6th September 2017 by Marc Saunders
Most design for additive manufacturing (AM) is not really worthy of the name – ‘adapt to AM’ would be a better description. Too often we design a component for function, and then ask someone else to think about how to build it. In other instances, we are forced to adapt an old design that was conceived with very different manufacturing methods in mind. Both cases inevitably lead to compromise in the form of a forest of 3D printing supports, which slow us down, cost us money and may impact on part quality.
When we are designing specifically for AM, we should aim to avoid relying on 3DP supports to make our builds feasible
We don’t have to accept this situation. As we move beyond our first forays into metal AM, we are learning to design for the process such that our components are mostly self-supporting. Failing this, our designs can include efficient sacrificial structures as part of their design. When we are designing specifically for AM, our goal should be to avoid relying on 3DP supports to make our builds feasible.
This post looks at the way that many AM processes are engineered today, which can be traced back to limitations in design for AM thinking. We will see how reliance on 3DP supports creates compromises that should be avoided wherever possible in industrial AM. We will look at some of the latest design tools that help us to design efficient components with minimal support structures. Finally, we will also look at some case studies that demonstrate an alternative design for AM approach.
Thanks once again to my colleague Kevin Brigden, on whose work this article is based. Kevin also contributed to DfAM strategy – create ‘design space’ for maximum AM impact.
Why do we use supports?
3D printing supports are often not really ‘supporting’ anything. In laser powder-bed fusion processes, they often perform the roles of tie-downs and heat sinks. They act mostly in tension to prevent the part from peeling upwards due to the residual stresses that are generated during the laser melting process. They also provide a pathway for heat to be carried away from thermally isolated areas of the part which would otherwise cool too slowly, with detrimental consequences for their surface finish and material properties.
Supports as an afterthought
3D printing supports are the extra sacrificial structures that we add to our part geometry to make it buildable. We can specify a wide range of support types including thin columnar, linear and buttressed structures, as well as trees and other angular shapes that enable us to support vulnerable regions whilst avoiding other structures lower in the build. Software tools to do this are getting ever more powerful and user friendly.
But this whole approach is tackling the symptom, rather than addressing the root cause. Adding 3DP supports after the design has been completed stems from a prototyping mind-set in which additive technologies are used as a craft process to realise whatever shape comes into a designer’s mind. In such situations, effectiveness trumps efficiency. Too often we think about the part geometry first, with buildability coming only as an afterthought, which can have messy consequences.
Adding supports in an unsupported environment
Build preparation is typically done in a specialised software application, outside of the CAD system and generally once we have converted our component into a dumb surface (STL) model. Using a ‘surface snapshot’ of the model means that we lose vital parametric data that expresses the designer’s intent. In the absence of this information, the engineers working on build preparation may not be aware of the most important performance characteristics of the component.
Also, because we are now outside of the product lifecycle management (PLM) environment, we have no ‘feature tree’ or similar design interrogation tool to control the supporting features that we are adding, leading to a further loss of traceability and dilution of design intent. 3DP supports themselves are typically generated as facet bodies and so are not parametric – if we make changes to our CAD model, then we may have to start again with our support generation.
Supports are typically added in a non-PLM environment where surface representations of the model lead to a loss of precision and design intent
What’s more, how the supports themselves connect to the part surface may affect the local component quality. Supports have to connect securely to the component, and so they are typically designed so that they extend into the component surface. Support structures will typically have different build parameters to the component since we are not aiming for maximum density in our supports. We will therefore have zones where we re-melt the ends of supports, and these zones will have a slightly different thermal history to their unsupported neighbours.
In summary, adding 3DP supports in ‘off the grid’ build preparation software can lead to an unwelcome loss of engineering control.
Parametric build preparation
Fortunately, this picture of build preparation isolated from the CAD / PLM world is changing. Integrated design tools are emerging that enable design for AM to be performed within the parametric CAD environment. For instance, Renishaw has collaborated with Dassault Systèmes to integrate its QuantAM build preparation technology within the 3DEXPERIENCE suite, which also includes advanced generative design tools to develop lightweight, efficient AM components.
