The 2014 Wohlers Report found that the AM market reached $3.07 billion in 2013, representing a 34.9% growth rate, the highest growth rate in 17 years.  And over the past 26 years, the average growth rate in worldwide revenue from AM was 27% [1].  In 2013, the McKinsey Global Institute released a report naming AM as among the technologies most likely to transform the world [2].  Without a doubt, AM is quickly becoming a strong segment of the manufacturing economy on a global scale; however, market penetration of AM products, specifically in aerospace markets, is limited by the lack of robust and mature inspection and validation technologies compared to traditional subtractive manufacturing parts.

Recently, NASA has been promoting the development of AM as a tool for the next generation of space flight.  In fact, astronauts aboard the International Space Station (ISS) have already begun printing parts, such as threads, springs, clamps, buckles, and containers using a 3-D ABS printer [3].  The use of 3-D printing in space overcomes a large logistics hurdle, removing the need to be reliant on launch facilities on Earth and the requisite launch window opportunities and risks associated with supplying replacement parts to astronauts aboard the ISS.  But more than just replacing a broken screw, NASA wants to push for even more AM in space, which could remove size and weight restrictions placed on satellites and structures built on the Earth.

The current process of launching material into space must take into account the tremendous forces applied by the cargo, and because satellites and probes can cost millions of dollars, there is an onus on the engineer to overdesign to ensure launch survival.  But the overdesign comes at a cost of a higher launch weight, and at the going rate of $10,000/lb to launch an object into space, adding extra material just to survive launch can quickly increase one’s launch cost.  NASA believes that AM in space can circumvent this issue by needing only to transport bulk material (such as that shown in Figure 1) used to build a structure in space that is optimized for the space environment, not the launch environment.  Nevertheless, due to the lack of ability to certify AM parts and critical structures, there is no desire to take a chance in the risk-averse world of space flight.

But NASA is making efforts to close the gap to take advantage of the benefits of AM.  NASA created its Nondestructive Working Group (NNWG) to help coordinate interagency cooperation on developing standards for AM inspections, including new standards produced by ASTM.  The NNWG helps researchers target information and technology gaps and directs resources to bridge these gaps.

Factors for Determining Inspectibility

Design Complexity
Before discussing the state of inspection technologies, let us first examine the types of parts that can be produced through AM, as well as different AM techniques and the defects seen in the AM process, all of which guide the inspection selection process.  Todorov et al. [5] defines a five-step evolution of design complexity that is based on the skill growth and increased technological comfort of the engineer designing a part.

Figure 2 charts the growth of the designer as he/she becomes more comfortable with AM.  Group 1 sees relatively simple parts that can typically be fabricated using traditional machining.  Parts produced in this group have surface features that can be easily accessed and can be served through traditional nondestructive evaluation (NDE) technology.  Often, parts in this group are produced as a proof-of-concept or rapid prototype, and because of the simplicity of the manufacturing techniques, these parts are not seen as economically viable when compared to traditional subtractive manufacturing parts.

Group 2 parts begin to take advantage of AM’s ability to produce more complex shapes and designs than traditionally fabricated parts without the need for complicated tooling processes.  The example in Figure 3 comes from a 2013 GE Aviation crowd-sourced competition to find ways to reduce the weight of a standard forged titanium engine mounting bracket [6].  The original bracket weighed 2,033 g (4.48 lbs), but the AM redesign was able to reduce the weight by 84% while maintaining an equivalent performance in lab tests.  Group 2 parts mark the start of cost savings from subtractive manufacturing by reducing the need for excess materials and complex tooling.  However, the addition of complexity comes with the cost of narrowing the technologies available to perform NDE on the part unless specifically made for the part.  Generally speaking, Groups 1 and 2 are not dissimilar enough from subtractive manufactured parts that they require any new or specialized inspection technologies from what is already available.

Group 3 AM components are defined as parts that cannot be manufactured through traditional subtractive manufacturing.  These parts feature internal structures such as tubes or channels that previously would have necessitated the part to be made through casting.  In a traditional setting, these parts would have multiple individual subcomponents manufactured and then an assembly phase to produce the final component.

