM-3D)

  • Material list — all materials of the project; each object points to one
  • Regular / Mixture — single phase or several phases
  • Phase — Add, Remove, Transformation, Grain
  • Tabs: Plastic · Elastic · Thermal · Diffusion · Dislocation · Hardness · Elec./Mag. · Advanced · Texture
  • Buttons: New, Delete, Load from lib, Save in lib, Import, Export, Copy prop., Unit conv.
  • Each property: Constant or f(Temp), edited in a table (pencil icon)
Material windowLibrary

Material library: load, save, import, export

Load from lib — choose class (steel, Al, Ti, Ni, Cu, die steels …), grade and temperature range
↓
Material copied into the project (.KEY / .DB) — edits stay local
↓
Save in lib — your own tested data for the next project
↓
Export / Import — material file (.KEY) between users, sites, versions
Library grades carry a validity range in the name, e.g. cold 20 °C or hot 900–1200 °C. A “cold” curve in a hot forging is a silent error.
DEFORM multiple operations of a disk forging

Multiple operations — heating, forging, trimming: one material, many temperatures (© SFTC)

Material windowRegular vs Mixture

Regular material or Mixture of phases?

RegularMixture
Phasesone2 … n (austenite, ferrite, pearlite, bainite, martensite …)
Dataone set per tabone set per phase + transformation data
Useforming, machiningheat treatment, quenching, carburising
Mixture rule: \( P = \sum_i \xi_i\,P_i, \quad \sum_i \xi_i = 1 \)
  • \(\xi_i\) volume fraction of phase \(i\), \(P_i\) its property (σ̄, k, E …)
  • Transformation button: kinetics (diffusion / martensite), latent heat, volume change, TRIP
Carburised helical gear martensite fraction

Carburised and quenched helical gear — martensite fraction (© SFTC)

Material windowUnits

Units: the C10100 trap — and which tabs you need

QuantityEnglish (C10100 above)SI in DEFORM
E16 679 ksi115 000 MPa
α9.44·10⁻⁶ /°F1.7·10⁻⁵ /°C
Tref68 °F20 °C
StressksiMPa = N/mm²
kBtu/(s·in·°F)N/(s·°C) = W/(m·K)
ρ·cpBtu/(in³·°F)N/(mm²·°C) = MJ/(m³·K)
ρlb/in³t/mm³ (7.7·10⁻⁹ for steel)

Unit system is set per problem; mixing an English library entry into an SI problem → use Unit conv.

SimulationPlasticElasticThermalPhase / grain
Isothermal forging●–––
Hot / warm forging●–●○
Springback, residual σ●●○–
Die stress (elastic die)–●○–
Heat treatment●●●●
Induction heating○○●● + EM

● required   ○ recommended   – not used

Elastic dataWhen needed

When does DEFORM need elastic data?

  • Rigid-plastic object: elastic strain neglected → no E, ν needed (most forging)
  • Elastic object: dies, die inserts, shrink rings → E, ν, α
  • Elasto-plastic object: springback, residual stress, distortion, heat treatment
$$\varepsilon = \varepsilon^{e} + \varepsilon^{p} + \varepsilon^{th} + \varepsilon^{tr}$$

elastic + plastic + thermal + transformation strain

Elastic deformation diagram

Elastic range: strain recovers fully on unloading

DEFORM die stress analysis

Die stress analysis with elastic dies (© SFTC)

Elastic dataE, ν

Young's modulus and Poisson's ratio

$$\sigma = E\,\varepsilon^{e}, \qquad \nu = -\frac{\varepsilon_{lat}}{\varepsilon_{ax}}, \qquad G=\frac{E}{2(1+\nu)}$$
MaterialE [GPa]ν
Carbon steel200 – 2100.29 – 0.30
Stainless steel (austenitic)193 – 2000.29
Aluminium alloys69 – 720.33
Titanium Ti-6Al-4V110 – 1140.34
Copper115 – 1200.33
WC-Co die insert500 – 6200.22
  • Springback ∝ σ/E: Al springs back ≈ 3× more than steel at equal stress
0.01 0.1 1 10 100 1000 Rubber 0.05 Low-Density PolyethyleneLDPE 0.2 Nylon 3 Oak (along grain) 11 Magnesium 45 Al alloys 69 Glass 72 Titanium 112 Copper 117 Steel 200 Tungsten 405 Tungsten CarbideWC (cemented carbide) 550 Diamond 1100 Young's modulus E [GPa] (log scale)

Young's modulus across materials (typical handbook values)

Elastic dataTemperature dependence

E(T) and thermal expansion α(T)

0 250 500 750 1 000 0 50 100 150 200 Temperature [°C] Young's modulus [GPa] Carbon steel, E(T) = k_E · 210 GPa

E(T) of carbon steel — EN 1993-1-2 reduction factor kE

Thermal expansion coefficient of steels vs temperature in DEFORM

α(T) of several steels in DEFORM (course lecture)

$$\varepsilon^{th} = \alpha(T)\,(T - T_{ref}) \qquad \text{DEFORM: “Thermal expansion” + “Reference temp.”}$$
  • E of steel falls to about 1/3 at 600 °C and almost zero above 1000 °C — use f(Temp) for quenching and hot die stress
  • α is the secant value from Tref; austenite expands more than ferrite (≈ 2.2 vs 1.4·10⁻⁵ /K)
Elastic dataHyperelastic

Hyperelastic materials: rubber pads and seals

  • Large reversible strains (100–500 %), nearly incompressible (ν → 0.5)
  • Stress from a strain-energy function \(W\) instead of E, ν
Neo-Hookean: \( W = C_{10}(\bar I_1 - 3) \)
Mooney–Rivlin: \( W = C_{10}(\bar I_1-3)+C_{01}(\bar I_2-3) \)
Uniaxial: \( \sigma_{nom} = 2C_{10}\,(\lambda - \lambda^{-2}) \),   \(G = 2C_{10}\)
  • Forming use: rubber-pad forming, flexible dies, seals in assemblies
  • Fit C10, C01 from uniaxial + biaxial tests — never from one test only
0.5 1 1.5 2 2.5 3 -6 -4 -2 0 2 4 Stretch λ = L / L₀ [–] Nominal stress [MPa] Neo-Hookean (G = 1 MPa) Linear elastic, E = 3G

Neo-Hookean vs linear elastic in tension and compression

Flow stressDefinition

Flow stress: \( \bar\sigma = f(\bar\varepsilon,\ \dot{\bar\varepsilon},\ T) \)

Engineering vs true stress-strain curve

Engineering vs true curve

Plastic deformation diagram

Permanent strain after the yield point

$$\sigma_t = \sigma_e(1+e), \quad \varepsilon = \ln(1+e)$$
  • Flow stress = true stress needed to keep plastic flow going
  • Depends on strain (hardening), strain rate and temperature; also on prior history
  • Rigid-plastic FEM uses only the plastic part — the curve starts at the yield stress
  • Engineering values are valid only up to necking (uniform elongation)
Flow stressT and ε̇

Temperature softens, strain rate strengthens

Flow stress vs temperature and strain rate

Stress–strain curves vs temperature and strain rate (course lecture)

RegimeT / TmDominant effect
Cold< 0.3Strain hardening, weak rate effect (m ≈ 0–0.05)
Warm0.3 – 0.5Both, recovery starts
Hot> 0.5 – 0.6Rate + T; hardening balanced by recovery / DRX (m ≈ 0.1–0.2)
  • Strain-rate sensitivity \( m = \partial\ln\bar\sigma/\partial\ln\dot{\bar\varepsilon} \)
  • Superplastic alloys: m ≥ 0.3–0.5
  • Adiabatic heating: \( \Delta T \approx \dfrac{\beta\int\bar\sigma\,d\bar\varepsilon}{\rho\,c_p} \), β ≈ 0.9
Flow stressTesting

Where flow curves come from

ZwickRoell universal testing machine

Tensile test
ε < 0.2–0.3, ε̇ ≈ 10⁻³–10⁻¹ s⁻¹ (© ZwickRoell)

Flow curve from a test

Compression (upsetting)
ε ≤ 1, friction correction

Gleeble thermomechanical simulator

Gleeble hot compression
up to ≈ 1700 °C, ε̇ to 10² s⁻¹ (© DSI)

▶ video: Introduction to Gleeble Systems

striker incident bar specimen transmitter bar strain gaugestrain gauge ε̇ ≈ 10² – 10⁴ s⁻¹ · stress from transmitted wave, strain rate from reflected wave

Split Hopkinson bar (SHPB)
ε̇ ≈ 10²–10⁴ s⁻¹

TestStrengthLimitationDEFORM use
TensionStandard, simple stress stateNecking limits strainSheet, cold forming start of curve
Compression / plane strainLarge strains, hotFriction, barrelling → inverse correctionBulk forming, forging
TorsionVery large strains (ε > 5)Gradient through radiusHot rolling, DRX
SHPBHigh ratesSmall strain, adiabaticMachining, impact (J–C)
Flow stressTabular data

Tabular flow stress \( \bar\sigma(\bar\varepsilon, \dot{\bar\varepsilon}, T) \) and interpolation

DEFORM flow stress graph and data table

DEFORM flow-stress table: curves per T and ε̇ (course lecture)

100 101 102 103 900 950 1 000 1 050 1 100 Strain rate ε̇ [s⁻¹] (log scale) σ̄ at ε̄ = 0, 20 °C [MPa] Linear interpolation in ε̇ Log interpolation in ε̇ 1028 MPa1012 MPaε̇ = 50 s⁻¹

Value at ε̇ = 50 s⁻¹ from the same four table points: 1012 vs 1028 MPa

  • Table = curves for several T × ε̇, each sampled in ε̄ — the most common format in DEFORM libraries
  • Between points DEFORM interpolates; strain rate spans decades → logarithmic interpolation in ε̇ is the physically sensible choice
  • Typical table: 5–8 temperatures, 4–6 strain rates (10⁻² … 10² s⁻¹), strains 0 … 2
Flow stressExtrapolation

Outside the table: extrapolation is a guess

10-1 100 101 102 103 104 900 950 1 000 1 050 1 100 1 150 Strain rate ε̇ [s⁻¹] σ̄ at ε̄ = 0.3, 20 °C [MPa] Material behaviour (Johnson–Cook) Table up to 10 s⁻¹, clamped above measured rangeextrapolated

Table measured to 10 s⁻¹, used at 10⁴ s⁻¹: flow stress 8.2 % too low

10-9 10-6 10-3 100 103 106 0.0 0.2 0.4 0.6 0.8 1.0 Strain rate ε̇ [s⁻¹] Homologous temperature T / Tm Tensile test Gleeble / hot compression Split Hopkinson Pressure BarSHPB Creep test Cold forging Hot forging / rolling Machining Extrusion Service creep

Process windows (solid) vs test windows (dashed) — schematic

Before running: read the post-processor ranges of ε̄, ε̇ and T and compare them with the table limits. DEFORM does not warn when it leaves the table.
Flow stressPower law

