Want to build accurate AM parts? No stress!

Residual stress is one of the biggest challenges associated with making precise metal components using laser powder bed fusion (LPBF). As the component is built up layer-by-layer, stresses induced by rapid, localised cooling can result in part distortion, or even fracture. Building precision parts to exacting tolerances means that we must consider potential residual stress in our component design, and also adopt methods to minimise its manifestation in production.

This article looks at the origins and magnitude of residual stress in different materials. We will explore the steps that we can take to diminish its impact through careful process design and optimal processing conditions. We will apply design for AM (DfAM) principles to reduce stress build-up, by adjusting our part design and build configuration. Where stress accumulation cannot be eliminated, we will mitigate its effects using post-process heat and surface treatment. Finally, we will look at techniques to compensate for any unavoidable stresses so that our components emerge with the correct geometry.

Origins of residual stress

Residual stress is a natural result of the rapid heating and cooling that is inherent to the LPBF process. Each new layer is created by moving the focused laser across the bed, melting the top layer of powder and fusing it to the layer below. Heat flows from the hot weld pool down into the solid metal below, and so the molten metal cools and solidifies. This all happens very rapidly; in a matter of micro-seconds.

As a new layer of metal solidifies and cools on top of the layer below, it contracts. The new metal is constrained by the solid structure below and so its contraction sets up shear forces between the layers.

Image above - laser melting of a new weld track on top of a solid substrate (left). As the laser moves along the scan vector, it melts the powder, which then cools mostly through conduction of heat into the solid metal below. Once it solidifies, the cooling metal contracts (right), setting up shear forces between it and the layer below.

Residual stress can be destructive. As we add layers on top of one another, the stresses build up and can result in distortion of the part, leading it to curl up at the edges and pull away from its supports:

In more extreme cases, the stress may exceed the strength of the part, leading catastrophic cracking of the component, or distortion of the build plate:

These effects are most pronounced in components with large areas of uninterrupted melt, as these tend to have longer weld tracks and so there is more distance over which the shear forces can act. Overhanging and protruding features can also be particularly prone to distortion as they are less constrained and there may be stress concentrations formed at their point of intersection with the main part.

As we will see, residual stress is affected by our product design and the processing methods that we use.

How much of a problem is this?

The magnitude of the residual stresses induced in the part and the impact of these on distortion is material-specific. Factors that drive the level of residual stress include the difference between the melting point and the temperature of the substrate below, as well as the cooling rate during and after solidification. How these stresses translate into deformation or fracture will be governed by the thermal coefficient of expansion of the alloy, its elastic modulus, its yield strength and its ductility in the 'as-built' state.

The part geometry and the level of constraint applied to the part by supports and the build plate will define the stiffness of the component and thus the level of distortion that will occur during the build itself. If the build assembly is very stiff, then stresses can reach a high level leading to component fracture. Less rigid builds may distort at lower stress levels, but still fail as a result of collision between the re-coater and the part.

During the build process, the residual stress is borne by the part, build plate and support structures as an assembly. When we separate the part from the build plate, the stiffness of the part is reduced and the stresses within it will find a new equilibrium, potentially resulting in further distortion.

A good way to assess the propensity for a material to exhibit stress and distortion is by studying a cantilever artefact. Here we build a horizontal beam with periodic vertical supports to tie it down to the build plate, sufficient to resist any residual stresses that are generated in the build. The thin supports are then severed using a wire EDM process, leaving the artefact attached to the build plate at one end only, so that it is free to distort in reaction to any residual stress. The amount of distortion can then be assessed for a particular material, processing technique and heat treatment.

Image above - cantilever artefacts following support removal. The amount of distortion is related to the level of residual stress and the stiffness of the material.

The z-displacement of the cantilever for each material is measured at three locations, as shown right. The maximum z-distortion is an indicator of the level of residual stress.

Which materials are most prone to deflection due to residual stress?

In this analysis, we have built identical cantilever artefacts in a range of common AM materials, using the same scanning strategy for all materials, but choosing typical melting parameters and layer thicknesses in each case:

We can see that Inconel shows the highest level of deflection due to the stresses built up during the LPBF process, followed by titanium, with aluminium showing the lowest displacement. This gives an indication of the level of tie-down supports that may be needed, and whether thicker build plates may be necessary.

Note that this does not imply that Inconel is the material most prone to cracking due to residual stress accumulation - this depends on ductility as well as stress. Of the materials tested here, Ti6Al4V is the most likely to fracture due to its low ductility in the 'as built' condition.

Minimising residual stress

We have seen how residual stress varies by material and that it can be significant. What steps can we take to minimise this phenomenon in our builds?

Melting strategy

One way to tackle residual stress is by varying our scanning strategy, choosing a method that is best suited to the part geometry. When we are filling in the centre of our part, an activity known as 'hatching', we typically move the laser back and forth. The pattern that we choose affects the length of the scan vectors and hence the level of stress that we are likely to build up in the component. Strategies with shorter scan vectors will generate less residual stress:

Image above - scanning strategies and their suitability for different part types. The two most common strategies are 'meander' for thin walled parts (also known as rastering), and 'stripes' for parts with thicker sections. 'Chessboard' or 'island' strategies can also be effective. Stripe and chessboard scanning keeps the lengths of individual scan lines shorter, reducing the build-up of residual stress.

