Hello. My name is Andreas. I just solve new, unique and difficult problems for defense and aerospace and then I train the teams how to do that. I feel a little bit out of place to be in an AI conference and talk about design for additive manufacturing but bear with me. This is a project a part of the that’s the second part training the team we you know we solve the problem with these are the recordings from the CDFAM computational design symposium held in Washington DC July 2026.

0:59 For more information about CDFAM, archives of previous recordings and details of future events around the world, visit CDFAM.com with I think you saw that slice on an earlier presentation, the three phase of matter. Okay. I’m going to talk about a little bit melting, you know, how in the heck we can simulate this and what are the benefits of doing something like that? I saw this nice parareto plot and it’s very exciting.

1:43 But we do parareto optimization for the last 30 years. Okay, there are tools like An’s workbench with design explorer or BMX with multiobjective design optimization but we don’t use the cloud. Okay, they’re all within the environment and I still use these tools and I also like the new tools they have they came out such as realtime simulation tools within the CAD environment. Okay, in the CAD environment nowadays, we have five viewing modes, wireframe, hidden line removal, shaded, shaded with images, and now we have another one, simulation results because they use GPU based solvers like I don’t know Creo simulation live.

2:39 You can see immediately the results while you’re designing the part. And of course, new capabilities in the cut system to generate latises, simulationdriven latises, generative design. And I know some of you say, “Oh, come on. Topology optimization is out for 30 years.” It is. And I’m one of the first users ever. But there’s a difference now. The results are very fast. Is anybody from Alter here? Okay.

3:13 In some project if you use let’s say other old systems takes four hours now it takes like 30 seconds. Okay. The other big deal is the results become CAD modelm they can come back to CAD and of course all of these are linked with the design optimization tools but let me come back to this. So really ouch. What happened when I have water and I have a bunch of ice cubes and I start heating it up, the temperature goes up and it reaches zero and it stays to zero till all of it melts.

3:56 And then after it melts, it goes up up up to 100° and then it stays 100° till all of it evaporates. The idea is can we take advantage of this for very high low flight duration vehicles projectiles they go very fast higher than the speed of sound but because the flight duration is small and there’s a lot of power electronics there we need to cool we can afford to put heat exchanger or heat pipes or other cooling systems We take the container with electronics and embed it in some phase change material like salt and then the heat being anticipated and it follows the it’s hard to point here but the the first flat portion go from solid to to liquid and say did anybody think of that?

5:03 Yes. Nature. This is for Matthew especially some there’s these antire beetles you have pockets of phase change material on them so they get hot on the sun and when it freezes they take advantage of that. Okay. So this is from again the second phase the training to the people. So how do we do that? One option is to there are two problems on phase change materials. Let me tell you the problems first.

5:40 The first one it has very low thermal conductivity like 0.1 and copper has 400. So we need really it will take forever for the heat to flow to the end of it so we can take advantage of the full melting pool. Okay. How do we solve that? We embed latis structures within. So we impregnate the the material with more highly conductive u matter so we can send the heat there.

6:16 Okay. You said two problems. Yeah. The second one it has high volume expansion. Actually some of my colleagues here which they work on that told me earlier. Oh yeah we put it in a container we close it. We heat it up and pop it popped out. So we need to capture the volume and improve the thermal conductivity. One thing we can do so is to put latice structure like what you see here if the heat was on the bottom and then we need to put some kind of energy absorbing material in the surrounding.

6:54 So if it expands, it can absorb the expansion. This requires design. Okay, I don’t know if I can use AI for it yet because this is trivial and also they give you half a day. You go in the morning say here is our problem. We need to design a container which it takes the face change material. It doesn’t break and it last at least delta t the amount of time.

7:25 Okay. So, thank God for the vendors with the provided material. They give us the data sheet so we can extract the heat storage capacity and thank God for the integrated CAD system simulation tools which you can go let’s say to Creoflow and enter the heat and then you can solve it within the CAD environment very fast. I told you real time simulation tools and also it needs to be simple.

8:02 Oh my gosh. It need to be this. It has two features. One is the cylinder and the other is a 2 and 1/2D latis structure a hexagon with drain holes. Okay. So you put just one latis you infill the volume and you get something like that. The the things you see there is the drain holes because we wanted to make sure you give it access to go sideways.

8:33 We found out it wasn’t important, but at least if it’s fast, it accelerates innovation because you’re not afraid to ask, “How about if I put drain holes? How about if I don’t put rain? How about if I put something else?” Okay. And then if you can see the PCM material, it will look like that. That’s the complimentary volume for the PCM material. And then because all these stuff now are interactive and some of you are users, you can generate the mess automatically.

9:11 You don’t need to even see it, but you see the first time. But this was conduction. So it’s not that super critical to have boundary layers, although it does. And then that’s the ultimate the holy grail to plot the temperature versus time of the PCM material. That’s the minimum temperature because at the close to the walls it’s hot. As you go farther away, it’s cold. So we plot the minimum temperature and we see on the bottom of that curve how long it takes to melt all of it.

9:45 And that’s the criterion. That’s the design constraint. We want to achieve certain amount of time till all of it melts because after that there is a thermal runaway. It goes like crazy. And of course you need to be able to see the results within. So you can see the melted material. The blue is the unmelted material. Okay. So of course what happens these big defense contractors you have an army of analysts and their perception is if it’s so fast it can be accurate man he uses vauil and he uses 256 GPUs and but you don’t argue you say yes sir so I want validation okay validation what is your validation code usually either nastra nances so you take the geometry to answers and you mess it.

