Critical Thinking in the Application of Additive Manufacturing Processes: A Review

Abstract

Additive manufacturing is a rapidly growing fabrication process. However, the metal Additive Manufacturing (AM) processes are not the best choice for all hardware. The various metal AM processes must be evaluated using critical thinking prior to selection for each and every part, just like all fabrication processes.

This article introduces the metal AM processes and discusses the potential sources of failure that are currently known, or should be anticipated, for metal AM processed hardware. In addition, the article discusses the effect of the metal AM processes on the failure analysis process, as well as the effect of the metal AM processes on specific failure analysis steps. Lastly, this article discusses areas of consideration when using metal AM processes including design, metallurgy, production, quality, supplier management, and failure investigation, as well as providing some potential solutions.

Introduction

Additive manufacturing (AM) processes are a fast growing group of manufacturing processes. Papers on the AM processes are a large part of many journals and conferences, and many industries are experimenting to determine how AM fits in their operational branches such as production, research, sales, etc. The AM processes are being used in many industries for a variety of reasons including production parts, spares, tooling, prototypes, mockups, and various other applications. The array of AM processes and the unique parts they can produce provide the engineer a new and agile production capability that adds to the standard wrought and cast product forms. The proliferation of the AM processes is a juggernaut for the simple reason that they can fabricate almost anything including enclosed features such as a shrouded impeller and operating mechanisms. Examples of additively manufactured polymeric materials are shown in Figs. 1 and 2. Note the enclosed feature of the ASM International Dome in Fig. 1 and the operating tank turret and crescent wrench in Fig. 2 which were fabricated in a single build and not assembled.

Fig. 1

ASM International dome

Fig. 2

Operating tank and crescent wrench

This article has been broken up into sections:

• (1) "Metal Additive Manufacturing Processes" section—introduces the metal AM processes,

• (2) "Sources of Metal Additive Manufacturing Failures" section—discusses potential sources of failure that are currently known or should be anticipated,

• (3) "The Effect of the Metal AM Processes on the Failure Analysis Process" section—discusses the effect of the metal AM processes on the failure analysis process,

• (4) "The Effect of the Metal AM Processes on Specific Failure Analysis Step" section—discusses the effect of the metal AM processes on specific failure analysis steps, and

• (5) "Metal AM Processes Considerations and Potential Solutions" section—discusses areas of consideration when using metal AM processes, including providing some potential solutions.

In addition, this article will concentrate on the metal AM processes of powder bed fusion and direct energy deposition.

Metal AM processes for engineering components are the logical progression of prior technologies as seen in Fig. 3. Prior to the 1960s, a common version of “additive manufacturing” was to “build up” components that were undersized due to wear in service by welding or plating processes followed by machining to return them to drawing dimensions. From the 1960s through 2000s, various technologies emerged and were developed such as powder metallurgy, rapid solidification rate processes, near net shape technologies, computer-aided design and manufacturing (CAD/CAM), and rapid prototyping technology. Metal AM processes are the logical progression and combination of these technologies.

Fig. 3

Timeline of manufacturing processes development

The metal AM processes proliferated in the early 2000s, especially in the aerospace and medical industries, and an argument could be made that their advancement was accelerated by the Great Recession circa 2007–2010 as businesses turned to smaller order quantities and more agile manufacturing. Since 2010, metal AM process technologies have exploded and have dominated the interest of most industries, journals, research, conferences, etc. At the same time, there has been a proliferation of AM consortiums, groups, experts, businesses, and business purchases as venture capitalists, business groups, universities, and companies position themselves in the AM universe. During this time, the R&D effort by academia and industry into the metallurgical characteristics of the AM processes has also grown. In 2017, 500,000 AM machines were sold worldwide, including more than 14,000 industry-grade machines [1].

The ISO/ASTM 52900 standard has defined AM as the “process of joining materials to make parts from 3D model data, usually layer upon layer, as opposed to subtractive manufacturing and formative manufacturing methodologies” [2]. The same standard organizes the AM processes into seven categories of vat polymerization, material jetting, powder bed fusion (PBF), directed energy deposition (DED), material extrusion, binder jetting, and sheet lamination.

The most common name for AM processes is 3D Printing. However, there are a multitude of names used for AM processes as companies attempt to brand their AM process in the marketplace. The range of metal AM technologies is increasing continuously. As of 2021, 18 different working principles are known with over 140 OEMs supplying machines [3]. Figure 4 provides a listing of the two dominant metal AM processes, PBF and DED, as of 2016 [4].

Fig. 4

Summary of metal AM processes and their commercial machine supplier names [4]

A more in depth review of the various AM processes can found in ASM Handbook, Volume 24, “Additive Manufacturing Processes” as well as other sources [5,6,7,8,9, 39].

Metal Additive Manufacturing Processes

The most common metal AM processes used for engineering components is the powder bed fusion (PBF) process, followed by the directed energy deposition (DED) process using metal powder or wire [6, 8, 9]. A generic illustration of these processes can be found in Figs. 5,6 and 7 [6]. The energy source can be electron beam, laser arc, etc.

Fig. 5

Generic illustration of a metal AM powder bed fusion process (Copyright 2014 by Springer Nature, used with permission) [6]

Fig. 6

Generic illustration of a metal AM powder feed process (Copyright 2014 by Springer Nature, used with permission) [6]

Fig. 7

Generic illustration of a metal AM wire feed process (Copyright 2014 by Springer Nature, used with permission) [6]

Another way to separate the various metal AM processes by comparing the absorbed power versus the beam velocity [10]. The DED AM processes have high absorbed power and low beam velocity. In comparison, the PBF AM processes have low absorbed power at approximately 10% of the DED AM processes, and high beam velocities at approximately 5× to 25× higher than the DED AM processes.

The metal AM processes have also been classified in terms of feed stock, build volume, and energy source as summarized in Table 1 [6].

Table 1 Metal AM processes

The unique benefits of metal AM processes during the design or redesign process are well documented and can include the following depending upon the specific application:

• 1. Reduced Product Development Time

• 2. Difficult/Intricate Part Fabrication / Functionally Graded Materials (FGM)

• 3. Customized/Personalized Parts

• 4. Part Number Reduction / Consolidation

• 5. Topological Optimization

• 6. Nesting

• 7. Reducing one of more secondary fabrication processes

• 8. Taking advantage of a fabrication process characteristic

• 9. Schedule Reduction / Remove input stock lead time / Fabricate at Final Destination

• 10. Cost Reduction

The first five reasons are the truly unique benefits of the metal AM process that are desired by a design team. Reduced product development time is, and has been, a goal of any manufacturing operation and the benefit was demonstrated with the introduction of rapid prototyping in the 1990s. Difficult and/or intricate part fabrication is a benefit unique to metal AM as shown in Figs. 1 and 2 showing an enclosed feature and operating mechanisms with no assembly. The metal AM processes can fabricate components that cannot be made by any other fabrication method or combination of methods. There are many advancements using bimetallic and multi-metallic methods to manufacture components in a single piece. Functionally graded materials (FGM) can be produced by metal AM processes to gradually transition from one material to another in order to provide a functional improvement by providing selected metallurgical properties in a specific area of the part. These solutions come with some risks such as a more complex AM process parameter development, and greater potential for undesirable carbide and intermetallic secondary phase formation with associated brittle and crack prone behavior. The fabrication of customized or personalized parts is another benefit unique to the metal AM processes. The medical and dental industries have taken advantage of this benefit by making medical implants or orthodontia specifically based on the unique physiology of the patient [11].

Part number reduction, or consolidation, has been another goal of manufacturing operations since superplastic forming was combined with diffusion bonding in the 1980s to reduce part count and eliminate secondary fabrication operations such as welding or joining. The metal AM processes take this to another level by integrating this benefit with difficult or intricate parts. Topological optimization is another benefit unique to the metal AM processes. This design approach uses both solid modeling and stress analysis software to analyze a predictably loaded part. This ensures that most or all fabricated sections of the part carry loads with the objective of reducing material to a minimum as seen in the metal hinge bracket in Fig. 8 [12].