We can also rotate the orientation of our scan vectors from one layer to the next so that stresses are not all aligned in the same plane. A rotation of 67 degrees is typically used between each layer to ensure that it is many layers before the scanning direction is exactly repeated.

We can illustrate the impact of scan vector alignment by measuring the deflection of cantilever artefacts that have been built with different scanning strategies. In this experiment, one CP-Ti artefact is built with long scan vectors longitudinal to the cantilever axis, another with shorter lateral scan vectors, and a third with a rotation of 67 degrees between each layer.

Unsurprisingly, we can see that the residual stress impact is greatest when the scan vectors are in the longitudinal / vertical orientation to the cantilever axis. These aligned scan paths result in 37% more strain than the baseline case (67 degree rotation).

Image above - deflection of CP-Ti cantilever artefacts in the 'as built' state. Three different scan strategies have been used: 67 degree rotation between each layer (left), all scans aligned perpendicular to the cantilever axis (centre) and all scans running along the cantilever axis (right).

When all the scan lines are lateral / horizontal, the cantilever deflection is reduced by around 30% compared to the baseline case in this material. Lateral scan lines can be useful to minimise distortion in vulnerable regions of a part. Of course, stresses will be increased in the perpendicular direction, so this is unlikely to be a good strategy to use on a whole component.

Layer thickness

Does layer thickness have an impact on residual stress accumulation? Not much. Thicker layers require a deeper melt pool to achieve fusion with the layer below, requiring us to input more energy and thus induce more residual stress in each layer. However, the lower number of layers offsets this, such that the total induced stress and thus the part distortion can be very similar, as shown below:

Image above - layer thickness has little impact on residual stress build up in cantilever artefacts made from Inconel 718 and Ti6Al4V. Each material is built in both 30 micron and 60 micron layers, with 67 degree scan line rotation between each layer. The same laser power (200W) is used in all cases, but the thicker layers use parameters that move the melt pool more slowly to increase the penetration of the laser energy. This increase in 2D energy density is offset by the reduction in the number of layers, resulting in very similar levels of total energy input, stress and cantilever deflection.

Processing conditions

Pre-heating of the substrate is another technique used to reduce residual stress. By reducing the temperature differential between the new melt track and the heated substrate, the layer-by-layer accumulation of residual stress is also reduced. Hence we are building up less stress in our part than before.

Furthermore, elevating and sustaining the temperature of the substrate during the build allows the stresses generated by the laser melting process to dissipate. Heating the component temporarily reduces its yield strength and increases its ductility, enabling zones within the part with residual stresses above this reduced threshold to yield or deform. This reduces peak residual stress to no more than the yield strength of the material in its heated state.

Image above - high temperature build volume enables processing at temperatures of up to 495 °C, sufficient to provide significant stress relief for many alloys. A water cooled jacket prevents heat soak into the AM machine structure. Another technique to achieve a similar effect is to pre-heat the substrate using a de-focussed beam prior to melting.

Generally speaking, the higher the pre-heat temperature the greater the stress relief effect. Atomic diffusion increases at elevated temperatures, and atoms in regions of high stress can move to regions of lower stress, which results in the relief of internal strain energy.

Higher bed temperatures also reduce the cooling rate of the melt track, and this lower thermal gradient affects the 'as built' microstructure. In Ti6Al4V, for instance, we see coarser alpha laths forming at higher build temperatures, whilst formation of the martensitic phase reduces with more build plate heating. This helps reduce residual stress, but also impacts on the strength and ductility of the 'as built' component.

Note that each material has a 'critical temperature' at which a change occurs in either the phase (e.g. solid to liquid) or the crystal structure (e.g. body-centered cubic to face-centered cubic). Stress relieving must be performed below this temperature to avoid a substantial and permanent change in 'as built' metallurgical properties.

The beneficial impact of high temperature processing is illustrated by the results shown below, where we compare distortion in cantilever artefacts made from Ti6Al4V that have been processed at two different bed temperatures. In this case, the distortion is reduced by up to 70%.

Image above - comparison of distortion of Ti6Al4V cantilever artefacts built with various scan strategies at two different bed temperatures: 170 °C (top set of graphs) and c. 500 °C (bottom set of graphs).

In this chart, a cantilever artefact has been built using the least favourable longitudinal / vertical scan strategy at each of a range of elevated build plate temperatures. As we would expect, the stress relieving effect increases with temperature.

There are practical considerations that may limit the temperature to which the build plate can be heated and the consistency of the stress relieving effect. For instance, bed temperatures may be limited by the choice of re-coater material. Furthermore, the reactivity of the powder will also be increased at elevated temperatures, potentially resulting in unwanted oxidation or greater affinity with the re-coater. At very high temperatures, the powder may become semi-sintered into a 'cake' requiring additional processing before re-use.

