Table of Contents
- What Is Thermal Stress? Constrained Expansion
- Governing Equations: Free Expansion and Constraint-Induced Stress
- One-Way vs Two-Way Coupling
- Sequential Coupling Workflow: Step-by-Step
- FEM Formulation: Thermal Expansion Strain
- Common Engineering Scenarios
- Software Comparison: Abaqus, ANSYS, Code_Aster
- Warping, Bowing, and Residual Stress in Manufacturing
What Is Thermal Stress? The Role of Constraint
Thermal stress arises from a deceptively simple physical mechanism: when a material is heated, it wants to expand; when it is cooled, it wants to contract. If this natural dimensional change is prevented by geometric constraints, bonded dissimilar materials, or non-uniform temperature distributions, the thwarted deformation becomes internal mechanical stress.
Two distinct mechanisms generate thermal stress in practice:
- External constraint: a structure is anchored at its ends and cannot expand freely when heated. Think of a steam pipe fixed rigidly between two walls — as the pipe temperature rises, the walls prevent elongation, and the pipe develops compressive axial stress.
- Internal constraint (differential expansion): different parts of the same structure heat up at different rates or to different temperatures. The hotter parts try to expand more than the cooler parts, and since they are physically bonded together, internal stresses develop — even in a completely unconstrained structure floating in space. A thick-walled cylinder with hotter outer surface and cooler inner surface, or the surface of a turbine blade first contact with hot combustion gas, are classic examples.
Key insight: a perfectly uniform temperature change in an unconstrained, homogeneous structure produces zero thermal stress. It deforms freely (expands or contracts uniformly) but no internal stress develops. Thermal stress is always a consequence of either non-uniformity or constraint — usually both.
So if I heat a steel rod uniformly in a furnace with no constraints, it just expands and there's zero stress? That seems counterintuitive — surely the heating does something.
Exactly right — and it really is zero stress for a free, unconstrained, homogeneous rod heated uniformly. The material wants to expand by \(\alpha \Delta T\) in every direction, and since nothing stops it, it simply does. Every atom moves apart from its neighbor by exactly the same amount; the spacing between atoms increases uniformly; and since stress is proportional to the deviation from the natural (stress-free) spacing, there is nothing to generate stress. The rod is longer, but completely stress-free. The confusion often comes from thinking about compressive stress in a hot pressure vessel or engine block — but in those cases, there's always either a temperature gradient (the inside is hotter than the outside, or one part heats faster than another) or a mechanical constraint (bolted flanges, press-fit components, weld residual stress). It's a useful mental model: ask "what's stopping this part from expanding freely?" Whatever that is — another material bonded to it, a rigid support, or a temperature gradient across its own cross-section — that's the source of the thermal stress.
Can thermal stress exceed the yield strength and cause permanent deformation without any mechanical load? I thought yielding only happened under external forces.
Absolutely — thermal stress can be severe enough to cause yielding and plastic deformation, and this is not rare in engineering practice. A classic case is quenching: when a hot steel component is suddenly plunged into cold water, the surface cools and contracts rapidly while the interior is still hot and expanded. The surface is under tension (the interior is preventing it from contracting fully) and if that tension exceeds the yield strength at the surface temperature, the surface yields. Then as the interior eventually cools and tries to contract, it finds the surface has already yielded and shortened plastically — so the final cooled part has residual compressive stress at the surface and tensile stress in the interior. This is actually exploited intentionally: controlled quenching creates beneficial compressive residual stress on the surface of gears and bearings, dramatically improving fatigue life. But in a poorly controlled quench, the tensile residual stresses in the core can cause delayed cracking. In turbine blades cycling between cold startup and hot operation, the thermal cycling drives low-cycle fatigue — stress reversals from thermal expansion alone, without any external mechanical loading, accumulate damage cycle by cycle until a crack initiates.
