34 ISE Magazine | www.iise.org/ISEmagazine

The ﬁnal step toward quality

in additive manufacturing

Analyzing and ﬁxing problems on the ﬂy in AM can smooth out defects, variances

By Prahalada Rao

The ﬁnal step toward quality

in additive manufacturing

Analyzing and ﬁxing problems on the ﬂy in AM can smooth out defects, variances

By Prahalada Rao

April 2019 | ISE Magazine 35

Humankind knows three ways to make things:

Eroding away material from a bulk raw material to

create a shape, called subtractive manufacturing or

machining; changing the shape of a bulk material

by either rolling, forging, extrusion, casting and the

like, called formative manufacturing; and joining

bits of material to make the ﬁnal shape, called additive manu-

facturing or 3D printing.

Additive manufacturing (AM) is a collective term for a host

of processes wherein the part is built typically by joining to-

gether material layer-upon-layer. There are four main ways in

which the layers can be stitched together:

• Supplying heat with an energy source, such as a laser, elec-

tron beam or ultrasonic vibration

• Gluing the material using an epoxy or photopolymer bind-

• Initiating a photochemical reaction in layers of thermoset-

ting plastic resins and biological materials

• Thermal initiation of polymerization of thermoplastics

This article summarizes the challenges and opportunities for

industrial engineers in the additive manufacturing domain to

address and reap the rewards from the technology given our

ability to understand, explain and mitigate its varied forms. In-

dustrial engineers, given our ability to identify, isolate, monitor

and control variation, have the unique opportunity to usher a

qualify-as-you-build paradigm in additive manufacturing. The

key idea is to manufacture a part with zero defects by monitor-

ing and correcting process faults through data collected from

multiple sensors inside the machine. If this concept of qualify-

as-you-build can be realized, it will take additive manufactur-

ing into the proﬁtable realm of industrial-scale production as

opposed to the prototype-demonstrator role the technology is

largely conﬁned to currently.

AM’s advantages, unique capabilities

From a scientiﬁc perspective, additive manufacturing is unique

because the microstructure of the part and its shape are cre-

ated incrementally and simultaneously. In other words, the mi-

crostructure obtained in an AM part is inﬂuenced by the part’s

geometry, as well as the process parameters; that microstructure

in turn inﬂuences the part’s functional integrity, such as fatigue

life. This singular ability to shape the geometry and structure of

the part comes with advantages and disadvantages.

The advantages of AM as an enabling technique stems from

the following so-called freedoms, some of which were ﬁrst enu-

merated by Hod Lipson with Melba Kurman in the book Fabri-

cated: The New World of 3D Printing.

• Freedom of design complexity. It takes the same effort

to make a square hole as a round hole, conformal internal

channels, steep overhang features and lattices. For instance,

Figure 1 examples (a) and (b) respectively show an Inconel

turbine blade and a titanium alloy spinal implant. The tur-

bine blade has a conformal cooling channel while the spinal

implant has a complex lattice-like structure.

• Freedom of material selection. A wide range of ma-

terials can be made on the same machine. For instance, it

is conceivable to switch from processing of a conventional

material, such as steel, to a high-temperature nickel-based

super alloy by adjusting a few parameters, such as laser pow-

er. In Figure 1, image (a) on the left shows the turbine blade

made using laser powder bed fusion (LPBF) with confor-

mal cooling channels; in (b), a titanium alloy spinal implant

made using LPBF shows complex lattice-like structure.

• Freedom to tailor the material structure to the ap-

plication. The structure and composition of material can

be varied across the part, called functionally gradient struc-

tures, leading to unusual bulk properties, such as negative

coefﬁcient of thermal expansion.

• Freedom of scale. No new ﬁxtures are needed if the part

has to be increased or decreased in size of features in the

design alterations.

• Freedom from waste due to worn tools and mini-

mized energy consumption. Unlike subtractive ma-

chining, the AM process does not require a cutting tool

to contact the material, which then wears due to chemical

and thermal phenomena. Consequently, there is no cost of

regrinding worn tools or pollution from use of coolants.

