3D printing is perfect for creating items perfectly fit for one person. It has already been applied successfully for prosthetic limbs, joint implants, orthopedic implants and inserts, orthotics, orthodontics, hearing aids, and many more advanced medical applications. Now that people with those rarer conditions have been helped, it’s time for 3D printing to move into applications that help the masses. 75–85% of women who wear bras are wearing the wrong size. By combining 3D scanning, custom software, and 3D printing, millions of women can live more comfortably. I am collaborating with two female industrial designers in Brooklyn who are working on just such a project, and I’m working on exploring a variety of other potential problems and solutions. Using scanning to determine size is very easy, but am taking on the much harder problem of matching and creating the desired curvature and structure. On September 21st, I showed an example design at World Maker Faire in New York that featured 3D-printed elastomeric cups. They are perfectly fit to the wearer using 3D scanning. In this case, the natural shape was ideal, and the function was only to support and preserve, with no shaping required. In cases where a sculpting of form is desired, there is an opportunity for benefit, but a fairly complicated geometry problem to achieve the perfect fit and comfort. I’m making good progress, and this project is tied in with others I’m working on, so expect to hear more about it very shortly.
In my earlier post “Voxels vs. Polygons” and in my post on engineering.com I mentioned a product from a Norwegian software company Uformia. They have a volumetric modeling plug-in for Rhino called SymVol that has some unique capabilities, enabling the creations of designs with certain characteristics that are very hard to replicate through traditional modeling. The stand-out features that set it apart are the capabilities to create smooth blends between objects, clean booleans (joining and subtracting of models), and generation of internal lattice patterns. All of these operations can be done in other software with some effort, but are almost always a source of problems. When done using a volumetric approach, the results are robust and reliable.
The main limitation I found in the demo version SymVol that I used was the fact that the geometry creation tools are based on parametric creation of geometric primitives. 3D modeling veterans will remember this approach being called CSG- Computational Solid Geometry, which was the predecessor to the more organic surfaces that form the solids in a modern CAD program. CSG is still popular in some circles because you can represent a 3D object with a very minimal set of instructions. Early ray-tracing (photo-real 3D rendering) software like POV-RAY created nearly all of it’s forms by addition and subtraction of spheres, boxes, cones etc. in various ways. OpenSCAD is probably the most popular implementation of CSG modeling and while its ability to control the generation of a model using parametric, script-driven control is impressive and useful, even the untrained eye will look at models made by CSG methods and label them primitive. They look like computer animation from the 1980′s, because that is the method that was being used at that time. Despite the amazing power to combine, blend, and shell 3D models, a design application is far less useful to me as a designer if it can only create blocky-looking models. Uformia is in the process of developing new approaches increasing the users control over form, but until then, only an experienced designer will be able to create models that don’t reveal the simple math behind their 3D definition. I have not tried the full software, which hopefully does not have any limitation about what type of geometry can be used as input.
The good news is that Uformia has taken a different track with MeshUp that brings polygon modeling into the volumetric domain so that we can enjoy the best of both worlds. They have launched a Kickstarter for a program called MeshUp that will take polygon models and convert them to volumes so the user can freely join, edit and hollow them without the normal limitations. This is not a panacea- there are many challenges left related to creating easy flawless output for 3D printing- but this solves most of the trickiest problems affecting the largest number of users, in one fell-swoop.
For those unfamiliar with what the challenges are and how they affect the experience of creating a 3D design, I’ll review these issues and how the volumetric solution addresses them.