Fig. 8

(a) Nacelle hinge bracket (b) AM-processed bracket with topographical optimization [12]

The next five reasons are common manufacturing goals that can be enabled by the metal AM processes. Nesting is an old concept for cost reduction in which the number of parts fabricated is maximized by utilizing as much of the input wrought material, or the build space for metal AM processes, as possible. Reducing one or more secondary fabrication processes is another manufacturing concept to obtain cost and/or schedule improvement. Some examples include using a braze process as the annealing cycle of the raw material, or direct precipitation aging a forging off the die without using the solution heat treatment step or near net shape fabrication to reduce machining. Taking advantage of a fabrication process characteristic is another common manufacturing concept. All fabrication processes have benefits, but they also have disadvantages that must be evaluated and considered during the part fabrication. If possible, the disadvantage can be turned into an advantage for some hardware. For example, most of the metal AM processes produce a rough surface finish in comparison to machined parts. This is usually considered a disadvantage as the parts are replacing a machined part, especially if the part is subjected to cyclic loading. However, many coating processes require a rough surface for adherence of the coating. For these parts, the rough surface can be considered an advantage.

Schedule reduction has always been one of the primary goals of any improvement to a manufacturing operation. Metal AM processes can achieve this goal by either removing the lead time of input stock, reducing part fabrication time or providing the ability to fabricate at or near the final destination of the part. Cost reduction would be the other primary goal of any improvement to any manufacturing operation. The metal AM processes can achieve cost reduction in a variety of ways including removal of unique and costly input stock, such as a forging, or the elimination of secondary processing such as machining.

However, the metal AM processes are not the best choice for all hardware. The various metal AM processes must be critically evaluated, or judged, prior to selection for each and every part, just like all fabrication processes. The engineer and/or design team should consider:

• (1) Part Design—Fit, Form, and Function

  • Dimensions, Dimensional tolerances, Surface finish, Mechanical properties, etc.

• (2) Metallurgical Considerations

  • Microstructure, corrosion resistance, residual stress magnitude and distribution, etc.

• (3) Secondary Manufacturing processes

  • Homogenization, hot isostatic pressing (HIP), heat treatment, distortion during thermal treating, machining, forming, welding, brazing, assembly to other components, etc.

• (4) Service Environment

• (5) Schedule

• (6) Cost

• (7) Supply Chain Considerations

  • Multiple suppliers, similar physical characteristics versus different metallurgical characteristics, spares, etc.

Sources of Metal Additive Manufacturing Failures

The potential sources of failures for metal AM-processed parts are separated into service, design, and fabrication processing.

Service Sources of Failures

There is very little published information regarding service failures of metal AM-processed components at the time of the writing of this article [13,14,15]. This will change as more and more AM-processed parts are put into service and companies are willing to go public with the failures. However, until the usage of the metal AM processes mature, and companies are willing to publicly document in-service AM part failures, failures will probably be dominated by the other two other sources of failure, design, and fabrication processing.

Design Sources of Failure

The first source of failure for metal AM processes is in the design phase. Poor design choice is usually the selection of the metal AM process, the material or the part to be fabricated by the metal AM process, or a combinatin of all three.

The first reason is common for every new fabrication process, Management Directive or “My Boss Told Me to Do it.” Company leadership sometimes makes a command decision to use a new fabrication process. Sometimes, this is done with deliberation and forethought involving the company engineering community. Other times, it is a knee jerk reaction as company leadership fears falling behind in a competitive industry. Unfortunately, this kind of thinking can sometimes lead to a poor selection for both the metal AM process, the material, and the part to be fabricated resulting in a part that cannot be fabricated or will not satisfy the design criteria or cannot be fabricated in a cost effective and schedule friendly manner.

The second reason is also common to a new fabrication process, using the metal AM process that the company already owns. Due to the substantial cost of the metal AM process fabrication equipment and its support equipment, many companies have invested a lot of money and resources to get involved with a specific metal AM process. Therefore, they wish to use that investment to its fullest in order to achieve a certain return on investment (ROI). However, in the last ten years, it has become well established that the various metal AM processes produce parts with different metallurgical characteristics. Depending on the specific physical and metallurgical characteristics of the part and the anticipated service conditions, not all metal AM processes will be a correct choice for all parts.

The use of topological optimization must be performed with the utmost care. The wording earlier in this article of a predictably loaded part is very important. Companies that choose to use topological optimization must completely understand the load magnitude and distribution in their hardware. This is no small task. This concept lends itself to replacement parts as there is hopefully a long successful service history and corresponding well-established load analysis. Topological optimization will be more risky on new parts as the load magnitudes and distribution are not as well established. In addition, during the design phase of many new parts, the load magnitude and distribution can change resulting in a new analysis and a new part design. Repetitive redesign and the accompanying drawing changes have become “persona non grata” for most companies due to the adverse effect on cost and schedule.

Lastly, topographical optimization may lead to an early failure if the design team does not realize the change in load magnitude or distribution. It would be advisable to perform a sensitivity analysis to determine the effect on the part if there was a shift in load magnitude or distribution. This would hopefully negate any potential failure if the loading changed in service, or a “domino effect” if another part failed during service which resulted in a loading change and failure of the topographically optimized part.

Probably, the most important reason for failure in the design phase is lack of attention to the metallurgical characteristics produced in the part by the metal AM process. During development of the metal AM processes in the early 2000s, the focus was on the development of the operation of the machines and the fabrication of the physical part shape. At conferences in the early 2000s, there was data presented showing up to 50% increase in tensile properties over comparable wrought products with discussions of isotropic and homogeneous microstructures, especially with the powder bed fusion and powder feed directed energy deposition metal AM processes. Isotropic and homogeneous microstructures are the ultimate metallurgical goal for any fabrication process and has been successfully achieved in some occasions such as powder metallurgy in specific applications such as bearings. However, as more research has been made public with the metal AM processes in the last ten years and actual hardware has been evaluated instead of subsize research samples, multiple reports have been published showing that metallurgical characteristics can vary greatly depending on the metal AM process employed and the quality of the build. For example, Fig. 9 is an electron backscatter diffraction (EBSD) map of an as built SLM metal AM-processed 316L stainless steel. In Fig. 9a, the build direction is vertical [16]. Figure 9b is taken from a plane perpendicular to the build direction. The microstructure was 100% austenitic and almost fully dense. Also note that the grains are heavily elongated in the build direction due to the heat removal through the build plate during solidification which will result in anisotropic behavior.

Fig. 9

EBSD maps (inverse pole figure: blue is 111, green is 101 and red is 001) (used with permission) [16]

Elongated or columnar grains in the build direction has been also noted in other alloys systems such as 17-4PH martensitic stainless steel, high silicon steel, and Inconel 718 [17,18,19,20,21,22].

The metallurgical characteristics will also vary with the part being fabricated due to the variation in build models and the use of support stilts. These metallurgical characteristics include microstructure, mechanical properties, surface finish, dimensions, dimensional tolerance, residual stress magnitude and distribution, corrosion resistance, etc. One interesting example that has been documented is the formation of laves phase in the nickel base superalloy Inconel 718 [51]. Laves phase is not a desirable phase but is a common phase in the cast form of Inconel 718 while not common in the wrought form of Inconel 718.

Other important metallurgical characteristics that are usually not considered by the design team but can cause secondary processing issues and service limitations include allotropic behavior, equicohesive temperature, recrystallization temperatures, and precipitation age hardening characteristics.

