0:01 Okay. Hello everyone. Thank you Dan. Thank you everyone for joining. My name is u Alessandro Zomparelli. I’m a computational designer. I have a background in architecture and engineering and since 2023 I’m part of AI Build. I work there as a senior geometry engineer as part of a geometry team. So, AIBU is a Londonbased software company. We develop an integrated all-in-one platform for manufacturing in different industries and we have a particular focus on a large format 3D printing.

0:39 So, while the background and the roots of the company are in architecture, we actually work with well with different industries. So we work with aerospace, automotive, construction, energy, marine art and design and many different industries. And as you can imagine, every industry have very different requirement, very different needs. And the software that we provide basically works in a modular workflow that allows to perform the slicing of a object to produce the tool path that you need at the end to print and is a hardware agnostic software.

1:16 So it means that you can use it with different setups. You can have from the daxto printer to big gun printers and the multiaxis robot and u after generating the slicing often time you have to also optimize the slicing different materials different machines they require many different tuning of how to make a very nice tool path for a machine. And then we allowed to simulate and see what actually happened with the machine and to monitor the the printing process while producing the part.

1:45 So we also added recently this feature for the thermal analysis and the problem is that with some material for example for example with metal you want to really keep an eye on what is the temperature while you print. So what we do is we do an analysis, a terminal analysis and based on the feedback that we get, we tune and we adapt the slicing path in order to slow down or wait or change some parameters according to which setup we are using.

2:13 So why is everything that complicated? Can we just use a normal slicing? Well, yes and no. Because what happened is that some clients they really like to do very complex things that are complex on both a geometrical point of view but also the material point of view and the workflow that we have to use they often become very complicated. Clients customer we work with we have very different shapes that we want to produce sometime is very simple but sometime can get complicated.

2:42 Or maybe sometime we have something that doesn’t look that complex but they want to print in a very specific way. They want the layers to be aligned according to certain features. They want to ensure that the quality at the end is optimal and up to their expectation or sometime we have something extremely complicated and there you have to do a lot of finetuning in order to find the optimal printing settings that gives you a super watertight object at the end or they work with complex material and technologies like DD metal printing and they want to print from a central part toward the external part of the object because this is how you produce this type of object or yeah other type of DD printing or for example they use WAM is a different process different parameters different constraint it’s nice to follow all these different cases but every one of them have very peculiar challenges or for instance you can combine and have an hybrid workflow where you do a little bit of printing so you take the advantage of printing in reducing the waste of material you just put the material where you need it and then you have CNC milling that allows to smooth the surface and have the the finishing surface that you usually have with milling.

3:59 So because of all the variety of different cases that we have to cover, we basically develop many different slicing method. So according to the user need, there is a specific slicing method. It works fine. You can use it and the software is kind of modular. Imagine like in grasshopper having many components. Here you have different operators that you can use to create a workflow and you know according to which type of challenge every customer has to to deal with.

4:27 We basically have already some workflow that show how you can achieve what you’re trying to do. And this looks fine from an engineeric point of view. But then from the design team they say yeah but isn’t it a bit complicated? Can we just simplify that? So we have this end goal that we are looking to which is the one click slice which is a very difficult challenge and this is actually a problem that we are currently attacking from many different direction and what I’m going to present and discuss here is the geometrical point of view of how to deal with this ideal oneclick slice.

5:07 So let’s imagine that we deal with a simple case. We have a bunny. No it’s not a bunny. Let’s imagine is in general a closed geometry. We want to print an object. We don’t want internal infield. We want to save material because we are talking about large format printing. And we want to just print the outer wall of this object. If you just use the regular horizontal slicing, what you usually see in 3D printing, you end up having gaps in the overhang area.

5:35 You have gaps at the top. You have the need of support probably. And with large format you don’t want support because support means extra material, extra time, extra labor to remove the support. So you want to avoid them as much as possible. So what we came up with was this judici slicing approach where basically is a slicing method that is more aware of the geometry of the object and try to guess what is the best orientation of the layers and it can work in non nonplannerly.