Governing Equations: Free Expansion and Constraint-Induced Stress
Free Thermal Expansion (Zero Stress)
For a linear elastic material heated uniformly from reference temperature \(T_\text{ref}\) to temperature \(T\), the thermal strain in each coordinate direction is:
Where \(\alpha(T)\) is the coefficient of thermal expansion [1/K or 1/°C], which is generally a function of temperature. For most engineering metals, \(\alpha\) increases slowly with temperature; for ceramics, it can vary strongly. The shear thermal strains \(\varepsilon_{ij}^{\text{th}} = 0\) for \(i \neq j\) in isotropic materials — thermal expansion is purely volumetric with no shear component.
Constrained: Thermal Stress in a 1D Bar
For a bar with both ends fixed (fully constrained), heated uniformly by \(\Delta T\):
The negative sign indicates compression for positive \(\Delta T\) — heating a constrained bar produces compressive stress (the bar tries to expand but is held in place, so it is squeezed). The magnitude can be surprising: for steel (\(E = 200\) GPa, \(\alpha = 12\times10^{-6}\) /°C), a temperature rise of just 83°C generates compressive stress of 200 MPa — close to the yield strength of mild steel.
General 3D Thermoelastic Constitutive Law
In 3D, total strain is decomposed into elastic and thermal parts:
Stress depends only on the elastic part:
Expanding for an isotropic material:
The last term is the thermal stress tensor — an isotropic hydrostatic stress that arises from thermal expansion being inhibited. Note that even in a completely unconstrained structure, if the temperature field is non-uniform, different elements have different \(\Delta T\), and the compatibility requirement (elements must deform consistently with their neighbors) generates internal stress from the non-uniformity alone.
The formula gives compressive stress when you heat a constrained bar. But I've seen examples where thermal stress is tensile — when does heating cause tension?
Heating causes compression when the hot part is constrained from expanding — it's being squeezed. Tension arises when a part that has cooled (or is cooler than its surroundings) is constrained from contracting. The surface of a quenched component during rapid cooling is a perfect example: the surface drops in temperature quickly and tries to contract, but the hot interior won't let it — so the surface is in tension. Another case: if you have two bonded materials, the one with the lower CTE (coefficient of thermal expansion) is effectively being "stretched" by the one with higher CTE during heating — it experiences tension. In a bimetallic strip, the low-CTE layer is in tension on the concave side (high-CTE layer side) after heating. In electronics, a copper trace on an FR4 PCB substrate has \(\alpha_\text{Cu} \approx 17\) ppm/°C while FR4 has \(\alpha_\text{FR4} \approx 14\) ppm/°C in-plane — relatively close, so stresses are modest. But in the through-thickness direction, FR4 has \(\alpha_z \approx 60\) ppm/°C, and copper's much lower CTE in that direction means solder joints and vias are under significant cyclic stress during thermal cycling — a primary reliability failure mechanism in electronics.
One-Way vs Two-Way Coupling
Thermal stress analysis involves two physics fields — heat transfer and structural mechanics — that must be linked. The choice of coupling strategy depends on whether the structural deformation significantly affects the thermal solution.
One-Way (Sequential) Coupling: Thermal → Structural
The most common approach for most engineering applications:
- Solve the thermal analysis independently to obtain the temperature field \(T(\mathbf{x},t)\)
- Apply the temperature field as a body load to the structural FEM model and solve for stresses and deformations
The thermal analysis is completely independent of the structural analysis — structural deformations do not feed back to change the temperature field. This is valid when:
- Deformations are small and do not significantly change heat flow paths or contact conductance
- Viscous heating in the structural material is negligible (no large plastic deformation)
- The structure remains in good thermal contact with its surroundings regardless of deformation
Most industrial thermal stress analyses — electronics cooling, turbine blade thermal fatigue, piping thermal expansion — use one-way coupling because it is far more computationally efficient and the coupling error is small.