• Freedom from set batch sizes. Because no special-

ized tooling has to be built, the ﬁxed cost per unit is small;

hence, the break-even point is smaller because production

does not have to be spread over a large batch. Furthermore,

because all machines are identical in an AM-oriented facil-

ity, there are no bottleneck machines, in the parlance of

theory of constraints that dictate production. If a machine

FIGURE 1

AM allows design complexity

An Inconel turbine blade (at left) and a titanium alloy spinal

implant were created with laser powder bed fusion, a process

in which metal powder is raked or rolled across the bed and

selectively sintered using a laser.

36 ISE Magazine | www.iise.org/ISEmagazine

Additive manufacturing builds opportunities for ISEs

breaks down, identical machines take the load without hav-

ing to adjust production.

• Freedom from waste of energy and materials. The

amount of material and energy needed is magnitudes small-

er. For instance, the buy-to-ﬂy ratio in AM – i.e., the ratio

of material processed to the ﬁnal weight of the part – is as

small as 7 to 1 compared to 20 to 1 with traditional machin-

ing.

• Freedom from assembly. The number of component

parts can be reduced by several orders of magnitude. The

General Electric LEAP engine nozzle built using laser pow-

der bed fusion has zero subassemblies compared to more

than 18 parts in the earlier design, while being 25 percent

lighter and reducing the number of welds from 25 to ﬁve.

• Freedom from variability due to breakdowns and

maintenance. Because there is no variability due to tool-

ing, controlling the input material is sufﬁcient to reduce

variability. Furthermore, the uniformity of machines, re-

gardless of part design, has a great impact on the reliability

and maintainability aspects of the production line. Instead

of maintaining several service parts and employing repair

technicians for various types of machines, a small group of

personnel and replacement items is sufﬁcient to keep pro-

duction at pace.

These advantages herald a paradigm shift and accompanying

challenges in a host of manufacturing-related domains, rang-

ing from modeling, materials, metrology, sensing, analytics and

logistics, among others. Indeed, there is a case to be made for

AM having a consequential and possibly disruptive impact on

the entire industrial engineering body-of-knowledge beyond

manufacturing.

Impending challenges of AM

Despite these groundbreaking abilities, an underlying challenge

to additive manufacturing is the uncertainty and variability in

the ﬁnal part properties. The uncertainty in part properties is

related in terms of Figure 2, which shows an identical part built

under various orientations with laser powder bed fusion (LPBF)

in which metal powder is raked or rolled across the bed and

selectively sintered using a laser.

Of the seven parts labeled A through G shown in Figure 2,

only D and G were built defect-free. The other ﬁve parts each

depict a unique build failure. This inconsistency and variation

in properties is the main cause hindering the application of AM

parts beyond the prototype-demonstrator roles.

The variability in AM processes can be narrowed to the fol-

lowing four reasons:

1. Material purity-related part inconsistency. In pow-

der-based processes, such as powder bed fusion and directed

energy deposition, foreign material inclusions and residue

from previous builds can contaminate the powder feed-

stock. As a consequence, the part will have a heterogenous

microstructure, thus affecting the functional integrity. In

Figure 3 (a) the panel shows an X-ray computed tomogra-

phy (XCT) scan of an LPBF Inconel 625 part with tungsten

impurities that appear as bright particles. These are poten-

tial sites for cracks.

2. Machine calibration errors. The spot size of the laser

relative to its position on the bed may vary in powder bed

fusion. Likewise, the powder might not be evenly distrib-

uted if the bed is inclined. Indeed, AM researchers often

report variation in identically shaped parts processed under

identical conditions across the build plate. In Figure 3 (b),

the LPBF part shown has failed to build due to contact with

the recoater mechanism as powder is raked across the pow-

der bed. This failure likely occurred due to deformation of

the build platen.

3. Improper selection of process parameters. Process

parameters such as laser power, hatch spacing, layer height

and scan speed in LPBF determine the mechanics of de-

fect formation. For instance, a high laser power will lead

to uniform, spherical-shaped pores due to vaporization of

material; this is called keyhole porosity. In contrast, if the

laser power is insufﬁcient to melt the material, uneven and

irregular-sized pores are observed; these are called lack-of-

fusion porosity. There are more than 50 different variables

in the LPBF process alone. It is not humanly possible to

screen and model all these variables using statistical design

FIGURE 2

Quality variance is a challenge

The throughput of AM can be severely limited due to the

sensitive nature of the process. Of the seven different orientations

of the same part geometry, and built under identical process

conditions, only two, D and G, were completed without any

visible defects. Each of the rest of the ﬁve parts had different

types of failures.