Example: Fillets in traditional CAD
When softening the edge of a model, either an inside or outside corner (a fillet or round), the best solid modeling CAD programs available will often fail to generate the blended surface if there are any unusual shapes converging. Specifically, if the blend moves far from the blended edge, it can envelop a nearby feature. From there, the algorithm has no clear solution about where the surface should be and how the remaining features should be connected. Companies have thrown some very sophisticated tools toward the problem, and despite the simple intuitive wish that the user might have -”Make it smooth!”- This is still an area where even an experienced designer with the best software available may still have to adjust their design to work around the limitations of the software. Siemens NX is arguably the most sophisticated professional CAD system on the market, and is used to create everything from cars to Apple products to medical devices. Their latest flagship product, NX 8.5, has a variety of tools for managing fillet intersections that you can read about in this excellent in-depth review of NX 8.5 at Develop 3D. Even then, significant skill and experience is required of the user to get the intended result. In contrast, solving this problem in the volumetric domain becomes trivial. Features can be engulfed by the blend and are gracefully absorbed into the form. Very large blend radii can be used, giving the designer complete freedom to shape the form as they wish without the software throwing out errors or generating bad surfaces.
Another tricky issue in the 3D world, at least in modeling software using polygons, has been the issue of combining models together and removing the overlap to create a single solid object. This is referred to as a Boolean operation, after the mathematician who came up with the idea of adding and subtracting 3D shapes as if they were numbers. Again this is another function that can create errors in even the most expensive and sophisticated software available. It is not a weakness of the software, it is a simple fact that if you are using what is essentially a set of 2D surfaces in 3D space, there are certain conditions that will create unpredictable results. This is especially true when the input model has some imperfections that the designer has missed, such as holes or double surfaces. Solid modeling does not suffer from this limitation, but the vast majority of modeling programs still use polygons. There are also vast libraries of existing content in polygon form, and volumetric conversion opens up the possibility that models that were created for visual-use only might now be 3D printed without too much hassle.
Lattices and Hollowing.
In the context of 3D printing, models are rarely a simple solid block of material. Since complexity is “free” but material and print time can be expensive, models are almost always hollowed-out in some way, and internal lattice structures are often used to maintain the strength of the part. Until now, this function has been built into the software that creates the final instructions that guide the printer. With filament extrusion based printers like those from MakerBot and Stratasys, the polygon model is sliced and filled with a cross-hatch pattern of material. The user might be able to choose things like fill density and the thickness of the outer shell, but this whole action is applied and considered after the design process is complete. Since it is completely automated, the designer does not have any control over many of the details, and so many variations of form and function are closed-off. What if I, as a designer, wanted to have thicker lattice near the outer walls and completely empty space in the vast middle? Not possible. What if I want a hole in the model so it can be a container, or relieve pressure, or some other reason? I can’t, because if the model is not closed, the algorithm will be confused, not knowing what is the inside and what is the outside. I have deliberately fed bad models to a Zcorp powder printer to see what would happen, and in some cases the result was a spray of voxel rays extending from the opening and impaling nearby models as the algorthm traverse the whole build chamber looking for the missing wall of the model. With Ufomia’s internal lattice generating capabilities, the shape, density, thickness, and placement of internal features is now under the complete control of the designer.
One of the most painful parts of learning to create 3D models for 3D printing is when a new user must learn how to hollow out their model or face absurdly high printing costs. Service bureaus like Shapeways count volume as one of the most important characteristics in setting the production cost, but unfortunately there is no simple solution that works in every case. Even basic solid modeling programs, like Autodesk’s free 123D, can often hollow a model using a “Shell” command, as long as it meets the requirement that this offset surface does not envelope a feature, making the software unable to calculate a result (the same limitation as with the filleting example). More often though, the user is starting with a polygon model that may have thousands of faces. When these faces are offset with the polygon equivalent of the shell command, the software will not show an error, but it can often go horribly wrong, generating a tangled mass of triangles that is unprintable and almost impossible to repair, even with many hours of tedious work. The only software that does this reliably does so by converting the model into a volumetric space, offsetting the surface, then converting the result back to polygons. Examples of this would be Materialise Magics ($13,000) or Rapidform ($42,000). One of the more accessible solutions has been available in recent versions of Pixologic’s Zbrush. The Dynamesh tool is a voxelization-based system that can convert the mesh into a volume of up to 1000 pixels on a side. That does cause a loss of quality for most types of models, but for detailed sculpture it is often the only realistic approach without excessive cost or labor. On the free end, there is also Uniform Mesh Resampling in Meshlab, which is a bare-bones implementation, but accomplishing the same thing, if you are patient enough to figure it out. None of the existing solutions are ideal, and this is a major sticking point preventing people from jumping right in with 3D printing. With MeshUp, Uformia intends to make this problem go away immediately and permanently. Hollowing a model in MeshUp should be extremely easy, and if implemented correctly, would avoid any loss of quality.