Figure 4 shows an injection molding tool (note especially the cooling channels moving through the component).  The tight channels within the part increase the cooling efficiency of the tool, allowing for faster production rates.  However, these embedded features represent a challenge to the inspectibility of the part and reduce the NDE technologies to those that can image the interior features.

Group 4 parts can potentially be produced through traditional methods; however, the cost and skill required to produce the designs make the operation economically unfeasible.  The engineer begins to incorporate organic and nonlinear shapes that place emphasis on performance rather than producibility.  In addition, the internal structure of these parts (as shown in Figure 5) can be complicated and produced without the need for traditional “line of sight” to create the features.  And the ability to inspect these parts is greatly reduced due to the presence of highly detailed and embedded features.  

Group 5 parts, which are almost entirely produced through AM, consist of extremely fine features.  Examples of these parts include metallic lattice structures (as shown in Figure 6).  These lattice structures can be tailor made to suit specific purposes and can include thousands of individual nodes in a relatively small space.  The complexity of these structures requires a long fabrication time, but this fact is offset by the structures’ potential to reduce material costs while maintaining the strength-to-weight ratio of bulkier forbearers.  A byproduct of the increased complexity of these parts is the lack of NDE technology that can provide a reliable validation of the part for use in larger systems.  Developing NDE technology for Group 5 parts would allow the parts to reduce costs overall at a system level, meaning the cost of manufacturing the entire final deliverable is reduced even with the increased cost of fabricating the Group 5 part.

AM Processes
AM covers a wide range of processes, depending on what type of material one is using.  Simple, inexpensive in-home 3-D printers tend to use spools of polymer wire that are melted and deposited layer by layer.  The plastic parts formed in these machines are often the only experience the general public has with AM.  Although these parts are certainly novel and exciting, they are not well suited for industrial or structural use.  For industrial purposes, the main form of AM for metals comes in the way of Powder bed fusion (PBF) systems, with the layers being joined either through the use of selected laser melting (SLM) or electron beam melting (EMB).  In both instances, layers of metal powder are deposited on the printing platform and then melted by either the laser or the electron beam, with the process repeating itself over and over until the part is completed.

While the two processes are similar, the subtle differences between SLM and EBM can impact the final product.  EBM has a higher energy density and scanning rate, and thus a faster build rate, with the tradeoff coming in the form of a poorer surface finish as compared to SLM.  Because EBM also requires the printing tray to be preheated prior to use, the thermal gradient in the part is minimized, resulting in a lower residual stress in the final product.  However, EMB is limited to standard metallic materials, while SLM ‘s range of materials includes metals, ceramics, and polymers.  Table 1 provides additional comparisons between the two systems.

Table 1. Comparison of Electron Beam Melting and Selective Laser Melting Traits [5].

Characteristic	Electron Beam Melting	Selective Laser Melting
Thermal Source	Electron Beam	Laser
Atmosphere	Vacuum	Inert Gas
Energy Absorption	Conductivity Limited	Absorptivity Limited
Scan Speed	Extremely Fast, Magnetically Driven	Limited by Galvanometer Inertia
Energy Costs	Moderate	High
Surface Finish	Poor to Moderate	Moderate to Excellent
Feature Resolution	Moderate	Excellent
Materials	Conductive Metal	Polymers, Metals, Ceramics
Beam Size	100–500 µm	100–150 µm
Powder Size	45–100 µm	20–50 µm

Defects Found in AM Parts
The PBF approach, whether laser-based or electron beam-based, is the most common form of AM manufacturing seen in the aerospace industry.  In PBF-manufactured parts, there are typically four classes of defects that can occur:  (1) volumetric defects, (2) cracking and delaminations, (3) balling, and (4) surface roughness.  These defects are typically the result of poor process controls, process parameters, or even the geometry of the part to be produced, though it should be noted that even the most stringent of process controls will not entirely prevent the formation of defects in AM parts.  As with traditional subtractive manufacturing, these defects can be detrimental to the performance of the part, and therefore there is a great importance placed on the inspection process to find the defects before the part becomes compromised.  Understanding the nature of the defect types is necessary to implementing the proper quality monitoring process and inspection technique for the finished part.