Power law — the cold forming workhorse

DEFORM: $$\bar\sigma = c\,\bar\varepsilon^{\,n}\,\dot{\bar\varepsilon}^{\,m} + y$$
Hollomon: \( \bar\sigma = K\bar\varepsilon^{\,n} \)   Swift: \( \bar\sigma = K(\varepsilon_0+\bar\varepsilon)^{n} \)
  • \(c, n, m, y\) may be f(T) → simple warm forming
  • \(n\) = strain-hardening exponent = uniform strain in tension
  • Pure Hollomon gives σ = 0 at ε = 0 → use \(y\) or Swift ε0
Material (annealed)K [MPa]n
Low-C steel DC04≈ 5300.22 – 0.26
C45 (normalised)≈ 9000.15 – 0.18
Austenitic stainless 304≈ 13000.40 – 0.50
Al 5xxx≈ 4000.25 – 0.30
0 0.2 0.4 0.6 0.8 1 0 200 400 600 800 Effective strain ε̄ [–] Flow stress σ̄ [MPa] c = 700 MPa, n = 0.10 c = 700 MPa, n = 0.20 c = 700 MPa, n = 0.35

σ̄ = c ε̄ⁿ for three hardening exponents

▼ Deep dive: 8 slides on hardening laws and the power law — press ↓
Power law ↓1 · The family of hardening laws

From Ludwik to Voce: one flow curve, several equations

LawEquationBehaviour at large εTypical use
Ludwik (1909)\( \sigma = \sigma_0 + K\varepsilon^{n} \)unboundedgeneral, with initial yield
Hollomon (1945)\( \sigma = K\varepsilon^{n} \)unbounded, σ(0) = 0sheet n-value, quick estimates
Swift (1952)\( \sigma = K(\varepsilon_0 + \varepsilon)^{n} \)unbounded, σ(0) = Kε₀ⁿsheet forming, prestrained material
Voce (1948)\( \sigma = \sigma_s - (\sigma_s - \sigma_0)\,e^{-\varepsilon/\varepsilon_c} \)saturates at σsAl alloys, large strains
DEFORM power law\( \bar\sigma = c\,\bar\varepsilon^{\,n}\,\dot{\bar\varepsilon}^{\,m} + y \)unboundedcold forging, heading, extrusion
  • All are empirical: they describe the shape of the curve, not the dislocation physics behind it
  • DEFORM's form = Ludwik + rate term: \(y\) plays the role of σ₀; \(c, n, m, y\) can be f(T)
  • Mixed forms (Swift–Voce: \( \sigma = w\,\sigma_{Swift} + (1-w)\,\sigma_{Voce} \)) are used for AHSS sheet
Power law ↓2 · Meaning of n

Strain-hardening exponent n and necking

Hardening rate: \( \dfrac{d\sigma}{d\varepsilon} = nK\varepsilon^{n-1} = n\,\dfrac{\sigma}{\varepsilon} \)
Considère (max. force, F = σA): \( dF = 0 \Rightarrow \dfrac{d\sigma}{d\varepsilon} = \sigma \)
Hollomon ⇒ \( \varepsilon_u = n \),   \( R_m = K\,n^{n}e^{-n} \)
  • n measures how long the material spreads deformation before it localises
  • High n (austenitic stainless ≈ 0.45) → large uniform elongation, strong work hardening in drawing
  • Low n (cold-worked, HSLA ≈ 0.1) → early necking, little extra strength from forming
  • Check of any fit: n ≈ uniform true strain of the tensile test
0 0.1 0.2 0.3 0.4 0.5 0 200 400 600 800 True strain ε [–] Stress [MPa] true stress σ = Kεⁿ hardening rate dσ/dε = nσ/ε engineering stress F/A₀ necking: dσ/dε = σ ⇒ ε_u = n = 0.24

Considère construction for K = 530 MPa, n = 0.24 (DC04-type steel)

Power law ↓3 · Measuring K and n

Determining K and n: the log–log plot

$$\ln\sigma = \ln K + n\,\ln\varepsilon$$
  1. Tensile test → convert to true stress and true plastic strain (subtract σ/E)
  2. Use the uniform range only, e.g. ε = 2 % … Ag (ISO 10275: n-value)
  3. Linear regression of ln σ on ln ε → slope n, intercept ln K
  4. Ludwik / DEFORM form: first fix y (σ₀ = yield), then fit ln(σ − y) vs ln ε
  5. Cold forging: add compression data to ε ≈ 1 (friction-corrected) — tensile data stop at necking

Not a straight line on log–log? The material does not follow a power law (e.g. Lüders plateau, saturation) — choose Voce or a table.

10-2 10-1 100 200 300 400 500 700 True plastic strain ε (log) True stress σ [MPa] (log) fit: K = 525 MPa, n = 0.237 slope = n, σ at ε = 1 → K

Synthetic tensile points (K = 530, n = 0.24) — regression returns n = 0.237

Power law ↓4 · Strain-rate exponent m

The strain-rate exponent m and the jump test

$$m = \frac{\partial\ln\bar\sigma}{\partial\ln\dot{\bar\varepsilon}}\bigg|_{\varepsilon,T} \approx \frac{\ln(\sigma_2/\sigma_1)}{\ln(\dot\varepsilon_2/\dot\varepsilon_1)}$$
Conditionm×10 in ε̇ gives
Steel, cold (20 °C)0.005 – 0.02+1 … 5 %
Al alloys, cold≈ 0 – 0.01≈ 0 … 2 %
Warm forming (0.3–0.5 Tm)0.03 – 0.1+7 … 26 %
Hot forming0.1 – 0.2+26 … 58 %
Superplastic≥ 0.3 – 0.5×2 … ×3
  • Jump test: change ε̇ by ×10 at constant ε and T → one specimen gives m
  • Positive m stabilises flow (neck strain rate ↑ → stress ↑) — basis of superplasticity
  • Steels at 150–350 °C: dynamic strain ageing → m < 0, serrated flow (“blue brittleness”)
0 0.1 0.2 0.3 200 300 400 500 True strain ε [–] True stress σ [MPa] σ(ε): ε̇ = 10⁻³ → 10⁻² → 10⁻³ s⁻¹ Δσ = 8.5 MPam = ln(σ₂/σ₁) / ln(ε̇₂/ε̇₁) = 0.012

Strain-rate jump test (m = 0.012, cold steel): small but measurable step

Power law ↓5 · The extrapolation trap

Same tensile data, four laws — and a factor of two at ε = 2

0 0.5 1 1.5 2 0 500 1 000 1 500 Effective strain ε̄ [–] Flow stress σ̄ [MPa] Hollomon: σ(2.0) ≈ 714 MPa Ludwik: σ(2.0) ≈ 1401 MPa Swift: σ(2.0) ≈ 1184 MPa Voce: σ(2.0) ≈ 667 MPa tensile dataextrapolated — cold forging range

All four laws fitted to the same points (ε = 0.02–0.20); scatter < 1 %

  • All four laws fit the tensile range with < 1 % error
  • At ε̄ = 2 (cold forging, heading): Voce 667, Hollomon 714, Swift 1184, Ludwik 1401 MPa
  • Press force and die stress scale with σ̄ → ≈ ± 35 % error in the load, just from the choice of law
  • Voce saturates (Al, Cu at large ε); power laws keep rising (steels with little recovery)
Rule: fit the law to data that cover the strain range of the process — compression or torsion to ε ≈ 1–2. Use tensile-only fits for sheet forming, not for cold forging.
Power law ↓6 · Temperature dependence

Power law with temperature: c(T), n(T), m(T), y(T)

  • DEFORM lets each constant be a table f(T) → the power law covers warm forming
  • With rising T: c ↓ (softer), n ↓ (recovery limits hardening), m ↑ (rate sensitivity grows)
  • Linear interpolation between the T points — give enough points around phase changes
  • Adiabatic heating in fast cold forging (ΔT 100–300 K) → even a “cold” simulation should be coupled thermal with c(T)
\( \Delta T \approx \dfrac{\beta}{\rho\,c_p}\int\bar\sigma\,d\bar\varepsilon \),   β ≈ 0.9
Tc [MPa]nmregime
20 °C9000.170.01cold
200 °C8500.15≈ 0 (DSA)avoid
500 °C6200.120.04warm
700 °C3300.080.10warm
900 °C1800.050.15hot → use sinh / table

Illustrative trend for a medium-carbon steel — use measured values for real work

Power law ↓7 · Applications

Where the power law is the right choice

ProcessWhy the power law
Cold forging, heading, cold extrusionHardening dominates, m ≈ 0; needs data to ε ≈ 2
Wire and bar drawingAccumulated strain over passes; n sets the drawing limit
Sheet stamping, deep drawingSwift / Hollomon with n and r-value; n → forming limit
Bending, roll formingModerate strains; springback needs E and hardening model too
Warm forgingc(T), n(T), m(T) tables
Keeler–Brazier (low-C steel): \( FLD_0\,[\%] = (23.3 + 14.13\,t)\,\dfrac{n}{0.21} \),   t in mm
0.05 0.1 0.15 0.2 0.25 0.3 0 20 40 60 80 Strain-hardening exponent n [–] FLD₀ (major strain, plane strain) [%] t = 0.7 mm t = 1.0 mm t = 2.0 mm

Higher n → higher forming limit in plane strain (empirical, low-carbon steels)

Power law ↓8 · Limits and summary

Limits, summary and questions

Limits

  • Empirical — no physics, so no safe extrapolation in ε, ε̇ or T
  • Unbounded hardening; Hollomon: σ(0) = 0 → use y (DEFORM) or Swift ε₀
  • Rate and T only through m and f(T) tables — not for hot forming or very high rates
  • No Bauschinger effect, no anisotropy — those come from the yield function and hardening rule (block 5)

Summary

  1. Ludwik, Hollomon, Swift, Voce: same data, different large-strain behaviour
  2. n = uniform strain (Considère); m from jump tests
  3. K, n from a log–log fit in the uniform range
  4. Fit to data covering the process strain — or the load can be off by ≈ ± 35 %

Check yourself

  1. Derive εu = n from the Considère criterion. F = σ·A; at maximum force dF = 0 → dσ/dε = σ. With σ = Kεⁿ: nKεⁿ⁻¹ = Kεⁿ → εu = n.
  2. Why does Hollomon need a y-term or Swift's ε₀ in FEM? Hollomon gives σ(0) = 0 (and an infinite initial slope), so undeformed material would flow at zero stress. The y term or Swift's ε₀ gives a finite initial yield stress.
  3. m = 0.15: how much does σ change if ε̇ rises 100×? σ₂/σ₁ = 1000.15 ≈ 2.0 – the flow stress roughly doubles.
  4. Your fit used tensile data to ε = 0.2. Can you use it for cold heading to ε = 2.5? No. That is extrapolation by more than ten times; laws fitted to the same tensile data differ by a factor of two at ε = 2. Use compression or torsion tests to ε ≈ 1–2.
  5. Which law saturates, and for which materials is that realistic? Voce saturates. Realistic for materials with strong dynamic recovery – aluminium and copper – and at elevated temperature; steels with little recovery keep hardening like a power law.

Press ↑ to return to the main lecture, or → to continue with Johnson–Cook.

