Evolutionary mechanical design

[tok-tokkie]

I would be interested in informed discussion of these three components.

Pictures come from a New Atlas article: https://newatlas.com/automotive/spyros- ... 4-92349157

A Greek 3000 HP (not a typo) "hypercar". Has a billet 4 litre V10 engine.
3D printed components carbon fibre for the con-rod & brake caliper. Ceramic for the piston.
From the article:

Generative design, as we've explored before, is a process in which a component's fixed points, stress loadings, materials and desired performance characteristics are placed into a CAD model, and then sent to a cloud processing service where they are randomly mutated over thousands, or even millions of generations in a process designed to emulate natural selection.

The better-performing designs are allowed to multiply and evolve while the worse ones die off, and eventually, complex but effective shapes emerge that could never have been designed by hand. They typically offer extreme strength, incredible lightweighting, and minimal use of materials in a manufacturing process that can only be achieved with 3D printing due to their strange, organic-looking forms.

So it is like evolution where natural selection selects the better variants and drops the lesser variants.

Steel pistons are used in F1. Does a ceramic piston not have the advantage of not absorbing heat to the same extent so running cooler. But, if it does get hot it would then be more difficult to cool. Not the strength issue but premature ignition - or is that not the issue with hot pistons?

The organic vegetative forms of the con rod & brake caliper surprise and intrigue me as a mechanical engineer. Not the forms I would have intutively chosen but seeing them I am impressed by their structural virtue.

[DiogoBrand]

I'm by no means an expert. I think this kind of AI design looks and sounds cool, I'm just curious to know if it actually works as well as advertised.

[e36jon]

My take is that the two designs presented here are more about visuals than actual mechanical optimization. No disrespect intended.

The optimization tools available to engineers can and do work, but like all software tools they need to be carefully managed by the user. Finite element analysis (FEA) used to be something only an engineer with a specialized PHD could do, and now it's built into almost every CAD package. Now 'anyone' can do FEA, and most of the decisions we used to worry about are buried behind an easy to use user interface. Anyway, long story short, the results are still only as good as the person doing the work.

Here's an example that I think used these design tools, the 'Radi-Cal' brake caliper by AP: https://apracing.com/race-car/brake-cal ... cal-cp5780. This site has a great collection of articles from the initial release: https://www.essexparts.com/news-blog/th ... ical-story



Looking at the F1 paddock I think the current Red Bull front suspension rockers were also designed using these tools.

Four or five years ago Force India made a radical change to their front uprights. Everything before had been pretty rational, with straight lines and flat planes, but their new design was very organic with no straight lines anywhere. Now all of the teams have similar designs. I still wonder what drove the change; new manufacturing tech, new design tech, or all of the above?

Regards,

Jon

[Tim.Wright]

These parts definately look completely different to what typically comes out of a non-AI driven topological optimisation. I'm interested to see if it actually works.

The output of a topological optimisation is only as good as the load and constraint inputs that you give it as a target. I've seen so many of these where it's obvious the boundary conditions were wrong. I don't know enough about ICE part design to comment on these parts though.

[Tim.Wright]

Four or five years ago Force India made a radical change to their front uprights. Everything before had been pretty rational, with straight lines and flat planes, but their new design was very organic with no straight lines anywhere. Now all of the teams have similar designs. I still wonder what drove the change; new manufacturing tech, new design tech, or all of the above?

I'd say it was the implementation of topological optimisation analysis tools. I did such a project on a motorsport upright about 5 years ago and the result was similar in form. Result was a big weight advantage ove the old design but was made using conventional 5-axis CNC techniques.

[Greg Locock]

For anyone who has FEA but doesn't have an optimisation tool, a few iterations of killing the blue elements and bumping up the red ones on a tetra mesh is not exactly unrelated to the methodology used. As such I'm not really convinced that the conrod is optimal ( I thought hollow tubular rods had advantages). It looks to have had the 'Giger' aka 'clickbait' option enabled.

[e36jon]

It looks to have had the 'Giger' aka 'clickbait' option enabled.

So, a 'Giger Counter' would read '2' then?

I probably over-attribute the use of 'optimization' tools because I see so many parts that I just can't imagine having been created any other way. No doubt a lot of those parts were ball-parked by people with a lot of experience (I'm looking at you Tim Wright) and then tweaked via FEA just as you suggest.

The parts shown have some questionable attributes that make them unlikely to be actual improvements on the current state of the art (Which has been steadily improved for 100+ years.). The hollow conrod (Honda) being an example, but even there it was only ideal for the incredibly high RPM's of the time.

[Greg Locock]

Back in the day there was much enthusiasm for optimising the stiffening webs on crankcases and the like for NVh and stiffness. So the boys had a lot of fun talking to the casting people about feasibility of various features, and then wasted a few weeks modelling them. Then a cynic (not me) meshed up a block that had the same volume of metal in it, but smeared out rather than in webs. Each had slight advantages in various parameters, but overall the uniform one was better. However, many of those ribs and webs are there for casting reasons.

[Zynerji]

When these Tensor processors get out more, this is going to become ubiquitous.

They are going to chew through these exact work types much quicker, so I would expect to see them more in the near future in F1.

[Jolle]

Looking at the pictures of the piston and brake calliper, looks like it’s more art then designed for strength. These curves will flex to some degree.

This almost looks like it be of those amazing engineering project that are actually fake investment schemes.

[Zynerji]

Looking at the pictures of the piston and brake calliper, looks like it’s more art then designed for strength. These curves will flex to some degree.

This almost looks like it be of those amazing engineering project that are actually fake investment schemes.

It looks like minimalist load-pathing to me. Organic structures like slime mold have been used to generate structures similar to this in the past. Maybe not in a 3d, point cloud system like this, but 2d for sure (Tokoyo Subway layout).

https://www.livescience.com/8035-slime- ... works.html

Why wouldn't this translate to 3d organic (generative) design?

[Greg Locock]

Yeah but Jolle's right, the main loads are axial, why aren't the fibres aligned directly axially?

[Rodak]

Cool looking con rod, but how does one actually make it? Fiber orientation is critical for strength and there is no way a 3-D printer can make a working con rod; looks great but tooling/layup would be impossible. I am an ex carbon fiber aircraft tool designer, I was involved with the C-17 and the B-2 bomber projects as well as civilian aircraft. Good luck with that.

[Andres125sx]

Yeah but Jolle's right, the main loads are axial, why aren't the fibres aligned directly axially?

Maybe to reduce flexion? With fibres that thin, if all of them were alligned axially flexion/torsion would probably be a problem

[Andres125sx]

Cool looking con rod, but how does one actually make it? Fiber orientation is critical for strength and there is no way a 3-D printer can make a working con rod; looks great but tooling/layup would be impossible. I am an ex carbon fiber aircraft tool designer, I was involved with the C-17 and the B-2 bomber projects as well as civilian aircraft. Good luck with that.

But that´s exactly what they´re claiming, a 3D printed con rod

They typically offer extreme strength, incredible lightweighting, and minimal use of materials in a manufacturing process that can only be achieved with 3D printing due to their strange, organic-looking forms.