Dr. Thomas Reiher, cofounder of AMendate, who is now director of Hexagon’s generative design division. (Picture courtesy of Hexagon.)
We catch up with Dr. Thomas Reiher, cofounder of AMendate, who is now director of Hexagon’s generative design division, in his office in Paderborn, Germany, to find out more about what MSC can do in terms of generative design—perhaps the most promising but overhyped design technology of the future.
MSC Software, the company best known for the most successful version of NASTRAN and at its peak, and the last generation’s leading simulation vendor with an unassailable leadership in the aerospace industry, has been undergoing decades of changes in product lines, new management and ownership.
In 2014, MSC seemed to enter a competition with itself by launching an alternative to NASTRAN, the Apex CAE product line.
In 2017, the company was acquired by Hexagon AB, a Swedish firm best known for metrology and also known for its aggressive pace of acquisitions. Hexagon bought Intergraph in 2010 and would go on to buy Bricsys in 2018. And bringing us to our story today, in June 2019 the company acquired AMendate, a German generative design startup.
Hexagon is all over the world. Paolo Guglielmini, our former CEO, lived in London. I think we do not have this one certain headquarters for MSC anymore. We have different subsidiaries, centers of excellence, which have different headquarters. For the additive manufacturing world, the senior technologists live in Hamburg and in Belgium.
How did the acquisition of AMendate come about and your role?
MSC did not have a generative design product for a long time. MSC needed this tool. They knew they needed something new in this area and especially something dedicated to additive manufacturing. They wanted to make sure they were providing it in addition to their competitors. I'm responsible for generative design now and AMendate, the company I cofounded, is the back-end intelligence behind MSC Apex Generative Design.
I was the founder and CEO with the idea for all of this, but I'm just an engineer and not in software development, so my colleague Steffan Vogelsang did all the coding.
Can you tell me what led to the naming of AMendate?
It comes from a Latin word to improve, optimize with A-M for additive manufacturing.
Did MSC have no generative design before AMendate?
No. It has a little bit of optimization inside NASTRAN. Apex is a CAE platform. Before, users had the Apex modeler and structures for standard CAE. Now, with AMendate, it has generative design as well.
AMendate was a stand-alone product?
Yes, it was stand-alone when we were a startup. It originated from my PhD thesis on the optimization of design for additive manufacturing, and together with my colleagues at AMendate, we developed the code.
The program was command-line driven. We believed a good generative design needed to be part of an automated engineered process with a close connection to the other simulation tools and with manufacturing—all part of a perfect design.
Now AMendate is part of Apex?
It is totally absorbed totally in Apex. You need to set up your model and put loads and supply the shapes that will be optimized using the Apex GUI, then let generative design run.
I think the NASTRAN engineer will get ideas when working with our solution because we’re doing a lot of things differently than the standard finite element simulation. They will be able to work as a CAD engineer rather than as a CAE engineer. We’re able to start with some rough geometry and turn on generative design tools to find the low stresses and other quantifiable goals. Maximum von Mises stress is an optimization goal but other are complexity.
Take thin struts, for example. You can specify one, three, four, six … struts. You can determine or drive the different designs by this value. The FEM engineer may think they have the right design and they will run a simulation, then [run] another one tomorrow if it is not the right design. We say instead run generative design today with your settings and let the program do it all in one day.
Apex is very strong in meshing, but we don’t need any meshing tools for our generative design. If you open MSC Apex Generative Design, you don’t see any meshing tools. Once you have a design candidate from generative design which you expect is perfect to use, you would go to NASTRAN and run the verification. You will be switching to the Apex GUI, which is more or less the same and familiar. You know where to find the important functions. There you have different functions available, like the meshing functions.
We have broadened our scope to include CAD engineers and designers who want to be creative, use the software to get new ideas. This is why we think that we can get new people into our universe.
Some of the other things we do are different from other generative design vendors. Each design that comes out of the software is feasible, geometrically and mechanically. We are going a different way based on what we know from years of the topology optimization. We have always given one specific geometry, which in the very beginning is very coarse and bulky, but over the iterations, more and more detailed design fills out, and it’s always correct. You can use each and every iteration always.