10:44 It looks ugly but it doesn’t need to be better. You need to have that’s where the human intelligence comes. AI is great but man can he make educated decisions based on experience what the element sizes you can do adaptive messing but here is different story. So then you do and you find similar results. This is on the bottom the minimum temperature on the PCM material and the other is the exterior and etc.

11:20 So and then you can see a little bit better the the melting pool whatever he melted. Okay. But then how is that the best lattice? What else do you think? You know the guy says porcupine. Okay. So and again the mission is simplicity. This is two features the cylinder and a latis structure. The latis structure has a unit cell which is a hexagon beam but the height of the unit cell is as high as the cylinder.

11:58 So that’s is when you see one and then you say okay on the bottom is so high the thickness and the diameter and so much on the top two freaking features okay simple keep it simple engineer or young engineer okay if it’s not simple men and it’s so complex and I rely only on AI I will be nervous and then you do that and you solve it and you find out if that was a good idea or a bad idea.

12:34 Okay. And of course, the real problem doesn’t look like that. You know, if it’s you, you can’t even show it. But the other option is can I if the heat was coming inside the cylinder, which is more realistic because it’s a container, a can, and you put the electronics in. Is there any better ideas to do that? And yeah, you can use again this is two features.

13:02 Okay, three because there’s a pattern but three features because it’s a variable section sweep with sinosoidal and it’s tapered and it’s okay. I forgot the fillet. Four four features. How long will take you to build the model like that? I don’t know two 3 minutes. So if the heat is inside, that will be a better solution to have radially spoke latises and they’re taper and corrugated and etc.

13:36 Okay. Of course, some of you say, can you print this? Yeah, we print it in quarters because they need to put the electronics and then the top. And I can show you that if you want. And then the next question will be how about if it was like the original and the heat was coming from the bottom. Can I use the same concept? Yes, you can. Maybe you need to blow some air so this hair or wig I mean worms bent so you can have more surface finish surface area embedded in the material.

14:15 So you it’s very easy to explore this and say oh can I do a sensitivity plot? What sensitivity? This is a discrete variable. So what you do within the modern cut systems today you can get a bar chart of all your design iterations. So you can say okay the first one was hexagon the second one was a porcupine the third you need to give him creative name so you can remember it the other one worm or whatever okay what you don’t like the term worm okay and then of course for printability sometimes it’s ah I keep touching the wrong it’s good to use TPMS triply periodic minimal surfaces like gyroids, primitive diamonds and etc.

15:05 Because you have larger interface area. You have continuous surfaces there you know not discrete serve corners better uniformity and you can retain well the volume because now you can capture some of the expansion by the gyroid. The gyroids or or TPMSs are great because first of all they’re printable. Second, the moment of inertia about X, Y, and Z is the same and they’re very strong in tension in pressure inside because they go from the left to right.

15:43 They’re braced. Okay, so great structures. And then you can bias the thickness based on the proximity to the heat source. So you can have really, you know, here and you can see how it looks like where you will need to pipe the heat as fast as possible up. But if your media is very low in thermal conductivity and you use aluminum instead, you need to have really big pipes to send it up.

16:15 So that’s sort of a design concept you can explore. Or you can explore a diamond which it will look oh no that’s a primitive which it would look like that. Again you bias the thickness but you embed that with that material so it makes it a little bit stronger structurally and or you can use you know another this was what diamond. So the the point I try to make the design iterations are very fast because if it’s one feature you go to the pull down which says no gyroid use diamond boom and then automatically because real time simulation tool it will give you the time or the temperature and etc.

17:05 So you’re not afraid to try. That’s where innovation comes. We have great ideas but if we can evaluate them quickly forget it but now we can and also we can generate automatically a bar chart for all the iterations and see which one wins okay I think I give you only 20 minutes for this so it’s unfair but you know so this is the other lat structure how many more minutes oh three great so this is the idea DPU and then conclusions men metal and high conductivity latises can use to enhance the effective thermal conductivity of a change materials.

17:56 Okay. The second one what’s which one is the best lattice I give you from the experience I didn’t you can use AI too but you know here’s from the experience use gyroids for maximum thermal conductivity for fast melting octed for light structures TPMS for additive gyroids and for anotropic heat steering like we need to send it radially you need to use beam oriented like The so the latises address that problem and also what we learned the hard way don’t tell anybody other than the people here 5 to 20% volume otherwise you don’t have enough PCM or you don’t have enough thermal conductivity so there is a nice tradeoff there so if we do a sensitivity plot we’ll plot volume fraction versus the time to melting.

18:55 Okay, so those are great things you can do. You can do them now with the tools and you need to create the vision which says, “Okay, I’m going to use these modern tools now.” Did they set me off? Yeah. Okay. You can use these tools now. Adapt it because it’s not easy to do that. Get training. Okay. I’m not advertising my training but anyway get training and then you need to adapt your process to be able to achieve it and it’s hard because you know when you go away if you need new horizon you need to have the courage to lose the sight of the sword I thank you what thank you these are the recordings from the CDFM computational design symposium held in Washington DC July 2026. For more information about CDFAM, archives of previous recordings and details of future events around the world, visit CDFAM.com. Test your mic.