Later in 2017, Spatial’s 3D ACIS modeller will enable users of Renishaw’s QuantAM build preparation software to import and manipulate parametric CAD models, avoiding the data loss associated with conversion to the STL surface data format. For more details, see Spatial and Renishaw Collaborate to Bring an Array of New Capabilities to QuantAM.
The hidden cost of supports
Whilst it may be possible to build just about anything if we use enough supports, that doesn’t mean that its sensible to try, especially if we have lots of parts to produce. Production designs should be much more considered.
Using 3DP supports has costly consequences. Not only have we consumed more material than we should have done to make the part, but we have also spent longer in doing so. Plus, we have given ourselves more work to do once the build is complete, removing attached support structures and cleaning up support regions, racking up yet more unwanted costs.
3DP supports are generally long and thin, which makes them prone both to thermal warping and to damage due to interaction with the powder dosing system. Warping occurs where there are excessive temperature gradients through a structure that is insufficiently stiff to prevent large strains or displacements. Problems may also occur where too much rigidity in one location causes load transfer to affect another, seemingly unrelated area of a structure, also leading to distortion. Such deflections may lead to supports, or even the part, catching on the dosing wiper as it moves across the bed, especially in cases where a rigid wiper is used. This can lead to us spending a lot of time adjusting variables, generally through time-consuming trial and error.
3DP supports are costly in terms of build time and post-processing, whilst warping and wiper damage can lead to failed builds
Whilst this sort of thing might be acceptable in a prototyping lab when we’re making one-off models, such inefficiency and non-conformance may not be tolerable in an industrial AM setting.
Can we avoid 3DP supports altogether?
Looked at in this way, any use of 3DP supports in production AM applications can be viewed as a shortcoming in our part design. The reasons that we need supports – overhanging regions, residual stress and thermal management during the build – are primarily functions of geometry, which are under our control. As designers, we can either design supports out altogether, or we can take care to design them in deliberately.
Designers can either design supports out, or design them in properly
But is it realistic to do away with 3DP supports altogether? Surely, many parts that we want to make additively just don’t have the sort of geometry that can be built unsupported, do they? Aren’t many businesses – notably in medical and dental applications – successfully using AM to produce components that rely on 3DP supports? Aren’t 3DP supports always going to be a fact of life?
Yes, yes and not necessarily.
It’s certainly the case that legacy designs that we have previously made using other processes, will require supports if we adapt them to AM. But the future of industrial AM does not lie in making legacy parts.
It is also true that medical implants that have to conform to the shape of patients’ bodies are always going to come in a range of awkward shapes that we will need to support and hand finish. 3DP supports may remain useful in these applications, although even here there is the opportunity to design part-specific supports.
However, the real power of AM lies in enabling tomorrow’s high-performance products and, in industrial applications at least, we are less at the mercy of nature and we can choose what we want to design. In this domain, it is often possible to do away with 3DP supports altogether.
Design your build, not just your component
If we are designing something new and we want to use AM for series production, then we can’t just design for function. We must take full control over both the function and the production of our parts. This means designing with the characteristics of the additive manufacturing process (both positive and negative) fully in mind. We must design the build at the same time as we design the part.
We should ideally avoid 3DP supports by designing our parts to be self-supporting. Where this is not possible, we should design sacrificial supporting structures in harmony with the part.
Wherever possible, we should avoid 3DP supports altogether by designing our parts to be self-supporting. With careful thought, the component can often be orientated and optimised to carry itself whilst using a minimum of material. Where constraints on the design make such a self-supporting design impossible, at the very least we should design any sacrificial supporting structures in harmony with the part. We can create carefully designed CAD supports that interface with the component in a planned manner, rather than add a random, Lilliputian forest of spindly props as an afterthought.
