1.

Introduction

Additive manufacturing (AM) fabricates the desired final part in a single step directly from the input digital computer-aided design (CAD) file by depositing and fusing layer upon layer of the source material with very little waste. An increasingly popular term for AM is “3-D printing.” In contrast, for thousands of years, traditional machining (referred to as subtractive manufacturing) removes the material in several steps, which is often laborious and can generate large amounts of scrap chips and turnings. For the aerospace industry (which typically must rely upon costly exotic alloys), AM offers the key advantage of the “fly-to-buy ratio,” which is the amount of raw material used to manufacture the desired finished part with the associated scrap. As a compelling example, each F-22 fighter plane started with 50 tons of expensive Ti-6Al-4V that were machined through conventional metal cutting to a net of 5 tons of final parts with 45 tons of waste chips,¹ giving a “fly-to-buy” ratio of 10%.

An important hallmark for AM is that it can make virtually any arbitrary shape or topology without the great cost, time, and penalty to make a complicated part by subtractive manufacturing. For AM, complexity is free because production cost is set primarily by the part’s simple mass, which controls the production time much more strongly than the part’s complexity. A complex part with intricate internal features may cost no more through AM than a simple part of similar mass. For AM, design engineering is liberated, and it becomes possible to design for build rather than design for manufacturing under the rule that “what you want is what you build” for AM, rather than “what you build is what you get” for traditional manufacturing.² AM, therefore, allows shapes and topologies not possible with traditional manufacturing.

AM is an example of a near net shape process in which the initially produced (gross) item is very close in dimensions to the final (net) shape. Other traditional near net shape processes (such as injection molding and die casting) lack the flexibility of AM and traditionally require machining a custom mold to make a specific one-of-kind part, whereas AM can make any arbitrary part unconstrained by any need for a pre-existing mold. Interestingly, one important application for AM is to actually make the intricate molds and dies for subsequent high-volume molding and casting because AM eliminates the great expense and long lead time of a highly skilled tool and die maker.²

2.

Advantages of Additive Manufacturing

Traditional manufacturing specialized for high production rates (such as casting and molding) may always be more economical than AM. AM can be cheaper and faster for complex low volume parts that otherwise require intricate machining. AM may be particularly well suited for the space industry’s low production volumes of satellites and launch vehicles. Elsewhere in the aerospace industry, General Electric is committed to implement AM for an annual production of 40,000 jet engine T-25 fuel nozzles because AM offers the potential for dramatic cost savings, which illustrates a cost-effective strategy at surprisingly high production rates.

It is currently common for design engineers to first use AM to make simple clones of existing parts. However, there is a revolutionary potential to replace several existing parts with a single optimal part with topology, which offers the prospect of cost savings in production and assembly. By reducing part count, there is a corollary revolutionary potential to eliminate the welds, bonds, and joints otherwise ordinarily needed with traditional manufacturing. In the case of the fuel nozzle, the cost savings are obtained by reducing the part count from 21 (for conventional manufacturing) to 1 (for AM) and eliminating the associated welding operations, along with the associated touch labor.^2,3

As an added benefit, a law of failure analysis (known as Pelloux’s law) states that “It always breaks where it’s welded.”⁴ By eliminating welds and joints, it will become possible to eliminate an entire large class of failure mechanisms, which will improve reliability. Eliminating welds will also eliminate the expense of their specialized nondestructive evaluation (NDE).

3.

AM for Metals: Key Concepts of the Production Process

The different AM production processes for metals are categorized according to (first) the material feed type and (second) the heat source used for fusion. Broadly, the two material feeds are either powder bed or beam deposition, and the three heat sources are either laser beam, electron beam, or (less commonly) plasma arc.

A traditional metal alloy with good weldability or good castability is a candidate for AM, and alloys that crack under high solidification rates are not good candidates.² Common AM metals include Ti alloys (such as Ti-6Al-4V), Al alloys (such as AlSi10Mg), Ni superalloys (such as Inconel 625 and Inconel 718), hot work tool steels (such as H13), stainless steels (such as 17-4 PH), and refractories (such as CoCr).⁵

3.1.