Metallurgists use the allotropic behavior of certain metals to achieve mechanical properties through heat treatment. The most common of these metals are:

• (a) the carbon and low alloy steels, the martensitic stainless steels, and the martensitic precipitation age hardening (PH) stainless steels which use the austenite transformation temperature; and

• (b) the titanium alloys, especially Titanium 6Al–4V and its derivatives, which use the beta transus.

Figure 10 is a side by side comparison of the metal AM-processed and wrought product microstructure for 17-4 PH martensitic stainless steel, both at 200x magnification and using Villella’s reagent. The wrought product has a fine and equiaxed low carbon martensite microstructure that usually has a small percentage (< ~ 5%) of delta ferrite. The larger and more irregular grain structure of the metal AM-processed microstructure was determined by X-ray diffraction to be a dual phase microstructure of 50–75% austenite and low carbon martensite. In addition, it was noted that the heat treatment characteristics and martensite start (M_s) temperature was altered due to the dual phase microstructure [22].

Fig. 10

17-4 PH stainless steel–Metal AM process (left) and wrought product (right) (Mag: 200×)

Titanium 6Al–4V as another example. Published work has demonstrated that the metal AM processes will melt a layer twice, or more, and will exceed the beta transus temperature four times, or more, as seen in Fig. 11 [6, 23]. The multiple thermal excursions beyond the melting temperature and beta transus will affect the final microstructure and therefore the resultant mechanical properties. A similar effect has been documented in Inconel 718 [51]. The design team must consider leaving the part in the as built condition or perform secondary thermal treatments to improve the microstructure at increased cost and schedule.

Fig. 11

Single-line, multilayer deposit on a thick plate for laser AM Ti–6AL–4V. Graph shows peaks in temperature-time history at various locations in the deposit [23]

It is well documented that residual stress/strain in a metal alloy can alter the precipitation kinetics of metals, usually by increasing the rate of the precipitation kinetics One example is the precipitation of hydrides in commercial pure (CP) titanium [24]. Another example can be found in the Time–Temperature-Transformation (TTT) diagram for the nickel base superalloy Inconel 718 as seen in Fig. 12 [25]. In the right side of the TTT diagram, it is noted how stress in combination with time and temperature promotes the formation of two undesirable phases, alpha chromium and sigma. Some of the metal AM processes have been documented to produce high residual stresses in the processed part which could lead to the formation of these phases during secondary thermal processing [26].

Fig. 12

Time–Temperature-Transformation (TTT) diagram for Inconel 718 (Copyright 1991 by The Minerals, Metals & Materials Society. Used with permission) [25]

It should also be remembered that most metallurgical databases such as TTT diagrams and phase equilibrium diagrams are for metals in an equilibrium condition. Since most metal AM processes produce microstructures that are formed by rapid solidification rates, these diagrams will be helpful, but will not be representative.

The design team should consider that the AM processes can affect other metallurgical characteristics of a metal. Some examples are physical properties such as density or magnetism; corrosion resistance, especially stress corrosion cracking (SCC) resistance; and fluid compatibility, especially to propulsion fluids. Other considerations would be phase stability when exposed to cryogenic or elevated temperatures and thermal treatment parameters such as martensite start (M_s) or martensite finish (M_f) temperatures.

Lastly, it has also been documented that the metal AM processes can produce mechanical properties that have greater variability, and can be inferior, to the corresponding wrought and cast products as shown in Fig. 13 [26,27,28]. The materials property space graph in Fig. 13a shows the variation of yield strength vs elongation for steels, nickel alloys, aluminum alloys, TiAl, and the high entropy alloy (HEA) CoCrMo made by metal AM processes and how their variation compares to the same alloys in the wrought product form as noted by the dashed lines.

Fig. 13

Materials property space for room temperature yield strength vs elongation of additively manufactured alloys and conventionally manufactured alloys (dashed lines) (a) Steels, Ni alloys, Al alloys, TiAl and CoCrMo (b) Ti–6Al–4V alloys (PBF). (Used with permission) [27]

The materials property space graph in Fig. 13b is more specific as it shows the variation of mechanical properties for titanium 6Al–4V made by different metal AM PBF processes and how their variation compares to the same alloy in the wrought and cast product form as noted by the dashed lines. The PBF metal AM processes show much greater variability than the wrought or cast product forms, and are not consistent with each other.

Reduction in fatigue life/strength and an increase in fatigue data scatter has also been noted due to inherent AM process characteristics such as rough surface finish, residual stress magnitude and distribution, internal defects such as lack of fusion or porosity, and microstructure [26, 29,30,31,32,33,34,35]. One example is the variation of the fatigue behavior of metal AM-processed titanium 6Al–4V noted in Fig. 14 [36]. Figure 14 shows summary data replotted of laser PBF SN fatigue behavior for Ti–6Al–4V tested at R = 0.1, along with Metallic Materials Properties Development and Standardization (MMPDS) data obtained for machined samples from the following products: 3-inch-thick cast material; 3-inch-thick annealed wrought material and 3-inch-thick solution annealed and aged wrought material. Although orientation-dependent fatigue behavior was found and some property improvements were achieved with machined and polished surfaces, the very poor performance in comparison to the other data summarized in this plot was assumed to result from process-induced defects. Hot isostatic press (HIP) processing of AM-processed hardware can help improve fatigue resistance of AM parts, but its effectiveness may be limited if the initial defect population and size of defect are large [37]. Therefore, improved optimization of the build parameters to reduce internal defect number and size would be a better approach.

Fig. 14

Summary of stress (S) vs. cycles to failure (N) (S–N) data for laser PBF, EBM PBF, and DED wire at R = 0.1. MMPDS data for cast, wrought machined data are shown for comparison. (Reprinted with permission from Annual Reviews. Metal additive manufacturing: a review of mechanical properties. Annual Review of Materials Research Vol. 46:151–186 (Volume publication date July 2016) First published online as a Review in Advance on April 21, 2016 https://doi.org/10.1146/annurev-matsci-070115-032024) [36]

Variation in fracture toughness has also been noted. Figure 15 summarizes the location-dependent toughness values along a tall, as built PBF (EBM) titanium 6Al–4V sample, in addition to the microstructure variations and significant differences in defect density along the build [38]. This figure is a good example of the mechanical property, microstructure, and internal defect variations found in as built metal AM-processed alloys. Although a HIP thermal treatment eliminated the process-induced defects, location-dependent toughness values remained which suggested that subtle detrimental changes to the microstructure were probably responsible.

Fig. 15

Location-dependent toughness values in an as built PBF (EBM) Ti–6Al–4V sample. Variations in microstructure and defect density are noted along the same sample. (Reprinted with permission from Annual Reviews. Metal Additive manufacturing: a review of mechanical properties. Annual Review of Materials Research Vol. 46:151–186 (Volume publication date July 2016) First published online as a Review in Advance on April 21, 2016 https://doi.org/10.1146/annurev-matsci-070115-032024) [38]

The design teams must consider all metallurgical characteristics and their potential effect on the service life of the part. For the specific metallurgical characteristics that are most critical to the part, accounting for the variability or developing controls to minimize the variability will be useful in maximizing the service life of hardware.

A positive step in the development of the metal AM processes will be the creation of databases such as mechanical property and corrosion resistance to assist the design team.

Fabrication Processing Sources of Failure

There have been reported various potential sources of failure for metal AM processes during the build process [6, 9, 26, 39]. For this article, the potential sources of failure have been separated into internal and surface, though some of the potential sources are listed in both groups.

