Transcript
From YouTube’s automatic captions, lightly cleaned; expect some errors. Each timestamp opens the video at that moment.
Read the full transcript · 3,927 words
0:03 Hello once again, sorry about the trouble a little bit earlier. I’m Rhushik Matroja, I am one of the co-founder and CEO of Cognitive Design Systems, based in south of France, Toulouse, to be specific, in the in the heart of the Aerospace Valley of of Europe. We are three years old startup, and we are, we have big dreams doing some crazy stuff, and let’s, let’s show you, show you some of those those things.
0:32 So first I want to start with the bottleneck of add manufacturing industries. So these are some some charts that I found online from materialized, saying that what is the the main concern with the with the with addit manufacturing, maybe many of you have seen it on LinkedIn, and we can clearly see that the the design for a manufacturing and data preparation are kind of as important as the the machine cost.
1:01 It looks like this, that was some other website showing that what are the 10 biggest challenges of of add manufacturing, and that’s where we we we saw that, okay, of course there are concerns with the the slow production rate from machine side, material development, manual postprocessing, but then there is also a lot of things about design. We need to to get the design right, right, get resign ready for the additive manufacturing processes. We also need to educate and the the workforce, also engineers and and and technicians who are familiar with traditional manufacturing processes, we need to to to bring the the the education on on additive side. Same with the developing a business case, and it it looks pretty obvious that from our perspective that additive is always a good good good case, but it’s not, we can’t do everything, it’s not that cheap either, so we need to find the right business cases. So, and there are many other things.
2:10 So we at CDS are focusing on six of those things, it’s, it’s a big task, but we want to make sure that all these challenges are reflected well in our our software suit, and that’s what we are bringing to the to the market. And one thing is, is the what are the biggest industrial challenges? That I can summarize in three questions that maybe companies are asking themselves. Is how can added manufacturing benefit my business, finding the identifying the opportunities, and what do we mean by opportunities, it’s a cost saving, lead time saving, carbon footprint reduction, weight optimization, etc.
2:56 Then there is also, once you have already a a large department of innovation department on addit manufacturing in large organizations, then the question also, how can I, how can I say, the store the knowhow from the the experts and redistribute it to the to the other people around the organization. That’s also a challenge that we believe is true, and that’s what we have seen it in in automotive and and aerospace industry that we work with.
3:25 And the third one is how can, how to think AM from the beginning while consuming a product, and and that’s where we need innovative workflows, where we can actually convert those opportunities into into into something more tangible, and make something happen out of it.
3:48 So, in order to do that, we started by creating a software which does the, the, which answers to the question: where can AM benefit my business, by identif in the opportunity for cost, lead time, carbon footprint. That’s once we have identified those opportunities, we, not all the parts that we can see in the in the industry are already ready to be manufactured in additive, it has to be sometimes adapted for the this purpose, and I’m sure there, there will be some other talks later on, we will be talking about it.
4:20 The second one is, once I have that know how on additive, I can, how can I propag it, for that we created a software called Cognitive Additive, where you can actually store your your knoow and redistribute it to the to the to the others who are maybe not as familiar as you in addage manufacturing. So that’s also a key point for us.
4:43 And the last one is, is something that we are going to present for the first time today, is Cognitive Design. It’s our computational design software, where we can rethink on how we are conceiving, but also adapt the interface to the expert, expert label on that, that is in front of of the software.
And in order to do that, we are creating industrial knowhow asset manager, where we can able to collect all those those those insights and redistributing in in through these different softwares. So we can design better, we can produce better, and we can make something out of additive, and and find the right business cases.
So it’s, it’s pretty simple, we, we collect those data, and we, we have this database, and then we redistribute it to the to the to the various designers, AM operators, technicians, so they can access to the to the understanding of how much feasible is my part, what is the the cost related with my part, what is the the the carbon foood print related with the the the manufacturing process itself. That’s our first step, on the screening side.
6:07 Of course, as I mentioned, cost saving, lead time saving and CO2 savings are, are key, we are able to to to take into account all the large assemblies, or, or the the the bomb of the of B of the of of organizations, to find the those those opportunities.
6:30 Same way, with the the Cognitive Additive, you can do very quickly, based on your database, based on your experts know how, do the feasibility check, costing check, sustainability check. The costing model is, is based on on on, how can I say, the industry standards, and that’s where we, we, we leveraged it to that.