MeshUp does a lot to address a handful of the biggest issues in making models for 3D printing, but there is still work to be done. As of right now, we still have the limitation that all models must ultimately be converted to polygons, and usually to the STL file format, which eliminates some of the potential advantages. Uformia is aware of this, and they are working to address it, but it is only by working with printer manufacturers that the input formats can be updated. They have done exactly that in at least one case. Neri Oxman, by teaming up with Uformia and printer manufacturer Objet, was able to print some seemingly impossible creations by side-stepping the mesh generation stage and generating the final slices that the printer would be applying at each layer. Indeed, the SymVol Rhino plug-in has a capability to export an image sequence of the slices of a model, a method which I think holds the key to opening up volumetric printing, since programs that work with 2D images and video have a solid foundation of code libraries that can be leveraged in the context of 3D printing with very little effort. What prevents this from happening is mostly the lack of standards and variations between hardware. If I have an image sequence only, I have no idea what the scale of each voxel is, how thick each layer should be, what material etc. There are plans to extend the AMF format to support volumetric models in the future, but the details of how they choose to implement it might make it unusable. I will go into more detail on that in a future post, but for now the best solution running is Unformia’s MeshUp and SymVol. I hope they are successful in bringing this product to market.
You can find the MeshUp Kickstarter page Here.
The ThreeForm Mask designs featured at Maker Faire are now available on Shapeways. There is still time to order for Halloween! I have offered them separately as well as grouped together for wholesale purchase. They are in different ranges of sizes to enable to match a predicted resale demographic (kids/women/men). Each design is available in the ThreeForm Shapeways Shop in packs of 5, 10, 15, or 20, as well as individually. Here is a link to a pack featuring one copy of each design in original scale: http://www.shapeways.com/model/734603/threeform-masks-daredevil-eagle-and-venetian.html
I’ve created a experimental method of generating composite materials with unique properties for 3D printing. This first example is called Macro Foam and you can download the data on thingiverse here: http://www.thingiverse.com/thing:27495
The idea of this material is that it will be very compressible because of the soft interconnections between the cells (shown in green). When the material deforms enough, the solid parts (shown in white) will collide and begin to resist further compression. By varying the density of the cells and the width of the connections, the material properties can be continuously controlled.
It’s no secret that 3D printing holds a lot of promise for improvements in materials and manufacturing. The most often stated advantages are that complexity is free, and unique shapes can be produced with no added cost. When applied to manufacturing, so far this has meant that a part can be made lighter-weight with less material, due to the 3D printing’s ability to easily form truss-like structures and hollows. Complex mechanisms can also be produced such as hinges and wheels, but so far the materials and precision are just not comparable to what we are used to in mass production.
What has not yet been addressed is composite materials. A few processes are capable of forming parts with multiple materials in very fine arrangements, but there are no examples of using the properties of the separate materials in a way that takes advantage of their spatial arrangement. Objet describes their material mixtures as composites, but these materials are homogenous (evenly distributed). In the current generation of printer, two resins and a support material can be applied in a single part. The machine uses dithering, just like a regular home inkjet printer. Objet is claiming that each possible proportional variation of the dithering pattern (4×4 voxels) is a unique material. Multiply that by the number of possible materials, and you get a very large number, so that your marketing materials can state you have 107 materials (and counting). The reality is that these materials behave in a homogenous way. That is, they behave no differently than if you had mixed the resins in the various proportions before putting them in the cartridge. The possibility that is being ignored is that of creating a true composite based on the neighboring relationship of the material elements.