Key Concept of the Powder Bed Process

In the powder bed technique, metal powder with a typical mean diameter of $50\mu \mathrm{m}$ is first spread to give a uniform layer with a typical thickness of $100\mu \mathrm{m}$. Second, the heat source traces out and fuses the shape of the desired part in the layer. Figure 1 shows the details of the powder bed process. On the left-hand side, the piston in the powder supply is raised by the desired layer thickness (say, $100\mu \mathrm{m}$). A roller or doctor blade then uniformly spreads the powder across the powder bed on the right-hand side. After the surface layer is fused to the solid part that is being formed beneath it, the piston in the powder bed lowers the part by the desired layer thickness. The process is then repeated from left to right by raising and lowering the two pistons in turn until the part is finished. In this way, a part height of 50 cm is built successively from thousands of individual layers.

Fig. 1

Illustration of the powder bed process with the powder supply on the left-hand side and the powder bed on the right-hand side.

Europe leads in the powder bed process. Prominent examples of the powder bed with laser heating are EOS (Germany), Concept Laser (Germany) (Concept Laser was acquired by GE in 2016), Phenix (France) [Phenix was acquired by 3D Systems (South Carolina) in 2013], and Renishaw (Great Britain). The powder bed with electron beam heating is manufactured only by Arcam (Sweden) (Arcam was acquired by GE in 2017). The laser beam is typically steered by galvanometer mirror scanners and the electron beam by electromagnetic deflection coils. Scan speeds are very fast for the electromagnetic coils for the electron beam but are limited by the galvanometer’s inertia for the laser beam. Representative scan speeds are $1000\mathrm{m}/\mathrm{s}$ for electron beam heating⁶ and $\text{7\hspace{0.17em}\hspace{0.17em}}\mathrm{m}/\mathrm{s}$ for laser beam heating.⁷ Consequently, deposition rates are faster for electron beam than for laser beam heating. As another difference, surface finish and feature resolution are generally worse for electron beam than for laser beam heating.⁵ Because the electron beam is more efficient heating, its energy costs are lower than for laser beam heating.

4.

Variability of AM for Metals

Both the complex physics and the transient nature of heating lead to variations in fusion during the build, which are responsible for part variability. In addition, the leading powder bed production machines currently operate in simple open-loop control without a process sensor rather than closed-loop feedback control with a process sensor. Open-loop control of a transient process is naturally very difficult,¹⁰ and poor control of the fusion process is ultimately responsible for part variability.¹¹ As a result, it is widely recognized that variability is displayed by a single production machine,¹² across similar production machines (such as two identical machines from the same manufacturer)¹³ and across different types of production machines (such as laser beam-based and electron beam-based machines). Interestingly, two consecutive runs on the same production machine will display build-to-build variability.¹⁴

The research frontier for the field is the development of process sensors with feedback control to make AM repeatable and consistent. The sensors will measure the temperature, diameter, and temperature gradient of the melt pool. By controlling the melt pool directly, the fusion process can be controlled in turn. Optomec and POM actually pioneered closed-loop feedback control with a process sensor in the early 2000’s for the beam deposition processes, and other manufacturers are now following for the powder bed processes. The development of feedback control for AM is an active area of research at original equipment manufacturers, contractors, universities, and national laboratories.

5.

Mission Assurance Considerations for AM

AM is based upon complex physical phenomena that are currently poorly controlled by the common open-loop architecture of production machines, which leads to part variability. As a result, AM has new flaws that differ from traditional parts, such as:

As with casting and welding, the melting and rapid cooling can lead to new residual stresses unseen in conventional machining of wrought stock. The residual stress can warp the part, and the tensile residual stress can inadvertently promote the part’s failure during assembly or service. Furthermore, the porous sintered skin presents a risk of foreign object debris (FOD), and the rougher surface finish worsens the AM part’s fatigue properties.¹⁵ All of these factors (flaws, residual stress, and surface finish) are the sources of the variability observed in AM.