Two-Way (Fully Coupled) Coupling: Thermal ↔ Structural
Required when structural deformation significantly affects thermal behavior:
- Contact conductance changes: if surfaces that initially touch may separate during heating (or vice versa), the heat flow path changes — a gap that opens reduces conductance dramatically. This requires full coupling to track the contact state simultaneously with temperature.
- Large plastic deformation with significant heat generation: in metal forming, crash simulation, or friction stir welding, plastic dissipation heats the material substantially, which softens it further and changes the deformation — a strong bidirectional coupling.
- Thermoelastic damping: in precision MEMS resonators, the coupling between stress waves and temperature waves (via reversible thermoelastic heat generation) affects the quality factor — requires simultaneous solution.
Fully coupled solvers use a monolithic approach: both the heat transfer and structural equations are assembled into a single large system and solved simultaneously at each time step. This is more accurate but roughly 4–10× more computationally expensive than sequential coupling.
For a turbine blade analysis with high temperatures, should I always use full coupling? The temperature gradients are very large.
In almost all turbine blade thermal fatigue analyses — even for the most challenging blades in modern aero engines — one-way sequential coupling is the standard industry practice. Here's why: the blade deformation during operation is typically a few millimeters at most, which changes the gas flow path and heat transfer coefficient marginally. The film cooling holes, trailing edge, and leading edge geometry are nearly unaffected by the deformation. The dominant physics is the temperature distribution creating differential expansion stresses — and that temperature field is determined by the aerothermal boundary conditions (hot gas temperature, film cooling effectiveness, conduction through the blade wall), not by the structural deformation. Full coupling would be required if, for instance, the blade were deforming so much that the cooling passages were changing shape and altering internal coolant flow — which doesn't happen in normal operation. Where full coupling does matter for turbines is in blade/disk contact: whether the blade root contact surface maintains or loses contact with the disk slot during centrifugal loading and thermal cycling, because the resulting contact conductance change affects how much heat flows from blade to disk. For that specific interface, a coupled approach is worthwhile.
Sequential Coupling Workflow: Step-by-Step
The sequential thermal-structural workflow is the standard approach for most industrial thermal stress analyses. Here is the complete procedure:
Step 1: Thermal Analysis (Temperature Field)
Model setup:
- Define material thermal properties: conductivity \(k(T)\), specific heat \(c_p(T)\), density \(\rho\) (temperature-dependent tables recommended for metals)
- Apply thermal boundary conditions: convection coefficients \(h\) and fluid temperatures \(T_\infty\), prescribed temperatures, heat fluxes, and radiation view factors
- For steady-state: solve \(\nabla\cdot(k\nabla T) + \dot{q} = 0\); for transient: solve \(\rho c_p \partial T/\partial t = \nabla\cdot(k\nabla T) + \dot{q}\)
Output: Nodal temperature field \(T(\mathbf{x})\) (steady) or \(T(\mathbf{x},t)\) (transient)
Step 2: Structural Analysis (Temperature as Load)
Model setup:
- Define structural material properties: Young's modulus \(E(T)\), Poisson's ratio \(\nu(T)\), CTE \(\alpha(T)\), and (if plastic deformation is expected) yield strength \(\sigma_y(T)\) and hardening data
- Read the temperature field from Step 1 as a predefined field or nodal temperature load
- Specify the reference temperature \(T_\text{ref}\) (the stress-free state temperature — typically manufacturing/assembly temperature)
- Apply mechanical boundary conditions: supports, symmetry constraints, contact definitions
- Solve: \([\mathbf{K}(T)]\{\mathbf{u}\} = \{\mathbf{F}_\text{th}\}\) where \(\{\mathbf{F}_\text{th}\}\) is the thermal load vector computed from \(\boldsymbol{\varepsilon}^\text{th}\)
Output: Displacements, thermal strains, thermal stresses, von Mises stress, reaction forces
Why do I need to specify a "reference temperature" in the structural analysis? What happens if I get it wrong?
