The Troubles with Topology Optimization for FFF - Part 1

Have you ever heard of Design for Additive Manufacturing (DFAM)? The term and associated technologies are constantly thrown around in the additive design and engineering world, challenging traditional design paradigms. What is it really, though, and why is it so novel? In this blog post, we’ll dive into one technology being widely touted as the next evolution for DFAM – Topology Optimization (TO).

DFAM is just a change of mindset - encouraging designers and engineers to consider the process used to manufacture a part when defining the topology (geometry). This is not the first time we have seen a change in design and engineer mindset though. I always like to draw a corollary to injection molding when talking about DFAM. When injection molding was introduced to industry, a similar shift in mindset had to occur. Traditional design techniques had to be adapted to take full advantage of the new process, and slowly new educational courses and software were put in pace to train the workforce for the shift.

When thinking about additive, the process itself can unlock new design possibilities which have never been thought of before, and maybe (potentially) which might not be conceivable to a human. Enter Topology Optimization and Generative Design - methods which use physical simulations to inform users which geometries would be best for a given set of performance criteria.

TO has been around for quite some time, but has entered more into mainstream with the large growth of additive, and the availability of generally inexpensive compute power. Its main advantage is that the geometries which are produced are generally "organic," and lend themselves to additive manufacturing - creating extremely lightweight designs which perform as desired.

Now let’s focus in on a specific 3D printing process – like extrusion (e.g. FFF or Bound Metal Deposition). In general, current commercial TO offerings assume the resulting geometry contains near isotropic material behavior. However, most (not all) break down once the resulting geometry does not contain near isotropic material properties – such as extrusion. Extrusion processes inherently create regions with orthotropic material behavior (see our previous posts on this topic here, here, here, and here), and TO cannot generally capture the behavior of the parts. There are 3 main reasons for this:

  1. Many TO rely on a fundamental assumption of isotropic materials (no directional dependence).

  2. Most (not all) TO do not consider the actual process in the optimization, they just ensure a part can be manufactured using “rules of thumb” (minimum thickness constraints, minimum hole size, maximum overhang angle, etc.).

  3. Most TO try to create the lightest design possible, but do not try to minimize manufacturing time (which are different).

For techniques such as SLS, DLMS, etc., the assumption of an isotropic material within the resulting geometry is generally accepted (although many will actually challenge this, and it depends on how rigorous you need to be), but for extrusion processes, this is not a safe assumption because the process imparts directional dependence into the geometry!

A Short Case Study for Explanation

Let’s look at a simple case study to explain why TO can break down for FFF. Consider a drone arm (picture below) which has been initially designed by an engineer using traditional techniques, and we’re interested in using Topology Optimization to lightweight the part (the lighter a drone is, the longer it can fly).

The Troubles with Topology Optimization for FFF

Figure 1. Original drone arm geometry.

Since TO relies on physics simulations (e.g. Finite Element Analysis), we’ll need to provide material properties. But what material properties do we use? The TO engine we have decided to use only accepts isotropic material properties, so we need to decide which set of properties we should be using for the TO. Assume we have properties for a printed specimen, and we intend to use the same print profile for the drone arm as the printed coupons. Let’s look at the mechanical properties for this specific material in the stiffest and softest directions.

The Troubles with Topology Optimization for FFF

We’re going to encounter a few major problems here:

  1. The as-printed material properties are orthotropic, but we’re constrained to isotropic material properties.

  2. The TO is assuming the part is solid… but we know in FFF that the part is not solid – there are walls, skins, and infill. All of these different types of regions can have very different stiffnesses and strengths. For example, the stiffness of a 20% infill region will be much less than the stiffness of a 70% infill region or a skin. This is some of the influence the process has on the performance of the part. We could 3D print the part as a solid, but the print times would be astronomical!

What can we do? In reality… nothing. We can make some engineering assumptions and move forward, but basically we are going to be manually iterating through this (defeats the purpose of TO, doesn’t it?). What we can do for this study is to bound the real solution by solving a TO with the stiffest and softest mechanical properties, which is a fun exercise. 
The TO results for the stiffest and softest mechanical properties are below. Remember, the TO is assuming the material is (1) isotropic, and (2) the part is solid!

The Troubles with Topology Optimization for FFF

Figure 2. Topology Optimization results for stiffest material direction (left) and softest material direction (right).

The actual solution for the TO, for an as-printed behavior assuming a solid 3D print, is somewhere between the two extremes shown above, but we do not know where (note the solution for the softest properties is the original geometry). We also do not know if either of these are optimal from a manufacturing time perspective - all we know is they are lightweight. There are likely better solutions by optimizing on the as-printed component, and trying to reduce manufacturing time.

Closing Remarks

Topology Optimization is an amazing tool which has potential to change the way we design for additive manufacturing, but we should be careful as to how to apply it, and understand its limitations.

Most commercial Topology Optimization offerings are not well suited for extrusion-based processes. The impacts of the slicing and 3D printing process on the performance of the part should not be disregarded.

Check out our next post, where we will explore an alternative to modifying the topology of the part, and instead optimize the process.

Subscribe to our newsletter