At this early stage in its development, AM has foundational knowledge gaps, such as:

The inspection needs are not yet fully defined because the types of defects possible in the AM process are not fully understood and the possible failure modes during service are not known. A catalog of AM defects is needed for both inherent and rogue defect populations. Inspection standards, inspection capabilities, and acceptance criteria are not mature, although a draft ASTM standard is being developed.¹⁶ The largest risks to mission assurance are the defects’ impact on material behavior.

The powder bed process only makes sense economically if the source powder is recycled and reused. Part variability due to powder reuse is not known and requires further investigation before defining limits and specifications.¹⁴ During reuse, the powder can inadvertently pick up unwanted oxygen or nitrogen or volatize alloying agents (such as the Al in Ti-6Al-4V), which will alter the alloy chemistry. For Ti-6Al-4V powder, the oxygen content increased from 0.08 to 0.17 wt. % and the Al content decreased from 6.47 to 6.36 wt. % after 11 reuses. In the printed Ti-6Al-4V samples, the Al content was 0.3 to 0.5 wt. % lower than the run’s source powder because Al also volatizes during the high temperature fusion process.¹⁷ The reused powder can also lose the small diameter fine particles and accumulate large diameter agglomerates.

The manufacturing instructions for the AM production machine, which is known as computer-aided manufacturing (CAM), are generated from the CAD model of the AM part, and the translation from CAD to CAM requires additional guidance. Historically, industry used two-dimensional print checkers, but no corresponding role exists for digital CAD files. An example of a printing error is incompatible surfaces constructed from the spline representation of CAD surfaces.

Identified risks for AM include:

Modern metallurgy is guided by the fundamental principle that processing controls microstructure, which in turn controls the metal’s properties (whether mechanical, electrical, thermal, or magnetic). As an example, heat treating of metals offers very fine control of both microstructure and properties and was developed in practice thousands of years before any theory of atoms and microstructure.

The heat treating practices for AM metals have not been established because the relationship between processing, microstructure, and properties is still being developed. AM metals may require new heat treating practices. One example is an Al-10Si-0.5Mg aluminum alloy used for AM that is based upon a traditional casting alloy. Although the heat treating practice for the very similar casting alloy has been published, the AM aluminum does not respond the same way to the standard T6 temper. Research is on-going to develop the modified heat treating practices for this AM aluminum.¹⁸

A leading approach to discover the relationship between processing, microstructure, and properties for AM metals is the process map (sometimes termed solidification map).² Physics-based finite element analysis of the AM fusion process can predict the temperatures and cooling rates of the melt pool for a set of process parameters (such as, laser power, scan speed, and scan spacing). The heat transfer analysis uses the nonlinear thermal properties of the metal (such as, thermal conductivity, diffusivity, specific heat, and heat of fusion) and the inert environment, which provides either a radiative (in the case of vacuum) or convective boundary condition (in the case of inert gas). The process map approach is potentially faster and more productive than the trial-and-error approach otherwise used to select the process parameters.

The microstructural development and evolution during solidification can be calculated by using independently characterized metallurgical thermodynamics and kinetics. The process map allows the selection of process parameters that yield a desired microstructure, such as columnar, equiaxed, or even single crystal, and also allows tailoring of the grain size. In addition, the physics-based analysis can predict the residual stress upon cool down and offer processing strategies to reduce or eliminate the residual stress.¹¹

6.

AM Part Design

The part should be designed to take advantage of AM, which should not be used if the design is easy to manufacture conventionally. The AM part should offer cost and/or schedule savings compared to conventional manufacturing for the production run. For powder bed AM, print time is slow. For beam deposition AM, dimensional tolerance is coarse. For all AM, surface finishing is expensive. The part must fit in the production machine’s build volume. Material properties are needed for relevant environments, including static, dynamic, acoustic, and fatigue loading, which is an added expense and need for schedule.

Good AM candidate designs have weight reduction holes and pockets that otherwise require machining in a conventional part. In addition, they have reduced part count, weld count, and touch labor, or are not possible with other conventional fabrication methods. A complex part may not be attractive if it requires the same number of setups to machine critical features as a conventional part. An excellent candidate for AM requires less machining.