The most common defects seen in AM parts are volumetric defects, either porosity (as shown in Figure 7) or a lack of fusion of the powder material.  Generally speaking, porosity is described as being spherical in shape while defects formed by a lack of fusion can be more irregularly shaped and may have unmelted powder material within them.  Gong et al. [9] found that beam power and scanning speed are the main drivers of porosity and lack of fusion in AM parts.  They discovered that at a given beam power level, a low scanning speed will produce porosity, while an excessively high scanning speed will produce a lack of fusion in the material.  Thus, to minimize the occurrence of volumetric defects in AM parts, operators must find the “Goldilocks” zone of scanning speed for a specific beam power and a specific material.  Fortunately, powder suppliers have conducted extensive research in this area and provide the necessary parameters to manufacturers to mitigate the risk of volumetric defects.

Cracks and delaminations make up the second class of defects.  These defects are more in line with the traditional defects seen in subtractive manufacturing and are the result of internal thermal stress gradients produced through the additive process (as shown in Figure 8).  As each layer of powder heats and cools, the thermal stresses can grow, leading to the AM part delaminating from the substrate or cracks growing between the layers.  This type of defect is more readily seen in structures with low geometrical stiffness, such as thin-walled tubes.  Of the two processes, delaminations and cracks are more often seen in SLM parts, as EBM systems use a heated production tray to reduce the thermal gradient in the part as it is being constructed.

Balling, the third class of defect (shown in Figure 9), occurs when instabilities cause the melt pool to break into thin spherical droplets.  This defect derives from problems of the liquid metal wetting in its solid form [11].  In these cases, the surface tension of the newly melted powder exceeds the wettability of the underlying layer, in much the same way that water beads up on a hydrophobic surface.  Because the molten powder resolidifies on the order of milliseconds, subsequent layers are built around the balling defects, leading to compounded defects as the part grows.  Moreover, layers built around the balling can experience interlayer loss of adhesion due to the reduced surface contact, while the volume occupied by the sphere itself can grow into a volumetric defect.

Although the last class of defect, surface roughness, is not inherently considered a defect in AM, it does have a bearing on the types of NDE that can be performed on finished parts.  AM parts are built by taking computer-aided design (CAD) models and then slicing them into consecutive layers, which are translated into reality through the 3-D printer itself.  Due to this layer stacking, any nonhorizontal, nonvertical face will be rough and give a stair-step-like appearance.  And the junctures of the stair step features can create sharp corners, which are ideal for stress concentrations that can lead to part failure.  Figure 10 compares SLM- and EBM-produced specimens to a traditional cast specimen.

Stroffregen et al. [13] found that when comparing AM parts against traditionally made test specimens of steel, the rougher surface of AM parts can be the site of initiation of fatigue cracks and the primary reason for fatigue failure in those parts (as shown in Figure 11).

Stroffregen also found that the mean deviation for surface roughness for AM parts (Ra) averaged 13.7 µm and the maximum height of the roughness profile (Rz) was 80 µm, compared to machined parts having respective roughness parameters of 0.2 µm (Ra) and 1.7 µm.  At 10⁷ cycles, the AM parts had a max stress of 219 MPa while the machined parts had a max stress of 49a MPa.  Surface roughness is a byproduct of the build process, and Figure 11 illustrates how much of an effect the build process can have on the overall performance of the final part.  Care must therefore be taken to minimize the surface roughness of AM parts, whether through tight process controls and slow build times or by post-production machining to refine the surface and eliminate crack initiation points.

Inspecting AM Parts

In-Situ Monitoring
The NDE of AM parts occurs in two forms, in-situ monitoring and post production inspection.  In-situ monitoring is important as a first-look capability for process control.  Unfortunately, in-situ monitoring is fairly limited in the types of systems that can be used.  The most widely used in-situ monitoring system involves using near infrared (NIR) cameras to capture the temperature gradient between the newest layer of melted material and the previously formed layers.  NIR cameras are able to detect areas where insufficient beam energy imparted on the powder bed has resulted in a “cold” spot where the powder has not completely melted.  As discussed previously, these locations of poor melt can produce volumetric defects in the finished parts.  NIR camera systems can be improved to include multiple cameras, real-time tracking, and feedback algorithms, which can help improve the weld consistency in AM (as has been seen in the manufacturing of stainless steel straight wall samples) [3].  Going beyond simple monitoring, the parametric information provided by the NIR cameras (temperature, shape, and cooling rate) can be analyzed in real time to create metrics for feedback and real-time control of the system.