To demonstrate this potential I’ve created an example of a composite material, generated using procedural 3d texture maps from an animation program. This example has not been tested, and would probably take some development to become useful, but the concept is fairly clear. Their are open spaces in the material, and the contacts between the touching materials are constrained and predictable. There is random variation in the finer structure, but overall the material will exhibit a regular behavior that can vary smoothly based on the density of the seed particles that generate the pattern. Simulations will eventually be used to drive the patterning of parts (and even architectural structures) made with 3D printing. This is an example of how the structures will be generated from the simulation results like strain, temperature, etc.
I’ve put together some nice footage of the show. This video and some of my photos have been picked up on Fabbaloo here and here, as well as a variety of other 3D printing blogs. The artist/model I worked with to create the corset design (Amandacera) was in Peru during the show, but amazingly Materialise worked with a modeling agency in Atlanta to find a model that fit the exact measurements of the design! Big thanks to Jamie Milas, Gary Mikola, and Mike Roosa for helping me get my designs into this amazing show, and I also want to thank the model, Stephanie (and her mom), for helping me to present the groundbreaking-but-complicated Seed of Life corset in the best possible light.
I now have a Twitter account for ThreeForm, my 3D printed apparel design company, where you can keep up to date on new developments.
I have a new design I’ve produced just in time for Valentine’s day. I designed this votive candle holder in 3D and have been hand casting and glazing them myself. You can buy them in red glazed porcelain on Etsy, and white glazed earthenware ceramics at Shapeways. Orders placed before Monday can be shipped before Valentine’s day.
I’ve just released a handy tool to help people get the most from their Next Engine 3D scanner. By placing this device on the scanning platform and scanning it along with your object, the scan alignment can be greatly improved. It is for sale on Shapeways.com, and for those who have their own 3D printer with at least a 6 inch build platform, you can download the non-hollowed version from Thingiverse.com and create one for free!
For best results you may want to paint the tracker with a light grey primer or matte paint. This will prevent light from diffusing within the material and increase the quality of the captured data. For prints made on a Fused-deposition Modeler like Rep-Rap, it helps to lightly sand the model before painting.
Simply place the scan tracker on the platform so that it is included is the scan. Do not trim the data out until after the scanning and alignment is complete, but do remove it before building a final mesh. When using manual alignment, be sure to click the correct reference points. The tracker has five-sided radial symmetry that could cause a bad alignment if your initial reference points are incorrect. I recommend using ten scans to build your object. The reference points will assist you in keeping track of your manual rotation as well.
The Cove Candle is an oil warmer with a natural theme. It was partly inspired by the Boboli Gardens in Florence.
The Cove Candle is designed to be made with ceramics or full-color 3D-printing. The ceramic material will hold up to heat well enough that paraffin tea lights can be used. The Full color material will have to use LED-based tea lights. As well as holding three candles, the design also holds six small statues. The ones I’ve designed for sale along with it are stylized mushroom and tree models. Any model with a 18mm base can be used.
The ceramic material is bright white and sealed with glaze. That will allow you to fill it with oil that will be gently warmed by the candles. The color material will not hold liquid at higher temperatures, so it is not suggested to use real candles in the color version. A diffusing reed can be placed in the central hollow, which continues the length of the trunk, down to the base.
3D printed jewelry is rapidly growing in popularity. My new Bottle Drawer gives collectors a suitable place to stash their additively manufactured baubles and other small, precious items. The form of the Bottle Drawer followed a long, circuitous path to it’s final re-birth as a functional piece of art. The shape is derived from a 3D scan of an antique pharmacy bottle. The broken bottle was fixed digitally, then simplified and sculpted into a container of polygonal drawers, further wrapped in a frame-like skin. 14 separate, removable drawers of varying sizes help you sort and store your rings, gemstones, dice or other small items.