0:00 Hello everybody. Thank you very much Dan for for the organization once again. I’m Rishik Batroa. I’m I’m here to to talk about how can we bring generative design to the concurrent engineering. We had some great talks yesterday and today being a relatively new kid on the on the street on the on the block could be intimidating but it’s inspiring to see that this whole shift that we can see from last year CDFM and this year big challenges are coming up and we have to get ready on it.

0:37 At cognitive design systems we are very simple and pragmatic people. We think to believe so and we want to solve a simple issue. We want to make parts which are performing, manufacturable at least cost and sustainable. And for that we looked into how engineers are creating this kind of geometries paths from end to end. And we end up finding that there used to be a concept in the the the European space field called concurrent engineering where we create a small task for task force with design engineer, simulation engineer, manufacturing engineer and the idea is to bring out the design which is or the path as fast as possible.

1:36 At Cognitive Design Systems. Just a little bit introduction, we are a small bunch of people in south of France, Tulus. Beautiful place if you have ever visited. We work a lot with European and Japanese customers and yeah, we are about 4 and a half year old. Our vision is quite ambitious. We want to transform engineering into cognitive creation. When I say cognitive creation, it also involves on how to make a process which is fluid which is very easy to to use engineering making it more easier but also not doing again and again same processes making the engineering process that has been already done in the past reuse it in a way that it can be leveraged in the best way possible.

2:28 And for that we want to break the silos. I see that a lot of companies presented just before us have also that ambition and we are all in that all in for that together. We want to accelerate the sustainable design. That’s a that’s a maybe very European thing to say but that’s a very important thing and I I truly believe that it has to be focus which is little bit shifting right now but it has to be focus of engineers for of tomorrow.

2:58 And we want to automate the innovation in the right way by using the the AI but in the right way to make it useful. For that we are relying on our technologic enablers. So we developed our own implicit modeling kernel which allows us enough flexibility to to to modify the geometry optimize the geometry in the best way possible. We combined it with datadriven design decision algorithms. So we do manufacturability analysis, we do cost analysis, we do carbon footprint analysis and we couple it with the with our optimization solvers.

3:39 So we can improve the the whole design much better. And at the end we have our UI and UX which is very user friendly where we we are able to re create a reusable engineering workflows and we can iterate faster and by doing that the whole idea is to to create a accelerated scalable intelligent engineering. That was that was not the real idea of creating any any assumptions on Illuminati but it looked like that a little bit.

So how we are bringing concurrent engineering to to our workflows. So for the concurrent engineering we start with engineers first. So engineers who want to start from some specifications could be functional regions could be what kind of material what kind of cost we want to achieve and we end up creating the best pass possible. On top of that that knowhow that engineer has created could be used by the engineering team.

4:53 It could be different engineers from manufacturing side from from simulation side and they can collaborate on the same product on the same project in the best way possible and at the end this whole workflow has to be captured as a engineering knowhow by the organization. So they can reuse those those engineering intelligence in other for other purposes for same family of paths for example and in order to accelerate the whole product development cycle.

5:33 So to do that first of all we started creating our own algorithms. I will talk a little bit more about it. Of course, topology optimization has been talked quite a lot, but we believe that topology optimization serves pretty good in certain specific fields but not for all use cases. So we we started creating new use cases, specific optimization algorithms. So I will present it a little bit in later how we do that.

6:02 Design exploration that is also something that has been focused today from from Alter from from Mentop from from Sinera also. That is also very very key but our approach is to make sure that each design that has been generated is manufacturable. We know the cost. We know the carbon footprint of it and in that way we can take the best decision while selecting the the the design and at the end of course our core pillar is the design for manufacturing.

6:37 Assessing the manufacturability early on in the design cycle is very important to make sure that we don’t bring out some artworks and nothing against the artwork but if the part is not manufacturable what’s the point of the design. So making sure of that is our key priority. And by doing this we create our engineering database. So here is the first optimization algorithm topology weaving. It’s a name that we gave it.

7:10 I’m I’m open to discuss about how can we name it better. It’s simple. We wanted to connect functional regions between them in a confined region like topology optimization. But we wanted to make sure that it’s completely parametric. So we can reuse we can modify geometries locally or we can modify those parameters to to refine or fine-tune the algorithm. So what we did is that we connected it with with a manufacturing analysis solver.

7:43 We connected it with a stress solver and we we assess that okay what’s the best way to to connect those functional regions. So we have simple connections but then we we assess that okay the stress is there so let’s modify it and we we modify the geometry accordingly and it goes through there and it all splines it could be splines it could be surfaces so you can come and modify also those geometries much better and you have full flexibility.

8:15 A small example just under the same path that we we designed with with different design intents. One to in additive manufacturing, another one in diecasting, another one in machining and you can see that the results have been come out of the same algorithm but as it has been driven by the the manufacturing insights it the design is completely different. I don’t know if it’s visible. Yeah, there you go.