The optimization works by starting with a simple shape, and given loads and restraints, removes material where it’s low stress and leaves it where it’s high stress. Is that correct?
Yes, it’s basically correct, but it can even create new material when needed. For example, you can start with the very thin part, and if there is too little material, it will add material. This is an important point that makes us different. You have to start somewhere—in the middle, whatever … the part’s material can move where it will do the most good. We can very easily determine which material we want in the end.
We always say that the software behaves like an engineer thinks.
We see a lot of results from optimization that are composed of thin structures. They’re stringy structures and it’s apparent that they’re only good in tension and may actually fail in compression. Can your software handle compression and buckling?
We are working on that. Our technology uses von Mises stress. We are going to change that to other failure criteria, like safety factor sometime in the beginning of next year. Then we can go for material that can handle compression, like ceramics. We are testing this.
Buckling is currently not considered as a failure mechanism in most topology optimization, correct?
Most generative design codes do not consider buckling failure. Standard optimization works to create linear aesthetic parts without considering buckling. It is the same for us.
Many generative designed parts appear to be the result of a specific to one load case. Has any software, including yours, been able to handle combinations of load cases?
Yes. I think this is technically possible in most of the generative design tools, as well as in ours. You always can set up different load cases at once. But arbitrary loads predict some crazy things [that] might happen, imagining loads.… I don’t know if this is possible with any software.
We still need human engineers?
Yes, that’s still something a human being needs to do.
You cannot usually think about lattices, as lattices are mostly not the best structural design. They might work in sandwich panels, but if you think about real structural parts, like a bracket, it’s better to have several really clearly defined load paths.
As we’re talking, I’m watching a building under construction out your window and the scaffolding is composed of long thin pipes and joints. It may not be an optimum shape but is a usable shape and easily manufactured. However, generative design does not take advantage of standard shapes like pipes, or other stock shapes. Another example, it can’t duplicate a bike frame with tubular members. The common diamond shape bicycle frame created by humans in the 19th century still has not been improved upon. The Golden Gate Bridge has not been improved upon [see Generative Design Takes on the Golden Gate]. Do you have any goals for Apex so generative could increase the versatility of generative design so it could improve on human designs?
At the moment, we’re really focusing on medical device manufacturing where this is not an issue. That’s a perfect match with generative design creating beautiful bionic shapes—shapes you could not imagine, shapes that you can manufacture only with additive manufacturing. It is really fitting perfectly together.
The Golden Gate Bridge will always be an issue. All the tools we have cannot have the engineering genius that thought of two spanning cables and then hanging the bridge on smaller cables along the span.
Would MSC take up our bicycle challenge?
It’s not that easy.
I know. I suppose it makes you realize the value of human beings over algorithms. It seems like the bike frame is stuck in time. We got the present diamond shape by welding round tubes together. Surely, shape optimization by computer can do better. Can’t a crude design like that be improved upon?
It’s really hard to optimize that. A lot of clever engineers have worked on it.
Yes, we have seen many clever optimizations, but most of them are for mountain bike frames. Mountain bike frames have more linkages, suspension, more varying loads … so there is more to optimize.
We did some optimization on a bicycle crankshaft in my university days. But the frame … with the simple round tube—it is the perfect shape from a mechanical point of view. A round shape is best for torsion. How can you improve on that?
What about manufacturability? Most optimized shapes we see can only be produced using 3D printing.
Yes. Marketing would say any shape is manufacturable. An engineer would say not in all cases. We are taking this into account. In addition to 3D printing, most of our designs are also able to be manufactured as investment castings. I don’t think that producing generative designed parts for milling is possible because milling has too many constraints.
What synergy will generative design have with other MSC software?
We can get all our loads from MSC Adams, our multi-body dynamic simulation software. It’s quite cool that you can have all the loads from all different directions that you can apply to the structure.
With Adams, you generate load cases and apply them at the very beginning of a design, like for the suspension systems of cars, and use the loads in generative design, for both design and manufacturing verification.
Thomas, thank you very much for joining us and letting us know about Apex Generative Design.