6:51 Okay, the, it’s not just printing, we have pre-processing, post-processing, that we need to to to to understand, because sometimes those cost can go up to up to 40 to 50%. Same with the sustainability, there is a lot of, now, new trend, and of course in in Europe it’s something more and more important to, there will be some standards which will come out, and on on the European Union side, so we need to be prepared for it, to to bring something out of it.
7:24 And the last is Cognitive Design, how to design better. So our approach, we are coming from the manufacturing industry, my background, my colleagues background, are also from manufacturing industry. We were producing parts, we were failing them, and we understood that, okay, where are the issues while creating, while designing the component. So we wanted to make sure that all the components which are coming out of my software has to be manufacturable, because if it’s not manufacturable, it’s an not, and that’s, that’s, that’s great, but it’s not all, always, always our purpose. Al, al, also in the large organization, industrial organization, that’s not always the purpose.
8:07 So we created this software where we automate DFM. So we will show you, we were able to create workflows using our geometry engine to automate the various different timec consuming processes and steps of of modifying a geometry into a manufacturable geometry. Second is how to lever the the opportunities that AM is providing us, lightweighting, part consolidation, M customization, also adding functionalities to the part is part of it.
8:39 And at the end we want to make sure that this, this, these large workflows, sometimes are created by experts, the expert engineers or or doctors, but it has to to reach to the masses. It has to, that anybody, as my friends at Trinkle mentioned, even the doctor, or the, or the the in in the dental studio, they should be able to design for their need. And that’s what we are bringing at the end.
9:09 So manufacturing during design, we are trying to bridge the design and manufacturing by doing this. And for that purpose we have two different categories of of of workflows: first is the the the product design cycle acceleration workflows, and the second one is the productized workflow for non-sp specialist.
9:31 For this purpose, the idea of the the product design cycle acceleration is to, how to make sure that all designs are manufacturable, how can we automate the TDS and repetitive task, to to accelerate them and take informed decisions while conceiving the product. The second one is, the, is how to adapt our UI to make sure that the anybody, from somebody on the on the shop floor, to the to the somebody in the in in in medical, how can I say, the bureaus, medical design bureaus, should be able to design without having to have a knowledge of computational design. So how can we package it in this way?
10:24 So let’s start with the the design cycle acceleration. So this is a workflow that was kind of produced by all the experience we got in automotive and aerospace industry, of course I have linearized it, but it’s not as linear as it looks like. We are able to, we, what we do is, we, we do the ideation on the CAD software, we sometimes we do shape or topology optimization, we do the the reconstruction of topology in CAD, we do the design for manufacturing, FA validation. There are some iterations between between these two, happens all the time, because we need to make sure that the products are, are also as, can, can sustain the the loads, and at the end we do the fine-tuning of of the of the cat, to to make sure that, okay, product is ready for the market.
11:18 In our experience, we, we saw those workflows at various different companies, which takes up to 600 to 800 hours of engineering time, that’s quite a lot. Most of the time is, is going on the the design for manufacturing, that’s where the the the biggest, because it’s not the same department, so the information is sent to to to another department, bring back, there is a going forth and back, and that takes a lot of time.
11:46 So we wanted to bring those insights into the into the CAT environment, to do that. And, in know, in order to do that, we, we created our our workflow, where we automate some of those those those time consuming processes. So with Cognitive Design, we are able to to easily create those, the ideation, so the the the rough CAD, we can also, once the topology optimization has done, do the reconstruction of topology optimization automatically, I will show you that, manufacturing dine, same way, convert the the the part into a manufacturable part this way, and at the end the fine manual optimization also can be done.
12:27 And we are, we, we demonstrated with our our customers that we were able to reduce the time by 90%, and these are the hard facts that we did with, in aerospace industry, with Saffron, with, with some other, in in Japan side it was Mitsubishi Electric, on the automotive side Nissan and Yamaha. Yeah, why there are so many Japanese names here? Because I, I used to work there for 8 years, so that’s why we continue working with with them, but we also believe that this, this workflows can, can apply to to to more consumer goods, and that’s what we demonstrated with Adidas and over Alp.
13:10 So Cognitive Design is a software which is engineered by engineers for their need. It’s a hybrid modeling engine with a visual coding interface, as you can see, there are multiple view, viewports, that you can, you can, can look into it, you can compare, it’s easy to compare with that, it, it, it has a its focus is on in on the automation of design. So we create this macro blocks or workflows, which are, which has focus to to automate the the tedious task of of designs, of fting, of Boolean operations, for that purpose.
