Residual Stress in Metal AM: Why Parts Warp and What to Do About It

A cantilever bracket, beautifully printed in Ti-6Al-4V, is wire-EDM'd from the build plate. The moment the last cut is made, it deflects upward by 0.8 mm. A turbine blade built in IN718 cracks along a layer boundary hours after removal. A thin-wall aerospace rib, stress-relieved and seemingly stable, distorts during final machining.

All three failures share a root cause: residual stress — internal stress locked into the part during the build process that is released, partially or completely, when constraints are removed.

Understanding residual stress in metal AM is not optional for anyone building structural parts. This article explains the fundamental mechanism, the process variables that amplify or reduce it, and the engineering options for managing it.

The thermal gradient mechanism

The dominant source of residual stress in LPBF and DED is the thermal gradient mechanism (TGM). It operates at every melt pool in every layer.

When the laser strikes a powder layer, it melts a small volume of material within microseconds. The temperature in the melt pool reaches 1500–3000°C depending on material. The surrounding solid material — just 100–200 µm away — is still near ambient temperature.

The hot newly melted material attempts to expand thermally but cannot expand freely because it is surrounded by cold, stiff solid material. It is forced into plastic compression — the thermal expansion is accommodated by plastic deformation rather than dimensional growth.

On cooling, the now-plastically-compressed material attempts to contract to its natural (stress-free) volume. Again, it cannot — the surrounding solid resists. It ends up in residual tension as it solidifies.

This sequence repeats for every melt pool, in every track, in every layer. The cumulative effect is a stress field with:

Tensile stress in the upper portion of the build (the most recently solidified layers)
Compressive stress in the lower portion of the build (older layers, constrained by the build plate and later deposits)
The build plate itself in compression due to the tensile stress in the part pulling it upward

The magnitude can be significant. In LPBF Ti-6Al-4V, residual stress parallel to the scan tracks can approach 600–900 MPa — close to the yield strength of the as-built material (900–1100 MPa). In IN718 LPBF, residual stress values of 500–800 MPa have been measured by neutron diffraction.

The bi-metallic strip analogy

The easiest way to visualise what happens when a part is released from the build plate is the bi-metallic strip analogy.

A bi-metallic strip consists of two layers of different metals bonded together. When heated, the two metals expand by different amounts, causing the strip to bend. In an AM build, the equivalent layers are:

The upper build layers: in residual tension (trying to shorten)
The build plate: relatively unstressed (constrained by the machine frame during the build)

When the part is cut from the plate, the tensile upper layers are released. They contract, pulling the ends of the part upward — a classic cantilever deflection. The amount of deflection depends on the part stiffness, the residual stress magnitude, and the distance from the neutral axis to the high-stress region.

For a simple cantilever LPBF beam, measured deflections at the free end after plate removal are typically:

Ti-6Al-4V (no stress relief): 0.3–1.5 mm per 100 mm length
IN718 (no stress relief): 0.2–0.8 mm per 100 mm length
316L SS (no stress relief): 0.1–0.5 mm per 100 mm length
AlSi10Mg (no stress relief): 0.05–0.2 mm per 100 mm length

Aluminium deflects less despite high build temperatures because its low yield strength means the thermally induced plastic compression is relieved more readily — less stress is locked in.

Scan strategy effects

The laser scan strategy — the pattern by which the laser traverses each powder layer — has a strong influence on the final residual stress state. Different strategies create different thermal gradient orientations and therefore different stress distributions.

Alternating stripe (bidirectional)

In alternating stripe, each layer is scanned in parallel tracks (stripes) from one edge to the other, alternating direction between layers. Residual stress is strongly anisotropic — higher in the direction perpendicular to the stripes and lower in the scan direction. If the part has a preferred loading direction, residual stress may align unfavourably with it.

Island (chess) scanning

The build area is divided into small square or hexagonal islands (typically 5–10 mm). Each island is scanned independently, and the order of island scanning rotates between layers. Island scanning creates a more spatially uniform residual stress distribution because no single scan vector runs across the full part cross-section. Peak stress is lower, and the stress pattern is less directional.

Island scanning is the default strategy in many commercial LPBF machines (EOS, SLM Solutions, Trumpf) specifically because it reduces stress and distortion.

Layer rotation (67° offset)

Many machines apply a rotation to the scan pattern between successive layers — the most common being 67°, derived from optimisation studies in the 2010s. 67° is not a round number by accident: it is approximately irrational with respect to 360°, meaning the scan pattern cycles through many orientations before repeating. This prevents layered anisotropy and distributes residual stress more uniformly through the thickness.

