Follow Along: Topology Optimization

After establishing the optimization objectives and constraints in our last lesson, the topology optimization has already run. Let’s take a closer look at this block, all its inputs, and the information that we see in the results. For now, we’ll collapse our optimization objective and our constraints and focus only on the Topology Optimization block.

The input for Max iterations is pre-populated with a value of 200. Because topology optimization is iterative, we can set a limit for the number of iterations attempted for more complex optimizations that may take longer to run. This should help users to reduce computation time, with the trade-off of accuracy.

Another way to reduce computation time, assuming the same trade-off, is by modifying this minimum objective change. Once the top op iterations converge to within this number, the process is considered complete. Therefore, the higher the minimum objective change, the faster the optimization will converge, and the lower this input, the longer it will take to run. Minimum density change is another convergence threshold, one that compares the optimized density between iterations. Again, we’ll see a slight trade-off between accuracy and computation time. The boundary penalty affects density values at the edges of the design space. An input of one will lower the density at the edges, while an input of zero increases the density. Therefore, lesser inputs will remove more inner material from the part and greater inputs will remove more material at the edges.

Save increment controls data storage and therefore affects file size. An input of zero will save only the first and last iterations. An input of one will produce the largest file size and save every iteration, and any input larger than one will be the increment by which iterations are saved. For example, an input of two would save every other iteration of the topology optimization. Filter size can be used to indirectly control the minimum feature size and remove the jagged checkerboard artifacts that we see in top op. If left blank, this will auto-generate based on the model size. Initial density can be any value between zero and one to give an initial guess at how the result will look before the process begins. If left blank, this will be set to 0.5.

I’ll go ahead and create a variable out of our deposit optimization, and I’ll pull in a Translate block so we can view our results side by side with the original part. I’ll right click, make a variable called top op, and I’ll pull in our Translate block to translate our initial design space.

Viewing these two parts beside one another, we see that a lot of this material from our original part has been removed during our top op process. Since topology optimization works by assigning a top op density value between 0 and one to each mesh element, we can use this topology optimization window to get a better understanding of how the calculated densities have affected our part and how the output contributes to our objective. The iso contour option shows the single surface with elements meeting the nominal threshold, which can be modified using this slider. The threshold equals zero, our initial part can be seen, and as it moves toward one, you see more and more removal of elements with lower top op densities. The iteration slider shows the evolution of the top op through each iteration. The thresholded elements option offers a similar visual representation but gives a more granular view of individual mesh elements and their top op densities. This can help us to visualize how and where material with different top op densities is removed throughout the optimization.

Opening up the settings in the results window, we can control the color palette of our elements, whether the gradient is continuous or banded, and the preferred number of color bands. We’ve now completed our topology optimization. Next, we’ll need to perform some post-processing of our results to generate useful geometric information in nTop.