• (1) Internal

  • a. Powder contamination

    • i. Oxidized/contaminated in storage

    • ii. Oxidized/contaminated or partially welded due to recycling powder

    • iii. Remnant powder from a previous batch

  • b. Densification

  • c. Anisotropic microstructure

  • d. Columnar or elongated grains

  • e. Lack of fusion

  • f. Porosity due to inert gas entrapment

  • g. Metastable phase formation

  • h. Detrimental phase formation

  • i. Hot tears

  • j. Delamination from residual stresses and/or unmelted areas

  • k. Swelling from the melt pool

  • l. Chemical composition change due to vaporization of elements

  • m. Alloy microsegregation

• (2) Surface

  • a. Weld tracks

  • b. Protruding unmelted or partially melted powder particles

  • c. Ejected molten droplets

  • d. Recesses

  • e. Swelling from the melt pool

  • f. Unfused powder

  • g. Keyholing

  • h. Balling

  • i. Warping or curling

  • j. Roughness

  • k. Lack of fusion

Figure 16 provides examples of defects that can form in SLM materials: (a) porosity formed in SLM Ti–6Al–4V (b) balling and (c) hot tears [40].

Fig. 16

Defects in SLM materials: (a) porosity formed in SLM Ti–Al–4V (b) balling and (c) hot tears (used with permission) [40]

Figure 17 shows three examples of lack of fusion in metal AM-processed 17-4PH stainless steel at various positions in the build.

Fig. 17

Lack of fusion in metal AM-processed 17-4 PH stainless steel

The ramifications of these potential sources of failure are twofold, production cost increase and failure in service. Both the internal and surface sources of failure promote a cost increase as jobs are build multiple times to meet production quotas. Currently, reported scrap rates for metal AM processed hardware has a wide range at 5–40%. For comparison, most mature production processes target a maximum scrap rate of ~ 2%. As with any new fabrication process there are many reasons for a large scrap rate. In addition, as with any fabrication process, scrap rate data is very dependent upon the selection of the metal AM process, the material and the part geometry. It has been well documented from other manufacturing processes that the internal and surface sources of failure noted above can cause a failure in service [41]. Prevention of service failures will be achieved by selecting the right application for any part which contains one or more of these sources of failure.

A positive step to reduce processing sources of failure would be to create an AM process control specification or standard. This can be done at the company level or at the industry level. In 2017, the FDA released a guidance document for AM processing of medical devices, and NASA Marshall Space Flight Center produced a standard and specification for the control of laser PBF AM-processed hardware [42,43,44].

Another positive step in the evolution of the metal AM processes will be the development of acceptance criteria for the build process. In conjunction, this will require development of nondestructive testing (NDT) methods for metal AM processes not only in inspection of the final build, but also in situ monitoring during the build [45]. Current NDT processes for the final build have proven adequate for the directed energy deposition (DED) metal AM process which are usually machined as a secondary fabrication step. Powder bed fusion (PBF) metal AM processes that are used with the as built surface are more of a challenge. Common surface inspection techniques such as dye penetrant and magnetic particle have proven not to be viable for the powder bed fusion metal AM processes due to the rough surface as seen in Fig. 18. Any changes in the magnetic properties of the metal must also be considered if magnetic particle inspection is used. Computed tomography (CT) has proven to be useful for internal examination due to the small part size and reduced wall thickness associated with parts made with the powder bed fusion (PFB) metal AM processes.

Fig. 18

Rough surface on a Titanium 6Al–4V PBF processed cylinder

In addition, research and development of the effect secondary processes after the build will further develop the metal AM processes. Development of thermal processes such as homogenization, hot isostatic pressing (HIP), solution heat treatment and precipitation age hardening specific to metal AM processes are currently in work. However, secondary fabrication processes such as welding, brazing, forming, etc. should also be evaluated in order to facilitate the usage of metal AM-processed parts into larger components or assemblies. Surface modifications and coatings should also be evaluated as they are used to extend the service life by increasing temperature capability, corrosion resistance, wear resistance, etc.

The Effect of the Metal AM Processes on the Failure Analysis Process

The failure analysis of hardware fabricated by metal AM processes follows the same general procedures outlined in the in the ASM Handbook Volume 11 [46]. This section will review the steps of a failure investigation and how these steps are affected by the metal AM process.

The four basic steps of a failure investigation include collection of background information, data, etc.; laboratory testing, evaluations, and examinations; analysis; and completion.

Collection of Background Information, Data, etc.

This step of the failure investigation is sometimes overlooked, but it is important. It consists of the preliminary evaluation of the failed part; a site examination and discovery of the circumstances of the failure; the selection of samples for laboratory evaluation, testing, dimensional inspection, etc.; collection of pertinent engineering drawings, specifications or codes; collection of all manufacturing records such as job travelers, etc.; collection of all quality records from fabrication or service; and fabrication scrap rate information.

The metal AM processes may affect this step of the failure analysis in the following areas:

• (a) Will there be enough material to obtain the amount of test samples required, and in the orientation of interest? Metal AM processes minimize the amount of excess material during the build. The use of topological optimization will minimize the amount of excess material even further. This practice translates into less, or no, material available for test samples. This is important because during a failure investigation it is best to examine and test material that is representative of the failed area of the part. Since each AM build could be considered unique to itself, this is a limitation to the failure analysis process.

• (b) Will the metal AM process build records be available for review? If yes, can the failure investigation team understand them, and be able to find any anomalies, errors or flaws in the build process?

• (c) Will the scrap rate information be representative of the process? For smaller parts, many operators may discard a part and make a new one instead of doing the paperwork to disposition a part for scrap.

Laboratory Testing, Evaluations and Examinations

This step in the failure analysis process consists of performing examinations and testing to understand the metallurgy and material property characteristics of the failed component. The examination and testing may include post-failure property determination of mechanical properties, friction characteristics, etc.; metallographic evaluation to understand the microstructure fracture surface examination; chemical analysis of the material, coating, corrosion or wear products, etc.; testing of exemplars; and testing under simulated-service conditions.

The metal AM processes may affect this step of the failure analysis in the following areas:

• (a) Will there be a material property database of metal AM processed material in which to compare the testing results? How will the engineer know if the test results are acceptable or unacceptable, or typical of the metal AM process?

• (b) In a similar manner, the metallographic evaluation will require a control sample for the microstructure. This may not be an easy task due to the large amount microstructural variation during the metal AM process build.

• (c) Once again, the fracture surface examination will require control samples to understand the failure mode and mechanism of metal AM processed material.

• (d) Testing of exemplars will mean sacrificing other components produced at the same time or making a new build. Once again, the question will be how representative of the failed part are the exemplar samples.

A simple approach would be to make another build, but is it truly representative of the failed part? It will not be the same powder or wire stock. It may not be the same build process, or even the same metal AM process machine.

A better approach for all of these issues would be to have a set of test bars in each build to determine basic metallurgical characteristics, such as mechanical properties and metallographic characteristics. The information from these test bars can become the required control samples for mechanical properties, microstructure, residual stress, fracture surfaces, etc. In addition, a few blank test bars could be easily created with each build and kept for any necessary tests later.

Analysis

This step in the failure analysis process consists of performing the required analyses of the component including a stress analysis, fracture mechanics analysis, quantitative fracture analysis, determination of failure mode and mechanism, etc. It also includes consultation with experts in other disciplines and the formation of root cause conclusions.

The metal AM processes may affect this step of the failure analysis in the following areas:

• (a) There may not be a mechanical property database for the stress analysis or fracture mechanics analysis.

• (b) There will not be a database to define the characteristics for the failure mode and mechanism, and the associated fracture surface, for metal AM processed hardware.

• (c) Can the failure investigation team find an expert on the specific metal and metal AM process used to make their component?

Completion

The last step in the failure analysis process consists of the engineer performing a synthesis of all the information collected, preparing a final report, and providing recommendations so the failure does not happen again. The recommendations may include a component change, a process change, a maintenance cycle change, etc.

The metal AM processes may affect this step of the failure analysis in the following areas:

• (a) There may be many unknowns in the information collected as noted in steps 1–3 above.

• (b) Will the failure investigation team be knowledgeable enough to suggest a build process change for the metal AM processed hardware?

