GE Bracket Part 2
Combining the strengths of topology optimization and high precision CAD engineering, we get the best of both worlds and directly arrive at optimized manufacturing ready designs. In this follow-up post to our GE Bracket design post, we try to answer some of the questions that ticked in.
First of all, thanks for the great feedback on our first blog post on high resolution topology optimization exemplified through the GE bracket challenge. A lot of people asked questions regarding the "manufacturing ready" design - what do we really mean? Several people also asked how we import the result back into CAD. I'll try to answer these questions by explaining our workflow:
In topology optimization you specify the design domain. This is the volume where the algorithm is free to place material. This volume is prepared in CAD.
You also specify solid volumes. These are volumes that always remain solid - the algorithm take them into account when performing the FE analysis, but it doesn't change them. This is where you define the interface to other parts in your assembly, e.g. threaded holes for mounting, plane surfaces etc. In these volumes you have the precision you normally work with in CAD. See example volumes below.
Finally, you specify the load cases (of which there may be multiple, especially, when you want include perturbations of your load case - a topic for a future blog post). This is when you press "start" and - depending on the problem size and the available computing resources - either grab a cup of coffee, go to bed or take a vacation. The GE bracket took 12,000 CPU hours - that's a long time on your personal computer, but manageable in a cluster setup with e.g. 500 cores. Roughly speaking, a CPU hour costs 0.10 euro in a cluster setup. So it's not free to perform high resolution topology optimization, but neither is a time consuming manual redesign by a skilled engineer.
When the optimization has finished, we perform a Boolean union, to join the optimized result with the solid volume: in this way you have e.g. 0.1 mm resolution on the optimized result, but keep the CAD resolution on the interface parts which requires machining, threading etc. Ideally, we don't bring the optimized result into CAD to make changes to it - it is already optimal and you shouldn't change it. We only need to plan the machining operations, which were already defined in the CAD of the solid volumes. However, we recognize that many companies require a CAD model for documentation or assembly modeling, so we also offer this functionality.
Some people have expressed a desire to e.g. subsequently move a threaded hole a few mm due to changes in the assembly. This is a valid request, but you really shouldn't do this as a post-processing step, since the optimized design is no longer optimal! Instead you go back to your design volume and solid volume definitions, make the changes and re-run the optimization. Luckily, it typically won't take 12,000 CPU hours this time, since you have only made a small change and your previous result is probably very close to convergence.
With the combined solid and optimized volumes, we are now ready to 3D print the part with the same resolution as the printer and post-process the part (machining, threading etc.) with the resolution of standard CAD. This is why we state, that the result is manufacturing ready. This should save you a lot of time and lower the entry barrier to topology optimization significantly.
Written by Klaus Loft Højbjerre ·