The AM design controls the part and includes:

For the space industry, the requirements for AM can be tailored based upon risk and desired reliability. Primary load bearing structure for which failure is catastrophic to mission or human life would receive the most stringent requirements and would require the highest structural margins. Redundant structure whose failure can be tolerated or secondary structure would have lower requirements and margins. In addition, the nature of the AM part can be judged to be high or low risk depending upon its intrinsic characteristics, such as build complexity and access for inspectability after build, both of which are potentially very different from traditionally manufactured parts.

7.

Tailored Approach to Qualify AM for the Space Industry

Figure 2 is a proposed top-level flow diagram for the development of AM for nonmanned space programs.¹⁹ Preliminary part design starts after the flow-down and derivation of design requirements, which arise from the program’s mission risk class (A, B, C, or D). Mission risk class A is an extremely critical mission with minimally practical risk and an operational payload. Mission risk class B is a critical mission with low risk and an operational or demonstrator payload. Mission risk C is not critical and accepts moderate risk with an exploratory or experimental payload. Mission risk class D is not critical and accepts higher risk with an experimental payload.

Fig. 2

AM part development program. Boxes filled in blue and numbered 1 to 6 indicate areas with specific requirements and specifications flowing down from the part’s classification as AM categories I, II, III, or IV.

The mission risk class and the AM process’ manufacturing readiness level (MRL) and rigor of the material properties together lead to the classification of the part as AM category I, II, III, or IV in block 1 of the flow diagram. AM is ranked from most mature, category 1, to least mature, category 4. Table 1 shows how the mission risk class, MRL, and rigor of material properties define the AM categories. MRL 4-6 is the capability to produce a prototype in a laboratory or production environment. MRL 7-8 is the capability to produce a system in a production or pilot line environment. MRL 9-10 is the capability to begin or demonstrate full rate production. Material properties with high rigor satisfy the requirements of the metallic materials properties development and standardization²⁰ (MMPDS) and are based upon extensive testing of the full suite of material properties with a sound statistical basis. Material properties with low rigor do not satisfy the MMPDS approach.

Table 1

Definition and usage guidance for AM categories I, II, III, and IV.

In Table 1, the loss of a mission critical part leads to direct loss of the primary mission because the part is a single point failure. The loss of a nonmission critical part does not lead to loss of the mission, either because the part has redundancy or represents a secondary function. Proof testing is the mitigation for low manufacturing maturity and to demonstrate that design allowables are met for material properties with low rigor. Once an AM part category is defined, an assessment should be performed to quantify the risk of AM part use in the mission.

Blocks 2 to 6 of the flow diagram define, use, and verify the associated AM requirements. The part’s AM category carries with it the corresponding AM requirements (block 2) and specifications (block 3). The requirements and specifications are tailored to satisfy specific program requirements, and compliance is tracked. Table 2 lists the structural requirements for National Security Space (NSS) missions by mission risk class.²¹

Table 2

Manufacturing, process, quality assurance, and testing requirements by mission risk class.

The detailed design of the AM part is ready to proceed under the complete set of tailored design requirements. Block 4 of Fig. 2 is an analysis to verify that the design satisfies specified requirements and the part can perform its designated functions. Structural and stress analyses are part of this process, using both interim and final geometric configurations, as well as the AM requirements and material properties. When necessary, allowance should be made for any expected variability in the AM material properties and manufacturing procedures. The verification analysis may impact the final AM part design and, as a result, this step is likely to be an iterative process.

The manufacturing of the AM part is the next step in this development process. Because AM manufacturing technology is still developing, careful monitoring of the process and definition of the material properties should be considered necessary. Block 5 of Fig. 2 is the verification of the AM parts, followed by the qualification process that entails analyses, testing, and inspections, as well as programmatic reviews and authorizations to proceed to this phase of the program. This is potentially another iterative step where adjustments to the manufacturing process, material properties verification, etc., might be necessary to complete part qualification.

