The process whereby an idea, or a scan of an object, is transformed first to a 3D software model, then to a printable file (which slices the planned physical object into printable layers), and finally to a finished object is complex: a blend of many skills and experience is necessary for success. In addition, additive fabrication, particularly in metal, requires integration with digital, shop floor, supply chain, and customer-facing processes (particularly in the case of mass customization). What follows is a cursory overview of the processes that can be involved. At the high level, initial choices include
Design rules for the primary structure itself are rapidly evolving. For a given material in a given fabricating machine, a number of considerations must be addressed, and numerous methodologies1 are emerging to do so in a systematic fashion. The following questions are a sample rather than a full methodology:
Apart from specialized cases such as sand castings, ceramics, and now a growing list of composites, 3D-printed material is usually some type of plastic, in both home and industrial applications, or metal. Those accustomed to consumer-grade filament priced at roughly $45/kilogram would be stunned by the costs and capabilities of high-end engineering polymers. To take but one example, polyetheretherketone (PEEK) can be used in environments reaching about 500 degrees Fahrenheit (250 centigrade), is strong enough to be used in pumps and ultra-high vacuum environments, and is durable enough to be used in aerospace, semiconductor fabrication, and medical implantation applications. It was developed in the early 1980s and can cost roughly $1,200 for a 1′′ thick 12′′ × 12′′ slab. Powder is substantially more expensive: a violin 3D printed out of PEEK in the Netherlands was said to cost 20,000 euros.2
The two classes of polymers used in additive technologies can be divided into thermoplastics and thermosets (which somewhat confusingly includes polymers that cure under light rather than heat). Thermoplastics can be melted and solidified, repeatedly, making many plastics recyclable. Traditional injection molding and the most common consumer 3D printers both heat up a solid thermoplastic (in the latter case, often a filament much like those used on string trimmers) to make it malleable, then form it either in layers or molds. In contrast, thermosets do not melt, but are cured with heat, light, and/or a catalyst. (Once exposed to heat, however, thermosets can lose structural integrity.) These polymers are used in everything from bowling balls to laminated countertops, and are not recyclable.
Ceramics, sand, and other materials can be printed as well. Materials can enter additive manufacturing as wire/filament, liquid/slurry/paste, or powder. Given the higher temperatures involved, metal-capable 3D printers are typically beyond the reach of home users given financial, power-supply, and workplace-safety considerations. New filaments are on the market, however, that mix a soft metal such as copper or brass with 11.5 percent polymer, making some desktop printers metal-capable under certain circumstances. Other composites are emerging. Some blend carbon fiber into a polymer filament, providing additional strength (see chapter 3).
One key attribute for additive materials is compatibility with human health. The US FDA has cleared a transparent resin for use in dental appliances such as retainers. PEEK (mentioned above for its high performance and high cost) as well as the titanium alloy Ti-6Al-4V have been approved for implantation. Considerations such as outgassing, toxicity, and the effects of oxidation or photosensitivity are also being assessed for products that could be worn or otherwise used in close proximity to people.
The metallurgy of additive manufacturing is a complex and fascinating field largely beyond our consideration here. It is rare for a metal to be formed of a pure element: even 18 karat gold contains 25 percent of its weight in other elements. Thus, most industrial processes are performed on alloys of multiple metals. For example, there are more than 3,500 types of steel: in its simplest form, iron and less than 1 percent carbon are heated and worked. Titanium is commonly used in additive manufacturing of aircraft parts in a family of alloys beginning with 6 percent aluminum and 4 percent vanadium. The point here is that with thousands of metallic alloys already in use, choosing how, when, and why to convert them to use in additive processes will take years of research, trial, and error. (A case in point: aluminum and titanium powders are pyrophoric, meaning they can be explosive when mishandled, including in post-processing.3 Aluminum and ferric powders trapped in a filter together can be particularly dangerous.4)
To date, most metals used in additive manufacturing were available in powder form for other purposes. Also, most alloys that were used for conventional manufacturing have not been successfully 3D printed, often because they are not readily available in the appropriate power form. It is expected that new alloys will be designed specifically for additive manufacturing, particularly in their resistance to cracking and lower tendency to develop porosity.5
In metal printing, grain behavior is critically important. Grains are multi-atom structures that are formed as molten metal (which has no crystal structure) is cooled, and are a key factor in a metal’s microstructure. Microstructure, in turn, determines most important properties of the metal: thermal and electrical conductivity, strength, elasticity, fatigue behavior, and so on. In traditional castings, outer edges, which cool first, are microscopically different from the interior, which cools last. Thus forging, which is a combination of sub-melting-point heat and physical forces, generally produces stronger steel than casting: the crystal structures that formed as the ingot cooled are strengthened by the forces applied to them in the reshaping process. A casting, in contrast, allows more freedom of shape because, as we have seen, molten metal lacks all structure so it can pour into many desired configurations. (Alloys behave differently from pure metals, complicating the grain issue further.)
The thermal history of a metal piece goes a long way toward determining its crystal structure, and therefore its physical properties. Because the intense point of heat in metal 3D printing behaves very differently from an entire vessel of molten metal, the crystal structures of additively produced metals are different from castings (or forgings for that matter).6 Exactly how, why, and when these differences emerge is still poorly understood.
Recent research suggests that taking advantage of the precise control of the laser or electron beam heat source can create many heretofore unobserved effects in a piece of metal by manipulating these grain structures. For example, the properties of a build can be varied by the local needs of the piece (such as strength, weight, smoothness, porosity, and ductility) without introducing a multi-material aspect to the build: merely managing the heating and cooling processes differently at different points in the build can affect the local performance of the material.7 Other researchers at Lawrence Livermore National Laboratory were able to double and sometimes triple the strength of stainless steel, once again with innovations in the laser heating and controlled cooling process.8 A third team found that doping metal powder with zirconium-based nanoparticles made it possible to print (and probably weld) in previously impossible grades of aluminum alloys.9 The recency, variety, and impact of these developments suggest that metal-based printing could get a lot better very fast.
One key alloy that has been adapted for use in additive manufacturing is Inconel. This class of twenty metals, based on a blend of nickel and chromium, is technically called a superalloy and was developed shortly after World War II for use in jet-engine turbine blades. These components must be light, precisely shaped, strong, and heat-resistant given the intense temperatures and stresses of the operating environment. Failure is expensive and dangerous given that blades from the early stages of the engine can get pulled through the remaining stages; catastrophic failure is common if a single blade breaks. Thus, the emerging industry needed metals that maintained their strength until nearly their melting point. Roughly fifty years later, Inconel was used relatively early in the research phase of what was then called laser manufacturing (circa 1990s), in part because of its importance to the aerospace and defense industries funding the research.
Going forward, it is expected that both metals and polymers will be designed for additive applications rather than being adapted from other, older fabrication methods. Experience with and further research into the many available materials will also continue to improve both the ability of additive techniques to supersede formative or subtractive manufacturing and define the cost/benefit tradeoff frontier.
Many physical builds have been discarded as software glitches are discovered and fixed. A new class of build management software that begins by checking and repairing the file that will drive the printer has emerged, exemplified by Netfabb from Autodesk and Magics from Materialise. These packages, along with other programs, can also generate support structure sizing and placement. (See below.)
Even though 3D modeling tools have been used for more than twenty years, 2D technical documents, often on paper, are still common. On many of these, measurements and other information taken off the 3D “master” represent approximations. Additive manufacturing is emerging in the workplace at the same time that new software packages, file formats, and process definitions attempt to break that 3D to 2D handoff with something called model-based engineering; instead, tolerances, dimensions, surface treatments, and other data all travel with the original design through the production workflow, and indeed the product lifecycle. As of 2018, technical workshops, new product launches, and company startups in this vein are widespread. The connections to additive manufacturing are obvious: if digital production of original 3D CAD files helps maintain the “digital thread” from original design all the way through spare parts replenishment, clerical errors should decline, process speeds accelerate, and quality improve.10
In addition to creating a 3D model in software as a natively digital design, it is also possible to scan a physical object, sometimes as part of reverse-engineering a spare part: the engineer may lack a CAD file, or even a blueprint, and so must specify a build to match an existing, possibly broken, component. Many tools can facilitate this process, from contact scanners that physically probe the object to be reproduced (at the risk of damaging fragile items such as historical artifacts), to CT and MRI scanners for body organs and structures, to many uses of light and lasers, some of them suitable for home use. Applications range from civil engineering (site modeling and analysis) and architectural modeling to crime-scene analysis to reverse engineering of existing parts to reproducing cultural artifacts to dental fixtures. 3D scans often replace molds or artisanal cut-and-try methods of production. These technologies also figure in the post-processing phase as they test the dimensionality of the build against the original specifications.
The STL data format (standing for stereolithography) was created in 1989 by 3D Systems and thus is quite dated relative to today’s computing, software, and imaging benchmarks. For example, although STL is bulky it does not include certain necessary information. Thus, different software packages must interpolate (that is, guess) how geometric triangles are connected rather than read this information from the file. This interpolation makes STL slow to process, and the guesses can be wrong.11 STL also does not support many new features of printers that are more modern than those available thirty years ago, and basic information relating to colors, materials, physical orientation of the build, and so forth is often not captured. OBJ (object) files, more commonly found in 3D graphics workflows such as video games, are also sometimes used in place of STL.
Two new additive manufacturing data standards have emerged over the past decade.12 The additive manufacturing format (AMF) came out of an ASTM (American Society for Testing and Materials) committee tasked with replacing the STL format. The first iteration of AMF was made public in 2011 and an improved version was approved by the International Organization for Standardization (ISO) in 2013.
More recently, Microsoft included support for 3D printing in Windows 8.1 and the then-upcoming Windows 10 in yet another format; the 3MF consortium was announced in 2015. Functionally, it is similar to AMF: both are based on the XML standard and are human-readable with regard to materials, tolerances, and other attributes. Microsoft started with a model of a paper print queue, so the interface design is similar: choose a file, select a printer from a list, choose from available options (high vs low resolution, color vs black and white, material choices), and push “print.”
A major question for the industry going forward relates to digital rights management (DRM). Much as Hollywood has engineered copy protection into Blu-ray and 4K video formats, holders of intellectual property rights for physical items may seek to enforce those rights in software. Just as a color copier cannot make super-accurate copies of currency without identifying the device, will 3D printers lock out certain classes of files, designs, or materials? Both AMF and 3MF include support for metadata including digital signatures; watermarking (to identify proprietary content in both software and the build) is included in the roadmap of future features to be added in AMF.
After the CAD file has been converted to STL, AMF, or 3MF, the software representation of the desired part to be built is often converted to G-code, or a machine-specific variant of G-code, which derives from the world of computer-controlled machine tools. (Some technologies use a different software translation to drive the printer head.) This language tells the laser, build platform, and other components how to operate to build the shape that has been specified.
All of these software steps, along with close monitoring of the actual build, can generate significant data volumes. According to one research paper,13 data was generated at eleven steps:
The author later said it was easily possible to generate a terabyte of data per additively manufactured part across forty different file formats. Information management is clearly an emerging area of concern as the technology goes mainstream.
While some characterizations of the technology might suggest that 3D printing is a matter of designing a part, choosing a material, getting access to a machine, and pushing a button, its reality is more complex. Post-processing, sometimes extensive, may be required, as we will see below. Primarily in metal-melting scenarios but also in plastic, intense heat can cause unintended things to happen to previously built layers, so care must be taken in planning and executing the build.
In the design phase, engineers have many details to attend to, many of them related to managing the effects of heat. For example, anyone familiar with welding will recognize the “potato chip” curling of sheet steel that is overheated. Given how much the physics and chemistry of additive manufacturing in metal share with welding, many concerns carry over. Getting the design axis and build plan correct usually involves a steep learning curve, particularly when a cluster of small and possibly unrelated parts are pooled into a single build.
Because of the layer-by-layer process of building up a shape, support structures are particularly important. In regard to gravity, certain geometries must have a form of scaffolding or other splinting to be able to “grow” without sagging or becoming stressed before a joint is sufficiently strong. The capital letter Y, if 3D printed in polymers, generally does not require support for the two upper forks, while a capital T and most capital Hs do require support structures under the horizontal spans. These are removed once the output is cured.14 In addition to gravity, support structures help manage heat, often serving to alleviate residual stress (sometimes by serving as a heat sink in metal builds, for example).
The use of support structures has evolved over the past twenty years or so, leading to a set of design principles for practitioners to follow. How much production time do the support structures require? How much powder or filament must be budgeted to build nonfunctional aspects that will be milled or dissolved away? How will the support structures be designed to be removed, minimizing surface damage and possible structural weakening? Many of these design rules are instantiated in software (such as Magics and Netfabb mentioned above) but opportunities remain for the engineer or machine operator to tweak the generic parameters.15
In the build plan, designers must consider that material strength in the z axis (the height created through adding layers of material) is generally weaker than in x or y. Sometimes items are built at a 45-degree angle to reduce the vulnerability to failure across this axis. At the same time, the support structures needed to position the build can be extensive, and removing them can leave surface deformities. Also, the support structures that are discarded will increase the cost of powder material consumed: with some shapes, more support material might be used than the section they are supporting. Given that they are often algorithmically generated, support structures can take on complex and aesthetically pleasing designs (see figure 2.1). Build orientation also affects surface quality in the form of the staircase effect resulting from layers being successively added. In the end, the build orientation choice must balance material strength, surface quality, cost, build time (more layers generally take longer), and other considerations.
After the design stage, machine operators, sometimes aided by cameras and other sensors in the build chamber, need to watch for several types of build failures. The technology for these sensors is rapidly improving, but the technical demands are substantial: build chambers can get extremely hot, the energy levels at the point of metal fusion can be intense, and the data must be captured at a high frequency. Getting data-savvy talent to analyze these sensor feeds is a further issue, given the vast shortage of capable people to handle “big data” in everything from biostatistics to advertising placement.
Build errors are more common in metal, but not unique to that class of materials. Because knowledge of how to avoid these errors is hard earned and a competitive advantage, little of it is being shared, which is slowing the pace of overall progress.16 Many parameters must be set and monitored during production. Some metal-based additive technologies occur in an environment of inert gas for both safety and quality reasons. Getting all the settings right involves considerable trial and error: for example, according to Yang and colleagues, “The optimum process parameters for an Inconel 718 turbine blade fabricated on an EOS M270 platform may not result in best fabrication qualities in a newer EOS M290 platform due to the improved inert gas flow control in the later system even though both are developed by the same manufacturer.”17 The angles of the printer head and/or the laser are sometimes adjustable, as is the tracing pattern of the print head over the build.
Specific design decisions and settings during the build can affect many aspects of the final part.
Surface smoothness can come from several sources. In directed energy deposition (DED), parts are overbuilt and then milled down to spec. In powder-bed methods, smaller sizes of metal grains and the lower thickness of the individual build layers contribute to surface smoothness, but both increase build time and cost. This balancing act is less critical than others because post-processing is generally required and can smooth most 3D printed surfaces relatively easily.
Many variables affect the presence or lack of cavities in the desired solid metal structure. Gas pockets in the powder feedstock can cause voids in the build, though this is relatively rare. Metal powder also needs to be sufficiently fine for the necessary precision of the feature being built and packed to the proper density in the chamber. If the laser is not sufficiently powerful, the metal may not fuse, much like a cold solder joint. In the other direction, too much power in the beam can spatter molten metal into adjacent areas. Labs and production facilities are learning a lot about maintaining adequate material density, including the use of hot isostatic pressing in post-processing, shaping the laser, varying the powder grain size, and so on. Especially in parts subject to fatigue from cyclic (rather than static) loads, adequate material density is an absolute necessity.18
As materials heat and cool, they expand and contract. The stresses of these cycles can exceed the tensile strength of the build, resulting in a variety of defects. Sometimes support structures are included to account for these stresses, but those then need to be removed in post-processing, and the removal comes with its own risks of damaging the part. Pre-processing software such as Magics (mentioned above) can help avoid some of these stresses, as can various strategies for planning laser scans during the fusing of each layer. Building a mirror image of the shape being built on the opposite axis can equalize stresses. In post-processing, heat-treating can also play a role.
Absent these countermeasures, metal builds can crack (especially in the early stages of getting accustomed to a new alloy), warp, and/or delaminate. Another set of constraints relates to strength versus surface smoothness: in tests involving FDM, lower extruder temperatures minimized surface roughness but also lowered the interlayer bond strength and thus the overall performance of the completed part. The conclusions pointed to a complex set of trade-offs: for every set of parameters that was optimized for one property, other properties typically suffered. Build speed, surface quality, material cost, overall part strength, particular feature strength (overhangs, curved surfaces, joints)—none of these comes without a cost somewhere else.19
The long-term behavior of additively produced metal is not yet well documented or understood. These metals’ grain structure, material consistency (that is, lack of unmelted powder and impurities), and overall density (lack of pores and other voids) cannot always be determined through nondestructive means. Furthermore, and more important, it is not yet known how metal fatigue will emerge after multiple cycles of mechanical loading. These cycles could be measured in hundreds of revolutions per minute or seasonally, as in a bridge beam. Given that what is known as high-cycle fatigue occurs after millions or possibly hundreds of millions of cycles, the current base of experience is limited. Because the thermal history of an additively manufactured part is so different from that of a conventionally machined item, attention is now being paid to the build process (for instance, the rate of cooling) with an eye toward improving long-term durability and, more important, predictability with regard to failure.20
Very rarely does a 3D-printed part come out of the printer ready for installation. The entire field of post-processing is evolving to include many techniques, ranging from compressed-air removal of powder to heat-treatment, such that the design of the part now potentially includes many steps after the item leaves the printer. Such a messy reality is at odds with the “printing” metaphor in which few of us have to do anything to the paper after it gets ink or toner applied to it in a home or office.
In fact, the practice of post-processing is very much in keeping with the workings of a factory or machine shop. Anything done to a printed part—curing, polishing, milling—is done to conventionally manufactured parts. Plastics can get coated in metal, metals get coated in polymers. Mounting holes are drilled, screw threads tapped, metals strengthened through various processes. The fact that post-processing is getting the same attention in additive manufacturing as it does elsewhere in the industrial facility is a signal that additive technologies are being fully integrated into the workflow of designers and engineers.
Machine makers are responding as well. Beginning in 2013, a generation of hybrid additive + subtractive machine tools has come to market from various vendors including Germany’s Hamuel, Mazak from the United States, and Germany’s ELB. The German-Japanese DMG Mori sells a six-function integrated machine for high-speed milling + 3D scanning + laser cladding [DED] + 3D inspection + deburring/polishing + laser marking. A key player is Hybrid Manufacturing Technologies, based in the United States, which partners with machine-tool manufacturers to implement its Ambit system of additive, subtractive, and inspection tool heads on traditional CNC equipment. The integration of additive and subtractive capabilities, along with measurement and validation, in one physical unit opens the way toward more complex process steps and newly available properties and geometries of parts.
Greg Morris, now a key figure at GE Additive, points out that because post-processing can be even more critical to the performance of a 3D-printed part than the actual build is, engineers are learning to design for an entire production path in which the additive build is but one step. In regard to metal parts, he lists a whole sequence of considerations.21
In polymers, post-processing works slightly differently. Support structures are more typically supporting horizontal elements such as overhangs than providing thermal relief as in metals. In both cases, removing support structures usually damages cosmetic aspects of the build. Joining two plastic pieces into a structure too large for a build chamber is accomplished with a cold weld: acetone can fuse ABS plastic parts together. Abrasives and fillers are used in metals as well as polymers, as are polishing, plating, and/or painting. Polymers can be smoothed in a chamber filled with acetone vapor: the resulting surface can look polished.
All of these powders, filaments, and fabrication machines will stand idle without an adequately trained workforce. Although efforts are underway, there are shortfalls all though the product life cycle: machine manufacturers need knowledgeable sales forces; production firms need designers, machine operators, and inspectors; service bureaus need informed clients to request their services. The breadth and depth of knowledge required to make 3D printing profitable is one reason for the sometimes slow adoption. We will return to this issue in chapter 6.
As the discussion of post-processing illustrates, additive techniques rarely stand alone in a modern factory, but designing workflows around the design capabilities, cycle times, and skill requirements of a 3D printer is still in its early stages. One key component of this exercise is cost optimization: knowing the break-even point where conventional manufacturing methods take over for additive prototype-like economics is often deceptively difficult. New project software features and the growing base of experience with 3D printing both help address this shortfall, but it will be many years until additive manufacturing is routinely specified and utilized with clear visibility into all of the associated constraints, costs, benefits, and risks.
We have seen how a variety of software applications, design and manufacturing expertise, material handling, capital investment, and sometimes significant operating expenses all come together to make an additively manufactured build. Nearly all of these contributing elements are in the early stages of the learning curve, with much still to be discovered and optimized. It is now time to catalog, at a high level, the basic types of 3D printing technologies that have emerged over the past thirty-plus years.