8:44 Another algorithm is generative enclosure. Something that was missing in topology optimization was that we can’t create all the time watertight geometries and those watertight geometries because inherently topology optimization will create some some generative design which has lot of holes looks beautiful but if you have to put some oil in it you have to put some some other stuff it’s not possible. So this algorithm was very simple.

9:18 We bring in the functional regions as it is and it will automatically create a surface which connects all the functional regions with the same thickness and this this allows to to create very fast a first geometry that you can iterate with simulationdriven design. So you can modify the thickness, you can add ribs, you can add stiffening elements to make it more performing on same time keep the manufacturability in check.

9:54 All these different algorithms of course we have also topology optimization but I’m not presenting it because it’s already already has been discussed quite a lot can we can generate lots of different concepts and this concept has to be analyzed and validated and for that we have our design explorer where you can actually map out all the concepts that has been generated you can compare cost carbon footprint you can compare different performance levels you and compare also various different kind of KPIs and select the best part which is suitable for your specific requirements at from the very beginning and this could be done completely automatically and could be reused.

10:39 So let’s talk about the yeah before that yeah the manufacturing driven design. So manufacturing during design we select first of all the process that we want to manufacture with could be diecasting, molding, machining, additive manufacturing, other processes. We assess the the feasibility of it. So we have our algorithms which allows you to assess where the risk with the manufacturing is and then we have some other algorithms which automatically modifies the geometry to make it more manufacturable.

11:12 Very simple but effective and that’s that that could be done very fast with full control from the user and all these workflows has to be done in a way that that it’s reusable so we can create reusable workflow not sure we can see it very well but yeah you can take one part one design space you can optimize it you used it for project A now this workflow has been created It could be more or less complex.

11:42 You can give it to your colleague who has some different requirements. He can also bring in another part which is completely different but he can reselect the the functional regions and apply the loads and regenerate the the model very fast. And this reusability of sharing those workflows with or without the 3D models inside is is very important and that’s what we see at our customer level that they are reusing it fantastically.

12:15 First first example here with from the aerospace industry is the the aircraft door bracket from Pes. It’s a supplier of DASO aviation of course France, France, Tulus we we turn around those those people very quickly. Idea was that okay we have a concept here we want to create the new door for the the Falcon 6X that is coming what is the right way to manufacture this part so let’s explore design for manufacturing so for five access machining or for additive manufacturing and by doing that we we we can know that okay the whole exploration time which was reduced dramatically We we were always aware of the cost and the manufacturability of those those components which was a a real advantage here and we were able to to bring out the the most optimized part which was 5x machining machining and not the edit manufacturing because of course the cost was not same and it was displayed at at Bourj last month.

13:21 Another example from the new space industry from T4i. It’s a it’s a satellite propulsion manufacturer. They wanted to to create this fuel tank pressurized satellite tank fuel tank. And they wanted to explore the possibilities of what they can do. 150 different lattice variants that they they used internally were analyzed in in half day or a little bit more than a more than a day and at the end we we found the the right one which has the best performing criterias which was relatively easy to to manufacture and which was also what they were looking for the the the stiffness level was the best.

14:19 All this now we are talking about our research part which is coming from by the end of this year. My CTO is here he he can confirm it. We are working on on the m multi-disiplinary optimization that makes a lot of sense because we want to automate the optimization for cost for carbon footprint and it has to be done automatically also we the user can always have the parameters that he can fine-tune but we also want to accelerate that by using surrogate modeling and by using our homegrown own optimization algorithms.

15:04 And the last point that I want to discuss is is also something that has been a bottleneck for the industry which is conversion of because all this implicit models and and mesh models they’re great but we had to bring it back to the CAD because 99% of the people and I’m sure people from Daso system here can confirm that if we can’t bring it back to the CAD we can’t bring it back to the PLM in the right way and that could be a problem.

15:36 So we are spending a lot of time doing mesh to CAD conversion automatically of course with the minimal user intervention. We use the mesh sorry the the recognition of of different features. So primitive features cylinders flat surfaces taurus etc etc. We recognize it on the mesh and then we recreate the the BREP geometries. So we can bring it back. We can reuse the parametric 3D model. Of course if the the surfaces are are curved or anything that remains in the in in the in the in in the patches like we can see here or but the the functional regions becomes much more cleaner.

This is it’s a kind of a little bit complicated to to explain on this part but yeah if you have a very simple parts that can be perfectly converted to to to the CAD the beta will be released end of this year and yeah if you are interested in it you can come to talk to me and my my colleagueuh at any time. Thank you very much for your time. To learn more about the CDFAM computational design symposium series, to see the archives of previous presentations, and to learn about future events, visit CDFAM.com.