The reference temperature \(T_\text{ref}\) defines the stress-free configuration — the state at which the thermal strain is zero. Thermal stress comes from the temperature difference \(\Delta T = T - T_\text{ref}\), not from \(T\) itself. If you set \(T_\text{ref}\) wrong, you'll compute \(\Delta T\) wrong, and your entire stress result will be off by a factor proportional to the error in \(\Delta T\). For example: an aluminum bracket is machined at 20°C and welded to a steel structure (still at ~20°C after cooling from the weld). When the assembly heats up to 150°C in service, the thermal loading is \(\Delta T = 150 - 20 = 130\)°C. \(T_\text{ref} = 20\)°C is correct. If you accidentally set \(T_\text{ref} = 0\)°C, you'd compute \(\Delta T = 150\)°C and overestimate thermal stress by about 15%. If the part was manufactured at a hot state — say, press-fit assembled at 200°C (heated for assembly) and then cooled to 20°C — the stress-free temperature is 200°C, and the structural analysis at room temperature should use \(T_\text{ref} = 200\)°C to capture the correct interference fit stress. Getting this right is especially important in manufacturing simulations like casting, welding, and heat treatment where the reference state evolves during the process.
Does the mesh for the thermal analysis have to be identical to the mesh for the structural analysis?
Not necessarily, but it depends on your software and workflow. In Abaqus, the cleanest approach is to use the same mesh for both analyses — you run a heat transfer analysis (step type HEAT TRANSFER) and then read the nodal temperatures directly into the structural analysis (step type STATIC) as a predefined field. Since the node IDs are the same, the mapping is exact with no interpolation error. In ANSYS Workbench, the "Thermal-Structural" linked system handles this automatically when you use the same mesh for both, but it also supports different meshes with automatic interpolation using radial basis functions or nearest-node mapping. The issue with different meshes is interpolation error: if your thermal mesh is coarse and your structural mesh is fine (to capture stress gradients), the temperature is only known at the thermal mesh nodes and must be interpolated to structural nodes. In regions of high temperature gradient, this interpolation can introduce error. Best practice: use the same or finer mesh for the thermal analysis as for the structural analysis, especially in high-gradient regions like near heat sources, cooling channels, or dissimilar material interfaces.
FEM Formulation: Thermal Expansion Strain
In the finite element implementation, the thermal load is introduced through the thermal strain vector. For an element with nodes at temperatures \(T_i\), the temperature field within the element is interpolated:
The thermal strain at any point in the element is:
(In Voigt notation for 3D: three normal strains equal, three shear strains zero, for isotropic expansion.)
The thermal load vector for element \(e\) is computed by integrating the thermal strain contribution through the element volume:
Where \([\mathbf{B}]\) is the strain-displacement matrix and \([\mathbf{C}]\) is the material constitutive matrix. This thermal force vector is assembled into the global load vector and the system \([\mathbf{K}]\{\mathbf{u}\} = \{\mathbf{F}_\text{th}\}\) is solved for displacements. Stresses are then recovered as:
Temperature-Dependent Material Properties
For accurate results, especially in high-temperature applications, \(E(T)\), \(\nu(T)\), and \(\alpha(T)\) must all be defined as temperature-dependent tables. The integral form for mean CTE used in practice is:
Some codes (Abaqus) require the "secant CTE" defined with respect to the absolute reference: \(\bar{\alpha}(T_\text{ref}, T)\). Consistent use of the same reference temperature in material definition and in the analysis step is essential to avoid silent errors.
My textbook defines CTE as \(\alpha = (1/L)\,dL/dT\). But Abaqus asks me to input "mean coefficient of thermal expansion." Are these the same thing?