And at the end we also have this manufacturing analysis view, where we can see that, okay, while conce, conceiving the part, is it manufacturable, where the issue is, where the the the the wall is too thin, where there is, there can be any any issue with the with the support structures. So those are, are, are able to to be done, so we can iterate it much easily in the same environment and faster.
14:15 We focused on six different workflows, these are the the the primary workflows, we will be continuing developing new workflows afterwards. But first workflow is designed for a manufacturing, how can we automate that. Second is the simulation driven lightweighting, or design in general, how can we modify the geometry to make it lightweight but still manufacturable. Topology optimization reconstruction, that is a a personal choice, I was spending too much time doing that while, while doing topology optimization on on various cut tools that I can’t name, but yeah, I wanted to automate it so we can actually spend some more time on on creating different variations. Then design for molding, how can we adapt any kind of mold design, I me, any kind of geometry into a moldable design, so we start with a two-path mold, simple one, we will go in in detail further after, how can we do much complex moldes.
15:15 We also do the the mold generation itself, because once we have the workflow setup, we can just put aart, and bring out the the the molds, and we can print with the cooling channels, which comes after, where you can, you can select that, okay, based on your stress, the thermal stress, where the cooling channel should pass through, and it creates automatically.
15:39 So the a morer, the idea is to not spend time on CAD while changing the geometry. Moreover, we are trying to reach complex geometries, those complex geometries are not always easy to be modified in CAD, because we need to to to find those structures everywhere. So this is a very simple example, but imagine a heat exchanger with lot of thin structures, and how can we modify those structures automatically, so we can, we can do, we can make it manufacturable easily, also add some ribs, if, if the this, if the the wall is too to thin, and and to to bring some stiffness to it.
16:26 Whole modification s, detect, the first of all, do the the feature detection on the mesh of the holes, and once it’s detected, convert them into into diamond, or or tear drop shape, so we can, we can print it easily, also add some over thickness easily. Same with the the smart hollowing, to to do the the the thermal management, if you have a data from the the the the the thermal stress data, how can we create the the void, but also create the the self-supporting structures internally, so we can, we are able to to print it easily without having to have internal supports. So that, that part is done with just one block, here you convert it into into that design pretty easily, something which we are working on, and it will be ready for the form.
17:18 Next is overhang optimization, also modify the overhangs mathematically, so we can, we don’t have to work on it manually, creating because of the the the geometry is coming out of topology optimization, it’s not always easy to to work on it. So that’s what we automate.
17:36 Second one is the the simulation driven lightweighting. So this example is, is, we used a lot in in aerospace industry, they wanted to have, keep the PO almost same, but they wanted to add some, reduce some weight. So based on stress, we Chang the the thickness of the of the geometry, and we, we reduce the the weight like that, and we were able to obtain with up to between 15 to 20%. It’s not as large as topology optimization, but it’s, it’s something which is a, which is a low hanging fruit, I would say, that needs to be to to be captured.
18:17 Then the the topology optimization postprocess, it’s pretty much all the bullan operations, all the the operations to to reconnect the zones which are disconnected, needs to be also automated. There you go, yeah. So what we did is that, we, we bring in the the design space, the, we extract the the functional regions automatically, and we put the the the raw topology optimization part which comes out of other softwares, and there is just one block which calls topology optimization postprocess, and in few few seconds you have your part ready, and to be, to be, to that, that can go into into manufacturing, ing, analysis process afterwards.
19:01 Another one is the mole generator, sorry, yeah, I think video, yeah. So the mole generator, we, we work with, with the, with the mold manufacturers, and but also with the, with the injection molding companies, to have a design of of their mold that can fit in in their machines for injections. So what we did is that we, we create, we bring in the PO, we select few few zones where we want ventilation, where we want ejector, ejector pins, where we want injection, and based on that we create a mold. You can see it’s, I think it’s accelerated two times, but not that, that much, you have, you have this two, two part of mold which is ready to be to bring it out.
19:54 All this is, is great, but we currently, our all the mesh that we are bringing out are, are mesh models, and we are working, I will talk about it later on, R&D, how to convert it into into a step model with the feature detection, so a little bit different than what we have before currently in the market.