Flow stressJohnson–Cook

Johnson–Cook — high rates, machining, impact

$$\bar\sigma = \underbrace{(A + B\bar\varepsilon^{\,n})}_{\text{hardening}}\underbrace{\left(1 + C\ln\dot\varepsilon^{*}\right)}_{\text{rate}}\underbrace{\left(1 - T^{*m}\right)}_{\text{softening}}$$
\( \dot\varepsilon^* = \dot{\bar\varepsilon}/\dot\varepsilon_0, \quad T^* = \dfrac{T - T_r}{T_m - T_r} \)
AISI 1045ABnCm
Jaspers & Dautzenberg553 MPa601 MPa0.2340.01341.0
  • Three independent factors — easy to fit, no coupling of T and ε̇
  • DEFORM form adds \(\dot\varepsilon^{*\alpha}\) and \((D - E\,T^{*m})\) for more freedom
  • Fails near Tm and for DRX-dominated hot forming
0 0.2 0.4 0.6 0.8 1 0 500 1 000 Effective plastic strain ε̄ [–] Flow stress σ̄ [MPa] ε̇ = 1 s⁻¹, 20 °C ε̇ = 10⁴ s⁻¹, 20 °C ε̇ = 10⁴ s⁻¹, 400 °C ε̇ = 10⁴ s⁻¹, 700 °C

Johnson–Cook, AISI 1045 (Tr = 20 °C, Tm = 1460 °C)

▼ Deep dive: 9 slides on Johnson–Cook — press ↓
Johnson–Cook ↓1 · Origin

A model built for impact codes

  • G. R. Johnson & W. H. Cook (Honeywell): “A constitutive model and data for metals subjected to large strains, high strain rates and high temperatures”, 7th Int. Symp. on Ballistics, The Hague, 1983
  • Goal: a model simple enough for explicit impact codes (EPIC) with constants obtainable from a few standard tests
  • Tests: torsion (quasi-static to ≈ 400 s⁻¹, large shear strains), static tension, Hopkinson-bar tension at elevated T
  • Validation: Taylor cylinder impact (strains > 2, rates > 10⁵ s⁻¹) — not used for fitting
  • Companion fracture model: Johnson & Cook, Eng. Fract. Mech. 21 (1985) 31
$$\bar\sigma = \big[A + B\bar\varepsilon^{\,n}\big]\big[1 + C\ln\dot\varepsilon^{*}\big]\big[1 - T^{*m}\big]$$
SymbolMeaningFrom test
Ayield stress at ε̇₀, Trquasi-static
B, nstrain hardeningquasi-static
Cstrain-rate sensitivitySHPB / torsion
mthermal softeningtests at T
ε̇₀, Tr, Tmreference rate (often 1 s⁻¹), reference and melting temperaturechosen / handbook
Johnson–Cook ↓2 · The three brackets

Three independent multipliers

10-3 10-2 10-1 100 101 102 103 104 105 0.6 0.8 1 1.2 1.4 1.6 1.8 ε̇* = ε̇/ε̇₀ [–] rate factor 1 + C ln ε̇* C = 0.0134 C = 0.025 C = 0.06 ε̇₀: factor = 1; below it < 1

Rate factor: linear in ln ε̇* — +12 % per 4 decades for C = 0.0134

0 0.2 0.4 0.6 0.8 1 0 0.25 0.5 0.75 1 T* = (T − T_r)/(T_m − T_r) [–] thermal factor 1 − T*ᵐ m = 0.55 m = 1.0 m = 1.7

Thermal factor: m < 1 softens early, m > 1 late; zero at Tm

  • Hardening [A + Bεⁿ]: Ludwik law — the flow curve at ε̇₀ and Tr
  • Rate [1 + C ln ε̇*]: below ε̇₀ the factor is < 1 (choose ε̇₀ = lowest test rate, or clip ε̇* ≥ 1); constant C → no upturn above ≈ 10³–10⁴ s⁻¹
  • Thermal [1 − T*ᵐ]: T* < 0 below Tr gives an undefined power for non-integer m → codes set T* = 0
Johnson–Cook ↓3 · Published constants

Constants from the original paper and later work

MaterialA [MPa]B [MPa]nCmTm [K]Source
OFHC copper902920.310.0251.091356J–C 1983
Cartridge brass1125050.420.0091.681189J–C 1983
Nickel 2001636480.330.0061.441726J–C 1983
1006 steel3502750.360.0221.001811J–C 1983
4340 steel7925100.260.0141.031793J–C 1983
2024-T351 Al2654260.340.0151.00775J–C 1983
AISI 10455536010.2340.01341.001733Jaspers & Dautzenberg 2002
  • Constants belong to one material condition (heat treatment, hardness) and one fitting range
  • Large scatter between published sets for the same grade — always check σ(ε) at your ε̇ and T before use
  • C is small (0.006–0.06): rate effects are moderate; m ≈ 1 means roughly linear softening with T
Johnson–Cook ↓4 · Calibration

Determining A, B, n, C and m

1 · A: yield stress at ε̇₀ and Tr (quasi-static tension/compression)
2 · B, n: \( \ln(\sigma - A) = \ln B + n\ln\varepsilon \) — log–log fit of the same test
3 · C: at fixed ε, \( \sigma/\sigma_{ref} = 1 + C\ln\dot\varepsilon^* \) → slope (SHPB at 10³–10⁴ s⁻¹)
4 · m: at fixed ε, ε̇: \( \ln(1 - \sigma/\sigma_{ref}) = m\ln T^* \)
5 · Check: correct SHPB curves for adiabatic heating; validate on an independent test (Taylor, cutting force)

Alternative: fit all five constants together by least squares (or inverse FEM of the tests) — better overall error, less transparent

-6 -3 0 3 6 9 0.9 0.95 1 1.05 1.1 1.15 ln ε̇* [–] σ(ε̇)/σ(ε̇₀) at ε̄ = 0.1 fit: slope C = 0.0137 quasi-staticSplit Hopkinson Pressure BarSHPB

Step 3 with synthetic points (C = 0.0134): regression gives C = 0.0137

Johnson–Cook ↓5 · Adiabatic heating

Thermal softening during fast deformation

0 0.2 0.4 0.6 0.8 1 0 500 1 000 Effective plastic strain ε̄ [–] σ̄ at ε̇ = 10⁴ s⁻¹ [MPa] isothermal, 20 °C adiabatic, β = 0.9: T rises to ≈ 278 °C

AISI 1045 at 10⁴ s⁻¹: plastic work heats the material by ≈ 258 K up to ε = 1

$$\rho\,c_p\,\dot T = \beta\,\bar\sigma\,\dot{\bar\varepsilon}, \qquad \beta \approx 0.9 \;(\text{Taylor–Quinney})$$
  • Above ≈ 10–100 s⁻¹ there is no time for heat conduction → adiabatic
  • J–C's thermal bracket then lowers the hardening curve as T rises
  • Hardening and softening compete → flat or falling curve → adiabatic shear bands, chip segmentation
  • In FEM: always a coupled thermo-mechanical run with ρ, cp(T), k(T); never fit J–C to SHPB curves as if they were isothermal
Johnson–Cook ↓6 · Limits and modified forms

Limitations — and the modified Johnson–Cook models

Known limitations

  • No coupling: rate sensitivity does not depend on T, hardening does not depend on ε̇ or T
  • Constant C → misses the upturn above 10³–10⁴ s⁻¹ (dislocation drag)
  • Accuracy falls at high T; no recovery, recrystallisation or phase change
  • Unbounded hardening (Ludwik) at large strains
  • BCC metals (yield strongly T-dependent) → Zerilli–Armstrong often better

Modified forms

DEFORM: \( \bar\sigma = (A + B\bar\varepsilon^{\,n})(1 + C\ln\dot\varepsilon^*)\,\dot\varepsilon^{*\alpha}\,(D - E\,T^{*m}) \)
Lin et al. 2010: \( \bar\sigma = (A_1 + B_1\varepsilon + B_2\varepsilon^2)(1 + C_1\ln\dot\varepsilon^*)\,e^{(\lambda_1 + \lambda_2\ln\dot\varepsilon^*)(T - T_r)} \)
Quadratic rate term: \( 1 + C_1\ln\dot\varepsilon^* + C_2(\ln\dot\varepsilon^*)^2 \)
  • Each adds coupling or curvature — and constants; the original stays the default in most codes
Johnson–Cook ↓7 · Fracture model

The Johnson–Cook fracture (damage) model

$$\varepsilon_f = \big[D_1 + D_2\,e^{D_3\eta}\big]\big[1 + D_4\ln\dot\varepsilon^*\big]\big[1 + D_5T^*\big]$$
$$D = \sum\frac{\Delta\bar\varepsilon}{\varepsilon_f} \;\ge\; 1 \;\Rightarrow\; \text{fracture (element deletion)}$$
  • \(\eta = \sigma_m/\bar\sigma\) stress triaxiality: shear 0, tension ⅓, notched bars 0.5–1
  • Same structure as the flow model: triaxiality × rate × temperature
  • 4340 steel (J–C 1985): D₁ = 0.05, D₂ = 3.44, D₃ = −2.12, D₄ = 0.002, D₅ = 0.61
  • Calibration: notched round bars (η), torsion (η ≈ 0), SHPB, hot tests
  • Used for chip separation, blanking, ballistic perforation; compare with Cockcroft–Latham (main slide 41)
0 0.5 1 1.5 0 1 2 3 4 5 Stress triaxiality η = σ_m / σ̄ [–] Failure strain ε_f [–] T* = 0.0, ε̇* = 1 T* = 0.3, ε̇* = 1 shearuniaxial tensionnotched bars

Failure strain falls steeply with triaxiality (4340 steel, J–C 1985)

Johnson–Cook ↓8 · Applications

Where Johnson–Cook is used

OperationWhy J–CTypical ε̇, T
Machining (turning, milling, drilling) — DEFORM-3D MachiningChip formation, cutting forces, tool temperature; J–C + J–C damage for segmentation10³–10⁶ s⁻¹, up to ≈ 1000 °C
Impact, ballistics, Taylor testOriginal purpose; robust in explicit codes10³–10⁵ s⁻¹, adiabatic
Crash of car structuresRate hardening of steel and Al sheet1–10³ s⁻¹
High-speed forming (electromagnetic, explosive, hammer)Rate + adiabatic heating10²–10⁴ s⁻¹
Blanking, shearing, punchingLocalised shear, damage model10–10³ s⁻¹
Hot forgingNot recommended — use sinh law or tables> 0.6 Tm
workpiece chip tool vc primary shear zone ε ≈ 1–5, ε̇ ≈ 10³–10⁶ s⁻¹ secondary zone (tool–chip) T up to ≈ 800–1000 °C orthogonal cutting (schematic)

Orthogonal cutting: extreme strain rate in the primary shear zone, high temperature at the tool–chip interface

Johnson–Cook ↓9 · Tips and summary

Practical tips, summary and questions

Tips for DEFORM

  • Choose the J–C entry in Plastic → Flow stress; set D = 1, E = 1, α = 0 to get the original form
  • Use Tr, Tm in the problem's temperature unit; set ε̇₀ to the reference rate of the published set
  • Run coupled thermal; enter k(T), ρcp(T)
  • Plot σ(ε) at the process ε̇ and T in the material window before running

Summary

  1. Three independent brackets: hardening × rate × softening
  2. Five constants from quasi-static, SHPB and hot tests
  3. Strong with adiabatic heating: shear bands, chips, impact
  4. Weak where T and ε̇ interact: hot forming, very high rates

Check yourself

  1. Which test gives C, and which gives m? C: tests at several strain rates (quasi-static + SHPB) at the reference temperature – slope of σ/σref vs ln ε̇*. m: tests at elevated temperatures at the reference rate – from ln(1 − σ/σref) vs ln T*.
  2. What does the rate bracket do below ε̇₀, and how do you avoid it? ln ε̇* < 0, so the rate factor drops below 1 (and turns negative at very low rates) – the stress is underestimated. Choose ε̇₀ = the lowest test rate, or clip ε̇* ≥ 1.
  3. Why must SHPB curves be corrected before fitting? At high rates ≈ 90 % of the plastic work becomes heat with no time to escape: the specimen heats adiabatically by up to hundreds of K, so the curve is not isothermal. Uncorrected, thermal softening would end up in n, C and m.
  4. Name two physical effects J–C cannot describe. Any two of: coupling of T and ε̇ (rate sensitivity is the same at every T); the steep stress upturn above ≈ 10⁴ s⁻¹; recovery and recrystallisation; saturation of hardening.
  5. How does triaxiality change the J–C failure strain? Failure strain falls as triaxiality rises (εf ∝ D₁ + D₂·eD₃η, D₃ < 0): high in shear (η = 0), lower in uniaxial tension (η = ⅓), lowest in notched bars.

