The Troubles With Topology Optimization for FFF - Part 2

In our previous post, we discussed why topology optimization (TO) is currently unsuited for fused filament fabrication (FFF) due to the inability of TO to understand how the manufacturing process affects the structural response of a part. We started investigating an example part and provided a thought experiment to demonstrate the dangers of using TO for FFF.

In this post, we will look at another method to optimize a component for structural requirements by modifying the process settings instead of the topology (geometry), which inherently captures the influence of the process on the structural behavior of the part while minimizing 3D print time.

Let’s again consider the same drone arm. Our objective is to minimize weight, material usage, and minimize manufacturing time while ensuring the part is structurally sound. In our previous discussion we used a TO engine to optimize the component assuming an isotropic material – we used two different sets of material properties to represent the stiffest direction of the as-printed material and the softest direction of the as-printed material which theoretically "bounds" the real solution. The results for the TO runs are shown below (note the TO result for the softest direction is the original geometry).

The Troubles With Topology

Figure 1. Topology optimization results for stiffest material direction (left) and softest material direction (right).

Our first objective will be to validate that the TO runs did indeed satisfy the requirements by simulating the as-printed component. Our requirements for the part are a maximum displacement of 3mm and a factor of safety of 2 (meaning the drone arm can sustain twice the applied load prior to yielding). The limiting requirement for the part is the displacement constraint, which means this is a stiffness driven design, and the factor of safety will inherently be met – so we will focus on the displacement requirement in this discussion. We can easily validate the as-printed behavior of the parts by using Teton’s SmartSlice environment within Cura.

The results of the simulation are presented in the table below. Remember, the parts need to be 3D printed as solid (100% infill) to align them as closely as possible to the TO. We also must align the skins (layers) with the direction of the material properties we used. So, the layers will be aligned with the length of the part for the stiffest direction, and perpendicular to the length of the part for the softest direction (see pictures under the table).

Table 1. As-printed results for the topology optimization runs for the stiffest material direction and softest material direction.

The Troubles With Topology

The Troubles With Topology

Figure 2. Layer orientations (line directions) for the stiffest material direction (bottom) and softest material direction (top).

Looking at the results in the table, we can draw some interesting conclusions:

  1. The TO result using the stiffest material properties does not satisfy the displacement constraint when we include the anisotropy of the process.

  2. The TO result using the softest material properties does satisfy the displacement constraint, but the resulting geometry is the original geometry (the TO did not remove any material) which means the print time is astronomical because it’s 3D printed as solid.

  3. As discussed in the previous blog post, the “optimal” result for a solid body is somewhere between the two results, but unfortunately, we do not know where…

What if we approached the problem by means of modifying the process itself, instead of the topology? For this, we can again leverage Teton’s SmartSlice environment within Cura to optimize the slice parameters for the structural requirements. This will allow us to explore a space by automatically varying wall counts, top and bottom layer counts and infill density at the global and local level. The objective would be to minimize print time while satisfying the structural requirements, and the result is included in the table and pictures below (as well as a comparison to the TO results).

Table 2. As-printed results for the topology optimization runs for the stiffest material direction and softest material direction and the optimized slice parameters from SmartSlice.

The Troubles With Topology

The Troubles With Topology

Figure 3. Optimized SmartSlice results showing global geometry and local reinforcement using modifier meshes.

The Troubles With Topology

Figure 4. Layer view of the optimized slice parameters from SmartSlice. Top is showing the slice view, and the bottom is shown with colors to differentiate between walls and infill.

Again, we can draw some conclusions:

  1. By optimizing the slice parameters, we can simulate the “as-printed” behavior of the part and take into account the impact of the process on the performance of the part.

  2. SmartSlice was able to find a solution to meet the requirements by optimizing global parameters and locally reinforcing the structure where needed, which minimizes material use and print time.

  3. The print time for the optimized part from SmartSlice is only 28 hours, which saves 15 hours on print time when compared to the TO solution which meets the requirements.

  4. There might be a TO answer between the two TO results which actually 3D prints faster than the SmartSlice optimized result, but at this point we would have to manually iterate on the topology to meet the requirements.

Closing Remarks

Topology optimization is extremely powerful and is definitely something which needs to continue evolving. Including the results of the process in TO would be a huge leap forward for DfAM, and we should continue exploring how to properly simulate the as-printed behavior of 3D printed parts.

Teton’s SmartSlice can validate and optimize the process for end-use requirements in a fraction of the time it takes to 3D print and validate. Optimizing the slice parameters to minimize print time while also meeting structural requirements allows us to have confidence that the 3D printed component will perform as desired, while reducing the overhead of using the machine – allowing an increase of throughput.

We look forward to eventually integrating the SmartSlice technology within topology optimization to continue evolving the state of the art for DfAM to finally marry design and manufacturing.

Subscribe to our newsletter