April 2019 | ISE Magazine 37

of experiments. In Figure 3 (c), a change in hatch spacing

during LPBF of the Inconel 718 parts led to appearance of

lack-of-fusion porosity.

4. Ill-considered part design. The part feature geometry

determines the direction and magnitude of heat ﬂow in the

part, called heat ﬂux. The thermal history of the part in

turn is directly responsible for the variation in microstruc-

ture, called microstructure heterogeneity; as a consequence,

the functional properties, such as surface ﬁnish of the part,

will vary. The titanium knee implant shown in Figure 3 (d)

has a steep overhang. Due to the poor transfer of heat in the

overhang section, the part shows coarse surface ﬁnish and

microstructure heterogeneity.

Of the four reasons for variability, two further stratiﬁcations

can be made based on the following reasoning. The ﬁrst two,

material and machine calibration errors, are closely related to

control and kinematics of the input materials and machine, re-

spectively. They can thus be termed as “small v” variation. In

contrast, the last two factors, related to the process factors and

part geometry, govern the process dynamics and can be called a

“big V” variation.

Unless both the big V and small v variations are mapped and

controlled, industries – especially those such as aerospace and

defense where adherence to speciﬁca-

tions is contractually mandated – will

remain reluctant to use AM-produced

parts in mission-critical assemblies. In

the early 1980s, the fatal crash of an

F/A-18 aircraft during test ﬂights was

eventually attributed to the fatigue fail-

ure of a high-temperature alloy com-

ponent in the jet engine that was made

using the then-newly adapted powder

metallurgy processing technique. As a

consequence, the use of powder met-

allurgy for making aerospace compo-

nents from this particular material was

drastically reduced, which in turn af-

fected the proﬁtability of the powder

metallurgy sector for more than half a

decade.

Such past incidents have understand-

ably instilled a degree of caution and

wait-and-see mindset for AM tech-

nology in strategic industries such as

aerospace and defense. As one produc-

tion manager from a leading defense

contractor said in referring to the 1986

Challenger space shuttle accident, “No

one wants to have the next 3D-printed

O-ring disaster.”

To overcome these quality-related challenges, the current

tack taken by manufacturers is to narrow down the process

parameters by building simple-geometry test artifacts and sub-

sequently characterizing their properties through ofﬂine tech-

niques, such as XCT and destructive materials testing. Given the

cumbersome nature of the AM process, the prohibitive expense

of metal powder and the time and cost of ofﬂine testing, such

an empirical marching-through-parameter-space strategy is not

feasible.

Unless this overly cautious status quo for certiﬁcation in a

high-value, low-volume industry such as aerospace changes, the

implementation of new materials tailored for AM is estimated to

take well over a decade despite the signiﬁcant advantages.

Lastly, an inherent consequence of the layer-by-layer ap-

proach to manufacturing is that if a defect is not detected and

corrected in the same layer it occurs, it is liable to be perma-

nently sealed in by subsequent layers.

AM opportunities for industrial engineers

A solution to overcome these quality-related impediments in

AM is to qualify the integrity of each layer as it is being built us-

ing data acquired from sensors built into the process. The 2013

National Science Foundation workshop in AM introduced the

term “certify as you build,” wherein the integrity of the part is

FIGURE 3

Causes of AM inconsistency

The four main reasons for variation in AM parts, illustrated in the context of laser powder

bed fusion. (a) An XCT scan of Inconel 625 part shows inclusion of unfused tungsten

particles. (b) A recoater crash due to deformation of the platen. (c) A lack of fusion

porosity in Inconel 718 parts due to change in hatch spacing. (d) Coarse microstructure

and surface ﬁnish in the part results in a steep overhang.

38 ISE Magazine | www.iise.org/ISEmagazine

Additive manufacturing builds opportunities for ISEs

seconded not through postprocess characterization but through

a record of sensor signature patterns acquired during the build.