Press ↑ to return to the main lecture, or → to continue with Zerilli–Armstrong.

Flow stressZerilli–Armstrong

Zerilli–Armstrong — physics of dislocations, BCC vs FCC

BCC: $$\bar\sigma = C_0 + C_1 e^{-C_3T + C_4T\ln\dot\varepsilon} + C_5\,\bar\varepsilon^{\,n}$$
FCC: $$\bar\sigma = C_0 + C_2\,\bar\varepsilon^{1/2}\,e^{-C_3T + C_4T\ln\dot\varepsilon}$$
  • Based on thermally activated dislocation motion; T in kelvin
  • BCC (steels, Ta, Mo): T and ε̇ affect the yield stress (Peierls barrier)
  • FCC (Cu, Al, Ni, austenite): T and ε̇ affect the hardening
  • DEFORM list: \(\bar\sigma = a + c_1\exp(-c_3T + c_4T\ln\dot{\bar\varepsilon}) + c_5\bar\varepsilon^{\,n}\)
200 400 600 800 0 200 400 600 800 1 000 Temperature T [K] σ̄ at ε̄ = 0.2, ε̇ = 10³ s⁻¹ [MPa] Body-Centred Cubic (crystal lattice)BCC (Armco iron) – thermal term on yield Face-Centred Cubic (crystal lattice)FCC (Oxygen-Free High-Conductivity (copper)OFHC copper) – thermal term on hardening

Zerilli–Armstrong (1987) constants for Armco iron and OFHC copper

▼ Deep dive: 10 slides on Zerilli–Armstrong — press ↓
Zerilli–Armstrong ↓1 · Origin and motivation

Why a dislocation-based model?

  • F. J. Zerilli & R. W. Armstrong, Naval Surface Warfare Center, J. Appl. Phys. 61 (1987) 1816
  • Goal: flow stress for material dynamics codes — impact, penetration, explosive loading
  • Empirical models (Johnson–Cook 1983) treat T and ε̇ as independent factors and use the same form for every metal
  • Experiments showed: in iron the yield stress depends strongly on T and ε̇; in copper mainly the hardening does
  • Idea: derive the form from thermally activated dislocation motion — and let the crystal lattice decide the form
Johnson–CookZerilli–Armstrong
BasisEmpirical fitDislocation mechanics
T–ε̇ couplingNone (product of brackets)Built in: \(e^{-C_3T + C_4T\ln\dot\varepsilon}\)
LatticeSame form for allBCC / FCC / HCP forms
Grain sizeNot explicitHall–Petch term \(k\,d^{-1/2}\)
TemperatureHomologous T*, needs TmAbsolute T [K]
Constants5 (A, B, n, C, m)5–6 (C₀…C₅, n)
Zerilli–Armstrong ↓2 · Physical basis

Thermal activation: dislocations jump over barriers

Dislocation position along glide plane Resisting stress τ applied τ* thermal fluctuation kT helps over the peak ↘ long-range barriers (other dislocations, grain boundaries) → athermal σG short-range barriers (Peierls hills in Body-Centred Cubic (crystal lattice)BCC, forest cuts in Face-Centred Cubic (crystal lattice)FCC) → thermal σth

Short-range barriers can be passed with the help of thermal vibration; long-range ones cannot (schematic)

$$\bar\sigma = \underbrace{\sigma_G}_{\text{athermal}} + \underbrace{\sigma_{th}(T,\dot\varepsilon)}_{\text{thermal}} + \underbrace{k\,d^{-1/2}}_{\text{Hall–Petch}}$$
Arrhenius rate of jumps: $$\dot\varepsilon = \dot\varepsilon_0\,\exp\!\left(-\frac{\Delta G}{k_B T}\right)$$
  • \(\Delta G\) = energy the stress must not supply; the rest comes from thermal vibration
  • Higher T → more help from kT → lower σth
  • Higher ε̇ → less time to wait for a fluctuation → higher σth
  • \(\sigma_G\): long-range stress fields, solutes, initial dislocation density
Zerilli–Armstrong ↓3 · Base mathematics

From the Arrhenius law to the ZA equation

① Thermal stress from thermal activation (activation area A)

$$\sigma_{th} = \frac{M\,\Delta G_0}{A\,b}\;e^{-\beta T}, \qquad \beta = \beta_0 - \beta_1\ln\dot\varepsilon$$

② BCC: A independent of strain (Peierls barrier)

$$\bar\sigma = C_0 + C_1\,e^{-C_3T + C_4T\ln\dot\varepsilon} + C_5\,\bar\varepsilon^{\,n}$$

③ FCC: A ∝ ε̄−1/2 (forest density ρ ∝ ε̄)

$$\bar\sigma = C_0 + C_2\,\bar\varepsilon^{1/2}\,e^{-C_3T + C_4T\ln\dot\varepsilon}$$

④ Common athermal term

\( C_0 = \Delta\sigma_G' + k\,d^{-1/2} \)
ConstantPhysical meaningUnit
C₀athermal stress + Hall–PetchMPa
C₁ (BCC)\(M\Delta G_0/(A b)\) — thermal yield amplitudeMPa
C₂ (FCC)thermal hardening amplitudeMPa
C₃ = β₀thermal softeningK⁻¹
C₄ = β₁T–ε̇ couplingK⁻¹
C₅, n (BCC)athermal strain hardeningMPa, –
k, dHall–Petch slope, grain sizeMPa·mm½, mm
  • M Taylor factor, b Burgers vector, ΔG₀ activation energy at 0 K
  • ε̇ in s⁻¹ (normalised by ε̇₀ = 1 s⁻¹); T always in kelvin
Zerilli–Armstrong ↓4 · Crystal structure decides

BCC vs FCC: the same temperatures, different curves

0 0.2 0.4 0.6 0 200 400 600 800 1 000 Effective strain ε̄ [–] σ̄ at ε̇ = 10³ s⁻¹ [MPa] T = 77 K T = 296 K T = 600 K Armco iron (Body-Centred Cubic (crystal lattice)BCC)

BCC: curves shift in parallel — T acts on the yield stress

0 0.2 0.4 0.6 0 200 400 600 Effective strain ε̄ [–] σ̄ at ε̇ = 10³ s⁻¹ [MPa] T = 77 K T = 296 K T = 600 K Oxygen-Free High-Conductivity (copper)OFHC copper (Face-Centred Cubic (crystal lattice)FCC)

FCC: curves fan out — T acts on the hardening

  • BCC (ferritic steels, Fe, Mo, W, Ta, Nb): high T- and ε̇-sensitive yield, low-temperature brittleness
  • FCC (Cu, Al, Ni, austenitic steels): yield nearly independent of T; hardening rate falls with T
  • HCP (Ti, Mg, Zr): both effects → combined BCC + FCC terms (Zerilli & Armstrong, 1990s)
Zerilli–Armstrong ↓5 · Sensitivities

Where the T and ε̇ sensitivity comes from

200 400 600 800 1 000 0 200 400 600 800 1 000 Temperature T [K] σ̄ at ε̄ = 0.2, ε̇ = 10³ s⁻¹ [MPa] total σ̄ (Body-Centred Cubic (crystal lattice)BCC iron) athermal C₅ε̄ⁿ (+ C₀) thermal part σthathermal part

Iron: thermal part vanishes with T; athermal part stays

10-3 10-2 10-1 100 101 102 103 104 0 200 400 600 800 Strain rate ε̇ [s⁻¹] σ̄ at ε̄ = 0.2 [MPa] Fe (Body-Centred Cubic (crystal lattice)BCC), 296 K Fe (Body-Centred Cubic (crystal lattice)BCC), 600 K Cu (Face-Centred Cubic (crystal lattice)FCC), 296 K Cu (Face-Centred Cubic (crystal lattice)FCC), 600 K

Rate effect from 10⁻³ to 10⁴ s⁻¹: Fe ≈ +350 MPa, Cu ≈ +100 MPa (296 K)

Rate sensitivity: \( \dfrac{\partial\bar\sigma}{\partial\ln\dot\varepsilon} = C_4\,T\,\sigma_{th} \)  — grows with T at first, then falls as σth → 0;   T–ε̇ trade-off: \( \Delta T \leftrightarrow \Delta\ln\dot\varepsilon \cdot \dfrac{C_4 T}{C_3 - C_4\ln\dot\varepsilon} \)
Zerilli–Armstrong ↓6 · Constants and variants

Published constants and model variants

MetalC₀C₁ / C₂C₃ [K⁻¹]C₄ [K⁻¹]C₅n
OFHC Cu (FCC)≈ 65*C₂ = 8900.00281.15·10⁻⁴–½
Armco Fe (BCC)k d−½C₁ = 10330.006984.15·10⁻⁴2660.289
Tantalum (BCC)30C₁ = 11250.005353.27·10⁻⁴3100.44

Stresses in MPa. *Cu: σG = 46.5 MPa + k d−½ with k = 5 MPa·mm½, d = 0.073 mm. Sources: Zerilli & Armstrong 1987; Banerjee 2005 (arXiv cond-mat/0512466); Ta set as quoted in arXiv 1105.5514

DEFORM (dropdown list)

\( \bar\sigma = a + c_1\exp(-c_3T + c_4T\ln\dot{\bar\varepsilon}) + c_5\,\bar\varepsilon^{\,n} \) — the BCC form

Modified ZA for hot forming (Samantaray et al. 2009)

$$\bar\sigma = (C_1 + C_2\bar\varepsilon^{\,n})\,\exp\{-(C_3 + C_4\bar\varepsilon)T^* + (C_5 + C_6T^*)\ln\dot\varepsilon^*\}$$
  • \(T^* = T - T_{ref}\), \(\dot\varepsilon^* = \dot\varepsilon/\dot\varepsilon_{ref}\) — fitted to Gleeble hot-compression data
  • Used for 9Cr-1Mo, Ti-modified austenitic SS (R ≈ 0.995, AARE ≈ 5 %), Ti alloys, Al alloys, roll steels
  • Loses the dislocation meaning — it is a semi-empirical fit; extensions (IMZA) add more couplings