20:19 Then the second workflows, our productized workflow, where we supercharge operators with computational design, so they don’t have to be specialist of computational design to use it, to leverage it. What we did is, we, we automate the the creation of of simplified UI directly from our our our computational design engine, so we can create that, okay, there is input, output, and and a a 3D model in the in the in in between, and so you can create workflows, you can share with your colleagues, and easily use them for that purpose.
20:56 Something which is not automated is the add-ons and plugins that we work with our partners, such as Cera grab cat from stasis, but also Autodesk. Some specific workflows that we, we automated are paper pulp mold, CMM fixtures, texturing from an image, dental aligners, surgical guides for for dental, and all this don’t need the computational design thing, you just import your your 3D model. It’s a import 3D model, output export 3D model, kind of workflow, with a few parameters that you can twick, so you are able to easily easily generate your models, and we let the the the experts do the computational design, so they can create the complex workflows, provide the the right inputs and outputs. So there is no, how can I say, so there is a flexibility there.
21:57 We also work on autoptic insults for, with the same purpose, we, we brought this, this these workflows and add-ons, on on, as I mentioned, with on grab cat CA, and and Fusion 360, on statuses. The the focus was that, okay, the machines are used in in on a shop floor, they are not used in in a in a design bureau, these machines for mass production are operated by technicians and operators, they should be, be able to to designs for their need, it can be jigs, fixtures, but also in in the in the dental industry, dental liners, surgical guides, they are not CAD operators there. So that’s what we are doing it here with grab cat, and we automate this pretty easily inside the grab cat environment, so they don’t have to go out of it with CA.
22:52 Similar kind of workflows but more complex designs, so simulation driven design workflows, design for molding, topology reconstruction workflows, and cognitive ad works are already present in in S, and you can leverage it with with other S tools, or senos partners tools, so you can create something much more valuable, much larger engineering workflows with that purpose, and Inf Fusion 60.
23:26 It’s a pretty simple, we started with the, with the jig gener, jigs generator, you put your pot, you have few parameters that you can, you can twick, what is the large of your jig, it’s a CMM jig, and then you, you click, okay, I guess, and in, it’s a real time, so you have, you have the the part ready. And this is something which is very simple, but it can be used by by operators veryy easy easily.
23:59 And the last point is what we are doing in research and development. So all this workflows, and they exporting mesh, and we all know how much industry love the mesh, STL, right, d. So what we are trying to do is, of course there is a alternative at 3mf that I have to mention, but we also want to bring it to the in in a B format. So we, we, we are, we did some really great research on it together with an engineering school in France, we were able to convert 2D drawings into a into a B models, of course automatically, we use various different machine learning algorithms to to to do that, and results are very promising, and by the end of the year or beginning of next year we should be able to to bring it to the market. And then, if you, if you want to partner with us, very happy.
24:59 Also, with the mesh, we have all the mesh coming out of our our our geometries, but we need to to make sure that it has to to bring back to the C. So, of course, there is this motorcycle graph kind of mesh generation with the patches, but we all know how difficult it is to use. So what we are trying to do is, we want to detect first the the feature detection, we detect the features on the on the mesh itself, and then we convert into into into the can, the the B features, and we create this, this model. We have some really great results at the moment, and we’ll be integrating into our our our solutions going forward.
So that’s, that’s us, we connect design engineering with the factory floor, basically in both way, and by bringing the insights of manufacturing to the designers and bringing the design tools to the factory floor. And if you want to to to have manufacturable designs, contact us, there are some free softwares available on our website. So also, the Cognitive Design, we can provide it, it for you, for free, for a short period of time, of course, and happy to to collaborate on that. Thank you very much.
More from Rhushik Matroja

Redefining mechanical engineering in the age of AI
CDFAM Barcelona 2026
Listen, search and cite
Listen to this talk as a podcast
More from CDFAM Berlin 2024

Democratizing Advanced Engineering Solutions from Industrial to Medical Applications
Moritz Valentino Huber · Hyperganic

Enhancing Computational Design with Real-Time Design Insights Using AI & Simulations
David Heiny · SimScale

Metal Additive Manufacturing in Power Electronics, Machines and Drives: Opportunities and Challenges
Nick Simpson · Bristol University

Computational Design and Machine Learning in MEP Systems for Large-Scale Architecture
Gabriel Garcia · Haskoning DHV

Implicit Modeling that is user-friendly and expressive, without limiting flexibility
Joseph Flynn · Altair

From Concept to Manufacturing: Evaluating Vision-Language Models for Engineering Design
Kristen Edwards · MIT

CDfAM at the Service of Emerging Technologies – Innovating with AM in Quantum Technologies
Manolis Papastavrou · Metamorphic