6:05 So you see a regular slice horizontal slicing and then what happened with a slicing method which is aware of which type of geometry you’re dealing with. So as we mentioned here we have gaps at the top. Normally you have yeah overhangs that are not dealt properly. And with jodesic you can cover entirely the surface maintaining the uniform distance between every layer and ensuring that is watertight at the end.

6:41 Also, we have some parameters that allows to tune the the layer planerization. And basically, the same slicing method works out of the box perfectly with many different cases. And all these cases usually were addressed with different slicing method, different input. You decide how to want you want to orient the layers, how do you want to print the object in order to achieve a better quality. And now we are able to achieve that just with a a single slicing method.

7:13 So how does it work? Well, it’s not that complicated. So we normally have an horizontal slicing, meaning that we have a vertical gradient according to the Z coordinate and our layers are just lines that have the same Z coordinate. While instead if you use jodesic fields meaning that you compute the distance from the bottom of where whatever you are starting to print and you reach the end of the object then you can produce those curves based on the same distance value from the base.

7:44 So basically what you want to achieve is of course to have a closer distance between the layers especially a uniform distance but you also care in this case about the tool orientation. We are not talking just about a gantry printer that print from the top. We can use robot we can use multiple axis so we can orient the tool in order to achieve the optimal printing quality.

8:06 So what you usually want is to have the tool that is oriented towards the previous layer. This gives you the best interface between consecutive layers. So considering both properties what happen is that the overhang analysis that normally would look like what we have on the left a lot of red bad comes up with a lot of blue good on the right. So the process is is not that complex.

8:35 Basically, you take a geometry, you analyze the mesh that you have, you have the triangles that allows you to compute and calculate the distance along the geometry. You store the distance information on the vertices. You create a nice gradient that represent the geodessic field and based on that you just produce iso curves contouring that generate basically the layers that you have to use. Now all these in theory work really well if you provide very nice meshes.

9:06 The problem is usually the customer doesn’t have a nice mesh. Sometime it doesn’t even know that there is a mesh at some point. It just provide a step file and the mesh that you get from a step file usually is not really nice. So what happen is that you have some glitches, some errors and you cannot ask the customer okay can you go back and make a nice mesh so it works nice.

9:26 No, you have to deal with that personally. So what happened is that you have original triangles that produce all those glitches. We have to do a meshing of a surface. The problem is yes, you have a nice triangles that are uniform, but still the isosurfaces they work on the level of detail that is present in the mesh. Meaning that if you have big triangles, you have glitches as well.

9:53 So what you have to do you have to increase the resolution of a mesh and have smaller triangles but then you have many triangles. So what you have to do is instead to do an adaptive remeshing that automatically optimize the distribution of the triangles in order to achieve good quality and good performances. At the end the user just have to care about the what is basically this expectation in the computation time.

10:15 If you just want a a very quick computation, you can just use low resolution or if you want a super high quality, then you can just increase that. So from a geometrical point of view, you say, “Okay, that’s solved. We are fine. Good.” No, it’s not that simple. Because in London, we have also the facility with a robot and people that is using that to print and they say, “Wait a minute.

10:35 It’s not that simple.” So what about the collisions? Oh, collisions. Yes. So what happen is that when you deposit the layers we say that the tool is oriented toward the previous layer but sometime you end up having collisions because basically in all the local tops the tool is oriented tangentially to the surface and then every time that you reach that bum collision. So how do we fix that?

10:58 Well why don’t we orient the tool just perpendicular to the surface so that in those areas we can print without worrying about those collisions. So now we provide a set of parameters that allows to control what is the correction area. This of course change and is affected by the type of hardware that you’re using. So according to how big and bulky is your extruder, you may need a different correction.

11:28 Then there is another problem. Not only the tool orientation, but because of the dimension of a tool because of how complex the layers can get, you may still have collisions. For that reason, we have also a way to plannerize the layers. So, it’s just about smoothing the judici field. This will produce nicer and prettier curves. But this has a small cost because this introduce variability in the layer height.