Case study #1 – fully self-supported design
Our first case study is an example where supports can be eliminated almost entirely through careful design. Here we are redesigning a seat post for a mountain bike using topological optimisation. The design software (ALTAIR’s solidThinking) creates a highly efficient load-bearing structure, although the first iteration proves to be not quite so efficient when it comes to the build preparation stage:
So the remaining design task is now to make this performance-optimised part design with a more efficient build process. Here we have to eliminate large overhangs, and also orientate the part such that the protruding arms can be built without supports. With some subtle geometry changes and a small reorientation, this proves to be possible:
This leaves us with a lightweight part (a 44% reduction in weight compared to the cast original), that can be built with minimal supports (yellow in the images below). These 3DP supports would benefit from conversion into CAD-designed supports to facilitate a controlled separation from the build plate.
For more details of this case study, see Making the world’s first 3D printed mountain bike.
Case study #2 – simple supports for an external lug
Now let’s look at an example where design constraints prevent a fully self-supporting design. In this instance, the majority of the shape self-supports, but an external lug is more problematic (see below).
This common situation can be easily dealt with. It is a well known design for AM rule that lateral circular holes in excess of 10 mm diameter require supports. You may also need to machine them once they are built to ensure that traces of the supports are removed and that you are left with a round hole.
So why make it a circle in the first place? If it is small enough then you can build it, but it won’t be very round. If you need it to be a more precise circle, then you will need to machine it anyway. Therefore, don’t build a circle, but do design a more efficient solution: a diamond shape that is self-supporting and will require milling to produce the final round hole. The tab on the right in the image above requires a single snap-off support. If you are really concerned about component weight, then the bottom corner of the tab could also be machined away when you are milling out the hole.
Case study #3 – designed supports for a pipe section with end flanges
This final example, a section of exhaust pipe, is another case where design constraints prevent a perfect design for AM solution. But it provides a useful demonstration for how there is often a way to avoid using 3DP supports if we think about it hard enough.
For performance reasons we need the inside of the tube to be as smooth as it can possibly be, whilst the flanges must form a good seal against the cylinder head and downstream pipework. We ideally want to avoid attaching any supporting structure (3DP or otherwise) to the internal surface of the tube. We must also be careful that the flanges don’t warp.
Firstly, we deconstruct the design into its essential parts – the tube and the fixings – by using surface offsets and the ‘thicken’ tool in our CAD software. When designing for AM, this is where we should spend the majority of our time, looking at the model in many different orientations to work out the best way to build the critical features and to link everything together.
Our overriding driver here is to avoid supports inside the tube, and this determines the orientation of the component in the build volume. Next we look at the detail design of the flanges. We know that we are going to need to support these as the distance of the fixing holes from the pipe aperture is just too big to integrate them into the main pipe. Knowing this, we can specify a small flat face to design our CAD supports up to. We will get a small stress raiser here which we can use to detach the CAD supports.
So let’s compare the original design with the more considered approach featuring CAD supports. In terms of time and effort, we are substituting time spent in our build preparation software manually adding and manipulating supports, for time spend in CAD optimising the part design. The advantage of the latter approach comes firstly in terms of shaving mass off the part through the lighter end flanges (part volume reduces from 11.9 cm3 to 9.2 cm3). The CAD supports are also substantially more compact than the 3DP supports that would be needed to build the original design, reducing process waste.
Too often ‘design for AM’ is about finding ways to build products that really haven’t been designed for AM at all. Heavy reliance on 3DP supports is a sign of insufficient up-front thinking about how a component is to be made. Applying 3DP supports as an afterthought means that a crucial stage of the design process not fully under control, and that part quality may be compromised.
With sufficient thought at the design stage, it is often possible to avoid relying on 3DP supports altogether
With sufficient thought applied in the design phase, it is often possible to avoid such wasteful behaviour altogether. Modern design tools support the creation of efficient structures, but intelligent design thinking is still needed to make the shapes that they create easy and efficient to build. Where design constraints make the self-supporting ideal unobtainable, we can still design sacrificial structures that maximise the efficiency of our designs.
Companies have a choice to make about the type of competitive play that they want to make with additive. Those companies that take additive most seriously – for whom AM is a business strategy – have their product designers fully engaged in designing products specifically for AM. By contrast, those firms that are asking manufacturing engineers to adapt unsuitable designs using 3DP supports, are acting tactically rather than strategically. Businesses that adopt a true ‘design for AM’ strategy will reap the greatest competitive rewards.