Topological optimization without making mistakes: when it’s worthwhile and what goes wrong

Topology optimization is the image that illustrates half of the additive manufacturing industry: a part that looks like a bone, full of nerves that no engineer would have drawn. It is also, by far, the most misused tool in this sector.

What it does is simple to explain. You define an available volume, specify where the loads will be applied and where the part will be attached, and indicate how much material you want to retain. The program then removes material from areas that don't contribute to rigidity until it reaches that target. The result is the shape that best utilizes the material for the specified load case.

That last sentence is what decides whether the project goes well or badly.

When is it worth it

Not always. Optimization costs engineering, simulation, and validation time, and that cost has to be recouped somewhere.

The six mistakes that ruin the result

1. Optimizing for a single load case. This is the classic and most expensive mistake. A real-world component experiences loads in service, during assembly, in transit, and on its worst day. A topology optimized for a single load case is rigid in that direction and surprisingly fragile in all others. You have to include all cases, including the worst ones.

2. Forgetting that the part needs to be manufactured. The algorithm doesn't understand what an overhang is. It returns shapes that would require supports in locations where a tool can't be inserted to remove them later. Almost all serious programs allow you to impose manufacturing constraints: minimum overhang angle, build direction, minimum thickness. These must be used from the beginning and not added later.

3. Forgetting that the part must be measured. An organic surface has no reference planes. If the part has functional dimensions, flat islands and cylinders must be preserved to serve as reference points for clamping and inspection. These are declared as untouchable areas before optimization.

4. Leaving channels that cannot be drained. The interior of an optimized part is often hollow. If these hollows do not connect to the outside, they remain filled with unmelted powder forever — dead weight, risk of contamination, and a certification problem as soon as someone asks. Drainage openings must be planned from the design stage.

5. Manufacture the mesh generated by the program. The raw output is a triangulated, rough mesh with no tangents. Manufacturing it as is results in a part with stress concentrators everywhere and a finish worse than desired. It must be rebuilt as a smooth surface—either by hand or using the program's own reconstruction tools—before manufacturing.

6. Do not simulate again. After rebuilding, the part is no longer the one the algorithm optimized. Validation is done on the final part, not on the proposed version. And for critical parts, testing is done afterward.

What almost no one talks about: the finish penalizes

An optimized part has a much larger surface area than the original, and a significant portion of that surface remains unfinished because it cannot be machined. This is important for parts subjected to cyclic loads: the roughness of a surface fresh from the machine reduces fatigue life compared to the same part after polishing, and this reduction is substantial.

The practical consequence: if the part is subject to fatigue, either a post-processing procedure is planned to reach the critical surfaces—and it must be verified that this can be achieved—or optimization is performed with a margin that assumes the rough finish. The information in post-processing and tolerances and finish applies here more than to any other part.

An alternative that often wins

Before optimizing, it's worth trying something more tedious and cheaper: removing material by hand in obvious places and consolidating the assembly into fewer parts. In a high percentage of cases, most of the weight savings come from the three or four obvious decisions an experienced engineer can make in an afternoon, and from bringing several parts together into one.

Topological optimization comes later, when the easy stuff has already been squeezed dry and weight still costs money.

In summary

It's an excellent tool, but it's not magic. It works when weight is a factor, when loads are well understood, and when the designer imposes manufacturing, measurement, and casting constraints from the outset. Otherwise, it produces parts that look spectacular in photos but disappointing on the test bench.

If you have a candidate part and want to know what it would cost to manufacture before investing in optimizing it, the quote calculator gives you the cost and time with the geometry you already have, and the design rules tell you which range to stay within.