The effect is meaningful: a 67° rotation between layers can reduce peak residual stress by 20–35% compared to a fixed scan direction, based on neutron diffraction and synchrotron X-ray measurements (Mercelis & Kruth, 2006; Parry et al., 2016).

Contour + infill

Most machines separate the outer boundary scan (contour) from the internal fill. The contour runs slowly at lower power to produce a good surface finish; the infill is faster and higher power. This means the surface layer is in a different stress state from the bulk. If the contour is scanned last (which is common), it re-melts the surface after the infill stress field is established — slightly reducing surface tensile stress.

Effect of pre-heating

Pre-heating the build platform reduces the temperature differential between the melt pool and the surrounding material, directly reducing the thermal gradient and therefore the TGM-driven stress.

EBM: The powder bed is preheated to 600–1000°C using a defocused electron beam before part scanning begins. The temperature during the build is maintained at these levels. The thermal gradient at the melt pool is 500–1000°C rather than 1500–2000°C. Result: EBM parts emerge from the build essentially stress-free relative to LPBF. Distortion on release from the build cake is negligible. This is a fundamental advantage of EBM for stress-sensitive geometries.

LPBF: Build platform heating in commercial LPBF is typically limited to 150–200°C due to machine constraints (seal materials, optics cooling, atmosphere control). Some high-temperature LPBF systems (Trumpf TruPrint with pre-heat option, SLM Solutions 280 HL) can reach 500°C for processing tool steels and intermetallics. The moderate heating available in standard LPBF reduces residual stress by approximately 15–25% compared to a cold build plate — useful but not transformative.

Practical pre-heat effect for standard LPBF:

Platform temperature	Typical residual stress reduction vs. cold plate
25°C (ambient, no heating)	Baseline
80°C	~10% reduction
150°C	~20% reduction
200°C	~25% reduction
500°C (high-temp LPBF)	~50% reduction
800°C (EBM)	~90% reduction

Layer thickness and melt pool depth

Thicker layers deposit more energy per unit area, creating a deeper melt pool with a steeper thermal gradient. Thicker layers therefore generate slightly higher residual stress per layer, partially offset by the fact that fewer layers are required to build the part.

The net effect on total part distortion of changing layer thickness from 30 µm to 60 µm is relatively small — typically 10–20% increase in peak stress magnitude. Scan speed and laser power (collectively captured by volumetric energy density) have a larger effect on melt pool geometry and therefore TGM intensity than layer thickness alone.

Melt pool depth matters because the TGM stress field extends approximately one melt pool depth below the surface. For standard LPBF parameters (30–60 µm layer, 100–300 W, 500–1500 mm/s), melt pool depth is 60–150 µm. Re-melting of previous layers is unavoidable — each layer is re-scanned by the next one's melt pool, creating a complex, accumulated stress history.

Stress relief annealing

Stress relief annealing is a sub-transformation-temperature heat treatment that allows residual stresses to relax by thermally activated creep, without significantly changing the alloy microstructure.

Mechanism: At elevated temperature, the yield strength drops and dislocation mobility increases. The stored elastic strain energy drives plastic flow at the stress concentrations, redistributing stress into a more uniform, lower-magnitude state.

Typical stress relief cycles for common LPBF alloys:

Alloy	Temperature	Hold time	Atmosphere	Notes
Ti-6Al-4V	650°C	2–4 h	Argon (must be inert)	Above 650°C, alpha microstructure begins to coarsen
IN718	870°C	1–2 h	Argon or vacuum	Below solution anneal — does not dissolve delta phase
316L SS	450–600°C	1–2 h	Argon or vacuum	Higher end gives more relief but reduces cold-work strengthening
AlSi10Mg	270–300°C	2 h	Air or inert	Above 300°C, Si precipitates coarsen and strength drops
17-4PH	480°C	4 h	Argon	Combined stress relief + H900 age for martensitic condition
CoCrMo	1050–1100°C	2 h	Argon or vacuum	Also partially anneals the as-built microstructure

When to do it: For LPBF Ti-6Al-4V and IN718, stress relief before build plate removal is strongly recommended for any part with a significant horizontal extent (> 30 mm) or walls taller than ~50 mm. Removing highly stressed parts cold from the plate causes immediate elastic spring-back that may exceed dimensional tolerance.