Recently, researchers at Penn State have examined the use of optical image analysis to perform layerwise in-situ monitoring of AM [10].  The research team focused on using the layerwise monitoring as a means to correlate the anomalies seen in post-production 3-D X-ray computed tomography (CT) scans to features seen in the images between layer melts.

Figure 12 demonstrates how the individual layer images can be stacked to produce a 3-D CAD model that maps the locations of defects.  This type of in-situ monitoring is useful in improving the process controls by identifying where in the build the defects generate, and therefore measures can be taken to eliminate the source of defects before they can affect the build process.

Moving beyond imagery techniques, one of the more promising techniques being developed is in-situ ultrasonic (UT) monitoring of the build process.  In-situ UT can be used to monitor the laser power in SLM machines, with the A-scans allowing an inspector to infer conclusions about the quality of the SLM process.  Reider et al. [14] describe the process of using in-situ UT when producing Inconel 718, a nickel alloy used for aero engine components.  The UT monitoring system used a four-channel transmitter and receiver system with a bandwidth ranging from 400 kHz up to 30 MHz, a sampling rate of 250 MHz and 14-bit resolution, the ability to perform 1,000 A-scans every second, and a temporal resolution of 4 ns. The SLM process was monitored in a layerwise fashion with simultaneous visualization of the radio frequency (RF) signals.  Because in-situ monitoring is still a relatively recent development, the parts manufactured for testing were simple test cylinders, with each one having intentional defects added to the build process in the form of spherical and half spherical voids made of nonmelted powder.  During the build, the voids were clearly seen in the scans, thus indicating that the SLM process can be used to fabricate calibration blocks.

In-situ UT can also be used to monitor the single-layer fusion process by comparing the time-of-flight of the ultrasonic signal and the build time.  This technique takes advantage of the ability of AM to produce nominally consistent layer thicknesses during the build time.  In this instance, the average layer thickness was 40 µm (see figure 13), meaning that for a part with a total thickness of 20 mm, the build time is approximately 90 min.

In another test, the researchers varied the laser power to monitor the effects on the microstructure of the Inconel test part.  Taking advantage of the high numbers of A-scans the system could record, researchers plotted the scans against the build time to view areas of low beam power, which resulted in areas of high and low porosity.

Verification of the in-situ UT monitoring was conducted in the post build phase using CT scans.  As seen in Figure 14, the aberrations seen in the UT B-scan align neatly with the porosity imaged by the CT scan.

A newer form of UT in-situ inspection is also in development using laser ultrasonics (LUT).  LUT works by using a pulsed laser beam to generate a transient ultrasonic wave in the solidified layer.  The waves then interrogate the layer for defects and arrive at the point of detection.  The resulting surface displacement is then detected with a separate laser-based receiver.  As the beams scan along the layer during production, the signal detected at each position is acquired, and the signals are combined to form a B-scan image that can be interpreted with advanced, automated signal and image processing algorithms to determine the integrity of each layer [15].

Figure 15 shows a defect and the corresponding signal used to create the defect profile. By applying a threshold level, seen as the yellow line in the right image, the detection of defects can be an autonomous process.  LUT in-line monitoring is still in development, but if the technology is able to mature, it has the potential to ensure that all finished AM parts will be qualified without the need for further inspections.

Post-Production Inspection
In-situ monitoring can be a powerful tool for monitoring process control and preventing large-scale batch poor builds; however, this monitoring does negate the need for post-build inspection of parts.  While most of the post-build inspection of AM parts is identical to the inspection processes of subtractive manufacturing, the method of inspection is often found to be a greater function of the complexity of the AM part.  For instance, penetrant dye testing (PT) is often used to find surface cracks in traditionally made parts.  However, because AM relies on the stepwise layer slice build-up of the part, the surface roughness is often greater than with subtractive manufacturing.