They are subtly different, and confusing them is a very common source of error. The textbook definition is the instantaneous CTE — the rate of length change per unit temperature change at a specific temperature. The mean (or secant) CTE that Abaqus wants is defined as the average over a temperature interval: \(\bar{\alpha}(T_\text{ref}, T) = \varepsilon^\text{th}(T) / (T - T_\text{ref})\), where \(\varepsilon^\text{th}\) is the total thermal strain accumulated from \(T_\text{ref}\) to \(T\). For materials where \(\alpha\) is genuinely constant (most metals over small temperature ranges), both definitions give the same number. The difference matters for materials with strongly temperature-dependent CTE — some ceramics, composites, or special alloys. If you input the instantaneous CTE from your textbook into Abaqus as if it were the secant CTE, the code will compute thermal strains incorrectly because it integrates the secant definition. Always check the material data source carefully, and if you have instantaneous values, convert them to secant CTE with respect to your chosen reference temperature before inputting into the code.
Common Engineering Scenarios
Bimetallic Strips and Actuators
The bimetallic strip is the canonical thermal stress problem: two metals with different CTEs (e.g., brass at \(\alpha = 19\) ppm/°C and Invar at \(\alpha = 1.2\) ppm/°C) are bonded together. When heated, the brass wants to expand far more than the Invar, but they are bonded — so the strip curves toward the Invar side. The curvature is approximately:
Where \(m = t_1/t_2\) is the thickness ratio, \(n = E_1/E_2\) is the modulus ratio, and \(h = t_1 + t_2\) is total thickness. This principle is used in thermostats, bimetallic circuit breakers, and temperature-compensated mechanical mechanisms.
Gas Turbine Blade Thermal Fatigue
Turbine blades in jet engines cycle between cold on-ground conditions and hot operating temperatures (1,200–1,700°C gas temperatures, blade metal temperatures of 900–1,100°C with film cooling). Each engine start-stop cycle imposes a full thermal cycle on the blade, and the non-uniform temperature distribution (hotter leading edge, cooler trailing edge, strong radial gradient, film cooling holes locally cold) generates complex alternating stress states. Low-cycle thermal fatigue (crack initiation after hundreds to a few thousand cycles) is the primary life-limiting failure mode for high-pressure turbine blades. Life prediction requires:
- Detailed temperature distribution from conjugate heat transfer CFD
- Sequentially coupled structural analysis with temperature-dependent elastic/plastic material properties
- Fatigue life prediction using strain-life (\(\varepsilon\)-N) methods (Coffin-Manson) for the cyclic plastic strain at the critical location
Electronics PCB Assembly (SMT Solder Joint Reliability)
Surface mount technology (SMT) solder joints between integrated circuit packages (BGAs, QFPs) and PCB substrates experience thermal fatigue during thermal cycling qualification tests (typically -40°C to +125°C). The mismatch between the package CTE (typically 6–9 ppm/°C for ceramic or 15–18 ppm/°C for plastic packages) and the FR4 PCB in-plane CTE (14–16 ppm/°C) generates shear stress at the solder joints. Solder creep under cyclic thermal load initiates cracks at the package corners where the offset from the package center ("distance to neutral point") is maximum. FEM analysis using viscoplastic solder models (Anand model) predicts accumulated inelastic strain per cycle, which feeds into Coffin-Manson life prediction.
For a BGA solder joint analysis, why do cracks always start at the corner joints and not in the middle of the package?
It's all about the distance to the package's "neutral point" — the center of the package where both the package and PCB expand and contract together. If the package and PCB had identical CTEs, every point would move exactly with its thermal neighbor and there would be no relative displacement across any solder joint — zero shear strain, zero fatigue. The CTE mismatch causes relative displacement, and that relative displacement scales linearly with the distance from the neutral point. Corner joints are the farthest from the center, so they experience the largest lateral displacement between the top and bottom of the solder ball during each thermal cycle. More displacement means more shear strain, more creep per cycle, and faster fatigue crack propagation. That's why the reliability specification for a BGA is often stated as "the critical joint is the outermost corner joint at the package corner farthest from the centroid." FEM confirms this analytically — the accumulated plastic work per cycle is 3–5× higher at the corner joints compared to the center joints, and observed failure locations in cross-sectioned boards after thermal cycling tests match the FEM predictions very well.