The Effect of the Metal AM Processes on Specific Failure Analysis Steps

There are also specific steps in the failure analysis process that will be affected by the metal AM processes including visual inspection, non-destructive testing, residual stress and metallographic examination, mechanical property testing, sectioning, and sample preparation.

Visual Inspection

A low magnification visual inspection is an important part of any failure investigation. The unaided human eye is the second most important tool of a failure analyst because under proper lighting it has exceptional depth of focus to see distortion, the ability to determine subtle texture and color changes, and to examine a large area quickly. The human eye is also attached to the most important tool of the failure analyst, their brain with their experience and critical thinking capability. However, it must be considered if the metal AM processes will bring changes to the metal surface roughness, optical reflectivity, fast fracture patterns such as radial marks and chevron marks, general or localized oxidation or corrosion behavior, etc.

Since cleaning is sometimes required, it must be considered if the standard cleaning processes are appropriate due to potential changes in material susceptibility by the metal AM processes.

Non-Destructive Testing (NDT)

NDT is a tool used by failure analysts to look for cracks during an investigation, sometimes for secondary cracks if the primary crack has been damaged. Previous sections in the article have discussed how the surface roughness of PBF metal AM processes will prohibit the use of surface inspection techniques such as dye penetrant or magnetic particle inspection. In addition, it must also be considered that if the magnetic properties of the metal have been changed by the metal AM processes, then magnetic particle inspection should not be used. This article has noted the change in the microstructure of 17-4 PH from martensite to austenite which should also change its magnetic behavior. The use of volumetric inspection techniques such as ultrasonic inspection or radiographic inspection will be influenced by the surface roughness and microstructural variations as well as the porosity and contamination. Lastly, the inspector probability of detection limits for NDT processes of AM processed hardware may also change which reduces the confidence level of finding an indication. As noted earlier in this article, computed tomography (CT) has proven to be useful for internal/volumetric examination due to the small part size and reduced wall thickness associated with parts made with the powder bed fusion (PFB) metal AM processes.

Sectioning and Sample Preparation

All failure investigations require sectioning and sample preparation of the various metallographic examinations and mechanical property testing. These techniques and practices are well established for wrought and cast products with specific cutting fluids, cutting wheels, polishing compounds and techniques, etching reagents, machining speeds and feeds, mechanical testing parameters, etc. As with any new microstructure produced by a fabrication process, new sectioning processes and metallographic preparation techniques may need to be developed.

Fractography

Interpretation of a fracture surface is an important and common part of any failure investigation, which takes years of practice to learn. Understanding a normal, or expected, fracture surface under specific loading conditions is important. In addition, since it is well known that fracture surfaces can be influenced by the microstructure, it should be anticipated that the metal AM processes may produce unique fracture surfaces. This is especially true for the many failure mechanisms that occur with loads in the elastic region of the stress strain curve where the influence of microstructure is significant.

Products fabricated using the metal AM processes have shown fracture surfaces different than the wrought product form. Figure 19 is an example of room temperature tensile test bars from wrought product demonstrating classic “cup-cone” ductile fracture behavior and “flat” brittle fracture behaviour [47]. Room temperature tensile tests of wrought titanium 6Al–4V usually produce a cup-cone failure similar to Fig. 19a.

Fig. 19

Macroscopic appearance of classic (a) “cup-cone” ductile and (b) “flat” brittle tensile test fractures [48]

For comparison, Fig. 20 shows fracture surfaces of tensile tests from as built SLM Ti–Al–4V specimens [48, 55]. The photograph in Fig. 20a is noted to be a cup-cone fracture, but it appears to demonstrate quite different cup-cone behavior than in Fig. 19a, and more like a tensile test from a Ti–6Al–4V cast product. In addition, the photograph in Fig. 20d is noted to be quasi-cleavage but appears more likely to be unfused power particles. Another example is the fracture of a metal AM-processed tool steel as seen in Fig. 21. The fracture in Fig. 21a exhibits a fracture surface and brittle behavior not characteristic of a tool steel. In addition, the fracture in Fig. 21b exhibits large amount of unfused particles at the edge of the sample. This can also be seen in Fig. 21a at the edges of the sample.

Fig. 20

Fracture surfaces of tensile tests from as-built SLM Ti–6Al–4V specimens: (a) cup-and-cone, (b) dimpled rupture, (c) and (d) quasi-cleavage facets (Used with permission) [49]

Fig. 21

Fracture surface of a metal AM-processed tool steel: (left) macro, (right) edge

Mechanical Property Testing

This article has discussed how the magnitude and variability of the mechanical properties of metal AM processed material are not the same as the wrought and cast product forms. In addition, failure analysts many times return to first principles during a failure investigation. Figure 22 is a notional stress strain diagram exhibiting a parabolic curve behavior for a metal with the specific material characteristics that are well established for wrought and cast products. A failure analyst will use typical stress strain diagrams, as well as other first principles curves, as a resource during a failure investigation. Similar diagrams will be required for the failure investigation of a metal AM processed part as curves from wrought and cast products will not be applicable.

Fig. 22

Notional stress strain diagram with the specific material characteristics noted

Metallographic Examination

The earlier sections of this article discussed the variation in microstructure possible by the metal AM processes. The variation in microstructure and the difference of AM product from the wrought and cast product forms must be understood by the failure analyst in order to properly understand a failure. Common testing such as a micro indentation hardness survey may be more difficult to interpret due to the finer and more complicated microstructure.

Residual Stress Magnitude and Distribution

The earlier sections of this article also discussed the variation in residual stress magnitude and distribution possible in the various metal AM processes. The variation in residual stress magnitude and distribution and the difference from the wrought and cast product forms must be understood by the failure analyst in order to properly understand a failure.

Metal AM Processes Considerations and Potential Solutions

As with any new fabrication process, engineers must attempt to decipher exactly what benefits and issues the metal AM processes may present to them. This critical thinking is part of the prevention component of failure analysis and prevention. This section will describe metal AM processes areas of consideration and then potential solutions.

The areas of consideration include the following:

• a. Limited working zone for PFB processes

  • i. Part size restriction

  • ii. Determination of actual working zone

• b. New mechanical properties, corrosion characteristics, surface finish, dimensional tolerances, etc.

  • i. Potentially lower than the wrought metal version and with higher variability

• c. No mechanical property, corrosion, etc. databases

• d. New process control criteria

• e. Metal AM processes may disappear

  • i. The metal AM process universe is changing due to companies being purchased or going out of business

• f. Production Rate limitations

• g. New metallurgical characteristics

  • i. Macrostructures, microstructures, corrosion, etc.

• h. Unknown effects of thermal exposure in manufacturing or service on metallurgical characteristics and distortion

  • i. Heat treatment, welding, forming, surface modifications, coatings, etc.

• i. Different or change in the residual stress magnitude and distribution

• j. New failure modes and mechanisms

• k. Added cost and resources necessary for proper powder control to prevent powder contamination, mixing of heats and fire hazard

The potential solutions are broken into design considerations, metallurgical considerations, production considerations, quality considerations, supplier management considerations, and failure investigation considerations.

Design Considerations

Design considerations are listed in Table 2. Limited working zone pertains mainly to the PBF metal AM processes as the working zone is relatively small as noted by the build volume in Table 1 and has not really increased much in the last five years. The choices are to use a machine with a larger working zone or combine metal AM Processes with a secondary fabrication process like welding. Finding a larger working zone from the perspective of least to most complex from an engineering context would be:

• (1) using a machine of the same metal AM process with a larger working zone,

• (2) using a different metal AM process machine with a larger working zone,

• (3) going outside your company to use a machine of the same or different metal AM process with a larger working zone.