To avoid legal implications, certify as you build is now supplant-

ed by the term “qualify as you build.”

Other incarnations of qualify as you build include, Born

Qualiﬁed from Sandia National Labs; In-process Quality As-

surance, a registered trademark of Sigma Labs in New Mexico;

and Certiﬁed Additive Manufacturing of Materialize Inc.

To realize the qualify-as-you-build paradigm requires fun-

damental understanding of each link in the AM process chain,

a research need that intrinsically encompasses the knowledge of

the industrial engineering profession: Process conditions, pro-

cess phenomena, process signatures, part microstructure (de-

fects), process control (rectiﬁcation) and part performance.

• Process conditions. Parameter settings, material contam-

ination, part design, machine errors

• Process phenomena. Vaporization, incomplete fusion,

melt pool instability, thermal gradients

• Sensor data signatures. Melt pool thermal proﬁle, melt

New 3D printing techniques

travel at the speed of light

Researchers at two U.S. universities have incorpo-

rated the use of light to reﬁne 3D printing process for

manufacturing.

A University of Michigan research team has devel-

oped a new printing process that lifts complex shapes

from a vat of liquid rather than building up plastic ﬁla-

ments layer by layer. The result creates products up to

100 times faster than conventional printing processes,

eliminating the need for expensive molds to create

identical items.

The method solidiﬁes liquid resin using two lights

to control where the resin hardens and where it stays ﬂuid. This enabled the team to solidify the resin in more sophisticated patterns and

create a 3D bas-relief in a single step rather than in a series of 1D lines or 2D cross-sections. They have produced a lattice, a toy boat

and a block “M” from the school’s logo design.

“It’s one of the ﬁrst true 3D printers ever made,” said Mark Burns, professor of chemical engineering and biomedical engineering who

co-led the program with Timothy Scott, an associate professor of chemical engineering.

“You can get much tougher, much more wear-resistant materials,” Scott said.

The university has ﬁled three patent applications and Scott is preparing to launch a startup company. A paper describing the research

will be published in Science Advances, titled, “Rapid, continuous additive manufacturing by volumetric polymerization inhibition pat-

terning.” (read about their efforts at https://link.iise.org/3Dlightprinting).

In a similar effort, researchers at the University of California Berkeley have devised their own no-layers additive manufacturing pro-

cess that relies on light to transform polymers into complex solid objects. As at UM, instead of building layer upon layer, the Berkeley

method concurrently prints all points within a three-dimensional object by illuminating a rotating volume of photosensitive material with

a dynamically evolving light pattern.

The device, Computed Axial Lithography (CAL), nicknamed “Replicator” by the inventors, selectively solidiﬁes a photosensitive

liquid within a contained volume. Light energy is delivered to the material volume as a set of two-dimensional images. Each projected

image propagates through the material from a different angle. The superposition of exposures from multiple angles results in a three-

dimensional energy dose sufﬁcient to solidify the material in the desired geometry.

CAL also is scalable to larger print volumes, and reportedly is several orders of magnitude faster than layer-by-layer methods. Re-

searchers can print features as small as 0.3 millimeter in acrylate polymers, as well as print soft structures with exceptionally smooth

surfaces using gelatin methacrylate hydrogel. The team has created a series of objects, including a tiny model of Rodin’s “The Thinker”

and a customized jawbone model. The device can make objects up to 4 inches in diameter.

“I think this is a route to being able to mass-customize objects even more, whether they are prosthetics or running shoes,” said

Hayden Taylor, assistant professor of mechanical engineering at UC Berkeley and senior author of a paper describing the printer.

The UC Berkeley research team’s 3D printer works by shining

changing patterns of light through a rotating vial of liquid.

Credit: Hayden Taylor, University of California Berkeley

April 2019 | ISE Magazine 39

pool shape, spatter pattern extracted from heterogeneous

in-process sensors such as thermal and high-speed cameras,

photodetectors

• Part microstructure/defects. Pinhole pores, acicular

pores or distortion caused by the above phenomena

• Process control. Changing the process parameters, scan

strategy, part design and support structures to compensate

for defects

• Part performance or quality. Fatigue life and surface

ﬁnish

The research challenges and needs are stratiﬁed per the

material, part, process and enterprise-levels, and enlisted in

Figure 4.