Software Comparison: Abaqus, ANSYS, Code_Aster
Abaqus: Coupled Temperature-Displacement Analysis
Abaqus provides three approaches for thermal stress:
- Sequential (two-step): Run
*HEAT TRANSFERstep, write temperatures to.odb, then run*STATICstep reading temperatures as*TEMPERATUREpredefined field. Most efficient for problems without strong coupling. - Fully coupled: Single step type
*COUPLED TEMPERATURE-DISPLACEMENT. Elements must be "coupled" elements (C3D8T, C3D10MT, CPE4T, etc.) that have both displacement and temperature DOFs. The monolithic system solves simultaneously. Required for contact conductance changes, viscous heating, and thermoelastic damping problems. - Adiabatic: Special case for very fast loading (explosive or impact events) where heat conduction is negligible within the event duration but thermomechanical coupling (heat generation from plastic work) is important.
ANSYS Mechanical: Thermal-Structural System Coupling
In ANSYS Workbench, thermal stress is handled via the Thermal-Structural linked analysis system:
- Create a Steady-State Thermal or Transient Thermal analysis system, define thermal BCs, solve
- Link its "Solution" cell to the "Setup" cell of a Static Structural analysis system — this automatically transfers temperatures as body loads
- The structural model inherits the thermal model's mesh if sharing geometry, avoiding interpolation error
- Material assignment: ANSYS Workbench's Engineering Data module holds temperature-dependent properties; enter \(E(T)\), \(\alpha(T)\), \(\sigma_y(T)\) as tabular data
- For full coupling: use the Coupled Field Static analysis type with PLANE223 (2D), SOLID226, or SOLID227 elements that carry both thermal and structural DOFs
Code_Aster: Free and Open-Source Multiphysics FEM
Code_Aster (developed by EDF, open source) is widely used in nuclear and civil engineering and provides robust thermal-mechanical coupling:
- Thermal analysis using
THER_LINEAIREorTHER_NON_LINEoperators (nonlinear for temperature-dependent properties) - Temperature field imported into mechanical analysis via
CREA_CHAMPand applied as aAFFE_VARC(variable field) - Full thermo-mechanical coupling via
STAT_NON_LINEwith coupled DOF formulation - Strong support for creep and viscoplastic material models — particularly relevant for high-temperature structural analysis (reactor components, pressure vessels)
- Freely available: an excellent option for academic work or organizations that cannot afford commercial licenses
In Abaqus, when I run a sequential thermal-structural analysis, I sometimes see stresses at unconstrained free surfaces even though they should be zero. What's going wrong?
That's a very common symptom and has a few possible causes. First, check your reference temperature: if you haven't specified *INITIAL CONDITIONS, TYPE=TEMPERATURE in the structural analysis to set the starting temperature consistently with the reference temperature in your *EXPANSION material definition, Abaqus uses 0°C as the default reference, and even a "uniform" temperature field will generate thermal strains because \(\Delta T = T - 0\) is non-zero everywhere. Second, check for temperature interpolation artifacts: if your thermal and structural meshes differ and you're interpolating temperatures, tiny temperature gradients introduced by interpolation errors can generate small spurious stresses at free surfaces. Third, and most subtle: even in a completely unconstrained body with a non-uniform temperature field, there genuinely can be residual stresses at free surfaces in the FEM solution that arise from the fact that the exact free-surface stress-free condition (\(\sigma_{ij}n_j = 0\)) is only approximately satisfied by the FEM interpolation. This is mesh-dependent and reduces with refinement. The diagnostic is: if you apply a perfectly uniform temperature everywhere equal to your reference temperature, all stresses should be exactly zero. If they're not, you have a problem with your reference temperature definition or material data setup.