Table 2 Design considerations

Using a machine of the same metal AM process with a larger working zone is the least risky of the choices, but there is still risk. As noted previously, it has been discovered in the last ten years that metallurgical characteristics of the metal AM processes vary greatly. It has also been reported that two machines for the same metal AM process built years apart may not produce the same results [49]. An appropriate analogy is cars. It is not reasonable to expect a 2008 Ford SUV to run identically to the same model 2020 Ford SUV. They are both produced by Ford, and the same model, but all models change over time. Therefore, there is a chance that when moving to a larger version of the same metal AM process the build model will not work and the physical and metallurgical characteristics of the part will be different. Hopefully, all it will take will be for the build model to be adjusted. Whether a new qualification of the part and testing is required depends upon the controlling AM process specification or standard.

Engineers must also recognize that there is a “ceiling” on part size. The metal PBF AM process machines have increased very little in size in the last 5 years. Recently, some of the metal AM process machine manufacturers have started to offer “customized” PBF AM process machines built to a customer’s larger size requirements. This issue does not apply as much to DED metal AM processes as they are usually designed to fabricate larger hardware of various configurations.

Using a different metal AM process with a larger working zone may alter the physical characteristics of the part and will most probably alter the metallurgical characteristics of the part. A new qualification of the part will probably be required. Continuing with the car analogy, moving from a five person Ford SUV to an eight person Chevrolet Large Family SUV provides the owner with a larger vehicle for transportation. However, these two SUV’s are made by two different car manufacturers and do not operate the same.

Going outside your company is the classic “make/buy” decision made by companies. A “make/buy” decision will compare the in house manufacturing operations to suppliers using factors such as capability, capacity, cost, and schedule. The technical ramifications of the previous paragraphs apply, but also perhaps more importantly, going outside alters the company’s control over the cost and delivery schedule of a part.

The second choice when the design team is confronted with a limited working zone is to combine the metal AM process with a secondary operation like welding. This will require a development and certification of the welding process and the weld characteristics. This can be a costly and long term commitment by a company, but the benefit is that it can then be applied to many parts. One issue with welding a metal AM processed part is that there might be an alteration of the microstructure at the weld creating a localized “metallurgical notch” in the part. A “metallurgical notch” is a localized microstructural variation which does not react to load in the same manner as the rest of the part.

The design consideration of new mechanical, physical, corrosion, etc. properties is directly related to the recent discovery that the physical and metallurgical characteristics can vary greatly depending on the metal and the metal AM process employed. The first solution is to determine whether the metal AM process properties are better or worse than the wrought or cast product properties it is replacing. The obvious ramifications are that if the AM properties are better, then all is well. However, if the properties are worse more work will be required by the design team. As with many fabrication processes, design teams may be able to be selective during the design process. They can choose a specific metallurgical characteristic, usually a mechanical property, which is improved by the metal AM process and pertinent to their design. As a result of this choice, the design team chooses to live with other metallurgical characteristics that are worse than in the other product forms.

In addition, it is well documented that certain metals are easier to fabricate using the metal AM processes. For example, the AM microstructure of titanium 6Al–4V, the most popular titanium alloy, is similar to the wrought product. In contrast, the AM microstructure of the nickel base superalloy Inconel 718, a popular superalloy, appears to be a cross between welding and cast product [50]. Lastly, 17-4 PH is a common martensitic PH stainless steel which usually has a small percentage (< ~5%) of delta ferrite phase in the microstructure. It has been documented that 17-4 PH produced using the PBF metal AM process will have 1-80% austenite phase in the as built product depending upon the powder used [51]. This represents the creation of an entirely different metal that will no longer be acceptable for the original functional requirements.

The second solution for the design team to consider is if the part being made is a new design or a replacement part. It is sometimes easier for new designs to select a different metal or use the new properties from the metal AM process without issue. In contrast, a replacement part has “already been designed” so the required physical and metallurgical characteristics may be set. Since revising part drawings is a cost issue, the metal AM-processed part may be required to satisfy the established requirements. In addition, as noted previously, the metal AM processors are discovering that certain metals are more suitable for the metal AM process. A new design would allow the selection of a metal suitable for both the design and the metal AM process.

The third solution is for the design team to consider if the part being made is a low risk secondary hardware or critical hardware. Low risk secondary hardware such as brackets, etc. are not highly loaded and usually have lower design and material requirements. Therefore, it should be easy for a metal AM processed part to replace the current one. This approach would also allow time to understand the nuances of the metal AM process machine, create a mechanical property database, and reduce the risk of using the metal AM process. Critical hardware are usually subjected to higher loads and have much more stringent design and material requirements. It may take much more time and engineering effort for a metal AM processed part to replace a critical part.

The last solution for a design team would be to establish new design property requirements. This might take some time and effort but in the long run will provide a long term solution that would be applicable to many parts. This solution will be discussed in more detail the following paragraphs.

The last design consideration is a small or absent mechanical property database for design. Most companies rely upon a mechanical property database which provides a statistically derived design allowable to ensure a margin of safety for the design process. At this time, there are no metal AM process mechanical property databases in the public domain. There are proprietary mechanical property databases being developed in private companies, but no database has been published.

Fortunately, the metal AM processes provide a simple solution which is easily incorporated in the build of a part. The best and most obvious solution is to create a limited, but statistically significant database to establish confidence in the metal AM process. One method to do this is to establish a statistically derived correlation to the existing wrought or cast product form of the selected alloy prior to implementation of the metal AM process to fabricate parts. Many companies would consider this approach to be time consuming and costly, but there are long term benefits to having such a database, especially if there is a critical material property in the design.

Another approach would be to perform enough testing to “prove the concept” of using the metal AM process for a part. Then, add X–Y–Z oriented test bars to every build, which is a simple matter. In fact, some build models are being developed with test bars already included in the build. These test bars have a twofold purpose. One, they provide mechanical property data for each specific metal AM-processed part. There is no need for a mechanical property database as each part provides individual mechanical property information. Two, the test bars can be used to continue the development of the mechanical property database.

Lastly, there are independent data management companies that could assist with the metal AM process build data and material property database. One example is the Additive Manufacturing Schema product by Granta Design. Granta Design is well established in material property data management, and this product focuses on AM process build data.

Metallurgical Considerations

Metallurgical considerations are listed in Table 3. New Microstructures, Texturing, Macrostructures, Grain Flow, etc. were discussed in depth earlier in this article in the Design Sources of Failure section of the “Sources of Metal Additive Manufacturing Failures” section. The current product forms have databases for workability, castability, hot and cold formability, brazeability, weldability, etc. There is a need for a database of material AM’ability.

Table 3 Metallurgical considerations

It would be advantageous that the variation of the metallurgical characteristics of the metal AM processes be completely understood prior to implementation of making hardware. A sensitivity study that varies the predicated range of part size, shape, and mass would be beneficial. In the last five years there has been a lot of work in the area of metal AM process control to predict the resultant microstructure in the as built part including process control maps such beam power vs scan speed maps, or temperature vs solidification maps as seen in Fig. 23 [51]. The process control maps provide guidance as to the expected microstructure in the as built part. These maps will also be important for failure analysis during the background data search. A review of the metal AM process build data of the failed part would indicate the position in the process control map, and therefore the expected microstructure. The build data should have enough detail that this analysis could be done at any point during the build and therefore at any position on the part.

Fig. 23

Solidification map showing the influence of solidification rate and temperature gradient on the solidification microstructure [52]

Each build should also have separate X–Y–Z oriented test bars to determine build to build consistency with control samples that provide empirical data on mechanical properties, microstructure, residual stresses, fracture surfaces, etc.

The effect of thermal exposure in manufacturing or service is another metallurgical consideration. There is a design temperature limit for all hardware, but either on purpose or by accident there are times during service or during the manufacturing process after the metal AM build in which this temperature limit will be exceeded. Wrought, cast, etc. product forms have data that provides “safe” thermal exposure limitations and can provide technical justification for an engineering disposition of these events. This data does not exist for metal AM-processed materials, but will be required to assist with the engineering disposition of hardware that exceeds its design temperature limit.