PT is based on using capillary action to draw the dye into the crack, whereby the excess dye is removed from the surface and an ultraviolet light is shone on the part, illuminating any dye that has become trapped in the cracks.  The surface roughness of the AM part presents multiple opportunities for small cracks to form between the layers as the part is built up, thus making it an almost insurmountable task to use PT on an as-built part, at least without first performing post-processing machining and polishing.

Beyond examining surface cracks with PT, AM parts can be inspected using Process Compensated Resonance Testing (PCRT).  PCRT is used in the automotive, aerospace, and power generation industries.  To conduct the test, the AM part is excited at its resonance frequency and the frequency shift is analyzed to determine whether or not the part is acceptable.  PCRT has been employed in evaluating in service engine blades.  If the mass and the stiffness of the part is known, the process is fast and reliable.  However, PCRT is considered a global test and does not provide the location of any defects, thus making it a good gatekeeper test with the ability to identify the parts that have no defects or the parts that need additional inspections.

The need for fast first-look testing is important given that the most widely used inspection method is X-ray CT.  Industry has been using CT inspections since 1972, and this method has proven its effectiveness by allowing inspectors to detect the exact position of the defect within the body of the part.  In a CT scan, a radiation source transmits X-rays through the part to a collector, where the images are compiled and reconstructed by a computer to create a 3-D image.  CT scans can be a powerful tool, able to reach further into a part than other NDE methods, no matter the complexity.  The resolution of the CT scan is dependent on the power of the scanner.  As the power of the beam increases, the depth of the beam penetration increases, which means that a more powerful beam is able to scan a denser part.  This is not to say, however, that a low-powered scanner should not be used in the inspection process. Inspection of less dense parts can employ a low-powered scanner with a radiation source with a small emitter size and can achieve resolutions down to the submicrometer scale [5].  For AM parts produced through powder beam methods, defects are expected to be on a smaller scale, and therefore submicrometer detection is a powerful asset.

Figure 17 shows an example of the power of CT scans.  Inspectors are able to detect and locate all instances of porosity in the test cube.  CT scans are considered the best post-production inspection method for AM parts up to Group 4 complexities; and if microfocused CT scans are employed, even Group 5 complexity parts can be inspected by this method.  However, the higher inspection capabilities do come at a cost.  CT scanning equipment is expensive and needs a radiation source to power the beam.  CT scans also produce high volumes of data and therefore need intense computing power to return results in a timely and useful manner.  A Group 2 part might take 10 min to process a few gigabytes of data using dual multicore processors, but as the parts become more complex and the number of welds becomes higher, the scan analysis quickly becomes an operation that can take hours to perform.  If complex AM parts are to become more prevalent in everyday use, the ability to inspect the parts quickly and accurately is going to be the limiting factor.

As with in-situ monitoring, LUT is also showing promise in post-production inspections.  The benefit of the LUT as opposed to CT devices is the lack of a radiation source and the subsequent infrastructure needed to support it.  This means the LUT systems can be less expensive and therefore more available to manufacturers.

Figure 18 demonstrates the work performed by Levesque et al. in applying LUT to post-production inspection of AM parts.  The Inconel piece was scanned from the substrate underside in the span marked by the arrows.  The scans detected a slight Heat Affected Zone (HAZ) as well as indications of possible discrete porosities in the thicker areas of the part.  LUT is still being developed but has the potential to work in conjunction with in-situ processes and other post-production inspection methods to decrease the time needed to perform inspections on complicated AM parts.

Conclusion

AM has not been widely adopted in the aerospace industry because there is a lack of standards and methods for easily and quickly qualifying parts for flight.  Simple AM parts can be qualified by the same methods as traditionally made parts; however, as the complexity of the part grows, the ability to inspect it becomes limited.  Current developments for in-situ monitoring seek to impose stricter process control as a first step to mitigating the formation of defects, while post-production inspections can provide a final certification of the part for use.  As more improvements in the methods are developed, the aerospace industry will be more likely to employ AM parts in greater numbers and increase the economic impact of 3-D printing through new and innovative designs and less material needed to produce those designs.

References: