New dimensions in printingTweet
3D printers are becoming more affordable and are breaking into new application areas. However, as Paul Schreier reports, we’re not yet at the point where files are as easy to print as with document printers – software for fixing and preparing files for rapid prototyping is still required
In the process known as rapid prototyping, also known also as additive fabrication or rapid manufacturing, users start with a 3D image on a screen and ‘print’ the object from materials such as plastic or metal. One obvious application is to make a prototype of a new part before going into production, but also imagine being able to make replacement parts for a pump that went out of production 50 years ago, using old CAD data. In medicine, an audiologist can ‘print’ a hearing aid customised for each patient, or a physician can create a physical representation of a tumour for examination prior to an operation. Architects are printing models of buildings instead of creating them with wood and paper. And, as prices drop and capabilities increase, new applications are popping up almost daily.
Reebok uses software supplied with 3D printers from Z-Corp to fix a CAD file before creating a prototype to examine during the product-development cycle.
Popularly known as 3D printers, these machines have become affordable with some models selling for little more than $10,000. In some cases, operating one can be as simple as running a document printer: simply hit the Print button. In other cases, though, the CAD data needs some modification to make it suitable for 3D printing. What if the CAD object has poorly located holes or incomplete contours? What if the object won’t fit within the printer’s build envelope? What if you want to add a texture or colour to certain areas? A number of software packages can help solve all these problems.
Understand the mechanics
To understand software requirements, you must first be familiar with some of the mechanics of 3D printing. These units create objects by setting down repeated thin layers of material, whether plastic or metal. The de-facto standard for sending data to 3D printers is with an STL file, which is a polygonal mesh format that approximates surfaces with tiny triangles. Users can export object designs from almost every 3D CAD package in STL format.
A Materialise screenshot shows an architectural drawing with too much detail (left) and the house section after shrink-wrapping (right).
The STL file must describe an object such that it is ‘watertight’ – that is, it can have no gaps, holes or missing surfaces. ‘The STL format is anything but perfect,’ says Terry Wohlers of Wohlers Associates, a company that closely tracks the rapid prototyping business. ‘Years ago, more than 50 per cent of the time users would have to do lots of repair. Things have improved, but even today the source data could still need work.’ About 95 per cent of the STL files that major CAD packages produce are 3D print ready, states Ismo Makela, managing director at DeskArtes, a supplier of 3D manipulation software. Katrien Lenaerts, product manager at Materialise, another supplier of such software, believes that most files still have some details that must be adjusted before the building process, so most still need minor fixing.
For almost every design, it makes sense to send a target STL file through a utility that verifies the object can indeed be built. The verified STL file then goes to a printer driver, which slices the object representation into layers, sets printer parameters and transfers the file. The larger 3D printer companies have developed utilities that add a ‘3D Print’ button to the menu bar of popular CAD packages to make this step almost transparent. As a rule of thumb, says Roger Kelesoglu, director of customer development at printer manufacturer Z-Corp, a 3D printer can create from 2 to 5 vertical centimetres per hour, and the cost of the material for creating the object runs in the order of $2 per cubic inch.
3Data Expert from DeskArtes provides a feature that allows users to split large objects into parts.
When problems arise
As noted earlier, STL files often need some minor tweaking, and the amount of work could be even greater for files generated with older versions of CAD software or files that come from specialised software for industries, such as architecture or digital content creation.
A mechanical CAD design is not intended to be a watertight model, explains Materialise’s Lenaerts, and the process of creating an STL file can cause bad edges (triangles not well connected to each other), double surfaces or identical triangles. Additionally, the triangles in an STL description have a normal and thus an inside and an outside. A CAD model doesn’t contain this information, so in the STL conversion triangle directions can be random, resulting in flipped triangles. For fixing these problems, STL repair software comes to the rescue. Popular products include Magics from Materialise Industrial Software, 3Data Expert from DeskArtes (and note that Z-Corp sells a version called Z-Edit Pro optimised for its printers), and Geomagic Studio from Geomagic. These packages check that there are no gaps or bad edges, that the surface cross-section has a closed contour, and that all surface normals face outwards. Either automatically or under user control, the software can invert flipped triangles and shells, stitch bad edges and smoothly fill complex holes.
Besides fixing such problems in CAD files, this class of software handles other tasks specific to 3D printing. For instance, Z-Edit has a Zcompensation feature that accounts for the overcure that causes extra material to build up on down-facing surfaces to ensure a part has the desired measurements.
3Data Expert from DeskArtes provides a feature that allows users to split large objects into parts, each one with mating features that allow precision assembly after individual parts are fabricated.
What if the object you want to fabricate doesn’t physically fit in the printer’s build envelope? Then it becomes necessary to split that object into smaller parts and assemble them afterwards. In the past, the most straightforward approach was to make straight cuts, but when joining the finished parts together it can be difficult to bond them while maintaining accuracy. Some software can create interlocking mating features to ease precision post-assembly. For example, Materialise reports that, with this technique, one user built and bonded an automobile bumper that measured 1.82m long, but was able to hold the dimensional tolerance to within 0.381mm over the part’s entire length.
The opposite situation might also arise, where you want to print multiple small parts all at once on the build platform to save time. Some packages provide utilities that help users place these objects for the best results.
For many applications, the object being printed might not have to be solid, but the CAD system creates an STL file that makes it look that way. Thus file-editing software can allow users to generate hollow parts to cut material costs and speed printing. In addition, hollowed parts can be more accurate because less internal stress is created. On the other hand, a CAD program might create a part where the material in some areas is too thick to be built properly or too thin to handle stresses. Here wall-thickness analysis helps identify such problem areas prior to printing.
Adding colour and texture
STL files describe a series of triangles, but they don’t contain colour or texture information. However, that’s starting to become a feature of 3D printers that makes them attractive in new application areas. The first to market with a colour 3D printer is Z-Corp, but that technology brings challenges with it. Explains Z-Corp’s Kelesoglu: ‘The STL format hasn’t been updated to what modern 3D colour printers require, and it is also missing obvious things such as scaling or units.’ The fact that STL files are colour-blind is not a major concern for most 3D printers, which are being used for fitting applications or patterns for moulds. But as systems mature, more people will want multicolour designs.
To address this need, Z-Corp’s printers read VRML files that do carry colour and texture information. VRML is just one of the newer file formats used to create objects in 3D printers, and among them are 3DS, PLY as well as proprietary formats developed by printer manufacturers. Meanwhile, 3D print-preparation software can work with a wide variety of formats so that, according to Kelesoglu, the choice of file format is largely irrelevant as long as it can encapsulate the geometry and include colour and texture maps.
A screenshot from Materialise (see above) shows an architectural drawing with too much detail (left) and the house section after shrinkwrapping (right).
More of an issue is whether the model is solid. He points to software used to design Hollywood sets where the buildings on a street have no backs; a 3D printer needs solid objects, not just surfaces, so that object must be corrected. Kelesoglu continues, ‘We can take beautiful 3D renderings of buildings and turn them into actual models. Some years ago, architects would create physical models with wood or paper, and today with 3D printing they can reconnect with that tradition of physical models.’ However, adds Materialise’s Lenaerts, in order to print an architectural file on a 3D printer, the model must be downscaled, which means that windows and walls might become too small to be buildable. Also, during the design process, the architect draws each wall, window and other features as a separate block, resulting in a STL file that is far from watertight and cannot be printed. For such cases, preparation packages often provide a shrinkwrap tool to fix these complicated models.
Similar issues arise with software for digitalcontent creation or for video games where users might want to print a small statue of a virtual character. Such software uses surface representations rather than solids, and files created with surface-based geometry software generally need repair. Thus many packages import IGES files with a surface representation and create the triangles for the STL format.
Medical Modeling’s study of a tumour in the palate started with CT data (above left), converted it to an STL file (above right) and then fabricated a model (above), so physicians could examine their task before going into the operating room.
Other sources for data
Besides taking data from various design packages, users of 3D printers are working with other sources of data. For instance, they are using laser scanners and digitisers to create models of existing objects. After scanning an object with an optical scanner the result is a point cloud – a collection of points describing the surface. A process called meshing then connects these points to create triangles. There are many sorts of optical scanners, and most come with software that can convert the scan data to STL format.
To study a tumour in the palate, Medical Modeling started with CT data (above left), converted it to an STL file (above right) and then fabricated a model (above) so physicians could examine their task before going into the operating room.
The medical profession is also showing an interest in 3D printing. Health-care organisations increasingly rely on 3D anatomical models for pre-operative planning, specialist consultation, implant fit and design, patient counselling and medical education. Andy Christensen, president of the service bureau Medical Modeling, explains that MRI and CT data today is generated in the DICOM (Digital Imaging and Communications in Medicine) format, which is similar in concept to a JPG or TIF file. The information comes as a series of slices, but they are not suitable for accurate 3D printing, so he relies on Mimics software from Materialise to convert DICOM images into STL format. His firm specialises in CT images for working on bone material; a CT image has roughly 4,000 grey levels, and on each slice the software must determine what is bone and ‘throw away’ everything else.