The last metallurgical consideration is the different residual stress magnitude and distribution that form during the build. It is well documented that residual stress can be the source of manufacturing issues and service failures; therefore, the residual stress magnitude and distribution must be understood. Modeling is one way to evaluate and document residual stresses and other thermomechanical factors during the build [52]. Part of the study discussed in the previous paragraph regarding New Microstructures, Texturing, Macrostructures, Grain Flow, etc. could be directed to understand residual stresses. Testing could be performed on the separately built X–Y–Z bars as well as the part. Once the residual stress magnitude and distribution is understood, an intelligent decision can be made to either use the part in the as built condition or perform a thermal treatment to stress relieve the part. If possible, the stress relieve could be incorporated into the build as with EB PBF.

Production Considerations

Production considerations are listed in Table 4. New production control criteria has been getting a lot of attention over the last five years. As the variability in metallurgical characteristics of the metal AM processes were better understood in the last ten years, the need for production control became apparent. The question becomes how much control and at what cost?

Table 4 Production considerations

The simplest approach which has been used in other product forms such as castings or forgings is to establish a “fixed process” agreement. A fixed process agreement basically stipulates that the supplier of the part cannot change anything about the fabrication process of the part without the consent of the buyer. The details of exactly what steps in the fabrication process are fixed is usually a negotiation as some steps are more critical than others. On the other hand, the supplier wants to have some freedom for process improvement and cost reduction. For metal AM processed parts, this approach seems to be a simple well established solution, however, there most certainly will be a discussion of how much of the build model will be fixed.

Other production considerations that usually go along with hardware fabricated using a fixed process include:

• (1) Qualification Testing—First Article Destruct Part(s)

  • a. Qualification testing is performed to verify the design and manufacturing process, and it provides a baseline for subsequent acceptance tests. One or more parts are destroyed during the qualification testing. Tests are performed in all critical areas and directions of the part. Recurrence of this test is usually determined by either a specified number of parts, or time basis, to ensure the fabrication process has not changed.

• (2) Acceptance Testing on Each Part

  • a. Acceptance testing is a test conducted to determine if the part satisfies the requirements of a specification or contract. It may involve chemical tests, physical tests, or performance tests. It also creates a correlation to the qualification testing performed as the acceptance tests are usually a subset of the qualification tests and done in the same areas of the part as the qualification tests.

  • b. The tests are usually performed on integral test bars that are sectioned off the part or separate test bars that are fabricated at the same time as the part. These two concepts are used in the forging and casting industry.

  • c. For metal AM-processed parts, the easiest concept would be:

    • i. Add X–Y–Z oriented test bars to every build of a PBF metal AM process

    • ii. Add integral test bars to every build of a DED metal AM process

• (3) Feedstock Control—Weld wire, powder, etc.

  • a. The necessity for the control of powder size, morphology, flowability, etc. has become established in recent years. Powder control has become an integral and costly part of the PBF metal AM process with various processes available [53].

  • b. Allow reusing of powder?

    • i. Reusing impacts powder morphology, density, flowability, and chemistry, especially oxygen pickup.

  • c. Allow mixing of feedstock metal lots?

    • i. Same chemical composition but different heat/batch number.

  • d. Allow mixing of feedstock lots of different chemical composition?

    • i. Example: Mix Titanium 6Al–4V extra low interstitial (ELI) powder with used Titanium 6Al–4V powder to bring oxygen level into specifications limits.

  • e. Allow different alloy powder feedstock in the same machine?

    • i. The question is whether it is possible to clean out the PBF metal AM process machine bin of all previous powder. The risk is high for contamination of the next part. If not, then there is a production constraint as a machine is now limited to one type of metal alloy.

    • ii. Powder contamination control could include requiring the same alloy to be used for fabrication tooling such as storage containers, funnels, etc.

    • iii. Should not be an issue with wire feed DED

Automated build programs have become more common in the recent years and are a production consideration. The automated build programs were designed to make using the metal AM processes easier. The programs can automatically make simple shapes like test bars while the more completed programs automatically incorporate stilts into the build models. Production control is about part to part consistency, but as these programs get smarter how much will they alter the production build?

The second production consideration is secondary processing such as heat treatment, welding, forming, machining, surface modification, coatings NDT, etc. The effect of the secondary processes on the metal AM processed part will have to be determined as well as the appropriate test to perform to show conformance. Some of these secondary processes such as heat treating and welding will require decisions as to the proper time to perform qualification and acceptance testing. The common practice today is to wait until all processes that can affect the metallurgical characteristics are completed prior to qualification and acceptance testing. These secondary processes may also affect the feedstock selected.

Quality Considerations

Quality considerations are listed in Table 5 and are closely related to the production considerations noted previously. The first and most important quality consideration is the definition of what is a lot. The definition of a feedstock lot, AM process lot, production lot, manufacturing lot, etc. is the basis for production control. More importantly to a failure investigation, the definition of a lot establishes the suspect or at-risk group of parts.

Table 5 Quality considerations

Lot definition is usually defined by each industry, sometimes each company, though many times the definition of a lot is not consistent. Due to its importance in determining the suspect group of hardware, the failure analyst must not assume what lot means, but determine it. The ISO/ASTM 52900 international standard “Additive manufacturing—General principles—Terminology” was released in 2015 replacing ASTM F2792 and provides a definition for powder lot and manufacturing lot [2]. Both of the specification were late to the AM universe and their usage may not be widespread to date. However, specifications of this kind prove to be useful when a common specification is desired by users and suppliers.

The definition of a feedstock lot for metal AM processes may have many derivatives. The first definition will be the original/virgin powder or weld wire lot supplied by the powder or weld wire manufacturer with a certificate of conformance. Feedstock lot definition, and therefore control, will become more complicated with the amount of reusing/recycling and blending/mixing of original/virgin powder lots, or blending of virgin lots with recycled lots, due to changes in chemical composition and powder characteristics as discussed in the production considerations section previously. Possible derivatives could be “recycle feedstock lot” or “blended feedstock lot.” The ISO/ASTM 52900 does not have a definition for feedstock lot, but has one for powder lot which is similar to first definition above, the original/virgin powder supplied by a powder manufacturer and has a certificate of conformance.

Another possible lot definition would be to define each build as an AM lot. This is a simple definition and would include all parts made in one build from one feedstock lot. Of course, an important part of the AM lot would a clear and detailed definition of the feedstock lot.

Production lots are a common concept and consist of all parts that are subjected to a secondary process at the same time such as heat treat lots, surface modification lots, coating lots, etc. A production lot will probably consist of many AM lots as most heat treat furnaces are large in comparison to the PBF metal AM process machines and therefore contain many AM processed parts during a production run.

A manufacturing lot is the most restrictive lot. This term would be all the parts made in one AM build from one feedstock lot that were subjected to all secondary manufacturing processes at the same time. This is an important lot designation for failure analysis as the parts in this lot would be the most similar in metallurgical characteristics if there were a failure. The ISO/ASTM 52900 standard has this same definition.

The second quality consideration are the controlling specifications. Current product forms such as wrought or cast product have material specifications for different product forms. In addition, there are forging or casting process specifications. There are also specifications for secondary processes such as heat treatment, surface modifications, coatings, etc. All of these specifications usually include process requirements and conformance testing requirements. The metal AM processes will require the creation of metal AM process specifications. In addition, the secondary process specifications will probably have to be revised to include metal AM processed hardware procedures. Many of the current controlling specifications that are in use are company proprietary specifications. Industry specifications will be revised or written once more data is in the public domain.

