AMStressForge

The failure-mode predictor of the AM stack. StressForge takes the thermal history, pulls local strength from GrainPath, and runs a quasi-static elasto-plastic FEM on the part — mirroring the build layer by layer, all the way through release from the build plate.

The solve mirrors the build: each layer is activated, loaded by its thermal strain, and brought to equilibrium with the layers below. The real residual state appears only at release, when the part is cut from the plate and springs back.

FIG.01 · stressforge_solve_pipeline layer activation · thermal strain · equilibrium · release · post-processing

Six features that decide whether a part ships or gets scrapped.

Layer-activated elasto-plastic FEM

Quasi-static elasto-plastic FEM, solved layer by layer as the build grows — exactly mirroring the physical process:

• Each layer activated at its build time
• Loaded by its thermal strain (from FusionCore's T(t) field)
• Brought to equilibrium with all previously-built layers
• Production-grade FEniCSx (DOLFINx) backend

Temperature-dependent constitutive model

Stress prediction is only as good as the material model behind it. StressForge runs a T-dependent, isotropic-hardening plasticity:

• State-dependent Young's modulus E(T) and CTE α(T)
• Temperature-dependent yield σ_y(T) with isotropic hardening
• Thermal strain integrated from FusionCore's T(t) history
• Stress-free reference at the activation temperature of each layer

ε = εel + εth + εp

εth = α(T)·(T − Tref)·I

f(σ, T) = σ̄ − σy(T, ε̄p) ≤ 0

E(T) · α(T) · σy(T) — T-dependent

Hall–Petch microstructure coupling

GrainPath's PDAS map flows directly into the local strength field — fine grains forge stronger metal:

• σ_y = σ_y0 + k_HP / √λ₁ per node
• Higher cooling rates → finer dendrite arms → higher local yield
• The columnar / equiaxed map shows up directly in the stress distribution
• k_HP calibrated per alloy from published EBSD–tensile measurements

σy(node) = σy0(T) + kHP / √λ₁

σy0 · base yield (T-dependent)
kHP · Hall–Petch slope [MPa·µm1/2]
λ₁  · PDAS from GrainPath [µm]

node 14201 → σy = 612 MPa

Two solve modes

Pick fidelity against the budget. Full-history mode for highest accuracy; inherent-strain mode (powered by FusionMap) for whole-part iteration inside the optimisation loop.

Full thermal

Layer-by-layer FEM

Highest-fidelity residual stress driven by FusionCore's full T(t). Best for validation, anomaly investigation, qualification.

~hours per part

Inherent-strain

Part-scale FEM

Reads an ε* map directly from FusionMap — no thermal solve needed. Enables whole-part iteration in the optimisation loop.

~minutes per part

A full failure-mode output set

StressForge writes the fields qualification reports actually cite:

• von Mises and principal stresses (σ₁, σ₂, σ₃)
• Displacement & warping — full 3D, true scale
• Springback — distortion captured at release from the plate
• Plastic strain (ε_p) and yielded-volume fraction
• Stress triaxiality for ductile-fracture risk assessment

von Mises σ̄ σ₁, σ₂, σ₃ warping springback ε_p yielded volume triaxiality

Validated materials & solver stack

Calibrated presets for the canonical LPBF alloys — including E(T), α(T), σ_y(T), and a per-alloy k_HP:

SS316L Ti-6Al-4V IN718 · soon AlSi10Mg · soon

All mechanical parameters are user-overridable through the Material Library.

FEniCSx · FEM J2 plasticity Hall–Petch coupling

StressForge ties the whole chain together — heat from FusionCore, strength from GrainPath, material card and constraints in. Its stress and distortion fields are both the target the FusionMap optimiser minimises and the gate CertifyAM checks against.

FIG.02 · stressforge_io_contract history + strength + material + constraints → stress · distortion · springback

StressForge is the second prediction branch alongside GrainPath — the one that produces the numbers customers actually qualify a part on — converging into CertifyAM.

Layer-activated FEM is heavy. SolidNetics elastically scales those solves across cloud workers so a full-part residual-stress run finishes overnight, not over a week.

Three steps from thermal fields to a full residual-stress prediction.

Built for LPBF process engineers killing distortion before metal is wasted, structural engineers qualifying parts against residual-stress limits, and qualification teams defending stress maps in regulatory audit.