Finally, if we are using a heated build plate and we are relying on the part to transmit heat to its top surface where the laser is adding new material, taller builds may not do this so effectively, reducing the stress relieving effect as the build progresses. An uneven temperature profile of this sort may also affect the material microstructure, with coarser grains forming at the bottom of the build where the highest temperatures are experienced for the longest time.

So, whilst elevated build temperatures can be helpful for stress relief, the technique used is important and may lead to other unwanted side-effects.

Component design

One of the best ways to minimise residual stress in our component is to design it out in the first place. The following design tips can help:

Whilst it is ideal to avoid large areas of uninterrupted melt, it may not always be possible to eliminate thicker horizontal sections from part designs - e.g. where a mating flange is required. The hybrid approach mentioned above is a useful technique to avoid thick sections at the base of a build, by incorporating the build plate into the component. We no longer have to build the flange but rather build on top of it, so we avoid all the residual stress that would otherwise be accumulated. This has the added benefit of reducing build time and component cost, but may require some post-process machining.

Images above and right - manifold manufactured using a 'hybrid' process where the build plate forms an integral element of the finished component. Post-process machining removes excess metal to create a mounting flange at the base of the component.

Mitigating residual stress

Despite our best efforts to minimise it through product and process design, some residual stress may remain at the end of our build, with possible adverse effects on the component's shape and strength. What can we do to mitigate these effects?

Post-process heat treatment

In a similar manner to heating the build plate, post-process heat treatments can relieve the stresses that have accumulated in the build. Once we have extracted the part from the build chamber and removed any un-melted powder, we are less constrained than we are in the build chamber in terms of the temperature cycles that we can apply.

If stress relief alone is our intent, then the heat treatment cycle will need to be carefully matched to the material to ensure that we get the correct stress reduction effect without inducing further metallurgical change. By contrast, higher temperature annealing and tempering processes that exceed the critical temperature of the alloy will create permanent modifications to the crystal structure, with consequential changes to hardness, strength and ductility. Hot isostatic pressing is a technique that combines high temperature and pressure to modify the alloy's microstructure and suppress any remaining porosity, as shown below for Ti6Al4V:

We can measure the effect of post-process heat treatment by putting our cantilever samples through a stress relieving cycle before the supports are severed. After heat treatment, we prepare the artefacts in the same way and measure the difference in distortion. Here we can see that heat treatment of these CP-Ti artefacts (850 °C in a vacuum furnace for 2 hours) leads to full stress relief.

Image above - cantilever artefacts measured in the 'as built' state (top) and after heat treatment (bottom). The heat treatment effectively removes the residual stress from the build. Note that the small negative displacements detected in the heat treated condition are within the measurement error and hence do not imply compression.

Post-process surface treatment

Part distortion is not the only concern raised by the presence of residual stress in our AM parts. The impact on component strength and fatigue performance may also be significant.

Residual stress can reduce component strength by reducing the additional load that can be born before the yield strength is breached. Tensile residual stresses are of particular concern, and so surface treatments such as shot peening can be helpful to introduce compressive stress into surface regions. This technique also benefits fatigue life by closing up and suppressing crack propagation from surface defects, although excessive coverage can have the opposite effect, so care is needed!

Measure and compensate

As we have seen, residual stress can be difficult to eliminate altogether. In some circumstances, it may be best to accept that stresses will be present and to attempt to compensate for their effect in the build.

One instance where this is a good strategy is when we are producing thin-walled Ti6Al4V lugs for a mountain bike frame that must be bonded to a carbon fibre tube. In this instance, an adhesive bond is form by mating a regularly shaped carbon fibre tube into a double-lap shear joint.

The joint comprises a pair of concentric thin-walled cylinders, with a narrow gap between them into which the carbon tube is inserted, to a depth of 20 mm (see image right, courtesy of Robot Bike Co). It is therefore critical that the mating features on the thin-walled metal lug are precisely round and parallel.

Since the lugs typically contain two such double-lap joints at customised angles to one another, it is not possible to build each mating feature in the ideal orientation to minimise stress. The webs between the joint geometries also exert forces on the part as the build proceeds, which can result in the joints becoming distorted. Careful selection of wall thicknesses, cross-bracing and internal supports can reduce this effect, but may not always eliminate it altogether if we are to keep component weight to a minimum.

Image above - Ti6Al4V lug set for a Robot Bike Co R160 customised mountain bike.

What can we do to eliminate this unwanted distortion? The answer is to measure the distortion and apply the mirror image of the error to the part geometry. This enables the counter-distorted component to 'pull itself straight', ensuring a precise fit with the carbon tubes.

Of course, a more sophisticated approach than trial and error would be to simulate such distortions and adjust the design before the first trial build. This will be the subject of a future article.

Summary

Understanding and minimising distortion due to residual stress is a key factor in successful LPBF builds.

Component precision depends on selection of an appropriate material; careful part design, build configuration and scanning strategy choice to minimise stress accumulation; use of optimum processing conditions to reduce stress during the build; post-process heat and surface treatment to alleviate stress; and compensation for unavoidable distortion using measurement and simulation.

Readers may also want to refer to Design for metal AM - a beginner's guide for more details about optimising part designs to minimise residual stress and to maximise feature precision.