Warping, Bowing, and Residual Stress in Manufacturing
Thermal processes in manufacturing — casting, welding, additive manufacturing (3D printing), heat treatment, injection molding — invariably leave residual stresses and permanent deformation in the finished part. These are among the most practically important applications of thermal stress analysis.
Welding Residual Stress
During fusion welding, the weld pool and adjacent heat-affected zone (HAZ) reach temperatures far above ambient. On cooling, the weld metal tries to contract but is constrained by the surrounding cold material. The result is tensile residual stress in and near the weld bead (axial and transverse), with balancing compressive residual stress in the base metal farther from the weld. These tensile residual stresses:
- Add directly to in-service loads, reducing the effective fatigue life
- Promote stress corrosion cracking in corrosive environments (pipelines, pressure vessels)
- Cause distortion (angular, longitudinal, transverse shrinkage) that affects dimensional accuracy
Post-weld heat treatment (PWHT) stress relief annealing reduces residual stresses by thermally relaxing the constraint at elevated temperature.
Additive Manufacturing (Metal 3D Printing)
Laser powder bed fusion (LPBF) and directed energy deposition (DED) processes involve very rapid heating and cooling cycles (cooling rates up to 10⁶ K/s in LPBF) layer by layer. Each deposited layer contracts on cooling and is constrained by the previously deposited layers, generating significant tensile residual stress in the build direction and compressive stress in-plane. This can cause:
- Part distortion and warping after removal from the build plate
- Delamination (layer separation) during printing if stresses exceed inter-layer bond strength
- Reduced fatigue life of the finished part
FEM simulation of AM thermal stress (using sequential thermal and mechanical analysis with incremental deposition steps) is an active research and engineering tool for support structure optimization, scan strategy selection, and distortion compensation ("reverse warping" of the CAD geometry).
In metal 3D printing, I've seen the part warp and detach from the build plate mid-print. Can FEM actually predict this and help prevent it?
Yes, and this is one of the fastest-growing application areas for thermal stress FEM. The basic idea is to simulate the AM process layer by layer: each new layer is "born" at elevated temperature, cools and contracts, and the stress state at the end of each layer deposition is carried forward to the next. The build plate support acts as a displacement constraint. When the computed stress at the base of a support structure exceeds the effective bond strength to the build plate, that's the predicted failure point — and you can use this to add more support material, change the part orientation, or modify the scan strategy to reduce the stress concentration. State-of-the-art tools like Simufact Additive, Amphyon (from Additive Works), and Autodesk Netfabb Simulation have automated workflows for this. They use inherent strain methods for speed (the per-layer plastic strain is precomputed from fast thermal simulations and applied as an equivalent distortion source to the structural model) rather than full thermo-mechanical simulation, which would take days for a complex part. The predicted warping accuracy is within 0.1–0.3 mm for typical titanium and Inconel parts, which is often sufficient to decide whether a print requires more support or a different orientation.
What's the "inherent strain method" you just mentioned? It sounds different from the usual thermal stress approach.
Inherent strain is a shortcut method that avoids simulating the full thermal transient for every laser scan line — which would be completely impractical for a complex part with millions of scan lines. The idea: experimentally or computationally calibrate the average plastic strain (inherent strain) produced in a unit volume of material by the AM process at a given parameter set (laser power, speed, layer thickness). This is a fixed strain tensor — typically compressive in-plane and tensile in the build direction for most metallic AM processes. Then, to simulate a full part, you just apply that inherent strain to each layer as it's "activated" in the structural model, without running any thermal analysis at all for the production simulation. It's essentially replacing the full thermomechanical simulation with a fast structural simulation driven by prescribed plastic distortion increments. The limitation is that it's calibrated for a specific machine, material, and parameter set, and it loses accuracy near geometric features (overhangs, thin walls) where the thermal behavior differs from the bulk calibration case. But for typical solid-block geometries, it gives good distortion predictions in minutes rather than days.