Implemented in

  • DEFORM (built-in) · LS-DYNA (*MAT_MODIFIED_ZERILLI_ARMSTRONG) · Ansys Explicit / Autodyn · Abaqus via VUMAT
Zerilli–Armstrong ↓7 · Calibration

How to determine the constants (BCC form)

1 · Test matrix: quasi-static (10⁻³–10⁻¹ s⁻¹) + SHPB (10³ s⁻¹) at 3–5 temperatures, 77 K … 0.5 Tm
2 · Athermal part: C₅, n from hardening \(\bar\sigma - \sigma_y\) vs ε̄ (log–log); C₀ from grain size or high-T plateau
3 · Linearise: \( \ln(\bar\sigma - C_0 - C_5\bar\varepsilon^n) = \ln C_1 - (C_3 - C_4\ln\dot\varepsilon)\,T \)
4 · Slopes for each ε̇ → plot slope vs ln ε̇: intercept C₃, slope C₄
5 · Check: AARE, R; simulate a Taylor test or the compression test in FEM (inverse correction for adiabatic heating)
0 200 400 600 1 2 3 4 5 6 7 Temperature T [K] ln(σ̄ − C₀ − C₅ε̄ⁿ) [ln MPa] ε̇ = 0.001 s⁻¹ ε̇ = 1 s⁻¹ ε̇ = 1000 s⁻¹ intercept = ln C₁slope = −(C₃ − C₄ ln ε̇)

Step 3–4: straight lines per strain rate (iron constants, synthetic test points)

Zerilli–Armstrong ↓8 · Applications

Where and in which operations ZA is used

OperationWhy ZATypical ε̇, T
Taylor impact test (model validation)Original validation of ZA for Cu and Fe; one test checks the whole ε–ε̇–T field10³–10⁴ s⁻¹, 296–1200 K
Ballistics, penetrators, EFP linersBCC Fe, Ta, W: strong rate effect at high rates10⁴ s⁻¹, adiabatic
High-speed machiningChip formation, shear bands in steels; compared with J–C10³–10⁶ s⁻¹, up to 1000 °C
High-speed forming (electromagnetic, explosive, hammer)Rate hardening of ferritic sheet and tube10²–10⁴ s⁻¹
Cold heading, crash, low-T serviceBCC yield vs FCC hardening; ductile–brittle transition1–10³ s⁻¹, 200–400 K
Hot forging, rolling (modified ZA)Semi-empirical fit of Gleeble data10⁻³–10 s⁻¹, > 0.6 Tm
rigid anvil before: L₀, D₀ v ≈ 100–300 m/s after: L < L₀, mushroom ε̇ ≈ 10³–10⁴ s⁻¹ Dmax compared with Finite Element MethodFEM

Taylor impact test: final length and mushroom diameter validate the model

Zerilli–Armstrong ↓9 · Limits and tips

Limits of ZA and tips for DEFORM

Known limits

  • Valid roughly below 0.5 Tm: no dynamic recovery, recrystallisation, phase change
  • BCC hardening is athermal → over-predicts hardening at high T
  • Above ≈ 10⁴ s⁻¹ dislocation drag (viscous) dominates — not in the model
  • Comparative study (Banerjee 2005, Cu): very good for room-T Taylor tests; over-estimates the T-dependence at high rates; larger errors at 1235 K
  • Original constants come from one material condition — grain size and prior cold work change C₀

Checklist for DEFORM

  • Choose the ZA entry in Plastic → Flow stress; constants a, c₁, c₃, c₄, c₅, n
  • Constants are fitted with T in kelvin and stress in MPa — check against the problem's unit system before typing published values
  • Use a coupled thermal simulation: adiabatic heating at high rates is essential
  • Plot σ̄(ε̄) at 2–3 temperatures and rates in the material window and compare with the test data
  • Hot forming above 0.6 Tm → prefer tabular data or sinh-Arrhenius
Zerilli–Armstrong ↓10 · Summary

Zerilli–Armstrong in five points

  1. Derived from thermally activated dislocation motion: σ̄ = athermal + thermal + Hall–Petch
  2. Thermal term \(e^{-C_3T + C_4T\ln\dot\varepsilon}\) couples T and ε̇
  3. BCC: T, ε̇ act on yield (A = const) · FCC: on hardening (A ∝ ε̄−½) · HCP: both
  4. Best for high rates below 0.5 Tm: impact, machining, high-speed forming
  5. Modified ZA = semi-empirical fit for hot forming; DEFORM has the BCC form built in

Check yourself

  1. What is the activation area, and how does it differ between BCC and FCC? The area a dislocation sweeps in one thermally activated jump. BCC: the obstacle is the Peierls barrier of the lattice → constant. FCC: the obstacles are forest dislocations → A ∝ ε̄−½, it shrinks with strain.
  2. Which constant couples temperature and strain rate? C₄ – in the term e−C₃T + C₄T ln ε̇.
  3. Why is ZA better than J–C for iron at 10³ s⁻¹ and −50 °C? Iron is BCC: its yield stress depends strongly and jointly on T and ε̇ (Peierls mechanism). ZA couples them through C₄·T; J–C has a constant C and independent brackets, and T* < 0 below the reference temperature.
  4. Why should you not use the original ZA for forging at 1150 °C? At 1150 °C steel is at ≈ 0.8 Tm: diffusion-controlled flow, dynamic recovery and recrystallisation and austenite govern – none of these is in ZA, and the athermal hardening is over-predicted. Use a table, the sinh law or the modified ZA.

Press ↑ to return to the main lecture, or → to continue directly with hot forming models.

Flow stressHot forming models

Hot forming: Zener–Hollomon, sinh-Arrhenius, Norton–Hoff

$$Z = \dot{\bar\varepsilon}\,\exp\!\left(\frac{Q}{RT}\right) = A\,[\sinh(\alpha\bar\sigma)]^{n}$$
$$\Rightarrow\; \bar\sigma = \frac{1}{\alpha}\,\mathrm{arsinh}\!\left[\left(\frac{Z}{A}\right)^{1/n}\right]$$
Norton–Hoff: \( \bar\sigma = K\,(\sqrt3)^{m+1}\,\dot{\bar\varepsilon}^{\,m} \),   \(K = K(T, \bar\varepsilon)\)
  • One parameter Z combines T and ε̇ — higher Z ⇒ higher stress, finer grain
  • C–Mn steel: Q ≈ 300–320 kJ/mol; Al alloys ≈ 150–160 kJ/mol
  • DEFORM: sinh form and power form \(\dot{\bar\varepsilon} = A\bar\sigma^{n}e^{-\Delta H/RT}\) in the list
900 1 000 1 100 1 200 0 100 200 300 Temperature [°C] Steady-state flow stress [MPa] ε̇ = 0.01 s⁻¹ ε̇ = 0.1 s⁻¹ ε̇ = 1 s⁻¹ ε̇ = 10 s⁻¹ ε̇ = 100 s⁻¹

Steady-state flow stress, sinh-Arrhenius (Q = 312 kJ/mol, n = 4.5, α = 0.012 MPa⁻¹)

▼ Deep dive: 10 slides on Zener–Hollomon and the sinh law — press ↓
Zener–Hollomon ↓1 · Origin and idea

Temperature-compensated strain rate

  • C. Zener & J. H. Hollomon, “Effect of strain rate upon plastic flow of steel”, J. Appl. Phys. 15 (1944) 22
  • Observation: raising ε̇ has the same effect on flow stress as lowering T
  • Both enter through one parameter:
$$Z = \dot{\bar\varepsilon}\,\exp\!\left(\frac{Q}{R\,T}\right)\quad[\mathrm{s^{-1}}]$$
  • Q activation energy [J/mol], R = 8.314 J/(mol·K), T absolute [K]
  • Same Z ⇒ same flow stress and similar microstructure (DRX grain size)
  • Example (Q = 312 kJ/mol): 1000 °C, 1 s⁻¹ → Z ≈ 6.3·10¹² s⁻¹; 900 °C, 10 s⁻¹ → 7.8·10¹⁴; 1200 °C, 0.01 s⁻¹ → 1.2·10⁹
10-3 10-2 10-1 100 101 102 103 600 800 1 000 1 200 1 400 Strain rate ε̇ [s⁻¹] Temperature [°C] Z = 10¹⁰ s⁻¹ → σ ≈ 34 MPa Z = 10¹² s⁻¹ → σ ≈ 82 MPa Z = 10¹⁴ s⁻¹ → σ ≈ 157 MPa

Lines of constant Z: a lab test at low ε̇ can stand in for a fast press at higher T (C–Mn steel)

Zener–Hollomon ↓2 · Physical basis

Why Arrhenius? Hot working as fast high-temperature creep

  • Above ≈ 0.5 Tm plastic flow is controlled by dislocation climb and cross-slip — both need vacancy diffusion
  • Diffusion is thermally activated: \( D = D_0\,e^{-Q_D/RT} \) → strain rate inherits the Arrhenius term
  • Hardening (dislocation multiplication) and softening (dynamic recovery, DRX) balance → steady state
  • Q of hot working ≥ Q of self-diffusion: alloying, solute drag and particles raise the apparent Q
$$\dot{\bar\varepsilon} = f(\bar\sigma)\,\exp\!\left(-\frac{Q}{RT}\right) \;\Longleftrightarrow\; Z = f(\bar\sigma)$$
MaterialApparent Q for hot working [kJ/mol]
Pure Al (self-diffusion ≈ 142)≈ 140 – 160
Al alloys (6xxx, 7xxx)≈ 140 – 200
Mg alloy AZ31≈ 130 – 170
C–Mn and low-alloy steels (30CrMoA: 262)≈ 260 – 320
Austenitic stainless steels≈ 400 – 460
Ti-6Al-4V: β / (α + β) field≈ 246 / 621
Ni-base superalloys≈ 400 – 600

Typical literature ranges; 30CrMoA from Zhang et al. 2020; Ti-6Al-4V from “Hot deformation characteristics of Ti-6Al-4V” (CityU)

Zener–Hollomon ↓3 · Three stress laws

Power law, exponential law — and the sinh law that joins them

Low stress (ασ < 0.8): \( Z = A_1\,\bar\sigma^{\,n'} \)
High stress (ασ > 1.2): \( Z = A_2\,e^{\beta\bar\sigma} \)
All stresses (Garofalo 1963; Sellars & Tegart 1966): $$Z = A\,[\sinh(\alpha\bar\sigma)]^{n}, \qquad \alpha \approx \beta / n'$$
  • sinh x ≈ x for small x → power law; sinh x ≈ eˣ/2 for large x → exponential
  • Typical: n ≈ 3–6, α ≈ 0.01–0.04 MPa⁻¹ (steels ≈ 0.01, Al alloys larger)
  • Power law = creep (“power-law breakdown” at high Z); exponential = high-rate hot working
0 100 200 300 0 5 10 15 20 25 Flow stress σ [MPa] log₁₀ Z [Z in s⁻¹] power law Z = A′σⁿ′ (ασ < 0.8) exponential Z = A″·exp(βσ) (ασ > 1.2) sinh: Z = A[sinh(ασ)]ⁿ transition