From qualify as you build

to correct as you build

The qualify-as-you-build approach integrating sensing and

closed-loop process con-

trol in AM is being in-

tensely studied. However,

a correction-based strategy

might be needed to tran-

scend process sensing and

control. Process control

due to inherent delays in

the feedback and param-

eter adjustment loop may

be too slow to counteract

the faster thermomechani-

cal phenomena that cause

defects.

Thus, despite process

sensing and control, a build

might still be scrapped due

to defects. Instead, a cor-

rective approach in the

LPBF is being studied at

University of Nebraska-

Lincoln by the author’s

team. The key idea of this

approach, called “correct

as you build,” is to cor-

rect defects such as porosity

by leveraging the intrinsic

phenomena of the process

to remelt previously depos-

ited layers.

For instance, once lack

of fusion porosity is de-

tected in a layer by sensors

built into the machine, the

subsequent layers can be deposited at a higher energy den-

sity, leading to melting of unfused particles. For defects such

as cracking and pinhole porosity not liable to be corrected

by remelting, a hybrid additive-subtractive approach can be

used.

The recently acquired hybrid metal AM systems acquired

by University of Nebraska-Lincoln (Matsuura Avance 25

and Optomec Hybrid system) have an integral subtractive

machining head, which can be used to remove a defect af-

ﬂicted layer. Through this hybrid AM approach it is possible

to envision a correct-as-you-build paradigm beyond qualify

as you build to ensure defect-free parts. 

Prahalada Rao is an assistant professor for mechanical and materials

engineering at the University of Nebraska-Lincoln. He is an IISE

member. The author thanks the NSF for funding his work through

the following grants CMMI-1719388, CMMI-1739696 and

CMMI-1752069 (CAREER) at University of Nebraska-Lincoln.

FIGURE 4

Challenges

Research needs

Material-level challenges and research needs

• Material cross-contamination.

• Heterogeneity of particle sizes.

• Proprietary compositions and single source

suppliers tied to each OEM.

• Change in powder material properties due to

recycling.

• Precise control of particle size, mixtures and

composition.

• Creation of functionally graded materials and

alloys.

• Safety benchmarks and standards for handling

powders and minimizing fire and inhalation

hazards.

Part-level challenges and research needs

• Freeform surface geometries are difficult to

measure with coordinate measurement

machines.

• Polishing and post-process machining of free-

form geometries is done manually and is

expensive.

• Large amount of point cloud data is generated

with optical measurement techniques – new

algorithms are needed to synthesize this data.

• Design rules and support structure

optimization.

• Non-destructive evaluation and post-process

quality assurance.

• Post-process finishing to improve surface and

geometric integrity of free-form surfaces and

aid in the removal of supports.

• Standardization of test procedures, geometric

dimensioning and tolerancing (GD&T) and

metrology of AM parts.

Process-level challenges and research needs

• Process speed is not par with mass production

technique. Laser and optics have a limited life

and replacement is expensive.

• Careful calibration of process machine

elements to ensure repeatability.

• Need to track multiple process variables.

• OEMs restrict changing of process parameters.

• Thermal physics is complex and modeling is

restricted to part deformation, as opposed to

microstructure.

• Process development to increase the speed and

volume of the part produced.

• Efficient and accurate modeling and

simulations to anticipate potential problems

• In-process sensing, monitoring and control to

ensure parts are produced to specification.

• Standardization of best practices, such as post-

process cleaning.

• Design of the optics and machine elements, and

automated generation of tool paths to increase

efficiency.

Enterprise-level challenges and research needs

• Suppliers, design bureaus and facilities with

AM systems are spread across different

regions. Systems types and capabilities vary

across facilities.

• Several systems are connected to the cloud or

the OEM servers.

• Technicians steeped in conventional

manufacturing are ill-versed in handling of

powders.

• Logistics and supply chain implications of AM

• Cyber security to defend against intrusion of

the digital thread in AM, and protection of

design intellectual property.

• Human-factors and safety implications of AM.

• New hands-on education and training

approaches to convey the principles of AM

processes to the next generation of users.

AM research challenges, needs