Supplier Management Considerations

Supplier management considerations are listed in Table 6 and start with the basic fact that the selected metal AM process company may disappear. The AM universe is constantly changing at this time. AM companies are being bought, or are buying other companies. Large corporations and venture capitalists are entering into the AM universe by buying up smaller AM companies or companies that supply feedstock. These companies acquire the business but do they acquire the expertise? In addition, some AM companies are going out of business. As with many manufacturing suppliers, it is a very distinct possibility that a metal AM process company could disappear for a variety of reasons.

Table 6 Supplier management considerations

The first solution, as with any fabrication process, would be to find multiple suppliers. If possible, the suppliers should all be using the same metal AM process for a potential consistency of product. If this is not possible, the next solution would be to bring the metal AM process in house. This is not a small consideration due to the cost of the metal AM process machines and support equipment. Interestingly, unlike many manufacturing processes like heat treatment and coatings, there has been a strong trend to bring the metal AM processes in house over the last ten years.

One cost and/or schedule benefit for companies that are replacing a part is that the metal AM process becomes competition for the conventional wrought or cast product it is replacing. Competition always leads to reduced cost and/or schedule.

Another supplier management consideration is that the metal AM processes produce small quantities only. The engineering and supplier management team must consider that the production rate of metal AM processes will be small and control the number of part numbers and quantities of each part number to ensure a consistent supply of parts. If a higher production rate is required, then there are three alternatives:

• (a) a second supplier using the same metal AM process could be used,

• (b) a second supplier using another metal AM process could be used,

• (c) the metal AM process could be brought in house.

There are engineering, cost, and schedule considerations for each of these three alternatives which have already been discussed.

The next supplier management consideration is that companies want multiple suppliers for a part. Sole source suppliers are persona non grata in most companies who wish to have at least three suppliers of any part for cost and schedule control. The issue is that the various suppliers using the various metal AM processes will produce parts with similar physical characteristics such as dimensions and surface roughness. However, the various metal AM processes have variability in metallurgical characteristics as noted previously in this article. This situation is very similar to the “Ship of Theseus” thought experiment.

Due to the variability of the various metal AM processes even mechanical properties will not be the same. For example, the metal AM processes will all satisfy the specification minimum for mechanical properties, but the typical mechanical properties will vary with the metal AM process. These variations in metallurgical characteristics could produce differences in service life. It will be up to the engineering specifications to ensure the different metal AM processes create parts that satisfy the design requirements and service life.

Therefore, a failure analyst must be aware that when the same part is supplied from multiple metal AM process suppliers, they will not have the same metallurgical characteristics. This issue also happens in other products such as fasteners, forgings, and castings.

The last supplier management consideration is cost and schedule variation. When the metal AM processes were first being evaluated and used for hardware in the early 2000s, the parts were being used in the as built condition for optimum cost and schedule. However, the metal AM industry has added more metals alloys to the metal AM processes, some of which have complicated metallurgy such as the PH stainless steels and nickel base superalloys. In addition, the AM industry wants more consistent and improved mechanical properties and metallurgical characteristics. The result of these engineering wants is the addition of secondary manufacturing processes such as homogenization, HIP, heat treatment, stress relieving, machining, etc. The result of adding these secondary manufacturing processes is cost increase and longer schedules.

Failure Investigation Considerations

Failure investigation considerations are listed in Table 7 and center on the basic concept of altered failure modes and mechanisms. Current product forms like wrought bar and plate, or castings have developed a large database on how these product forms fail under tensile, compression or torsional loading at various temperatures and in various environments. There is information and databases for overload, fatigue, wear, or corrosion failures. In addition, the fracture surfaces associated with these failure modes and mechanisms are well established and documented.

Table 7 Failure investigation considerations

As noted throughout this article, the established databases for the other product forms do not appear to apply to metal AM-processed parts. As noted earlier, simple tensile tests of AM processed Titanium 6Al–4V did not produce fracture surfaces that are the same as the wrought form [49, 54]. In addition, the variation in microstructure of the metal AM processed parts may alter their susceptibility to specific failure mechanisms and/or service environments. Therefore, when a failure of a metal AM processed part occurs, how does the failure analyst know what is “expected or normal” and what is not?

As with all failure investigations, the failure analyst must ensure that the part was fabricated correctly. The common sources of failure for wrought or cast products are well established and documented. Once again, this is not true for the metal AM processes.

The first option is to speak with the metal AM process supplier or the company that made the metal AM process machine. Hopefully, they will have answers to some questions or assist in finding the answers:

• (1) How should parts made from the metal AM process fail?

  • a. They should have some knowledge of specific sources of failure, failure modes and failure mechanisms associated with their metal AM process.

• (2) What are the potential flaws? And what are the fatal flaws?

  • a. All manufacturing product forms have flaws, imperfections, etc. specific to them. And some of these flaws are fatal to the product. Metal AM processed parts have specific sources of failure such as densification, swelling, balling, unfused powder, etc., and the supplier should be able to identify the fatal flaws associated with their specific metal AM process.

• (3) How do you control your feedstock?

  • a. As with all product forms, the potential sources of failure go back to the metallurgical characteristics of the part [41]. For metal AM processed hardware, this will start with the feedstock.

• (4) What NDT methods were employed to inspect the part?

  • a. Is the NDT method appropriate to find the fatal flaws of the particular metal AM process or was it used because that is what was done on the previous product form?

  • b. What is the probability of detection of the specified level on inspection for the metal AM process?

Failure analysts working with metal AM processed parts must understand the new failure modes and mechanisms for the metal AM processes. It is well documented that for current product forms like wrought bar and plate there is a range of the sizes and shapes which will have similar failure modes and mechanisms. It is also well documented what sizes and shapes do not have similar failure modes and mechanisms. This inherent knowledge will have to be developed for the metal AM processes. The key will be to discover which metal AM processes produce similar microstructures because failure modes and mechanisms are influenced by microstructure, especially ones that occur at lower stress levels like fatigue. The worst case scenario, and a distinct possibility, is that each metal AM processed part is unique.

Once again, the metal AM processes provide a simple solution which is easily incorporated in the build of a part by the simple addition of X–Y–Z oriented test bars to every build. As noted before, these test bars would provide data for the development of the mechanical property database, residual stress database, microstructure database, etc. They could also be used to provide information on the fracture surfaces of an exemplar tensile sample. If this concept is taken to the next step, additional samples could be added to each build that would be allocated for part specific testing based on the anticipated failure mechanisms of the part. This would provide test data and exemplar fractures of the anticipated failure mechanism.

Lastly, the failure analyst must determine the X–Y–Z orientation employed during the build cycle for the failed metal AM processed part. This is a simple question, but should not be overlooked. There is a general consensus, but that does not mean everyone follows it. In addition, the orientation of any test bar data must be confirmed. Once again, ISO/ASTM 52921 defines this orientation, but many published papers do not follow the specification [55].

Summary

Additive manufacturing is a new technology and the same critical thinking and basic concepts for a failure investigation must be used. The questions to be answered include, but are not limited to:

Was the manufacturing process appropriate for the design and application?

• Was design appropriate for the application?

• Were the operating conditions properly anticipated?

• Were the capabilities and limitations of the process recognized and planned for in the design?

• Did AM build raw material meet specifications?

• Did the AM build and secondary manufacturing meet specifications?

These same questions apply to all failure analyses, but as this article has tried to inform, are not precisely the same for AM-processed hardware as for traditional wrought and cast materials. Thus, critical thinking about the AM process during, and prior to, the failure investigation process is paramount to successful use of AM and successful failure investigations of AM-processed hardware.

This article has discussed potential sources of failure that are currently known or should be anticipated for AM-processed hardware. In addition, the article discussed the effect of the metal AM processes on the failure analysis process, as well as the effect of the metal AM processes on specific failure analysis steps. Lastly, this article has discussed areas of consideration when using metal AM processes and provided some potential solutions.