Same constants: the sinh law follows the power law at low σ and the exponential at high σ

Zener–Hollomon ↓4 · Base mathematics

Inverting the law and what follows from it

Flow stress as a function of Z

$$\bar\sigma = \frac{1}{\alpha}\ln\!\left\{\left(\frac{Z}{A}\right)^{1/n} + \left[\left(\frac{Z}{A}\right)^{2/n} + 1\right]^{1/2}\right\}$$

Strain-rate sensitivity

$$m = \frac{\partial\ln\bar\sigma}{\partial\ln\dot\varepsilon} = \frac{\tanh(\alpha\bar\sigma)}{n\,\alpha\bar\sigma}$$

Temperature sensitivity

$$\frac{\partial\ln\bar\sigma}{\partial T} = -\,m\,\frac{Q}{R\,T^2}$$
  • Example: 1000 °C, 1 s⁻¹ → σ = 110 MPa, ασ = 1.32, m ≈ 0.15; −10 K ≈ +3 % stress
0 1 2 3 4 0 0.1 0.2 0.3 0.4 ασ [–] m = ∂ln σ / ∂ln ε̇ [–] n = 3 n = 4.5 n = 6 m → 1/n (power-law creep)high Z: m → 1/(nασ) → 0

m falls from 1/n (creep) towards zero (high Z)

Zener–Hollomon ↓5 · Calibration

Determining α, n, Q and A from hot compression tests

1 · Tests: Gleeble compression, 4–6 T × 4 ε̇ (e.g. 900–1200 °C, 0.01–10 s⁻¹); peak or steady-state σ
2 · n′ and β: slope of ln ε̇ vs ln σ (low σ) and ln ε̇ vs σ (high σ) → α = β/n′
3 · n: \( \ln\dot\varepsilon = \ln A - \dfrac{Q}{RT} + n\ln\sinh(\alpha\sigma) \) → slope of ln sinh vs ln ε̇ = 1/n
4 · Q = n R × slope of ln sinh(ασ) vs 1/T (at constant ε̇)
5 · A from the intercept of ln Z vs ln sinh(ασ); check with R and AARE

Example 30CrMoA (Zhang et al. 2020): α = 0.011 MPa⁻¹, n = 4.40, Q = 262 kJ/mol, A = 3.67·10⁹ s⁻¹, R = 0.983

-4 -2 0 2 -2 -1 0 1 ln ε̇ [ε̇ in s⁻¹] ln[sinh(ασ_p)] 900 °C 1000 °C 1100 °C 1200 °C slope = 1/nspacing ∝ Q/(nR)·Δ(1/T)

Step 3: parallel lines, one per temperature (synthetic points from the C–Mn constants)

Zener–Hollomon ↓6 · Whole flow curves

From one stress value to the whole flow curve

0 0.25 0.5 0.75 1 0 100 200 300 Strain ε [–] Flow stress σ [MPa] Z = 10¹⁰ s⁻¹ (ε_p ≈ 0.13) Z = 10¹² s⁻¹ (ε_p ≈ 0.25) Z = 10¹⁴ s⁻¹ (ε_p ≈ 0.50)

Typical DRX-type curves: peak, then softening to steady state (illustrative, σ_p from the sinh law)

Strain-compensated Arrhenius model

\( \alpha, n, Q, \ln A = f(\bar\varepsilon) \): polynomials of 4th–6th order, fitted at ε = 0.05 … 0.8
  • Constants solved at each strain → smooth polynomial → full curve σ(ε, ε̇, T)
  • Widely used since Lin et al. 2008 (42CrMo); accuracy typically R ≈ 0.98–0.99, AARE ≈ 3–6 %
  • Downside: 20–30 coefficients, no physics in the polynomials, poor outside the fitted strain range

Physics of the curve

  • Peak stress σp and steady state σss both rise with Z
  • Peak strain grows with Z and d₀: \( \varepsilon_p = a\,d_0^{\,h}\,Z^{\,c} \), DRX starts at \( \varepsilon_c \approx 0.8\,\varepsilon_p \)
Zener–Hollomon ↓7 · Microstructure and processing maps

Z controls grain size — and the safe processing window

DRX grain size: \( d_{DRX} = a\,Z^{-m_d} \),   md ≈ 0.1–0.3
Steady-state stress ↔ grain size: \( \sigma_{ss} \propto d_{DRX}^{-k} \), k ≈ 0.7–0.8
  • Same equations sit in DEFORM's Grain data (JMAK, block 8)
  • Finish forging at higher Z (lower T, faster) → finer grain, higher strength

Processing maps (Prasad, dynamic materials model)

Efficiency: \( \eta = \dfrac{2m}{m+1} \)
Instability: \( \xi = \dfrac{\partial\ln[m/(m+1)]}{\partial\ln\dot\varepsilon} + m < 0 \)
  • DRX domains: η ≈ 30–55 % (pure Al: max ≈ 55 %) · instability: shear bands, cracking
109 1010 1011 1012 1013 1014 1015 100 101 102 Zener–Hollomon parameter Z [s⁻¹] Dynamic RecrystallisationDRX grain size [µm] d_DRX = a·Z^(−0.25) (a = 10⁴ µm) high T, low ε̇ → coarselow T, high ε̇ → fine

DRX grain size falls with Z (illustrative constants for a C–Mn steel)

Zener–Hollomon ↓8 · Related laws and DEFORM

Norton–Hoff, Hensel–Spittel — and how DEFORM takes the data

Norton–Hoff (viscoplastic power law)

\( \bar\sigma = K(T,\bar\varepsilon)\,(\sqrt3)^{m+1}\,\dot{\bar\varepsilon}^{\,m} \) — e.g. FORGE; m ≈ 0.1–0.2 hot

Hensel–Spittel (empirical, many terms)

\( \bar\sigma = A\,e^{m_1T}\,\bar\varepsilon^{\,m_2}\,\dot{\bar\varepsilon}^{\,m_3}\,e^{m_4/\bar\varepsilon}\,(1+\bar\varepsilon)^{m_5T}\,e^{m_7\bar\varepsilon}\,\dot{\bar\varepsilon}^{\,m_8T}\,T^{m_9} \)
  • Common in FORGE, QForm, Simufact libraries — no Z, but the same data source (hot compression)

DEFORM flow-stress list

sinh: \( \dot{\bar\varepsilon} = A\,[\sinh(\alpha\bar\sigma)]^{n}\exp[-\Delta H/(R\,T_{abs})] \)
power: \( \dot{\bar\varepsilon} = A\,\bar\sigma^{\,n}\exp[-\Delta H/(R\,T_{abs})] \)
  • ΔH ≡ Q; DEFORM uses absolute temperature — check the units it expects for ΔH and R in your unit system
  • Alternative: compute σ(ε, ε̇, T) from the law and enter it as a table — easy to check, and strain hardening at small ε can be added
  • Plot the curves in the material window at the process T and ε̇ before running
Zener–Hollomon ↓9 · Applications

Where Z and the sinh law are used

OperationUse of Z / sinh law
Hot closed-die forgingFlow stress for die fill and load; Z → DRX grain size in the part
Hot rolling, pass schedulesMean flow stress per pass; Z → recrystallisation between passes
Aluminium extrusionSinh law is the classic model (Sheppard & Wright); exit temperature and speed limits
Isothermal / hot-die forging of Ti, Ni alloysLow ε̇, near-creep regime; m from the sinh law
Process design and scale-upEqual Z: transfer Gleeble results to the press
Processing mapsChoose safe T–ε̇ windows (DRX, no shear bands)
Creep of dies and hot toolsSame Arrhenius structure, low-stress (power-law) branch
DEFORM hot forging sequence of a disk

Hot forging sequence of a disc (© SFTC)

DEFORM ring rolling

Ring rolling — Z changes along the roll gap (© SFTC)

Zener–Hollomon ↓10 · Limits and summary

Limits, summary and questions

Limits

  • Steady-state (or peak) law — needs strain compensation or a DRX model for the full curve
  • Q, n, α valid only in the tested T–ε̇ window; phase changes (α/γ, α/β) change Q
  • Not for cold or warm forming: below ≈ 0.5 Tm use power law or J–C
  • Adiabatic heating at high ε̇ distorts test curves → correct before fitting

Summary

  1. Z = ε̇·exp(Q/RT): one parameter for T and ε̇ (T in K)
  2. Physics: thermally activated, diffusion-controlled flow
  3. Sinh law joins power and exponential laws; σ = (1/α)·arsinh[(Z/A)1/n]
  4. Z also sets DRX grain size and processing windows

Check yourself

  1. Compute Z for 1100 °C, 5 s⁻¹ and Q = 300 kJ/mol. T = 1373 K, Q/RT = 300 000 / (8.314 · 1373) ≈ 26.3 → Z = 5 · e26.3 ≈ 1.3·10¹² s⁻¹.
  2. Why does the sinh law work at low and at high stresses? sinh x ≈ x for small x → power law (creep range); sinh x ≈ eˣ/2 for large x → exponential law (power-law breakdown). One formula covers both.
  3. Which plot gives n, and which gives Q? n: ln ε̇ vs ln sinh(ασ) at constant T (slope = n). Q: ln sinh(ασ) vs 1/T at constant ε̇ – Q = R · n · slope.
  4. A forging should get a finer grain. Raise or lower Z — and how in practice? Raise Z: lower temperature and / or higher strain rate – a colder, faster last blow. dDRX ∝ Z−(0.1…0.3).
  5. Why is a strain-compensated model risky beyond ε = 0.8? α, n, Q and ln A are polynomials fitted only up to ε = 0.8; outside that range polynomials diverge and give nonsense, while forging strains (flash, corners) easily exceed it.

Press ↑ to return to the main lecture, or → to continue directly.

Flow stressBeyond equations

Microstructure-based, AI and user-defined flow stress

DEFORM flow stress model dropdown

Flow-stress options in the Plastic tab (course lecture)

OptionIdeaWhen
Dislocation densityσ̄ = σ₀ + αMGb√ρ; ρ evolves with ε, ε̇, TMulti-step hot forging, DRX
Texture / crystal plasticityGrain orientations → anisotropic flowTi, Mg, sheet, earing
DNN predictorNeural network: chemistry → flow curveNo test data, early design
User routineOwn FORTRAN subroutine compiled into the solverResearch models
ConversionEquation → table; test data → fitted modelLibrary building

The Material Suite (SFTC) fits flow stress from compression data, computes TTT diagrams from composition and JMAK constants.