In the space industry, and especially for the development of launch vehicles, it is common to require a first flight as final demonstration of system performance and functionality before the fleet production is undertaken. Spacecraft systems are typically acquired in small blocks or as single units, and the “first flight” item may take the form of an engineering model that is subjected to various flight environments in the laboratory to verify performance requirements.

Once the fleet production is approved, flight units are manufactured. However, flight approval requires that flight units be verified and subjected to acceptance screening according to program requirements. This step is shown in block 6. Acceptance screening may require extensive proof testing of AM hardware in addition to ongoing monitoring of the manufacturing process and tracking of material properties until the AM process matures.

8.

Case Studies to Develop and Qualify AM for the Space Industry

Three case studies are presented on the development and qualification of AM for the space industry. All three examples (NASA Marshall Space Flight Center, USAF Space and Missile Center, and the Defense Production Act Title III Program) share the common goals of reducing variability for AM, making AM repeatable and reliable, and maturing the AM process.

9.

Discussion

AM of metals is at an early stage in maturity and will naturally improve as the technology rapidly develops. With development, the part variability currently observed should naturally be reduced. Until that reduction is seen in practice, it may be sensible in the near future to use AM for secondary or redundant structures rather than for primary load bearing structure whose failure may lead to loss of mission. As a related point, an early failure may unnecessarily throttle the field. While AM makes possible topologies, a sound AM design at this stage of maturity should recognize the needs to inspect and proof test the part. “Design for AM” should consciously include “design for inspection and proof testing” until the field develops further.

Two advancements for the field are anticipated in the coming years. First, closed-loop feedback control should reduce the part variability. Second, AM’s layer-by-layer build inherently allows access to the part’s interior during fabrication, and real-time in situ inspection of flaws will be possible during the build, rather than after. Further, the inspection of the layer-by-layer build will conceivably allow a repair of the flaw by going back to fuse the flaw again before resuming the lay down of the next layer. This anticipated advancement should naturally reduce variability, as well as the burden placed upon conventional NDE methods traditionally used after production.

The AM field lacks a database of material properties similar to the MMPDS published for conventionally produced metal. In addition, the field does not know how the AM process parameters control microstructure and properties. Sharing of data and processes across industry, national laboratories, and universities may be the path to advance the field as quickly and cheaply as possible. The database will allow a new vendor to qualify its production by showing statistically that its production is in-family with accepted properties from accepted processes.

The aerospace industry is in the early stages of developing and publishing the qualification and certification standards for AM, which will be the consensus of industry, government, and the governing bodies for standards. The broad acceptance of the best practices for qualification will mature the field more rapidly and will help the space industry by improving quality and reducing the expense of qualification. As an ultimate goal, qualifying AM as a process should reduce the need for extensive qualification part by part, which would otherwise be expensive, duplicative, and time consuming.

10.

Conclusion

AM (commonly called “3D printing”) fabricates the desired final part directly from the input CAD file by depositing and fusing layer upon layer of the source material with very little waste. AM offers the revolutionary potential to eliminate welds and joints and, as a result, an entire class of associated failure mechanisms. New engineering designs are possible in which a single optimized part with topology can replace several traditional parts. For the AM of metals, the complex physics of metal deposition leads to variations in quality and to new flaws and residual stresses not seen in traditional manufacturing, which makes mission assurance challenging. The field also has knowledge gaps in material properties, NDE, and process control. Mission assurance will require: qualification and certification standards; sharing of data in handbooks; predictive models relating processing, microstructure and properties; and development of closed-loop process control and in situ NDE to reduce variability. Mission assurance can be tailored to account for mission risk class, MRL, and rigor of material properties. Proof testing of the AM part can be the mitigation for a low maturity process. AM parts for the National Security Space programs illustrate the potential for part count, cost, and schedule savings and the challenges to mature the technology. Three case studies for space programs share the common goals of improving the reliability and reducing the variability of AM, which can be achieved through DOE to optimize the process and SPC to monitor the process and properties. The fields need a breakthrough, which is a new process combining both high build rates and high accuracy and resolution.