Play God with 3D Printing? Getting There, Says UK Startup

A London-based startup, Additive Flow, aims to address some of the problems of additive manufacturing. Champions of 3D printing have been on the verge of claiming the technology could impart God-like powers (3D printing organs, for example) to engineers, but the technology has not even been able to become a mainstay of manufacturing. Enough barriers to additive manufacturing exist to keep the technology somewhat of a novelty.

Ask engineers or manufacturers why they are not using a technology that they must have heard about—often in breathless tones—one that promises the lightest-weight parts and is tailor-made for a digitalized workflow, and you will elicit wearied sighs. They won’t know where to start. They may ask, “What is the problem that 3D printing is trying to fix?” And if traditional manufacturing is not broken, why try to fix it? Or why use a new unproven technology when other manufacturing technologies have withstood the test of time? Or is it because what they have tried to 3D print has failed? Or is it that materials don’t reassemble as they should? Or is it that the right material is not available?

Additive Flow plans to deliver software that addresses the latter failings of additive manufacturing by accounting for not only the right shape, internal and external, but also the variation of material properties of the part. Variation due to anisotropy (variation in the z-direction, for example, from building a part layer by layer) is being addressed for the first time, according to Alexander Pluke, CEO of Additive Flow, with the ability to inject different materials into the part as well as to account for the result.

Alexander is not an engineer, but engineering runs in his blood. “I grew up doing a lot of engineering with my father in his workshop,’” he said.

Turn on God Mode

A 3D printer is able to inject different materials in different regions of apart. Think of an inkjet printer head with inks of different colors, but instead of different colors, you have different materials. Different properties of the materials can be taken advantage of. Think of a thermally conductive material that could provide a built-in cooling channel—or an electrical path—or the energy absorbing heel of a running shoe with a stiff midsole.

If a multitude of design innovations is not boiling over in your brain, you are not paying attention. This would involve any mix of material you desire, each forming inside a part, each perfectly suited for a given purpose.

This was a design option previously available only to God.

Topology optimization, another overhyped technology, is but one step toward playing God, allowing you to material in the right place. You could consider topology optimization two-material modeling if you consider one material to be solid and the other to be air. A binary solution, either with material or without—with a single material being 1 and no material being 0. True multi-material(ity) would fill in between the 1 and 0 binary solutions with a range of values. A 0 could be air, 0.1 a light foam, 0.5 a plastic, and 0.9 a high strength steel, for example.

The Multi-Material Reality

We met with Alexander Pluke via Skype, with me in our San Francisco office and Pluke in his London flat. It was two weeks before lockdowns were imposed by the coronavirus pandemic. The affable young CEO was only too glad to discuss the ills of and the cures for additive manufacturing, and to endure rants about all that is wrong with 3D printing, as well as bring us down (graciously) from the idea of playing God.

“You can’t mix anything with anything,” said Pluke.

Why not? Nature has been making multi-material parts forever. Are we not ourselves multi-material? So why do we struggle with what nature has done so well and for so long?

That’s an analogy that my CTO loves talking about (Pluke is referring to Charles Fried, who has a background in generative design and artificial intelligence, and over a decade of experience in 3D printing). Charles is always talking about the bone as the classic multi-material. We’ve got different geometrical shapes, different tensile strengths. That’s complexity.

But industry executives are always telling us complexity is free with 3D printing. Is it really, though? They use as examples that parts within parts were hitherto impossible outside of skilled jade carvers or 3D puzzle masters. Other parts are complex in their geometry, uneven, tree like, complex enough to defy mathematical description, and therefore, unable to be modeled or analyzed. And yet, we have only succeeded in making lattices.

Lattices are a third priority for us. First are materials, second is parameters.

Like a Dog with a Bone

God or nature (you pick) didn’t design the cell structure inside a bone. The cell structure created itself, putting material where it was needed, removing it where it was not. The part is being optimized on the fly. These are irregular, nonuniform cells. These are not those neat, perfect lattices.

I think it’s always important to say what you can’t do. Though we have capabilities and deep data structures, and we have done some testing and have simulations about how different structures perform, it is not limitless. Empirically, we have this simulation data. Our software is treating a region with a material property without any predetermination what the structure (material, lattice or cell) inside each region.

The question is, does any material need to be in a space or not? If so, can the design have that space? If we can tell our software what we want to achieve, can we optimize our production, our design and our material simultaneously? At the end of the day, you’re going to have physical parts on the table. It can be additive or nonadditive.

How do you handle different strengths in the Z-direction?

The layers created by 3D printing are like a whole bunch of different physical parts that are interacting with each other. We’ve modeled what is happening in between layers and at different scales.

Materials are not the same in additive manufacturing, correct?

They are not. In additive, the material will have multiple properties, even though the bulk material is the same. The final material properties of the end part are a function of the build parameters and the process. With FormFlow, we minimize the variation in material properties.

Making New Materials

At what scale does Multiflow operate? If it is mixing materials on a small enough scale, is it not essentially able to make a new material? Like the mixing of metals to make alloys?

If I may just zoom out just a bit, let’s look at different manufacturing techniques. We have several, including welding, cold spray and metal cladding. On the polymer side, same story, we have several, including FDM and polyjetting.

Additive Flow’s specialty really is around this combination of the topology optimization and multi-materials.

What we’re able to do is look at either known blends of materials. There are compatibility issues. You can’t have anything buttressing up against anything. There’s primary physics and material science coming into play. We’re working with a number of experts on that.

Actually, generating a new material, or determining how a new material behaves, enters into another tier of simulation/modeling—more around the molecular structures and how they will generate new material properties.

We’re working with third parties to create algorithms that connect a new material’s molecular structures to its material properties. The algorithms will be used within our software.

Then, in the end, has the 3D printer created a new material by mixing materials at the voxel level?

Yes, if we talk about material in general terms.

Let’s say we’re looking at a jetting technology and polymer materials, say, liquid A and liquid B being mixed. Different ratios of liquid A and liquid B will result in different kinds of properties. They may be linear differences or nonlinear differences, and that will be a function of the materials that are being mixed. This mixture would result in a new material behavior. If you’re working with liquid or powder metals and using heat, like a laser or electric beam to join the metals, that’s different. You may have a melt pool. A new metal is created in a sense, because you’ve got two different component parts. The mixture could vary linearly with the gradient mixture, or you could have a sharp transition.

There’s a lot of complexity in that. This is an area of accretive research. It takes a very different stimulation stack chemically, molecularly than if you were thinking about simply mixing liquids, which is not so simple with polymers.

Different organizations will be at different points of their material R&D cycle. Some may be much more experimental in the materials science. Some might be more on the application side. They just want to get something into production.

It will also depend on what the machine will be able to do currently. So, a combination of what the machine is able to do and what the materials science is able to predict. In both cases, the result is not fully known. These are still any unexplored combinations of materials.

We have to be mindful that once we start talking about multi-material, we risk people thinking we can mix anything with anything. That’s not true. Understanding which materials will mix well together, what ration will work best, what creates the kind of material properties desired. Some of those combinations are known, and some of those can be simulated digitally.

How does FormFlow work?

We start with the physics, such as the forces, thermal requirements, vibration, electric fields, etc. Then we begin with optimization. Within each optimization loop, there is an allocation of the different material type. This can be a preset material type.

How can FormFlow be AM-process-agnostic, as you claim, when processes are so dissimilar to each other? Shouldn’t you be process specific?

Additive Flow software is process-agnostic in the way that we handle data. Our algorithms are looking at certain elements of material properties and processing characteristics which are manufacturing process-agnostic because we’re looking at how the material behaves in that region in the geometry.

A multitude of processes can arrive at that material distribution within such a geometry. The way that it can become transferable, if you will, is by taking that framework and then making it specific for a number of particular manufacturing techniques and material systems.

We don’t use agnostic to mean you can directly transfer everything to everything, but that the framework of our software structure is able to work within a number of specific fields. For example, consider powder metals. We’d be looking at those kinds of material transitions and how materials are being allocated differently.