Flow stressChoosing a model

Match the model to the process

Which process? Cold forming T < 0.3 Tₘ Power law σ = c εⁿ ε̇ᵐ + y strain hardening Warm / hot forging T > 0.5 Tₘ Tabular σ(ε, ε̇, T) or sinh-Arrhenius rate + T sensitive Machining ε̇ = 10³–10⁶ s⁻¹ Johnson–Cook (or Zerilli–Armstrong) high rate, adiabatic Long hold at T hours Creep model (Norton, Perzyna) time dependent Sheet / springback anisotropy Hill 48 / Barlat + kinematic hardening R-values, unloading Always: check the measured ε, ε̇ and T range against the process window

Decision guide used in this course

Yield & hardeningEffective stress

From the stress tensor to one number: effective stress

Stress tensor on a volume element

Nine stress components on a volume element (course lecture)

$$\boldsymbol\sigma = \begin{bmatrix}\sigma_{xx} & \tau_{xy} & \tau_{xz}\\ \tau_{yx} & \sigma_{yy} & \tau_{yz}\\ \tau_{zx} & \tau_{zy} & \sigma_{zz}\end{bmatrix}, \quad s_{ij} = \sigma_{ij} - \sigma_m\delta_{ij}$$
$$\bar\sigma = \sqrt{\tfrac{3}{2}\,s_{ij}s_{ij}} = \sqrt{\tfrac{1}{2}\left[(\sigma_1-\sigma_2)^2+(\sigma_2-\sigma_3)^2+(\sigma_3-\sigma_1)^2\right]}$$
  • Hydrostatic part \(\sigma_m\) does not cause yielding of dense metals
  • Flow curve (uniaxial) is compared with \(\bar\sigma\) → yield when \(\bar\sigma = \sigma_f(\bar\varepsilon, \dot{\bar\varepsilon}, T)\)
  • The yield function defines how \(\bar\sigma\) is computed — Plastic tab, “Yield function type”
Yield & hardeningIsotropic yield

von Mises vs Tresca — isotropic yield criteria

σ₁/σ₀σ₂/σ₀ von Mises (R = 1) Tresca

Plane-stress yield loci (σ₃ = 0)

Tresca: \( \tau_{max} = \dfrac{\sigma_1-\sigma_3}{2} = \dfrac{\sigma_0}{2} \)
von Mises: \( \bar\sigma = \sigma_0 \)
  • Max. difference: 15.5 % in pure shear (\(k = \sigma_0/\sqrt3\) vs \(\sigma_0/2\))
  • Experiments on most metals lie closer to von Mises
  • von Mises is smooth → convenient for FEM; DEFORM default
Yield function type dropdown in DEFORM

Yield function types in DEFORM (course lecture)

Yield & hardeningAnisotropy

Hill 1948 — orthotropic anisotropy

$$F(\sigma_{yy}-\sigma_{zz})^2 + G(\sigma_{zz}-\sigma_{xx})^2 + H(\sigma_{xx}-\sigma_{yy})^2 + 2L\tau_{yz}^2 + 2M\tau_{zx}^2 + 2N\tau_{xy}^2 = 1$$
  • 6 constants F, G, H, L, M, N; axes = rolling (x), transverse (y), normal (z)
  • F = G = H = 1/(2σ₀²), L = M = N = 3/(2σ₀²) → von Mises
  • DEFORM: enter FGHLMN directly or R-values (next slide)
From R-values (σ₀ in RD): \( \dfrac{H}{G} = r_0,\quad \dfrac{H}{F} = \dfrac{r_0}{r_{90}},\quad \dfrac{N}{G} = \left(r_{45}+\tfrac12\right)\left(1+\dfrac{r_0}{r_{90}}\right) \)
Hill quadratic yield function dialog in DEFORM

Hill quadratic (FGHLMN) dialog in DEFORM (course lecture)

Yield & hardeningR-values · Barlat

Lankford R-values, earing and Barlat 91

$$r = \frac{\varepsilon_w}{\varepsilon_t}, \quad \bar r = \frac{r_0 + 2r_{45} + r_{90}}{4}, \quad \Delta r = \frac{r_0 + r_{90} - 2r_{45}}{2}$$
  • High \(\bar r\) → good deep-drawability (DC04: r̄ ≈ 1.6–1.9)
  • \(\Delta r \ne 0\) → ears on drawn cups
  • Al alloys: r < 1 → Hill 48 fails (“anomalous” behaviour)
  • Barlat 91: non-quadratic, exponent a = 6 (BCC) / 8 (FCC) — better for Al
0° (Rolling DirectionRD): ear 45°: trough Cup top view, Δr = (r₀ + r₉₀ − 2r₄₅)/2 > 0

Four ears from planar anisotropy

σ₁/σ₀σ₂/σ₀ von Mises (R = 1) Tresca Hill 48, R = 1.8 (steel Cold-rolled deep-drawing steel grade (EN 10130)DC04) Hill 48, R = 0.6 (Al alloy)

Hill 48 normal anisotropy, plane stress

Yield & hardeningIsotropic hardening

Isotropic hardening — the yield surface expands

Work hardening diagram

Work hardening: reloading needs a higher stress (course lecture)

Isotropic Kinematic surface grows uniformly (radius ↑) α surface moves (back stress α), size constant

Isotropic (left) and kinematic (right) hardening in the deviatoric plane

Isotropic: \( f = \bar\sigma(\boldsymbol\sigma) - \sigma_f(\bar\varepsilon^{p}) = 0 \) — surface radius grows with ε̄p
  • Default in DEFORM; exact for monotonic loading (most forging)
Yield & hardeningKinematic hardening

Kinematic hardening — the yield surface translates

0 0.02 0.04 -600 -400 -200 0 200 400 Strain ε [–] Stress σ [MPa] Forward loading Reverse – isotropic hardening Reverse – kinematic hardening |σ| = 200 MPa (Bauschinger)|σ| = 400 MPa (isotropic)

Load reversal after 5 % strain: σ₀ = 300 MPa, linear hardening

\( f = \bar\sigma(\boldsymbol\sigma - \boldsymbol\alpha) - \sigma_0 = 0 \)
Prager / Ziegler: \( \dot{\boldsymbol\alpha} = c\,\dot{\boldsymbol\varepsilon}^{p} \)
Armstrong–Frederick: \( \dot{\boldsymbol\alpha} = c\,\dot{\boldsymbol\varepsilon}^{p} - \gamma\,\boldsymbol\alpha\,\dot{\bar\varepsilon}^{p} \)
  • Bauschinger effect: after tension the material yields earlier in compression
  • Matters for bending–unbending over die radii and drawbeads, cyclic loading
  • Mixed hardening: fraction isotropic + fraction kinematic
  • Data: tension–compression or cyclic shear tests
Hardening rule dropdown in DEFORM

Hardening rule selection (course lecture)

Yield & hardeningYoshida–Uemori

Yoshida–Uemori two-surface model and springback

  • Two surfaces: inner yield surface moves inside a bounding surface
  • Captures: early re-yielding, transient Bauschinger, permanent softening, work-hardening stagnation
  • Plus E-modulus degradation with plastic strain: \( E = E_0 - (E_0 - E_a)\,[1 - e^{-\xi\bar\varepsilon^p}] \)
  • 7 material constants (Y, C, B, Rsat, b, m, h) from cyclic tension–compression tests
  • Standard for AHSS springback (DP, TRIP, CP steels)
Springback check list: elasto-plastic object · E(ε̄p) · kinematic / Y-U hardening · anisotropic yield · ≥ 5 elements through thickness
DEFORM bullet casing drawing

Multi-stage drawing of a casing — unloading after each stage (© SFTC)

CreepBasics

Creep: plastic flow at constant stress over time

0 200 400 600 800 1 000 0 0.03 0.06 0.09 0.12 Time t [h] Creep strain ε_c [–] σ₃ (lowest) σ₂ σ₁ (highest) III steady state× rupture σ₁× σ₂

Creep curves for three stress levels (schematic, computed)

Creep specimen and creep curve stages

Stages I, II, III (course lecture)

DEFORM creep model dialog

Creep option in the Plastic tab (course lecture)

  • Significant above ≈ 0.4 Tm; strain grows at constant load — I primary, II steady (minimum rate), III tertiary → rupture
  • In DEFORM creep is added to the elasto-plastic strain: \( \dot{\boldsymbol\varepsilon} = \dot{\boldsymbol\varepsilon}^{e} + \dot{\boldsymbol\varepsilon}^{p} + \dot{\boldsymbol\varepsilon}^{c} + \dot{\boldsymbol\varepsilon}^{th} \)
CreepModels

Creep models available in DEFORM

ModelEquationTypical use
Power law (Norton)\( \dot\varepsilon^{c} = A\,\bar\sigma^{n} \)Steady-state creep, relaxation
Bailey–Norton\( \varepsilon^{c} = A\,\bar\sigma^{n}\,t^{m} \) (m < 1)Primary + secondary creep
Perzyna\( \dot\varepsilon^{c} = \gamma\left\langle \dfrac{\bar\sigma}{\sigma_y} - 1\right\rangle^{p} \)Viscoplasticity above yield
Soderberg\( \dot\varepsilon^{c} = A\left[\exp(\bar\sigma/\sigma_0) - 1\right] \)Low-stress creep
Exponential\( \dot\varepsilon^{c} = A\,e^{B\bar\sigma}\,e^{-Q/RT} \)High-stress creep, T-dependent
Tabular\( \dot\varepsilon^{c} = f(\bar\sigma, T, \varepsilon^{c}) \)Measured data
  • Norton exponent n: 3–8 for dislocation creep; ≈ 1 for diffusion creep
  • Fit constants to the stress and temperature range of the process — creep data extrapolate even worse than flow stress

Model list: DEFORM V14 manual, 10 Material Data – Creep

CreepApplications

Where creep shows up in forming

0 20 40 60 80 100 0 100 200 300 Time [h] Remaining stress [MPa] Bolt / shrink-fit stress, Norton creep

Stress relaxation under fixed strain, Norton n = 5 (computed)

  • Stress relaxation: shrink-fit die rings and bolts at temperature lose preload
  • Isothermal / hot-die forging of Ti and Ni alloys at low ε̇ (10⁻³–10⁻² s⁻¹)
  • Superplastic forming of Ti-6Al-4V, Al 5083 (m ≥ 0.5)
  • Hot sizing, stress-relief annealing: residual stress drops during holding
  • Heat treatment: creep during slow cooling and tempering
Dead weight creep testing frame

Dead-weight creep frame (© DoITPoMS, Univ. of Cambridge)

Thermal dataThermal tab

The Thermal tab

DEFORM thermal tab

Thermal tab: conductivity, heat capacity / specific heat, emissivity, density (course lecture)

$$\rho c_p\,\frac{\partial T}{\partial t} = \nabla\cdot(k\nabla T) + \beta\,\bar\sigma\,\dot{\bar\varepsilon} + \dot q_{tr} + \dot q_{EM}$$
PropertySI unit in DEFORMControls
Conductivity k(T)N/(s·°C) = W/(m·K)Heat flow inside the part
Heat capacity ρcp(T)N/(mm²·°C)Heating per unit energy
Emissivity ε–Radiation to environment
Density ρt/mm³Mass (hammers), specific-heat option
Thermal dataConductivity

Thermal conductivity k(T)

0 250 500 750 1 000 0 20 40 60 Temperature [°C] Thermal conductivity k [W/(m·K)] Carbon steel (European Norm (European standard)EN 1993-1-2) Austenitic stainless (≈15 → 30)

k(T): carbon steel per EN 1993-1-2; austenitic stainless (typical)

Material, 20 °Ck [W/(m·K)]
Copper≈ 390
Aluminium alloys120 – 200
Carbon steel45 – 54
Hot-work tool steel H13≈ 25 – 29
Stainless 304≈ 15
Ti-6Al-4V≈ 7
  • Low k (Ti, Ni, stainless) → chilled skin at die contact, local overheating in shear bands
  • Carbon steel: k halves from 20 °C to 800 °C — always f(Temp) for hot forging
Thermal dataHeat capacity

Heat capacity or specific heat?