What it does not do: Stress relief at sub-transformation temperatures does not close porosity, does not change the grain structure significantly, and does not improve mechanical properties. It is strictly a geometric stability treatment.

HIP as combined stress relief and densification

Hot Isostatic Pressing (HIP) applies both temperature and isostatic gas pressure (typically 100–200 MPa argon) simultaneously. For LPBF parts, HIP serves two distinct roles:

Densification: The isostatic pressure drives creep closure of spherical gas pores and shrinkage porosity, achieving >99.9% density. (Note: HIP does not close planar lack-of-fusion defects — see the post-processing guide.)
Stress relief: The high temperature (typically 900–1200°C for most alloys, well above standard stress relief temperatures) combined with extended hold time produces thorough stress relaxation. The residual stress in HIPped LPBF Ti-6Al-4V is typically below 50 MPa.

The trade-off: HIP temperatures are high enough to cause grain growth and microstructural changes. For IN718, HIPping above ~1000°C dissolves strengthening precipitates — a post-HIP aging treatment is required to restore strength. For Ti-6Al-4V, the alpha+beta microstructure coarsens, which slightly reduces tensile strength but improves ductility and reduces fatigue scatter.

For aerospace and medical fracture-critical parts, the HIP cycle is increasingly mandated. Use the HIP cycle designer to specify cycle parameters for your alloy.

Simulation tools

Residual stress prediction is an active area of commercial software development. The available tools range from process-scale finite element models to simplified distortion predictors.

Full physics simulation (layer-by-layer FEA):

Ansys Additive Print / Additive Science
MSC Simufact Additive
Autodesk Netfabb Simulation

These solve the thermal and mechanical problem for every scan track or layer. They can predict distortion contours and residual stress fields with reasonable accuracy (±20–30% on distortion magnitude for well-calibrated models). Computation time is hours to days for part-scale models.

Inherent strain methods (fast distortion prediction):

Amphyon (Additive Works)
3DSIM exaSIM
Materialise Magics Simulation module

Inherent strain methods extract a characteristic strain field from a calibration build, then apply it analytically to predict distortion at part scale in minutes rather than hours. Accuracy is lower than full simulation but sufficient for comparing orientations and identifying high-risk features.

What to validate against: Simulation results should be calibrated against distortion measurements from a known benchmark geometry (e.g., a cantilever bridge, an Inconel benchmark as per NIST AM-BENCH). Without calibration, absolute stress predictions are unreliable.

Design mitigations

Beyond heat treatment, several design choices reduce the residual stress problem at source.

Topology and wall thickness transitions: Sharp changes in cross-section create stress concentration during solidification — analogous to a notch in fatigue. Smooth tapering transitions at section changes reduce peak stress magnitude. A transition from a 3 mm wall to a 10 mm wall should be spread over at least 5–10 mm of height.

Avoiding long unsupported bridges: Horizontal spans without intermediate support accumulate layer-by-layer distortion. A 60 mm horizontal bridge in Ti-6Al-4V LPBF can accumulate 0.3–0.6 mm of mid-span deflection by the time it is fully built. Either add support below the bridge, split the span with a central support strut, or re-orient to bring the bridge closer to vertical.

Orientation selection: Residual stress and distortion are lowest in the direction perpendicular to the build plate (Z). Thin features built vertically (fins, ribs, walls) accumulate far less distortion than equivalent features built horizontally. If the part has a thin, flat section that is the most distortion-sensitive, orient it vertically.

Gussets and ribs: For thin-wall parts that must be built horizontal, adding temporary stiffening ribs (to be machined off post-build) can constrain distortion during the build. This is common practice in sheet-like aerospace AM parts.

Use the thermal distortion tool to estimate distortion risk before committing to a build strategy. The orientation advisor helps identify the orientation that minimises distortion for a given part envelope.

Summary: residual stress mitigation by severity

Risk level	Typical scenario	Recommended action
Low	Short, compact geometries; EBM; AlSi10Mg	Island scanning, mild plate heat, remove cold
Medium	Medium complexity LPBF; stainless or aluminium	Island + 67° rotation, 150°C plate heat, stress relief before removal
High	Ti-6Al-4V or IN718 LPBF; long cantilevers; thin walls	Aggressive stress relief (650°C / 2h for Ti), pre-heat, simulation before build
Critical	Aerospace fracture-critical, tight tolerance post-machining	HIP + microstructural treatment, distortion simulation with calibrated model