11:51 This is not a problem in some processes. Usually polymers are quite forgiving, but when you use metal, this becomes a problem. So, depends really on the type of process you’re using. Then we realized that a lot of product that works well with this type of process are air duct and because you have a cave geometry you have just the outer wall and you can make very complex shapes.

12:17 But what happen in this case is that sometime because of the type of finishing or recognition that you need you want to have the layers to be aligned exactly with the naked edges of your geometry with the opening. So this is usually not something that you can control directly because what you get normally is what you see on the left. You get just the lines that at some point arrive at the edge and sometime they are just trimmed.

12:43 So we had to introduce a way to control how you can align the layers to the naked edges in those in those parts and you can control how much you want to move the transition from the top towards the geometry. How does it work? Also that one is not that complicated. You have normally a geometry let’s say this particular node with some naked edges. You normally generate the slicing from the bottom.

13:08 So you have the gradient all the layers and on the edge you have some trimmed poly lines. So what you do you start from the jodic field from the bottom which is not always the bottom but where you want to start to print and then you process another one that start from the naked edges. And you have those two gradients. And what you do, you combine them and then you get a nice gradient that provides you a transition from the horizontal base towards all the naked edges.

And this works fine with this type of object. And those two object that normally would require different inputs in order to have the optimal printing quality. In this case, they are just working fine because we recognize those features. And here you can see something a bit more extreme just to see how no matter how many holes and openings you have, you can still produce a slicing from the bottom to the top.

14:04 Good luck printing that. So what happened is that when you plan when you apply a correction along the naked edges you are messing up the ideally perfect judici field that you have. So we realized that some clients we we were getting some variable layer height that I mean it’s just part of how it works. But they wanted to have a very specific control again different material different processes they have different requirements.

14:34 So sometime they want to keep the average layer height or sometime they want to maintain a minimum layer height. So we have to understand exactly how we want to distribute those curves in order to ensure that the layer height variation is within the parameter that we need for printing optimally. And from that we derived a very specific and simplified application which is called base mode which basically just work every time that you have two naked edges it creates the slicing from one to the other.

15:02 So this automatically works of course with different geometries. Doesn’t matter the meshes, doesn’t matter the parameters, doesn’t matter if they’re planner, non-planner. It provides you the layers that you need to print them. And without that with all the slicing method that we had before, we had every time to specify different input to provide the perfect slicing for them. This is a video in collaboration with generative machine where we were printing using Josic a lot of small tests just simple sphere but still you don’t have the gaps and you have a perfectly closed sphere or for example yeah just a a very regular ways.

15:43 Okay, nothing special about this. It’s just horizontal because the shape is horizontal. But here then you have an inclined top. So you it automatically detect that and orient the layer in order to reach exactly the end of the object. Or if you have more axis like here, then you can go crazy and do something non-planner and print a geometry that would definitely require some support, but you can avoid them because you can just print exactly following the shape.

16:10 You may not see a big advantage in a small object like this but imagine a big object where the support are money. Okay. So all this was about how to handle the surface of the object but now we are working on implementing a volutric approach of that. So normally you have the surface and you produce iso curves. If you just process the volume, you can produce iso surfaces and with them you can perform a complete slicing with multiple walls in field of the geometry.

16:41 So here if you see the bunny what happen is that for most of the layers at the base they are kind of horizontal. We are just adapting a little bit but then you see that they start bending when they need more support to provide the perfect sequence of layers until you see the ears and the layers are oriented according to the shape that you have there. A branching shape, similar behavior.

17:07 The layer are automatically aligned and they follow nicely the shape and they provide all the internally in field 100% density and whatever. Or you have maybe more complex shapes like this and again all the layers are oriented and they produce those non planner in field that provide the optimal deposition for this object. So now yes we have probably don’t have yet the one click slice but as you can see you can also tune the plannerization of a layer so you can move from something extremely non-planner towards just a sequence of planes along the object.

17:45 So with that we hope that we like make a small step toward a single slicing mode to rule them all. Thank you. 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.