0 250 500 750 1 000 0 1 000 2 000 3 000 4 000 5 000 Temperature [°C] Specific heat c_p [J/(kg·K)] Specific heat c_p [J/(kg·K)] α → γ transformation(latent heat, ≈ 735 °C)

cp(T) of carbon steel — EN 1993-1-2 (peak at the α → γ transformation)

$$\underbrace{\rho\,c_p}_{\text{heat capacity}}\;[\mathrm{N/(mm^2\,°C)}] = \rho\,[\mathrm{t/mm^3}]\times\underbrace{c_p}_{\text{specific heat}}\,[\mathrm{mm^2/(s^2\,°C)}]$$
  • Steel at 20 °C: 7850 kg/m³ × 440 J/(kg·K) ≈ 3.45 N/(mm²·°C)
  • Above 900 °C: cp ≈ 650 J/(kg·K) → ≈ 5.1 N/(mm²·°C)
  • Option Specific heat → DEFORM multiplies by density
  • Latent heat of transformation: in the peak (Regular) or in Transformation data (Mixture) — not both
Thermal dataEmissivity · density

Emissivity and density

Radiation + convection at a free surface: $$q = \varepsilon\,\sigma_{SB}\,(T^4 - T_\infty^4) + h_{conv}(T - T_\infty)$$
SurfaceEmissivity ε
Oxidised / scaled steel, hot0.7 – 0.9
Bright machined steel0.2 – 0.4
Polished aluminium0.04 – 0.1
Oxidised titanium0.5 – 0.7
  • At 1200 °C radiation dominates heat loss in transfer from furnace to press
  • Density: mass for hammers and screw presses, specific-heat option, volume change ρ(T) in heat treatment
DEFORM cogging simulation

Open-die cogging of a hot ingot: hours of radiation cooling between passes (© SFTC)

Thermal dataInterface HTC

Die–workpiece heat transfer and the Inverse HTC module

Inverse HTC step 1

1 · Test with thermocouples

Inverse HTC step 2

2 · Simulated vs measured T(t)

Inverse HTC step 3

3 · Fitted HTC (© SFTC)

\( q = h\,(T_w - T_d) \),   h = f(pressure, time, lubricant, scale)
  • Not a material property — set in Inter-object, but sits next to k in every heat balance
  • DEFORM defaults: forming contact ≈ 11 N/(s·mm·°C) (= 11 kW/(m²·K)), free surface convection ≈ 0.02
  • Inverse HTC: DEFORM varies h(t) or h(T) until simulated thermocouple curves match the measured ones
  • Same idea used for flow stress: inverse analysis of the compression test
Beyond mechanicsGrain · JMAK

Recrystallisation and grain size: JMAK (Avrami)

10-2 10-1 100 101 102 103 0 0.25 0.5 0.75 1 Holding time after deformation [s] Recrystallised fraction X [–] static RecrystallisationRX, 1100 °C static RecrystallisationRX, 1000 °C static RecrystallisationRX, 900 °C

Static recrystallisation, Avrami n = 1.5, t0.5 from 0.1 s to 40 s (illustrative)

$$X = 1 - \exp\!\left[-\beta\left(\frac{t}{t_{0.5}}\right)^{n}\right],\; \beta = \ln 2$$
$$t_{0.5} = a\,d_0^{h}\,\varepsilon^{p}\,\dot\varepsilon^{q}\,e^{Q/RT}$$
DRX grain: \( d_{drx} = a\,d_0^{h}\,\varepsilon^{n}\,\dot\varepsilon^{m}\,e^{Q/RT} \)
  • DRX starts at \(\varepsilon_c \approx 0.8\,\varepsilon_p\)
  • Data: Phase → Grain button
DEFORM microstructure module turbine disc

Grain size in a forged Waspaloy disc (© SFTC)

Beyond mechanicsPhases · hardness

Phase transformation kinetics and hardness

Diffusional (ferrite, pearlite, bainite)

\( \xi = 1 - \exp(-b\,t^{n}) \),   b, n from TTT + Scheil additivity

Martensite — Koistinen–Marburger

\( \xi_M = 1 - \exp[-\Omega\,(M_s - T)] \),   Ω ≈ 0.011 K⁻¹

Hardness — Hardness tab

\( HV = \sum_i \xi_i\,HV_i(\text{chem.}, \dot T) \)
  • TTT / CCT from tests or from composition (Material Suite, JMatPro)
DEFORM residual hoop stress in a disk

Residual hoop stress after quenching a disc (© SFTC)

DEFORM nut forming progression

Cold heading progression — hardness from strain hardening (© SFTC)

Beyond mechanicsDiffusion · EM

Diffusion (carburising) and electro-magnetic data (induction)

Diffusion tab

$$\frac{\partial C}{\partial t} = \nabla\cdot(D\,\nabla C), \quad D = D_0\,e^{-Q/RT}$$
  • Carbon in austenite: D₀ ≈ 2.3·10⁻⁵ m²/s, Q ≈ 148 kJ/mol
  • Surface: carbon potential + mass-transfer coefficient
  • Carbon content → Ms, hardness, TTT shift

Elec./Mag. tab

  • Electrical resistivity ρe(T)
  • Relative magnetic permeability μr(T, H) — drops to 1 above Curie point (≈ 770 °C)
Skin depth: \( \delta = \sqrt{\dfrac{2\rho_e}{\omega\,\mu_0\mu_r}} \)
  • Above Curie: δ grows ≈ 10× → heating zone moves inward
DEFORM induction hardened spur gear

Induction-hardened spur gear (© SFTC)

Beyond mechanicsFracture

Ductile fracture: normalised Cockcroft–Latham

$$D = \int_0^{\bar\varepsilon_f}\frac{\langle\sigma_1\rangle}{\bar\sigma}\,d\bar\varepsilon \;\ge\; C$$
  • Only tensile max. principal stress counts — compression does no damage
  • C is a material constant, calibrated by simulating a test to its crack strain
  • Typical C: 0.2 – 0.6 (cold steels) — depends on inclusions, T
  • Advanced tab: damage model choice, element deletion threshold
  • Others in DEFORM: Oyane, Rice–Tracey, Brozzo, McClintock, Lemaitre …
σ₁ > 0(hoop) compression Free bulging surface: D reaches C first → surface cracks

Upsetting: tensile hoop stress at the bulge

DEFORM cross wedge rolling

Cross-wedge rolling — risk of central cavity (Mannesmann effect) (© SFTC)

Checklist: which material data does my simulation need?

ProcessFlow stressYield / hardeningElasticThermalExtra
Cold forging, headingPower law / table, ε to 2–3von Mises, isotropicFor die stressk, ρcp (heating)Damage C
Hot closed-die forgingTable σ(ε, ε̇, T) or sinhvon Mises, isotropic–k(T), ρcp(T), ε, HTCGrain (JMAK)
MachiningJohnson–Cook / ZAvon MisesToolk, ρcpDamage, wear
Sheet / springbackSwift / HollomonHill 48 / Barlat, kinematic, Y-UE(ε̄p)–FLD
Heat treatmentPer phase, f(T)Isotropic (+ TRIP)E(T), α(T)Per phaseTTT, latent heat, ΔV, HV
Induction hardeningPer phaseIsotropicE(T), α(T)k, ρcpρe(T), μr(T)
  1. Check units (SI / English) · 2. Check validity range vs process · 3. Check temperature dependence · 4. Validate one result (force, T) against a measurement

Quick quiz & sources

  1. Write the general flow-stress function and name its three variables. σ̄ = f(ε̄, ε̇̄, T): effective plastic strain, effective strain rate and temperature.
  2. Why can a rigid-plastic forging simulation run without E and ν? In bulk forming the plastic strains (hundreds of %) dwarf the elastic ones (fractions of a %), so the rigid-plastic formulation neglects elasticity. E and ν are needed only for elastic dies, springback / residual stresses and heat treatment.
  3. Which flow-stress model for machining, which for hot forging — and why? Machining: Johnson–Cook (or Zerilli–Armstrong) – very high strain rates and adiabatic heating. Hot forging: a measured table σ(ε, ε̇, T) or the sinh-Arrhenius law – thermally activated flow where T and ε̇ interact.
  4. What happens if the process exceeds the strain-rate range of a table? DEFORM extrapolates without a warning – it holds the last value or extends the last slope – so the flow stress there is a guess (e.g. a table to 10 s⁻¹ used at 10⁴ s⁻¹ is ≈ 8 % too low). Compare max ε, ε̇, T from the post-processor with the table range.
  5. How do R-values enter the Hill 48 criterion? What does Δr predict? R-values at 0°, 45° and 90° set the ratios of Hill's constants (H/G = r₀, H/F = r₀/r₉₀, …). Δr = (r₀ − 2r₄₅ + r₉₀)/2 predicts earing of drawn cups: Δr > 0 → ears at 0° and 90°, Δr < 0 → ears at 45°.
  6. Isotropic vs kinematic hardening: what happens to the yield surface? Isotropic: the yield surface grows uniformly around a fixed centre (radius = current flow stress). Kinematic: it keeps its size and moves by the back stress α → early reverse yielding (Bauschinger effect), important for springback.
  7. Heat capacity vs specific heat — what does DEFORM expect in SI units? Heat capacity is per volume (ρ·cp), specific heat per mass. DEFORM's default field expects heat capacity: steel ≈ 7850 kg/m³ × 440 J/(kg·K) ≈ 3.45 N/(mm²·°C) = 3.45 MJ/(m³·K). Typing 460 there makes the billet practically never heat up.
  8. How is the Cockcroft–Latham constant C obtained? By calibration: simulate a test (upsetting, tension, notched bar) to the strain at which the real specimen cracked; the damage D at the crack site at that moment is C. It depends on the heat of material, cleanliness and temperature.
  • SFTC: DEFORM V14 manual, Ch. 10 Material Data; Material Suite; deform.com
  • Necpal, M.; Sobota, R.: Modelovanie a simulácia procesov tvárnenia. STU, 2023
  • Kobayashi, Oh, Altan: Metal Forming and the FEM. 1989
  • Johnson & Cook (1983); Jaspers & Dautzenberg, J. Mater. Process. Technol. 122 (2002)
  • Zerilli & Armstrong, J. Appl. Phys. 61 (1987)
  • Sellars & Tegart (1966) — sinh law; Hill (1948); Barlat & Lian (1989), Barlat et al. (1991)
  • Yoshida & Uemori, Int. J. Plasticity 18 (2002)
  • EN 1993-1-2: thermal and mechanical properties of steel
  • Images © SFTC, Gleeble/DSI, ZwickRoell, DoITPoMS — linked for teaching

Next lecture: preprocessing — mesh, boundary conditions, friction